Impact of Water Users Associations on Water and Land Productivity,
Equity and Food Security in Tajikistan 
Ted Horbulyk and Soumya Balasubramanya
Final Report Volume II: Technical Appendices
September 2018
ii
Prepared for the United States Agency for International Development, USAID Grant Number AID-BFS-G￾11-00002.
Recommended Citation:
Ted Horbulyk, Soumya Balasubramanya. July 2018. Impact of Water Users Associations on Water and 
Land Productivity, Equity and Food Security in Tajikistan. Final Report Volume II: Technical 
Appendices. Colombo, Sri Lanka: International Water Management Institute.
IWMI Contact:
Dr. Soumya Balasubramanya
International Water Management Institute
PO Box 2075
Colombo, Sri Lanka
Telephone: +94 11 288 0119
Email: S.Balasubramanya@cgiar.org
The International Water Management Institute (IWMI) is a non-profit, scientific research organization 
focusing on the sustainable use of water and land resources in developing countries. IWMI works in 
partnership with governments, civil society and the private sector to develop scalable agricultural water 
management solutions that have a real impact on poverty reduction, food security and ecosystem 
health. Headquartered in Colombo, Sri Lanka, with regional offices across Asia and Africa, IWMI is a 
CGIAR Research Center and leads the CGIAR Research Program on Water, Land and Ecosystems (WLE).
www.iwmi.org
Front cover photograph: Madeline Dahm/IWMI.
iii
Contents
Appendix 1...................................................................................................................................... 1
Appendix 2.................................................................................................................................... 38
Appendix 3.................................................................................................................................... 69
Appendix 4.................................................................................................................................... 92
Appendix 5.................................................................................................................................. 113
Appendix 6.................................................................................................................................. 129
iv
1
Appendix 1 
Balasubramanya, S.; Price, J.; Horbulyk, T. 2018. Impact assessments without true baselines: 
assessing the relative effects of training on the performance of water user associations in 
Southern Tajikistan. Water Economics and Policy (online first). Available online at:
https://doi.org/10.1142/S2382624X18500078
2
Title: Impact assessments without true baselines: assessing the relative effects of training on the 
performance of water user associations in Southern Tajikistan
Keywords: water user associations, performance, difference-in-difference, propensity scores, Tajikistan
Abstract: Conducting rigorous evaluations of whether the process of creating new institutions affects 
their performance of mandated duties presents several challenges. Not only is assignment to process 
often not random, but when the process of creating new institutions starts, outcomes and other 
performance-influencing covariates are not measurable because the yet-to-be created institutions are 
not functioning at baseline. This paper compares the performance of 74 ‘treated’ water user 
associations in Tajikistan that were created using a longer training process with 67 ‘control’ water user 
associations that were created using shorter training, to assess the impact of training on WUA 
performance of mandated duties. First, propensity scores were constructed to estimate the probabilities 
of being ‘treated’ by treatment status. These results guided the application of the difference-in￾difference technique with right-hand side covariates in a context where field measures of outcomes and 
other performance-influencing covariates were made after the new institutions were created and 
functioning. The first measures were taken within 12-18 months of the new institutions being functional 
and the second measures were taken 24 months after the first. This choice of methods introduces a bias 
due to measurement error causing an underestimate of the treatment effects, while controlling for 
biases due to time-invariant and time-varying unobservables. An alternative method that only compared 
the differences in outcomes at a single point in time after the new institutions were created would have 
provided an inaccurate estimate of the effects of the intervention. This is a context where methods such 
as synthetic controls are impossible to employ due to the nature of the intervention, other 
macroeconomic structural changes, and severe data restrictions. The methodology employed here 
generates evidence that, while biased towards generating an underestimate of effect, can still be useful 
and informative for policy and management purposes, and for evaluating the impact of process on the 
functioning of new institutions in transition settings. 
1. Introduction
Institutions for water management are rapidly transforming in Central Asia, as larger political and 
economic changes take place. Water user associations (WUAs) are now the legally mandated form of 
organization that provides irrigation services to agricultural land in Tajikistan (Republic of Tajikistan, 
2006). Some of these WUAs were created using a longer and more gradual process than others. The 
3
recent literature on institutional reform, which has focused on process, rather than design (Denyer Willis 
and Mota Prado, 2014) suggests that WUAs that have been created through a longer process may be 
more successful in performing their legally mandated functions than those created with a shorter 
process (Yap-Salinas, 1994 and Mukhtarov et al., 2015).
Quantitative evidence on whether longer training affects performance is important for shaping 
government programs that are creating new and strengthening existing WUAs (e.g., the ongoing 
Tajikistan Second Public Employment for Sustainable Agriculture and Water Resource Management 
Project).1 However, testing the effects of training on WUA performance in Tajikistan is complicated by 
the fact that the length of training was not randomly assigned. Observable characteristics of the 
geographies where WUAs were created; and unobservable characteristics of the same geographies that 
either are constant or are varying over time may have determined selection into longer and shorter 
training, consequently confounding the effects of training. 
While well-established approaches for constructing an appropriate counterfactual in the absence of 
randomized assignment to an intervention are available, they require treatment and control units to 
exist in the pre-intervention period. Two more popular methods consist of synthetic controls (Abadie 
and Gardeazabal, 2003) and propensity scores (Rosenbaum and Rubin, 1985). Both methods rely on 
using pre-intervention data (either covariates and outcomes, or covariates alone) to select control units 
that look similar to the treatment units on observed characteristics. For example, Abadie et al. (2010) 
studied the effects of a tobacco control program in California by constructing a control group comprised 
of a weighted average of other US states that did not have any tobacco control programs using 19 years 
of pre-program data on covariates and outcomes. Rubin (2001) used propensity scores to design studies 
to estimate the causal effects of smoking and tobacco-company misconduct by matching smokers and 
never-smokers on pre-intervention covariates that were not affected by the intervention. 
When treatment and control units do not exist during the pre-intervention period, experimental 
economics (especially laboratory experiments) could be used to construct counterfactuals; however, 
these methods are generally used to study strategic behavior in response to exogenous or endogenous 
policy changes. For example, Tellez Foster et al. (2017) conducted laboratory experiments to examine 
groundwater pumping decisions under alternative energy subsidy scenarios. In their study, all 
 
1 Project documents may be found at : http://projects.worldbank.org/P133327/tajikistan-second-public￾employment-sustainable-agriculture-water-resources-management-project?lang=en.
4
participants play a status-quo round first, after which ‘treatment’ participants were randomly assigned 
to play one of the alternative situations, and ‘control’ participants were randomly assigned to play 
another status-quo round.2
Fischer et al., 2003 constructed ‘treatments’ to examine altruistic behavior 
under differential resource growth rates in the presence/absence of intergenerational links in laboratory 
settings. 
This paper examines whether WUAs created via longer training processes perform their legally 
mandated functions as well as, or better than, those created via training processes with a shorter 
duration. This occurs in a non-strategic setting and context where WUAs did not exist before training 
began. The paper considers the population of WUAs in 20 districts of Southern Tajikistan and employs a 
difference-in-difference (DID) technique with right-hand side covariates. Since WUAs did not exist in the 
pre-training period, the DID technique is used with both sets of data collected in two time periods after
the WUAs were created. This modified DID approach departs from the standard approach where one set 
of observations would have been collected at the time when WUAs were created and the other later. 
Though the modified approach introduces a bias in the estimated effects of training (explained further in 
methods section), this approach is able to control for time-invariant selection effects. This would not 
have been possible if the standard approach were used in this context, since all measures of WUA 
performance at the time of WUA creation would have taken values of zero.3
Including covariates that 
affect performance on the right-hand side may help control for unobservable effects that vary over 
time. 
This choice of methods is guided by the context of the intervention in question. Synthetic controls and 
experimental methods could not be used here due to the non-existence of pre-training data on 
covariates and outcomes; and due to the non-strategic setting. WUAs created with longer and shorter 
training components were established in geographies with gravity irrigation schemes of similar 
characteristics that were not influenced by the treatment;4 however, they were created in different river 
basins (see results section). This assignment of WUA program groups across different river basins likely 
 
2 Tellez Foster et al. (2016) also conducted field experiments and compared results between the lab and field 
settings.
3
Since WUAs did not exist at the start of the intervention, they were not performing any mandated functions. 
4 These gravity irrigation schemes were constructed before 1991 in the Soviet era. Since the collapse of the Soviet 
Union, no expansion of gravity schemes has been undertaken. Tax revenues are not sufficient to cover the costs of 
such an expansion, and any expansions would require loans. Development banks such as the World Bank and the 
Asian Development Bank have been keen to increase the efficiency of existing surface water schemes before new 
schemes are constructed, or existing schemes are expanded.
5
lowered the costs of the WUA creation program for the government (which created WUAs using a 
shorter training period) and the United States Agency for International Development (which created 
WUAs using a longer training period); but needs to be controlled for when assessing impacts, in order to 
address selection effects. 
This paper contributes to the literature in three key ways. First, there are many cases in natural resource 
management where units of treatment are created by a new program or intervention; consequently, 
units did not exist in the pre-intervention era. Such cases are often not considered as plausible 
candidates for quantitative assessments due to difficulty in controlling for selection effects and 
constructing a baseline. Consequently, many assessments of institutional transformations are qualitative 
(Theesfeld, 2004; Zhou, 2013;Mustafa et al., 2016). Quantitatively assessing performance of new 
institutions in their early years can yield important evidence that contributes to adaptive management, 
but also supports construction and testing of theories of change for the future. The modified DID 
technique with right hand side covariates used in this paper can be used to estimate impacts (albeit with 
an underestimate of effects) when treatment and control units do not exist before the intervention 
begins.
A second contribution of the paper is to explore the size and direction of the resulting bias in 
performance estimates that is associated with the (requisite) delay in collecting the first set of data. 
Since both treatment and control units did not exist in the pre-training era, most outcomes of interest 
take values of zero at the start of the intervention. With the modified DID technique, the resulting 
estimate is biased. However, it is less likely to be confounded by time-invariant selection effects than a 
simpler estimate that only examines the difference in the performance of the treatment and control 
groups at one point in time after the intervention is completed. 
A third contribution of this paper is the choice of performance indicators it implements. Rather than 
using standards of performance such as efficiency and equity that may be more suitable while assessing 
established institutions, the paper uses indicators to reflect the functions that the WUA law (Republic of 
Tajikistan, 2006) requires WUAs to perform. This approach draws from Nagrah et al. (2016) who 
examine the correlation between farmers performing mandated tasks, and scheme characteristics and 
management perceptions. An examination of mandated functions in the case of new institutions can 
yield important information for adaptive management with longer term effects.
6
The results suggest that WUAs created using longer duration training processes perform mandated 
functions better than WUAs that were created using shorter training processes. Perhaps the most 
significant empirical result is that WUAs created using longer training recovered membership fees from 
19% more of their members; and they were also 10% more likely to hold board meetings for planning 
activities before the start of the irrigation season. These results contribute to the literature on 
institutional design by providing new evidence on the relationship between length of training and 
performance of mandated functions. From a policy perspective, longer training can increase the 
functionality of new WUAs that are being created in Tajikistan. 
A limitation of this paper is that it is not possible to test formally whether the outcomes for WUAs 
created with longer and shorter training would have followed the same trend in the pre-training period. 
WUAs did not exist before the training began, and consequently were not functional. The modified 
difference-in-difference technique followed in this paper to control for time-invariant unobservables; 
and the use of covariates that affect performance on the right-hand side to control for time-varying 
unobservables, is employed in a context where many other standard methods of controlling for 
selection effects are unsuitable. The empirical methods used here are easy to apply during the early 
years following creation of new institutions. They provide timely data-driven information on the way 
institutions are functioning, especially in contexts where it is not possible to separate the intervention 
from the units of intervention. 
The remainder of the paper is organized as follows. Section 2 provides a context for examining WUAs in 
Tajikistan. Section 3 presents a conceptual framework and reviews literature concerning related aspects 
of WUA performance globally and in Central Asia. Section 4 describes the methodology, while Section 5 
presents the summary statistics and results of the quantitative assessment. Section 6 aggregates the 
results and discusses them in relation to the existing literature and the broader context of evaluation of 
water institutions. 
2. Context
Land-locked Tajikistan lies in the semi-arid region in Central Asia. Over 70% of its population lives in rural 
areas (United Nations, 2015), with almost one-third living below the national poverty line (World Bank, 
2015). Agriculture—especially cotton cultivation—remains the main source of gross domestic product. 
Around 95% of crop production takes place on irrigated land (FAO, 2012). Cotton and wheat cultivation 
is predominant in Southern Tajikistan; in the Soviet era, this was Central Asia’s main cotton-producing 
7
hub (FAO, 2012). Tributaries of the Amu Darya River flow into the fertile Khatlon Province in the 
southwest, home to the majority of the country’s population.
The introduction of WUAs in Tajikistan is best considered in the context of distinct periods that 
characterize the country’s history. During the Soviet era, large collective institutions dominated the 
political and economic scene. An extensive system of irrigation canals was constructed as part of the 
Soviet economic plan to provide water to collective farms (cooperative and state enterprises; O’Hara, 
2000). When the Soviet Union collapsed in 1991, ‘Tajikistan did not have chance to become a state 
before it descended into political violence and civil war’ (Matveeva, 2009). The Civil War (1992-1997) 
undermined public services and infrastructure, especially irrigation services.
After the civil war, collective farms were decollectivized into dehkan (private) farms. The district 
irrigation departments—called the vodkhozes—which were responsible for providing irrigation services 
to the Soviet collectives were not able to cope with the challenge of providing water to the thousands of 
private farms. A lack of Soviet subsidies and the departure of Russian irrigation specialists compromised 
clarity on jurisdiction and roles, budgets for maintenance and operations, and the availability of 
technical skills (Gunchinmaa and Yakubov, 2009; Shahriari, 2009). As one response, the government 
enacted the WUA Law (Republic of Tajikistan, 2006) and named WUAs as the institution that would 
henceforth be responsible for delivering irrigation services to the dehkan farms. International assistance 
to create these new institutions was also requested. Consequently, WUAs were piloted in the late 2000s 
by several international organizations, and a countrywide program to create new WUAs gained 
momentum in 2011-12. 
In Southern Tajikistan, WUAs were created in 2012 and 2013.5 The WUA boards and their members 
were provided training from international agencies and the national government in water and financial 
management and in conflict resolution. However, one key difference among these new WUAs is the 
duration of time over which they were trained when they were created (USAID, 2014; ADB, 2012). 
Specifically, WUAs created through a longer training process were sponsored by USAID, while those 
created through a shorter training process were sponsored by other international agencies. On average, 
USAID WUAs were created with 1.89 years of training, whereas non-USAID WUAs were created with 
 
5 All WUAs were officially registered by the agency creating them, about one to two years before the community￾level process of creating them began. Consequently, the registered age of the WUA may be greater than the 
number of years of its physical existence. 
8
0.73 years of training (Balasubramanya et al., 2016: 44).6 The content of the training provided to both 
groups of WUAs was similar, in part because USAID-designed WUA training materials were used by the 
government as well.
3. Conceptual framework 
3.1. Review of performance assessment literature
Irrigation systems provide a prime example of common pool resources (CPRs). The constituent users of 
irrigation systems—water and infrastructure—have the defining characteristics of being difficult to 
exclude, and are subject to competition among themselves when seeking to access the subtractable 
resources (Hamidov et al., 2015). Rivalry among water users may be due to inadequate water supply, 
issues of water distribution or timing of water delivery. Given the distinct features of CPRs, scholars have 
theorized about which forms of institutional design are most likely to enable community-level 
institutions to perform their legally mandated functions successfully (Kazbekov et al., 2009). 
In her seminal work, Ostrom (1990) offers a blueprint of eight core design principles that would enable 
CPR institutions to exist independently among local users, without being privatized or managed by 
central government. While acknowledging this body of literature (including Meinzen-Dick et al., 2002), 
assessing the extent to which different groups of WUAs fit Ostrom’s CPR institutional design principles 
would be beyond the scope of this paper.7
Such analysis would likely require techniques such as 
ethnographic observation, which involve considerable time. New WUAs may have to be closely observed 
over several years to gauge if the core design principles are correlated with successful performance. 
New institutions for resource management in quasi-democratic and liberalizing political economies 
often faces a number of challenges (Skaperdas, 2001). New bodies may not be perceived as legitimate, 
and low social capital may hamper cooperation and trust (Sehring, 2007; Hamidov et al., 2015). For 
example, in Tajikistan, government officials have underlined the difficulty in introducing monetary 
mechanisms, such as water-related fees, because of a so-called ‘Soviet mentality’ that expects and may 
 
6 Before WUAs were created, irrigation water for farms was supplied by the state irrigation departments; the latter 
cannot serve as an appropriate control group to examine WUA performance of mandated duties because the role 
of supplying irrigation water to the new dehkan farms has been changed and legally handed over to WUAs. There 
are more than 300 WUAs across Tajikistan. 
7
In Tajikistan, all WUAs may align with a number of Ostrom’s design principles. For example, user participation in 
decision-making, monitoring by elected members and accessible conflict resolution mechanisms are key design 
features of WUAs, especially those created by USAID (USAID, 2014).
9
encourage non-payment for services including water (Sehring, 2007; Matveeva, 2009). New bodies such 
as WUAs may find it difficult to perform their legally mandated functions. 
Barcellini et al. (2015) stress that the duration of the process for creating participatory institutions 
warrants investigation because it can influence performance of mandated functions by young 
institutions. This focus on process partially stems from the idea that rapidly applying a universal 
‘blueprint’ design may not fully appreciate the important contextual nuances of community water 
management (Smith, 2008; Mukhtarov et al., 2015; Thiel et al., 2015; Ricks, 2016). Longer training 
periods may foster productive relationships and improved functioning, leading to stronger collective 
action and more serviceable institutions (Nagrah et al., 2016). Empirical studies have also demonstrated 
the importance of training on WUA performance. Batt and Merkley (2010) attributed the shortcomings 
of WUA performance in Egypt to inadequate training. This echoes Mukhtarov et al.’s 2015 case studies 
in Turkey and Azerbaijan, where WUAs were deemed unsuccessful in performing their functions due to 
inadequate preparation for taking over irrigation management responsibilities. In Uzbekistan, Wegerich 
(2000) and Hornidge et al. (2013) reviewed the limited or unclear duration of training provided, which 
they correlated to unreliable water delivery and poor WUA performance. By contrast, Yap-Salinas (1994: 
126) concluded that ‘the longer the project, the greater the chance of success,’ after an eight-year WUA 
project in the Dominican Republic. Also, Johnson and Stoutjesdijk (2008) drew attention to WUAs in 
Kyrgyzstan, which after receiving training over four years were able to improve performance of 
mandated functions, improving water delivery and fee collection. 
3.2. Hypotheses
Short-term analysis of the performance of community based irrigation management organizations is 
usually assessed by considering mandated functions (Yakubov, 2011). For example, collection of 
irrigation and/or membership fees and dispute arbitration functions of WUAs have been examined in 
Jordan, China and Turkey (see Çakmak et al., 2004; Huang et al., 2010;Mustafa et al., 2016). Nagrah et 
al. (2016) consider performance indicators such as maintenance of irrigation watercourses, dispute 
resolution, collection of water charges and success of monthly meetings when explaining functioning of 
water user groups in Pakistan. 
Building upon this literature, this paper considers the following indicators, and introduces a number of 
hypotheses regarding the effects of longer training. A list of indicators is provided in Table 1.
10
a. Water delivery planning: WUAs in Southern Tajikistan are supposed to create water delivery 
schedules, preferably by irrigation season. The WUAs are also supposed to hold at least two 
board meetings every calendar year, where delivery schedules are finalized and other important 
decisions such as setting of membership fees and other water deliver-related challenges are 
discussed. WUAs with longer training are hypothesized to have a greater probability of having 
seasonal water delivery schedules and of holding two annual board meetings than those with 
shorter training.
b. Repair and maintenance of irrigation infrastructure: WUAs are supposed to conduct routine 
cleaning and maintenance of secondary canals (distributaries) and tertiary canals (watercourses) 
at the start of each irrigation season (there are two cultivation seasons every calendar year). It is 
expected that WUAs with longer training are more likely to conduct two pre-irrigation season 
rounds of maintenance annually.
c. Collecting irrigation service fees and transferring them to the government: The WUA is 
mandated to collect irrigation service fees from all members seasonally; and transfer all 
irrigation fees to the government. Longer training is likely to increase the share of members 
from whom irrigation fees are recovered, and the share of collected irrigation fees that is 
transferred. 
d. Collecting WUA membership fees: WUAs are authorized to collect annual membership fees from 
all members to finance day-to-day operations. Longer training is likely to increase the share of 
members from whom the WUA recovers membership fees. 
e. Arbitrating disputes: WUAs are authorized to arbitrate disputes between members, and it is 
expected that WUAs with longer training have a higher probability of being the body that settles 
disputes. 
4. Methodology
The difference-in-difference (DID) technique calculates the effect of an intervention by comparing the 
average change over time in the outcome variable for the treatment group, to the average change over 
time for the control group, in effect controlling for differences in the starting points and common time￾trend effects in one or more measure of outcome between the two groups. 
4.1. The DID estimator treatment and control units exist in the pre-intervention period
11
Consider the (standard) case where outcomes are measured at the start of the intervention. To describe 
the estimator mathematically, standard notation that uses two time-periods (where time is regarded as 
a discrete variable) is used. 
Let the policy maker be interested in understanding the effects of the intervention two years after it was 
implemented (short-term effects).8
Let 𝑡 refer to the number of years since the intervention, where 𝑡 ∈
{0, 2}. Let 𝑇𝑡 be a categorical variable that denotes the two years, with 𝑇2 = 1 if 𝑡 = 2 and 𝑇0 = 0 if 𝑡 =
0. 
Let 𝑑 refer to treatment state, where 𝑑 ∈ {0,1}, where ‘0’ indicates the non-treated/control state, and 
‘1’ indicates the treated state. In this paper, the treatment is the longer duration of training for some 
WUAs. Let 𝐷𝑖 be a categorical variable that denotes the treatment status of WUAi, with 𝐷𝑖 = 1 if 𝑑 = 1
and 𝐷𝑖 = 0 if 𝑑 = 0. 
Let 𝑖 denote a specific WUA, where: 𝑖 ∈ {1, 2, 3, … , 𝑘} when 𝐷𝑖 = 1; and 𝑖 ∈ {𝑘 + 1, 𝑘 + 2, … , 𝑘 + 𝑚}
when 𝐷𝑖 = 0; that is, there are 𝑘 treatment WUAs and 𝑚 control WUAs. 
Consider the following equation: 𝑌𝑖𝑡 = 𝜇 + 𝛾𝐷𝑖 + 𝜃𝑇𝑡 + 𝛽(𝐷𝑖 ∗ 𝑇𝑡
) + 𝜗𝑖𝑡 (1)
9
where 𝑌𝑖𝑡 refers to a specific outcome variable for WUA i at time 𝑡; and 𝛽 reports the causal effect of 
the treatment, which controls for pre-existing differences in outcomes at the baseline.10 Following the 
standard approach (Woolridge, 2002) to describe the difference-in-difference estimator 𝛽̂ : 
𝛽̂ = {𝐸( 𝑌|𝐷 = 1, 𝑇 = 1) − 𝐸( 𝑌|𝐷 = 0 , 𝑇 = 1)} − {𝐸(𝑌|𝐷 = 1, 𝑇 = 0) − 𝐸(𝑌|𝐷 = 0, 𝑇 = 0)} . (2)
Therefore: 𝛽̂ = (
1
𝑘
∑ 𝑌𝑖2
𝑘
𝑖=1 −
1
𝑚
∑ 𝑌𝑖2
𝑘+𝑚
𝑖=𝑘+1
) − (
1
𝑘
∑ 𝑌𝑖0
𝑘
𝑖=1 −
1
𝑚
∑ 𝑌𝑖0
𝑘+𝑚
𝑖=𝑘+1
) . (3)
Alternatively:
 
8
In development practice, short-term effects are generally studied two years after the intervention was 
completed. This is a commonly observed norm, and is not necessarily supported by any theoretical model.
9 The error structure is assumed to follow: 𝐸 (𝜗𝑖𝑡| 𝐷 = 1, 𝑇 = 1) = 𝐸 (𝜗𝑖𝑡 |𝐷 = 0, 𝑇 = 1) = 𝐸 (𝜗𝑖𝑡 |𝐷 = 1, 𝑇 =
0) = 𝐸 (𝜗𝑖𝑡 |𝐷 = 0, 𝑇 = 0) = 0 . This is because the identifying assumption is that, by explicitly accounting for , 
the errors are uncorrelated with 𝐷 and 𝑇 . 
10 The difference-in-difference technique eliminates 𝜇 and 𝛾 because it involves using the difference(𝑌𝑖2 −
𝑌𝑖0 ) ∀ 𝑖. The coefficient 𝜃 controls for any year-related effects.
12
𝛽̂ =
1
𝑘
(∑𝑌𝑖2
𝑘
𝑖=1
− ∑𝑌𝑖0
𝑘
𝑖=1
) −
1
𝑚
( ∑ 𝑌𝑖2
𝑘+𝑚
𝑖=𝑘+1
− ∑ 𝑌𝑖0
𝑘+𝑚
𝑖=𝑘+1
). (4)
The DID technique controls for time-invariant unobservables (𝜇), such as any WUA-specific, area￾specific, or agency-specific fixed effects that are constant over time but may drive differences in level of 
performance.
4.2. The DID estimator when treatment and control units did not exist in the pre-intervention 
period
To demonstrate the DID technique when outcomes for treatment and control groups take values of zero 
at the start of the intervention, time is now defined as a continuous variable 𝑡𝑟
, where 𝑡𝑟 ∈ (0, ∞).
Let 𝐴(𝑡) represents outcomes for the treatment unit, which vary with time, 𝑡. Let 𝐶(𝑡) represent 
outcomes for the control unit, which also vary with time. Due to the nature of the outcome, both 
𝐴(𝑡) ≥ 0 and 𝐶(𝑡) ≥ 0. 
Let the difference in magnitude of outcomes between the treatment and control units at any time 𝑡𝑟 be 
defined as 𝛿𝑟 = 𝐴(𝑡𝑟
) − 𝐶(𝑡𝑟) ∀ 𝑡𝑟 ∈ (0, ∞).
Let the intervention be implemented at 𝑡0 for the treatment unit. Since treatment and control units did 
not exist at the start of the intervention, 𝐴(𝑡0
) = 𝐶(𝑡0
) = 0. Therefore, 𝐴(𝑡0
) − 𝐶(𝑡0
) = 𝛿0 =
0. (5)
Let the policy maker be interested in measuring subsequent performance at time 𝑡𝑒
, where 𝑡0 < 𝑡𝑒
. For 
the purposes of a clearer demonstration, one unit of observation from the treatment group and one 
unit of observation from the control group are considered.
(a) First, consider the case where the treatment unit performs better than the control unit after 𝑡0; 
presented in Figure 1. 
The standard DID estimator would measure changes in outcomes for the treatment and control group 
between 𝑡𝑒 and 𝑡0. That is: 𝐷𝐼𝐷 = (𝐴(𝑡𝑒
) − 𝐶(𝑡𝑒
) ) − (𝐴(𝑡0
) − 𝐶(𝑡0
)) = 𝛿𝑒 − 0 = 𝛿𝑒
. (6)
13
Since treatment and control units did not exist before 𝑡0, and 𝐴(𝑡) ≥ 0 , 𝐶(𝑡) ≥ 0, and 𝐴(𝑡0
) = 𝐶(𝑡0
) =
0, it is not clear if the standard DID technique would eliminate the time-invariant unobservables 
represented by 𝜇 in equation (1).
Suppose now that the first measures of outcomes are made at 𝑡1 > 𝑡0. In this case the ‘modified’ DID 
estimator is: 𝐷𝐼𝐷
′ = (𝐴(𝑡𝑒
) − 𝐶(𝑡𝑒
)) − (𝐴(𝑡1
) − 𝐶(𝑡1
)) = 𝛿𝑒 − 𝛿1 . (7)
Then 𝐷𝐼𝐷 − 𝐷𝐼𝐷
′ = 𝛿𝑒 − (𝛿𝑒 − 𝛿1
) = 𝛿1 > 0. (8)
This implies that 𝐷𝐼𝐷 > 𝐷𝐼𝐷′; that is, the modified DID underestimates a positive effect of the 
intervention. 
Generalizing, let the first measurement of outcomes be made at 𝑡𝑛 such that 𝑡0 < 𝑡𝑛 < 𝑡𝑒
. Then, 𝛿𝑛 =
𝐴(𝑡𝑛
) − 𝐶(𝑡𝑛
). Then, the ‘modified’ estimator is
𝐷𝐼𝐷𝑛 = (𝐴(𝑡𝑒
) − 𝐶(𝑡𝑒
)) − (𝐴(𝑡𝑛
) − 𝐶(𝑡𝑛
)) = 𝛿𝑒 − 𝛿𝑛. (9)
Let the error 𝜀(𝑡) = 𝐷𝐼𝐷 − 𝐷𝐼𝐷𝑛 = 𝛿𝑒 − (𝛿𝑒 − 𝛿𝑛
) = 𝛿𝑛 > 0. (10) 
Since 𝐴(𝑡) and 𝐶(𝑡) are diverging, this implies that 𝐷𝐼𝐷 > 𝐷𝐼𝐷𝑛
. (11)
As 𝑡𝑛 → 𝑡0 , 𝛿𝑛 → 0 . Therefore 𝜀(𝑡) → 0 as 𝑡𝑛 → 𝑡0 .
In this case, if one wishes to assess the effects of the intervention at 𝑡𝑒 with baseline outcomes 
measured after 𝑡0, then measuring farther from 𝑡0 would further underestimate the actual effect of the 
intervention. 
(b) Now, consider the case where the control unit performs better than the treatment unit after 𝑡0; 
presented in Figure 2. 
The standard DID estimator is 𝐷𝐼𝐷 = (𝐴(𝑡𝑒
) − 𝐶(𝑡𝑒
) ) − (𝐴(𝑡0
) − 𝐶(𝑡0
)) = −𝛿𝑒 − 0 = −𝛿𝑒
. (12)
As in case (a), since treatment and control units did not exist before 𝑡0; 𝐴(𝑡) ≥ 0 , 𝐶(𝑡) ≥ 0, and 
𝐴(𝑡0
) = 𝐶(𝑡0
) = 0. Therefore, it is not clear if the standard DID technique would control for time￾invariant unobservables that may be driving performance of the two groups.
14
Consider the ‘modified’ DID estimator: 𝐷𝐼𝐷
′ = (𝐴(𝑡𝑒
) − 𝐶(𝑡𝑒
)) − (𝐴(𝑡1
) − 𝐶(𝑡1
)) = −𝛿𝑒 − (−𝛿1
) =
−𝛿𝑒 + 𝛿1 . (13)
Similar to case (a), and demonstrated in equation (8): 𝐷𝐼𝐷 − 𝐷𝐼𝐷
′ = −𝛿𝑒 − (−𝛿𝑒 + 𝛿1
) = −𝛿1 < 0.
In this case, 𝐷𝐼𝐷 < 𝐷𝐼𝐷′; that is the modified DID underestimates the magnitude of a negative effect of 
the intervention.
Generalizing, let the first measure of outcomes be made at 𝑡𝑛 such that 𝑡0 < 𝑡𝑛 < 𝑡𝑒
. Then, 𝛿𝑛 =
𝐴(𝑡𝑛
) − 𝐶(𝑡𝑛
).
𝐷𝐼𝐷
𝑛 = (𝐴(𝑡𝑒
) − 𝐶(𝑡𝑒
)) − (𝐴(𝑡𝑛
) − 𝐶(𝑡𝑛
)) = −𝛿𝑒 − (−𝛿𝑛
) = − 𝛿𝑒 + 𝛿𝑛. (14)
Similar to case (a) and as demonstrated in equation (10), 𝜀(𝑡) = 𝐷𝐼𝐷 − 𝐷𝐼𝐷𝑛 = −𝛿𝑛 < 0. 
Since 𝐴(𝑡) and 𝐶(𝑡) are diverging, this implies that 𝐷𝐼𝐷 < 𝐷𝐼𝐷𝑛
. (15)
As 𝑡𝑛 → 𝑡0 , 𝛿𝑛 → 0 . Therefore 𝜀(𝑡) → 0 as 𝑡𝑛 → 𝑡0, similar to the result found in case (a).
In this case, if one wishes to assess the effects of the intervention at 𝑡𝑒 with baseline outcomes 
measured after 𝑡0, measuring farther from 𝑡0 would also underestimate more greatly the magnitude of 
the actual effect of the intervention. 
(c) Adding across case (a) and (b)
The net effect of the intervention will be biased under the modified DID technique, with the direction of 
bias unknown, since the modified DID approach underestimates both positive and negative effects. In 
this context, if some WUAs experience a positive effect of longer training duration and others 
experience a negative effect, then summing across WUAs provides biased aggregate measure of 
program effect, where the direction of bias is not known. However, as 𝑡𝑛 → 𝑡0 , 𝜀(𝑡) → 0. Therefore, 
measuring as close as possible to 𝑡0 would minimize the bias, and would address time-invariant 
selection effects. 
5. Data 
15
There are four river systems within Tajikistan—Vakhsh, Kafarnihon, Pyandj (which are tributaries of the 
Amu Darya Basin) and the Syr Darya. Irrigated agriculture is practiced on only 4% of Tajikistan’s land.11
To design a study that was representative of WUA populations in gravity irrigation schemes, ten such 
schemes that supply water to 164 jamoats12 in 20 districts were identified.13 A census of WUAs was 
conducted in late 2014 in these 164 jamoats, and 150 legally registered and functioning WUAs were 
identified. The chairs or managers14 from 141 of the 150 WUAs agreed to participate in the study. Of 
these, 74 WUAs were created with a training period of 22-24 months; and 67 WUAs that were created 
with a shorter training period of three months. 
The chair/manager was requested to donate around 1.5 hours of their time twice to provide the survey 
team with data. The first round of data described the 2014 calendar year and was collected in early 
2015. The second round of data described the 2016 calendar year and was collected in early 2017. 
Respondents were not compensated monetarily for their participation. Instead, manuals on better 
agronomic practices, such as the use of improved seeds and application of fertilizers, were provided to 
be shared with the larger community. 
Questions in the 2014 survey were repeated in the 2016 survey; the difference being the year about 
which the respondent was asked to report. The performance indicators for which data were collected 
reflect the various functions that WUAs in Tajikistan are mandated to perform by the Water User 
Association Law (Republic of Tajikistan, 2006; Table 1). 
In addition to these performance indicators, data on covariates that might affect performance of 
mandated functions were also collected for the 2014 and 2016 calendar years. Data collected included: 
the physical attributes of the WUA such as the type and magnitude of irrigation infrastructure; the area 
covered by gravity and lift irrigation infrastructure;15 the service area under irrigation; the area of 
cotton, wheat and fodder cultivated; the number of members in the WUA; the size of the WUA board 
 
11 About 96% of the land area is covered by mountains, where irrigated agriculture is not practiced.
12 A jamoat is an administrative area within a district (in other countries, the equivalent of a jamoat would be a 
sub-district). A collection of jamoats makes a district, and a collection of districts make a province. Tajikistan is 
divided into four provinces, and 400 jamoats. 
13 Considerable shares of area in other agricultural jamoats were irrigated by lift irrigation schemes, the 
operational costs of which are much higher than gravity schemes. This study covers almost every WUA in a gravity 
scheme that was legally registered and functioning in 2014.
14 If these individuals were not available, then the WUA engineer or treasurer was asked to respond. 
15 Very small areas in only a couple WUAs were irrigated by lift irrigation. This was as expected because the study 
was designed to focus on gravity irrigation schemes.
16
and its gender composition; the number of disputes; and perceptions about the quality of the main, 
secondary (distributary) and tertiary (watercourse) canals. Measures of covariates pertaining to type 
and magnitude of irrigation infrastructure, and the area covered by gravity and lift irrigation 
infrastructure remained constant between the two surveys.16 Measures of other covariates generally 
varied between the two surveys. 
6. Results
Since assignment of WUAs to shorter versus longer training periods was not random, propensity scores 
were constructed by regressing each WUA’s training status (1= longer training; 0=shorter training) on a 
set of covariates. These included: location of the WUA (head/middle/tail) on the canal; the number of 
main, secondary and tertiary canals within the WUA’s command area; the share of the command area 
covered by gravity and lift irrigation infrastructure; the number of pumping stations within the 
command area; the number of drainage collectors within the command area; the number of irrigation 
wells within the command area; the number of years the WUA has been registered (but may not have 
existed on the ground); the district the WUA was located in; and the river that was the source of water 
for the irrigation system. These covariates were chosen because they are unlikely to have been 
influenced by the WUA program. Irrigation infrastructure was developed in the Soviet era, and has not 
expanded since 1990. This is reflected in the fact that in 1994, 720,000 hectares of land were irrigated, 
which marginally increased to 742,000 hectares in 2009 (FAO, 2012).17
Figure 3 presents results for the frequency distribution of the propensity scores associated with longer 
and shorter training, which were constructed using all covariates, except the district and the river. WUAs 
with longer and shorter training had overlapping distributions (indicating common statistical support). 
This implies that both groups of WUAs looked similar in terms of: location (head/middle/tail) on the 
canal; the number of main, secondary and tertiary canals within the command area; the share of the 
command area covered by gravity and lift irrigation infrastructure;18 the number of pumping stations, 
drainage collectors and irrigation wells within the command area; and the number of years the WUA 
had been registered (but may not have existed on the ground). Propensity scores were then constructed 
 
16 No new irrigation infrastructure was developed nor was existing infrastructure expanded between the two 
surveys.
17 The latest year for which official data are available is 2009. 
18 There is almost no use of groundwater, because aquifers are located in deep, hard rock layers, which increases 
pumping costs.
17
by including all of those attributes plus dummies for the districts in which the WUA was located (Figure 
4). The extent of overlap between the frequency distributions is still considerable, but less than in Figure 
3. Finally, when propensity scores were constructed by also including dummies for the river that was the 
source of water (Figure 5), the two frequency distributions had limited common support. Figures 3-5 
suggest that while WUAs with longer and shorter training are not different in terms of their physical 
characteristics, they differ in terms of the river basins in which they were created. Since WUAs with 
longer and shorter training were created by USAID and the government respectively, program costs 
would have been lower with both agencies concentrating their efforts on specific river basins rather 
than each agency covering every river basin. 
The following sections report the results from the modified DID technique with right-side covariates, 
listed in Table 2. There were minor changes between years in the share of irrigated area; the share of 
irrigated area under wheat, cotton and fodder production; and perceptions on the quality of the 
primary, secondary and tertiary canal. Covariates such as the number of registered members in the 
WUA, the size of the WUA board, and the number of disputes varied considerably between 2014 and 
2016. Covariates such as the physical infrastructure, the share of command area served by gravity and 
lift irrigation, the district the WUA is located in, and the river that is the source of water did not vary 
between the two years. Only covariates whose measures varied between 2014 and 2016 are reported, 
because those whose measures did not vary between 2014 and 2016 (such as the physical 
infrastructure, river and district) are eliminated by the DID technique. 
Tables 3 through 7 report the effects of having longer training duration on WUAs’ performance of 
mandated duties, while including covariates (Table 2) that may also affect performance. In each table, 
the reported values express the modified DID estimate as a percentage difference, followed in 
parentheses by the standard error.
6.1. Water delivery
WUAS with longer training were 4% more likely to have a seasonal water delivery schedule in 2016, and 
were 9% more likely to have at least two board meetings in 2016, but these results are not statistically 
significant at the usual levels (Table 3). A 1% increase in area under cotton cultivation between 2014 and 
2016 was associated with a 28% increase in the probability of having a seasonal water delivery schedule 
in 2016 (p <0.05). This is consistent with the observation that cotton is a water-intensive crop that is 
cultivated in the dry season (summer). Also, a 1% increase in the area under wheat production between 
18
2014 and 2016 was associated with a 27% decrease in the probability of having a seasonal irrigation 
schedule in 2016 (p < 0.1). This is consistent with the observation that wheat is a winter crop, which is 
mostly rainfed, with minimal supplemental irrigation. One additional member on the board between 
2014 and 2016 increased the probability of having a seasonal water delivery schedule by 1% in 2016 (p < 
0.05). 
6.2. Routine cleaning and maintenance of irrigation infrastructure
WUAs with longer training were 6% more likely to conduct pre-irrigation seasonal maintenance of 
secondary canals in 2016, and 3% more likely to conduct the same for tertiary canals, but these results 
are not significantly different from zero at the usual levels (Table 4). A 1% increase in the area under 
fodder cultivation between the two years reduced the probability of conducting pre-irrigation seasonal 
maintenance of tertiary canals by 21% in 2016 (p < 0.1), consistent with the observation that fodder 
cultivation in Tajikistan is mostly rainfed. 
6.3. Irrigation fee collection and transfers
WUAs with longer training had 9% more members paying their irrigation fees in 2016 than in 2014, as 
compared to WUAs with shorter training, although these results are not significant at the standard 
levels (Table 5). However, longer training has no effect on the change in the share of irrigation fees 
transferred to the government. WUAs with longer training also experienced an increase in indebtedness 
between 2014 and 2016 of TJS 27,691 (~$ 3,146),19 which effect was not statistically significant from 
zero at the usual levels. A 1% increase in the cultivated area of wheat was associated with a reduction of 
debt of the magnitude of TJS 25,952 (~$ 3,377) reflecting the fact that wheat cultivation requires 
considerably less water. A 1% increase in area under fodder cultivation resulted in 53% increase in the 
share of members paying their irrigation fees (p< 0.05) consistent with the fact that since fodder 
cultivation is mostly rainfed, low (zero) water use results in low (zero) irrigation charges, which more 
farmers could afford to pay.
6.4. WUA membership fees collection
 
19 All numbers adjusted for inflation, and converted at the exchange rate of 1 Tajikistan Somoni (TJS) ≈ 0.13 US$ in 
2016-17.
19
WUAs with longer training were able to increase the share of members from whom they recovered 
membership fees by 19% between 2014 and 2016 (p<0.01; Table 6). In addition, a 1% increase in 
irrigated area between 2014 and 2016 was associated with a 37% increase in the share of members 
from whom the WUA was able to recover fees during the same period (p<0.1). Adding an additional 
WUA member reduced the share of members from whom the WUA would be able to recover fees 
(p<0.01), reflecting that having more members makes monitoring more difficult; however, the 
magnitude of this coefficient is very small (0.01%). 
6.5. Dispute arbitration
WUAs with longer training were 6% less likely to be the arbitrator of disputes between 2014 and 2016 
(Table 7); however this result is not significant (p=0.46). A 1% increase in the area under cotton 
cultivation increased the probability of the WUA arbitrating disputes from 2014 to 2016 by 28% (p <0.1), 
consistent with the higher demands on irrigation associated with cotton cultivation. 
7. Discussion
Experimental methods such as randomized controlled trials in the field and laboratory experiments have 
the ability to control for selection effects and to provide unbiased estimates of effects of interventions, 
but have limitations related to the generalizability of results under different circumstances. When an 
appropriate control group is difficult to identify due to the scale of the intervention, synthetic controls 
can be used to construct a counterfactual, but they require the treatment unit and the potential control 
units to exist and be functioning for a considerable duration before the intervention commenced. 
Assessment of the impact of large-scale development programs, such as those that create new 
institutions for community based resource management, is challenging for two reasons. First, the 
intervention often creates the units of interventions themselves. This means the units of intervention 
did not exist—and consequently their covariates or outcomes were not measurable—before the 
intervention, thus ruling out the application of methods such as synthetic controls. Second, assignment 
to treatment is not random, and because all covariates and outcomes took values of zero at the 
baseline, a standard difference-in-difference approach may not eliminate time-invariant unobservables.
In this paper, the difference-in-difference technique is employed in a context where measurements of 
outcomes and covariates were made in two time periods after the intervention was implemented, 
20
because, at baseline, both took values of zero. This modified difference-in-difference technique with 
right hand size variables eliminates bias due to time-invariant unobservables, and controls the effect of 
time-varying unobservables, but introduces a measurement error. Despite this bias, the estimates 
provide empirical evidence on how the intervention influences the performance of new institutions in 
their early days, which could be highly valued for adaptive management. An alternative available 
estimate—the measured difference between outcomes in the later period—is likely to be less 
informative because it cannot control for selection effects that influence outcomes in each group. 
A note of caution is warranted when using the modified difference-in-difference technique when the 
treatment and control units do not exist before the intervention. In this paper, the outcomes being 
examined—service delivery and financial functions—are typically slow to change, especially when 
institutions are new. In cases where the outcomes respond more rapidly to an intervention, the 
measurement bias may outweigh the reduction in bias achieved through the elimination of time￾invariant unobservables. Practitioners using the modified DID technique would be advised to consider 
the nature of the outcomes under consideration. 
In this paper, the results demonstrate that longer training enabled WUAs in Southern Tajikistan to 
perform their mandated functions better between 2014 and 2016. This was particularly the case for 
recovering membership fees from dehkan farm members, where WUAs with longer training increased 
their membership fee recovery by 19% in the two-year period. The existing literature on common-pool 
resources and institutional reform suggests that the length of the training process may be an important 
factor in determining institutional performance. The results in this paper add support to that claim, 
providing some evidence that additional training may allow WUAs to perform functions better. Follow￾up measurement in subsequent years may present a more complete picture of the differences in this 
effect.
In countries such as Tajikistan where political and economic transitions are giving rise to new 
institutions, understanding performance in the short term can provide timely evidence for adaptive 
management. In this case, the results suggest that the government should focus on longer training if it 
expands its WUA program. Equally usefully, data-based comparisons such as those reported here, can 
indicate conditions where there are no radical differences in response to the program intervention that 
might necessitate immediate action. Where some analysts may be reluctant to conduct empirical 
analysis in the absence of well-established treatment and control groups, such as with the formation of 
21
WUAs, quantitative methods of assessment can be applied in less-than-ideal conditions to generate 
evidence that can prove useful for policy purposes. 
Acknowledgments:
This research is funded through a project titled ‘Impact of Water Users Associations on Water and Land 
Productivity, Equity and Food Security in Tajikistan’ funded by the United States Agency for International 
Development under USAID Grant Number AID-BFS-G-11-00002. The opinions expressed herein are those 
of the authors and do not necessarily reflect the views of the U.S. Agency for International 
Development. The authors would like to thank Sahana Balasubramanya and Marie-Charlotte Buisson for 
suggestions; Sohrob Aslamy and Katherine MacDonald for supervising data collection; and two 
anonymous referees for extensive comments that helped us improve the paper. Any errors are ours. 
Bibliography
Abadie, A., Diamond, A. and Hainmueller, J. (2010). ‘Synthetic control methods for comparative case 
studies: estimating the effect of California’s tobacco control program’. Journal of the American
Statistical Association, volume 105, pp. 493-505.
Abadie, A. and Gardeazabal, J. (2003). ‘The economic costs of conflict: a case study of the Basque 
country’. American Economic Review, volume 93, pp.113-132.
Asian Development Bank (ADB). (2012). ‘Tajikistan: Irrigation Rehabilitation Project completion report’. 
Validation report, Independent Evaluation Department. Manila: ADB
Balasubramanya, S., Buisson, M.-C., Saikia, P., MacDonald, K., Aslamy, S., Horbulyk, T., Hannah, C., 
Yakubov, M. and Platonov, A. (2016). ‘Impact of water user associations on water and land productivity, 
equity and food security in Tajikistan’. Baseline report, prepared for the United States Agency for 
International Development (USAID), Colombo: International Water Management Institute, pp. 1-127. 
Available at: 
https://agrilinks.org/sites/default/files/resource/files/Feed%20the%20Future%20Tajikistan%20Water%
20User%20Associations%20Impact%20Evaluation%20-%20B.pdf (last accessed June 15, 2017)
Barcellini, F., Prost, L. and Marianne, C. (2015). ‘Designers’ and users’ roles in participatory design: what 
is actually co-designed by participants?’ Applied Ergonomics, volume 50, 31-40.
22
Batt, H.A. and Merkley, G.P. (2010). ‘Water management and user association analysis for irrigation 
improvement in Egypt’. Irrigation and Drainage, volume 59, pp. 150-160.
Çakmak, B., Beyyri˙bey. M. and Kodal, S. (2004). ‘Irrigation water pricing in water user associations, 
Turkey’. Water Resource Development, volume 20, pp. 113-124.
Denyer Willis, G. and Mota Prado, M. (2014). ‘Process and pattern in institutional reforms: a case study 
of the Police Pacifying Units (UPPs) in Brazil’. World Development, volume 64, pp. 232-242.
Food and Agriculture Organization of the United Nations (FAO). (2012). ‘AQUASTAT: Tajikistan’. 
http://www.fao.org/nr/water/aquastat/countries_regions/TJK/. (last accessed 9th January 2017).
Gunchinmaa, T. and Yakubov, M. (2009). ‘Institutions and transition: does a better institutional 
environment make water users associations more effective in Central Asia?’ Water Policy, volume 12, 
number 2, pp. 165-185.
Hamidov, A., Thiel, A. and Zikos, D. (2015). ‘Institutional design in transformation: a comparative study 
of local irrigation governance in Uzbekistan’. Environmental Science & Policy, volume 53, part B, pp. 175-
191.
Hornidge, A.K., Oberkircher, L. and Kudryavtseva, A. (2013). ‘Boundary management and the discursive 
sphere – negotiating realities in Khorezm, Uzbekistan’. Geoforum, volume 45, pp. 266-274.
Huang, Q., Wang, J., Easter, K.W. and Rozelle, S. (2010). ‘Empirical assessment of water management 
institutions in Northern China’. Agricultural Water Management, volume 98, 361-369.
Johnson, S.H. and Stoutjesdijk, J. (2008). ‘WUA training and support in the Kyrgyz Republic’. Irrigation 
and Drainage, volume 57, pp. 311-321.
Kazbekov, J., Abdullaev, I., Manthrithilake, H., Qureshi, A. and Jumaboev, K. (2009). ‘Evaluating planning 
and delivery performance of water user associations (WUAs) in Osh Province, Kyrgyzstan’. Agricultural 
Water Management, volume 96, pp. 1259-1267.
Matveeva, A. (2009). ‘The perils of emerging statehood: civil war and state reconstruction in Tajikistan’. 
Working paper, number 46, series 2. London: Crisis States Research Centre, pp. 1-59.
23
Meinzen-Dick, R., Raju, K.V. and Gulati, A. (2002). ‘What affects organization and collective action for 
managing resources? Evidence from canal irrigation systems in India’. World Development, volume 30, 
number 4, pp. 649-666.
Mukhtarov, F., Fox, S., Mukhamedova, N. and Wegerich, K. (2015). ‘Interactive institutional design and 
contextual relevance: water user groups in Turkey, Azerbaijan and Uzbekistan’. Environmental Science 
and Policy, volume 53, pp. 206-214.
Mustafa, D., Altz-Stamm, A. and Mapstone Scott, L. (2016). ‘Water user associations and the politics of 
water in Jordan’. World Development, volume 79, pp. 164-176.
Nagrah, A., Chaudhry, A.M. and Giordano, M. (2016). ‘Collective action in decentralized irrigation 
systems: evidence from Pakistan’. World Development, volume 84, pp. 282-298.
O’Hara, S.L. (2000). ‘Lessons from the past: water management in Central Asia’. Water Policy, volume 2, 
pp. 365-384.
Ostrom, E. (1990). Governing the commons: the evolution of institutions for collective action. Cambridge, 
United Kingdom: Cambridge University Press, pp. 1-294.
Republic of Tajikistan. (2006). ‘Water user association law’. Law of the Republic of Tajikistan, pp. 1-12. 
Ricks, J.I. (2016). ‘Building participatory organizations for common pool resource management: water 
user group promotion in Indonesia’. World Development, volume 77, pp. 34-47.
Rosenbaum, P.R. and Rubin, D.J. (1985). ‘Constructing a control group using multivariate matched 
sampling methods that incorporate the propensity score’. The American Statistician, volume 39, pp. 33-
38. 
Rubin, D.J. (2001). ‘Using Propensity Scores to Help Design Observational Studies: Application to the 
Tobacco Litigation’. Health Services and Outcomes Research Methodology, volume 2, pp. 169-188.
Sehring, J. (2007). ‘Irrigation reform in Kyrgyzstan and Tajikistan’. Irrigation and Drainage Systems, 
volume 21, pp. 277-290.
24
Shahriari, H. (2009). ‘Activities, water and gender in Tajikistan’s rural sector: a social assessment of 
Konibodom, Bobojon Ghafurov, and Yovon.’ Social Development Department, Europe and Central Asia 
series. Washington, DC: World Bank. Available at: 
http://documents.worldbank.org/curated/en/364511468117242593/pdf/514450WP0TJ0So10Box34202
8B01PUBLIC1.pdf (last accessed June 15, 2017). 
Skaperdas, S. (2001). ‘The political economy of organized crime: providing protection when the state 
does not’. Economics of Governance, volume 2, pp. 173-202.
Smith, J.L. (2008). ‘A critical appreciation of the ‘bottom-up’ approach to sustainable water 
management: embracing complexity rather than desirability’. Local Environment, volume 13, number 4, 
pp. 353-366.
Tellez Foster, E., Rapoport, A. and Dinar, A. (2016). ‘Alternative policies to manage electricity subsidies 
for groundwater extraction: a field study in Mexico’. UCR SPP Working Paper Series WP#16-05. 
Riverside, California: School of Public Policy, University of California-Riverside.
Tellez Foster, E., Rapoport, A. and Dinar, A. (2017). ‘Groundwater and electricity consumption under 
alternative subsidies: evidence from laboratory experiments’. Journal of Behavioral and Experimental 
Economics, volume 68, pp. 41-52. 
Theesfeld, I. (2004). ‘Constraints on collective action in a transitional economy: the case of Bulgaria’s 
irrigation sector’. World Development, volume 32, issue 2, pp. 251-271.
Thiel, A., Mukhtarov, F. and Zikos, D. (2015). ‘Crafting or designing? Science and politics for purposeful 
institutional change in social-ecological systems’. Environmental Science & Policy, volume 53, pp. 81-86.
United Nations (2015). ‘Tajikistan’. http://data.un.org/CountryProfile.aspx?crName=TAJIKISTAN (last
accessed 22nd June, 2017). 
United States Agency for International Development (USAID). (2014). ‘USAID Family Farming Program -
Quarterly report: year four – first quarter.’ Unpublished document, pp. 1-46.Washington, DC: USAID.
Wegerich, K. (2000). ‘Water user associations in Uzbekistan and Kyrgyzstan: study on conditions for 
sustainable development’. Occasional paper – Water Issues Study Group, London: School of Oriental and 
African Studies, University of London, number 32, pp. 1-23.
25
Woolridge, J. (2002). Econometric Analysis of Cross Section and Panel Data. Cambridge, Massachusetts: 
The MIT Press. 
World Bank (2015). ‘Tajikistan’ http://data.worldbank.org/country/tajikistan (last accessed 22nd June, 
2017).
Yakubov, M. (2011). ‘Assessing irrigation performance from the farmers’ perspective: a qualitative 
study’. Irrigation and Drainage, volume 61, issue 3, pp. 316-329.
Yap-Salinas, L.H. (1994). ‘Converging factors in the successful transfer of irrigation management 
responsibilities to water users’ associations in the Dominican Republic’. Paper presented at the 
International Conference on Irrigation Management Transfer, 20-24 September, Wuhan, China, pp. 117-
127.
Zhou, Q. (2013). ‘Decentralized irrigation in China: an institutional analysis’. Policy and Society, volume 
32, pp. 77-88.
26
Figure 1: The standard and modified DID technique when the treatment unit performs better than the 
control unit
27
Figure 2: The standard and modified DID technique when the control unit performs better than the 
treatment unit
28
Figure 3: Propensity scores using time-invariant physical characteristics (excluding district and river)
29
Figure 4: Propensity scores using time-invariant physical characteristics (including district)
30
Figure 5: Propensity scores using time-invariant physical characteristics (including district and river)
31
Table 1: WUA performance indicators
Water delivery
WUAs had a delivery schedule
WUAs had at least two board meetings per year
Routine cleaning and maintenance of irrigation infrastructure
WUAs conducted two pre-irrigation maintenance sessions for secondary canals per year
WUAs conducted two pre-irrigation maintenance sessions for tertiary canals per year
Collect irrigation fees and transfer to government
Share of members paying irrigation service fees
Share of collected irrigation fees transferred to the government
Collect WUA membership fees
Share of members paying membership fees
WUA arbitrates disputes
Likelihood WUA arbitrates dispute between farmers
Source: Authors
32
Table 2: Summary statistics for covariates that could affect performance (included as right-hand-side variables in 
the DID)
2014: survey 1 2016: survey 2
Indicator (Y) Treatment Control Treatment Control
N Mean (Std Dev) N Mean (Std Dev) N Mean (Std Dev) N
Mean (Std 
Dev)
% WUA service area irrigated 73 0.96 (0.10) 67 0.92 (0.17) 73 0.94 (0.12) 67 0.94 (0.12)
% irrigated area producing 
wheat 73 0.23 (0.19) 58 0.28 (0.29) 73 0.23 (0.21) 58 0.25 (0.20)
% irrigated area producing 
fodder 73 0.07 (0.16) 58 0.07 (0.13) 73 0.08 (0.06) 58 0.10 (0.13)
% irrigated area producing 
cotton
73 0.47 (0.24) 58 0.42 (0.42) 73 0.49 (0.20) 58 0.34 (0.25)
# members in the WUA 74 240.62 (230.08) 67 285 (295.21) 74 687.41 (3424.94) 67 584.22 
(1468.40)
# disputes 73 24.22 (31.53) 67 32.61 (44.67) 73 14.66 (22.06) 67 33.91 (68.76)
# board members 73 5.28 (1.18) 67 4.10 (2.08) 73 9.32 (7.22) 67 7.17 (6.65)
% of female board members 73 0.07 (.11) 67 0.08 (0.15) 73 0.09 (0.13) 67 0.09 (0.14)
Quality of main canal (1-5) 74 2.91 (0.58) 67 2.98 (0.67) 74 2.54 (0.92) 67 2.57 (0.96)
Quality of secondary canal (1-
5) 74 3.08 (0.52) 67 2.87 (0.72) 74 3.03 (0.92) 67 2.69 (0.85)
Quality of tertiary canal (1-5) 74 3.09 (0.48) 67 2.79 (0.66) 74 3.12 (0.72) 67 2.81 (0.78)
Source: data collected by the 
authors 
33
Table 3: Effects of longer training on water delivery
WUA had a WUA had at least
delivery schedule two board meetings
in 2014 and 2016 in 2014 and 2016
Longer training 0.04(0.07) 0.09 (0.06)
% WUA service area irrigated 0.33 (0.21) -0.22 (0.20)
% irrigated area producing wheat -0.27 (0.13)* -0.02 (0.12)
% irrigated area producing cotton 0.28 (0.13)** 0.15(0.12)
% irrigated area producing fodder -0.00 (0.22) -0.16 (0.20)
# members in WUA 0.00 (0.00) -0.00 (0.00)
# disputes -0.00 (0.00) -0.00 (0.00)
# board members 0.01 (0.00)** 0.00 (0.00)
% of female board members 0.23 (0.20) 0.28 (0.189)
Quality of main canal (1-5) -0.04 (0.03) -0.02 (0.03)
Quality of secondary canal (1-5) -0.03 (0.03) 0.00 (0.03)
Quality of tertiary canal (1-5) 0.01 (0.04) 0.04 (0.04)
Constant 0.78 (0.06)*** 0.77 (0.05)***
N 131 131
R-squared 0.14 0.11
F-statistic F(12,118)=1.59 F(12,118)=1.29
Prob > F 0.10* 0.29
The reported values express the modified DID estimate as a percentage difference, followed in 
parentheses by the standard error.
*** denotes that the difference is significant at 1%; 
** denotes that the difference is significant at 5%; 
* denotes that the difference is significant at 10%
34
Table 4: Effects of longer training on routine maintenance of irrigation infrastructure
WUA conducted WUA conducted
pre-season maintenance pre-season maintenance
of secondary canals of tertiary canals
Longer training 0.06 (0.04) 0.03 (0.04)
% WUA service area irrigated -0.11 (0.15) 0.07 (0.12)
% irrigated area producing wheat 0.05 (0.09) 0.05 (0.08)
% irrigated area producing cotton 0.06 (0.09) 0.11 (0.08)
% irrigated area producing fodder -0.20 (0.15) -0.21 (0.12)*
# WUA members -0.00 |(0.00) 0.00 (0.00)
# disputes 0.00 (0.00) 0.00 (0.00)
# board members 0.00 (0.00) 0.00 (0.00)
% of female board members 0.01 (0.14) 0.04 (0.12)
Quality of main canal (1-5) -0.01 (0.02) -0.01 (0.02)
Quality of secondary canal (1-5) -0.00 (0.02) 0.00 (0.02)
Quality of tertiary canal (1-5) 0.03 (0.03) -0.02(0.03)
Constant 0.05 (0.03) 0.03 (0.03)
N 131 131
R-squared 0.06 0.09
F-statistic F(12,118)=0.60 F(12,118)=1.01
Prob > F 0.84 0.44
The reported values express the modified DID estimate as a percentage difference, followed in 
parentheses by the standard error.
* denotes that the difference is significant at 10%
35
Table 5: Effects of longer training on irrigation service fee recovery and transfer
Share of members Share of collected 
from whom irrigation irrigation service fees 
service fee is recovered transferred
Longer training 0.09 (0.07) 0.00 (0.11)
% WUA service area irrigated 0.21 (0.22) -0.08 (0.33)
% irrigated area producing wheat -0.18 (0.17) 0.33 (0.26)
% irrigated area producing cotton 0.08 (0.13) -0.02 (0.18)
% irrigated area producing fodder 0.53 (0.22)** -0.24 (0.34)
# WUA members -0.00 (0.00) 0.00 (0.00)
# disputes -0.00 (0.00) -0.00 (0.00)
# board members -0.00 (0.00) -0.01 (0..01)
% of female board members -0.07 (0.19) 0.24 (0.31)
Quality of main canal (1-5) 0.01 (0.03) 0.01 (0.05)
Quality of secondary canal (1-5) -0.02 (0.04) 0.02 (0.06)
Quality of tertiary canal (1-5) 0.04 (0.04) -0.04 (0.07)
Constant 0.08 (0.06) 0.94 (0.09)***
N 101 102
R-squared 0.13 0.05
F-statistic F(12,88)=1.09 F(12,89)=0.38
Prob > F 0.38 0.97
The reported values express the modified DID estimate as a percentage difference, followed in 
parentheses by the standard error.
*** denotes that the difference is significant at 1% 
** denotes that the difference is significant at 5%
36
Table 6: Effects of longer training on WUA membership fee recovery
Share of dehkan farms
from whom WUA
membership fee is recovered
Longer training 0.19 (0.06)***
% WUA service area irrigated 0.37 (0.19)*
% irrigated area producing wheat -0.14 (0.13)
% irrigated area producing cotton 0.07 (0.12)
% irrigated area producing fodder 0.21 (0.21)
# WUA members -0.00 (0.00)***
# disputes 0.00 (0.00)
# board members 0.00 (0.00)
% of female board members 0.12 (0.18)
Quality of main canal (1-5) -0.01(0.03)
Quality of secondary canal (1-5) -0.06 (0.03)
Quality of tertiary canal (1-5) 0.02 (0.04)
Constant -0.27 (0.05)***
N 115
R-squared 0.26
F-statistic F(12,102)=2.77
Prob > F 0.00***
The reported values express the modified DID estimate as a percentage difference, followed in 
parentheses by the standard error.
*** denotes that the difference is significant at 1%; 
* denotes that the difference is significant at 10%
37
Table 7: Effects of longer training on the likelihood of a dispute being arbitrated by the WUA
WUAs settled 
disputes in 2014
and 2016
Longer training -0.06 (0.08)
% WUA service area irrigated 0.24 (0.27)
% irrigated area producing wheat -0.07 (0.17)
% irrigated area producing cotton 0.28 (0.16)*
% irrigated area producing fodder -0.27 (0.27)
# WUA members 0.00 (0.00)
# disputes 0.00 (0.00)
# board members -0.00 (0.01)
% of female board members 0.06 (0.25)
Quality of main canal (1-5) 0.00 (0.04)
Quality of secondary canal (1-5) 0.01 (0.05)
Quality of tertiary canal (1-5) -0.00 (0.05)
Constant 0.35 (0.07)***
N 131
R-squared 0.05
F-statistic F(12,118)=0.48
Prob > F 0.92
The reported values express the modified DID estimate as a percentage difference, followed in 
parentheses by the standard error.
*** denotes that the difference is significant at 1%; 
* denotes that the difference is significant at 10%
38
Appendix 2 
Balasubramanya, S. 2018. Farm participation in water user associations in Southern Tajikistan: 
effects of longer training and the role of gender. In review at Agricultural Water Management.
39
Title: Farm participation in water user associations in Southern Tajikistan: effects of longer training and 
the role of gender
Keywords: participation, training, gender, difference-in-difference, propensity scores, Tajikistan
Abstract: This paper examines whether longer training increases farm participation in community￾managed water user associations, in a context where assignment to training duration was not 
randomized and none of these institutions existed before training began. We also examine whether 
participation is affected when farm managers migrate and leave farm operations to other workers, in a 
context where only managers have been directly trained, almost all managers are male, and females are 
increasingly operating farms. We collected microdata from 1,855 farms in Southern Tajikistan, where 
farm managers in 40 subdistricts received longer training, while those in the other 40 received shorter 
training. These ‘treatment’ and ‘control’ subdistricts were selected by constructing propensity scores 
and matching without replacement to address observable selection effects that may affect assignment 
to training duration. Farms were then selected from a census using a stratified random sampling 
process. A difference-in-difference technique with right-hand-side covariates is employed, where both 
sets of data were collected after training was completed. This choice of econometric methods controls 
against farm-level selection effects, but introduces a potential bias due to measurement error. Longer 
training has a causal effect on increasing participation in WUAs. Results also demonstrate that when 
male workers not directly trained operate farms, participation is not affected; however, participation is 
negatively affected when female workers operate farms. These results provide evidence for designing 
irrigation management programs to target female workers directly, in order to strengthen institutions 
whose success depends on active farm participation.
1. Introduction
In Tajikistan, water user associations (WUAs) are legally mandated to bring publicly provided irrigation 
water to the farm gate (Republic of Tajikistan, 2006). WUAs in Tajikistan serve dehkan (meaning private) 
farms, and legally, dehkan farms (not farmers) are eligible members of a WUA. These WUAs are 
participatory institutions; this implies that the participation and cooperation of representatives of 
member-farms is needed for WUAs to perform their mandated duty of water delivery successfully 
(Beresford, 2010). The dehkan farm is headed by a manager—a legally recognized position that is listed 
on the title of the dehkan farm. The farm title also lists the workers of the farm; these are individuals 
40
with a legal claim to work on the farm, who also have a stake in the outputs of the farm. The farm 
manager and the listed workers typically belong to the same family, but not necessarily the same 
household.1 The manager is the operator of the farm and represents the farm at the WUAs, when 
physically in residence.2
The WUA Law (Republic of Tajikistan, 2006) specifies the roles of the members. Dehkan farms are 
supposed to pay an irrigation service fee (toward the expense of water provision), which is levied 
seasonally; and a WUA membership fee (toward the expenses of WUA services provided), which is 
levied annually. They are also expected to contribute to (uncompensated) pre-irrigation season repair 
and maintenance of canals that is coordinated by the WUAs through the provision of labour of the 
workers of the farm (listed on the title) for the task. Dehkan farms are encouraged to sign a contract 
with the WUA at the start of the year. In the contract, water needs are stated for planning purposes. 
Representatives of member-farms are encouraged to attend WUA meetings. Participation can be 
assessed by examining these member-mandated functions in WUAs (Yakubov, 2011). 
Some WUAs in Tajikistan were created using longer and more gradual training processes. Consequently 
the managers of the farms served by such WUAs received longer training on how to participate and 
cooperate with the WUA for more successful irrigation management. The institutional literature on 
farmer participation in community-based participatory management organizations suggests that longer 
training can increase the likelihood and extent of participation and cooperation, which is needed for 
such organizations to function successfully (Yap-Salinas, 1994; Kazbekov et al., 2009; Mukhtarov et al., 
2015). 
Estimating the impact of longer training using quantitative econometric methods can provide important 
information to target efforts to strengthen participation and functioning of new institutions.3 However, 
conducting such evaluations in Tajikistan is challenging for at least two important reasons. First, 
assignment to longer training was not random, introducing potential selection bias. The second 
 
1 A family is defined as a set of individuals related through blood or through marriage. A household is defined as a 
set of individuals who consume food cooked in the same kitchen. 
2 When the manager is physically present, he is the operator for the farm. When the manager is not physically 
present, this could be because either they are deceased, or they have migrated to Dushanbe or overseas. In these 
cases, another member of the farm (listed on the title) became the operator. In cases of migration, the name of 
the manager is not legally changed on the title of the farm. 
3 While estimating the impact of longer training on water delivery is a worthy question, the lack of gauges at inlets 
and outlets to measure water flows in distributary canals and watercourses complicates addressing this question 
(for examples of such work, see Fernández-Pacheco et al., 2015.
41
challenge pertains to the establishment of a baseline. Dehkan farms were created in the mid 2000s 
through the de-collectivization of the Soviet-era collectives. The government enacted the WUA Law in 
2006 and mandated WUAs to be responsible for delivering water to farms. Even so, there was no 
participatory management irrigation organization in place until 2012. The process of creating the WUAs 
only began in 2012 through coordinated international assistance and government policy. At the start of 
the training, consequently, WUAs were not functioning, and there were no functional WUAs in which 
farms could participate. WUAs were established (and the managers of the farms were trained) in gravity 
schemes in Southern Tajikistan, the breadbasket of the country, by USAID and by the government, 
during 2012 and 2013. Managers of farms in WUAs created by USAID received 20-24 months of training, 
while those in WUAs created by the government received three to six months of training 
(Balasubramanya et al., 2018).4
Training in participatory irrigation management in all WUAs—whether created using the longer or 
shorter process—has mostly been directed at managers of the farms. Since 98% of dehkan farm 
managers are male (FAO, 2018), this implies that training in participatory irrigation has mostly been 
provided to males. Migration of (mostly) males to either urban areas or overseas is rather common in 
Tajikistan. A study in 2013 estimated that 28% of households had at least one migrant, with migration 
greater in locations with rural and poorer households (Danzer et al., 2013). Another study, with a study 
area matching that in this paper, estimated that in 2015, 48% of rural households in Southern Tajikistan 
had at least one migrant (Buisson et al., 2016). A consequence of migration is that farm workers who 
were not directly trained in participatory irrigation are increasingly operating farms, with female farmers 
constituting an important share of non-trained individuals operating farms. 
Whether training can increase participation in circumstances when individuals not directly trained are 
managing the farm would depend on the diffusion of information from those who were directly 
trained—the farm managers in this case—to those who were not trained but are now taking on the role 
of operating the farm. The literature on whether (and to whom) information diffuses from farmers 
directly trained (and produces associated behavior changes), suggests that the evidence is mixed and 
depends on farmer and farm characteristics of trained and other farmers (e.g. see Clausen et al., 2004; 
Feder et al., 2004).5
 
 
4 Training was directed only at the managers of the farms, not at the listed members of the farm. 
5 This literature has focused on pest management, seed choice, and fertilizer application.
42
This paper examines whether longer training has a causal effect on increasing the probability of 
member-farm participation in WUAs in a context where no participatory organization existed before the 
training began. The paper considers a sample of 1,855 member-farms in 80 subdistricts6 of Southern 
Tajikistan, where member-farm managers in 40 subdistricts received longer training (the treatment 
group) and those in the other 40 subdistricts received shorter training (the control group). These 80 
subdistricts were selected from a population of subdistricts by constructing propensity scores and 
conducting a 1:1 matching without replacement to control against selection of observable confounders 
at the sub-district level (Rosenbaum and Rubin, 1985). Member-farms in these matched subdistricts 
were then selected from a census of member-farms through a stratified random sampling process to 
arrive at a representative sample of farms in Southern Tajikistan. Since there was no organization for 
member-farms to participate in before the end of the training , controlling against selection effects at 
the farm level entailed using a difference-in-difference (DID) technique with time-varying and time￾invariant right-hand-side covariates, with data collected in two time-periods after the training of 
managers of member-farms was completed (Balasubramanya et al., 2018).7 This approach introduces a 
potential bias in the estimated effects of training due to measurement error, but controls against 
selection effects (as demonstrated in Balasubramanya et al., 2018, and explained further in the 
methodology section of this paper). 
The paper also considers whether participation is affected when the farm is operated by a person who 
was not trained, and whether any effect depends on the gender of that person. This is motivated by the 
observation that an increasing number of farms (~ 50% in this paper) are being operated by individuals 
who were not directly trained. To keep training costs low, most training programs train lead farmers 
(managers in this paper) (Anderson and Feder, 2007), who are often male (as is the case in this paper). 
However, diffusion of information to vulnerable individuals in the community (e.g., see Alesina and La 
Ferrara, 2000) and to females (e.g., see Kumar and Quisumbing, 2011; Beaman and Dillon, 2018) has 
often not been observed when lead male farmers are trained. In the context of Tajikistan, where almost 
everyone who was trained was male; where irrigation is regarded as “a man’s job” (Mukhamedova and 
Wegerich, 2018); and where those directly trained are decreasingly managing their farms, an 
 
6
In Tajikistan, a subdistrict is called a jamoat. 
7 This is in contrast to the standard DID approach, where one set of observations would have been collected at the 
time the training commenced, and the other after the training was completed.
43
understanding of how participation is affected when untrained males and females operate farms can 
provide information for targeted training. 
The context of the participatory irrigation intervention renders the application of other methods of 
establishing causal effects impossible. Using a method of synthetic controls (Abadie and Gardeazabal, 
2003) was not possible due to the non-existence of data on participation from before the starting of the 
training. Using experimental methods (Tellez Foster et al., 2017) was not possible due to the non￾strategic setting. Since data on subdistricts were available, they were used to select matched pairs of 
treatment and control clusters, in order to control for selection on observable confounders. This 
modified DID technique with the inclusion of time-varying right-hand-side covariates controls for 
selection on time-invariant and time-variant unobservable confounders. 
This paper contributes to the literature in three ways. First, rather than using cross-sectional data (e.g., 
Nagrah et al., 2016; Qiao et al., 2009); or conducting qualitative analysis (e.g., Aydogdu et al., 2015) to 
analyze participation, panel data are used. These data control for several types of selection effects and 
generate evidence needed for adaptive management and targeted investments in the early years of 
institutions whose success depends on member participation. The second contribution is to understand 
how participation, and thus the functioning of these participatory institutions, is likely to be affected as 
the gender composition of the agricultural workforce changes due to labor market adjustments. The 
third contribution is to contextualize the results presented in Balasubramanya et al. (2018) that 
examined the effects of longer training on WUAs’ performance of mandated functions. That paper 
demonstrated that WUAs with longer training were able to recover membership fees from 19% more 
members, and were 10% more likely to hold meetings with members for planning purposes. The current 
paper uses a different primary dataset that was collected using a different study design that was 
specifically tailored for testing the effects of training duration and other determinants of farm 
participation such as gender.
The results in this paper demonstrate that dehkan farms whose managers received longer training are 
8% more likely to pay their membership fees; 20% more likely to sign a water contract with the WUA, 
and 9% more likely to attend the WUA meetings. Dehkan farms whose managers received longer 
training contributed seven more person-days of labor towards routine pre-irrigation repair and 
maintenance of canals. Participation was not affected when the farm was operated by a male worker 
who was not directly trained. However, participation was negatively affected when the farm was 
44
operated by a female, with such farms 9% less likely to pay their membership fees than farms operated 
by males, 11% less likely to sign a water contract, and 3% less likely to attend the WUA meetings. 
A limitation of the paper is that it is not possible to test formally whether participation outcomes for 
farms with longer and shorter training would have changed in the absence of the intervention. This is 
because participatory organizations, with longer and shorter training, came into existence at the same 
time (2012), which coincides with the commencement of the intervention under consideration in this 
paper. Before 2012, no participatory water management institutions existed, and consequently, pre￾2012 participation data do not exist. The choice of methods in the paper is based on the observation 
that standard methods to control for selection effects cannot be readily employed on this context. 
The remainder of the paper is organized as follows. Section 2 provides a context for how participatory 
WUAs evolved in Tajikistan. Section 3 presents a conceptual framework and reviews literature 
concerning length of training and participation in WUAs globally and in Central Asia, along with 
literature on the diffusion of information from trained to other farmers. Section 4 describes the 
methodology, while Section 5 presents the study design and Section 6 contains details on data. 
Summary statistics and results are presented in Section 7. Section 8 aggregates the results and discusses 
them in relation to the existing literature and the evaluation of participatory institutions. 
2. Evolution of participatory water user associations in Tajikistan
Within the former Soviet Union, Tajikistan was designated as Central Asia’s main hub of cotton 
cultivation, which was practiced on collective farms (FAO, 2012). Only 4% of the land in Tajikistan is 
agricultural, with 95% of crop cultivation on irrigated land (FAO, 2012). Southern Tajikistan is the most 
populous part of the country, where cotton and wheat production are dominant. 
WUAs were created in Tajikistan in response to the de-collectivization of the collective farms into 
dehkan (private) farms, which began in the mid-2000s. With the departure of Russian irrigation 
specialists and the lack of Soviet subsidies, irrigation departments—called vodkhozes—were not able to 
deal with the challenge of providing water to thousands of private farms (Gunchinmaa and Yakubov, 
2009; Shahriari, 2009). The government enacted the WUA Law (Republic of Tajikistan, 2006) and named 
the WUA as the institution responsible for delivering water to the farms. International assistance for 
creating these organizations was requested. WUAs were piloted by several international organizations 
and a countrywide program to create new WUAs gained momentum in 2011. More than 300 WUAs are 
45
now functional in Tajikistan, with the service area typically in the range of 1,400 ha to 1,600 ha 
(Balasubramanya et al., 2016). All WUAs require their member-farms’ participation and cooperation for 
irrigation service delivery. 
Participatory governance and civic engagement are not new to Tajikistan (Cieslewska, 2010). For 
instance, even though state-directed institutions dominated the agriculture and water sectors during 
the Soviet era (O’Hara, 2000), traditional neighbourhood (mahalla) councils, which were elected by the 
households in the neighbourhood, played an important, though informal, role in addressing local 
problems (including water-related disputes) in the community. Such neighbourhood councils continue 
to play an important role today.
3. Conceptual framework
3.1. Participation in community-managed resource systems
Key among Ostrom’s (1990) eight principles for managing subtractable common pool resources is the 
concept of participation, and the arrangements that facilitate such participation. Participation is viewed 
as a way of reconnecting disengaged citizens with the decision-making process in contexts of 
‘democratic deficit’ (Pratchett, 1999: 619), as well as improving the quality of those decisions (Martin, 
2009). Beyond mere participation, Gurung (1992: 32) states how it is important that users ‘abide by the 
terms of agreement before, during, and after the implementation process of the participatory 
management program.’ The collective management of common pool resources such as water critically 
depends on users continuously following the rules (Cleaver, 1999). 
Among decentralized participatory institutions such as WUAs, external professionals may deploy an 
enabling logic, but the users in fact perform the service task for themselves (Bovaird et al., 2015). For 
example, WUAs coordinate routine cleaning and repair of irrigation canals before the start of the 
irrigation season, but member-farms must contribute labour (of the workers of the farm) in order to 
improve irrigation service delivery for all member-farms. Accordingly, users may be viewed as being in 
partnership with their organizations, as they participate to co-produce services (Beresford, 2010). 
3.2. Training members in participatory management
In the literature on institutional reform and service provision, it has become widely accepted that 
participation of users is required for institutions to be fully effective (Bovaird et al., 2015). However, 
46
when newly established, members of participatory institutions, such as managers of member-farms in 
the case of WUAs, may need training to comprehend how they need to participate and cooperate with 
the institution (Nagrah et al., 2016). 
Since these (new) water users may be constrained in participating and cooperating effectively due to a 
lack of knowledge (Hu et al., 2014), the length of the training period provided can be an important 
variable determining user participation (Yap-Salinas, 1994). For example, Nagrah et al. (2016) note that 
without a sufficiently long period of training, ‘farmers in Pakistan may not be ready or even interested in 
the task’ of participating in a WUA and following its rules. In Turkey, Aydogdu et al. (2015) found that 
‘farmers lacked sufficient knowledge regarding WUAs’ and needed more training to perform the WUA 
functions independently. In contrast, a ‘high degree of understanding about water user associations’ 
was found to be an important factor determining respondents’ satisfaction based on surveys in Inner 
Mongolia (Qiao et al. 2009: 822).
Concerning post-Soviet states with a similar recent history to Tajikistan, service user satisfaction and 
willingness to pay among Armenian WUAs was compromised by insufficient, unreliable and untimely 
delivery of water, due in part to inadequate training of WUAs and their users (Alaverdyan and Houston, 
2004: 11). In Kyrgyzstan, improved water delivery was brought about after a lengthy four-year period of 
training ‘encouraged member participation’ (Johnson and Stoutjesdijk, 2008: 311). By contrast, training 
time was limited or nonspecific in Uzbekistan, leading to poor rates of payment of fees and participation 
in WUA governance (Wegerich, 2000). 
3.3. Diffusion of information from trained farmers to other farmers
Farmer-training programs are typically designed to improve performance by providing technical 
information to increase human capital (Anderson and Feder, 2004). Since farmers often rely on other 
farmers for information about agricultural practices (Rees et al., 2000); such programs usually train 
head farmers or village heads, who in turn share that technical information with other farmers 
(Anderson and Feder, 2007). This also keep costs of programs low (Feder et al., 2004).
The literature on whether information diffuses from directly trained farmers to other farmers and 
encourages changes in behaviors is mixed. For example, Feder et al. (2004) found that while Indonesian 
farmers directly trained in pest management reduced pesticide use, farmers not directly trained did not 
experience an improvement in either knowledge or a reduction in pesticide use. In contrast, in Uganda, 
47
Clausen et al. (2017) found that farmers directly trained in pest management were also able to reduce 
the pesticide use of neighboring farmers.
These differences in diffusion of information, and associated changes in behaviors, are likely to depend 
on several factors. The complexity of the knowledge to be shared (e.g., Rola et al., 2002), and the 
strength of interpersonal networks (e.g., Tripp et al., 2005) are two such factors. Also important are the 
social status and gender of persons chosen for training (e.g., see Pemsl et al., 2006; Beaman and Dillon, 
2018); and idiosyncratic characteristics of farmers and their farms (e.g., Fuglie and Kascak, 2001). 
Using low cost options to diffuse information, which often involves using pre-existing networks, may 
have unfortunate distributional consequences. For example, with respect to gender, Beaman and Dillon 
(2018) find that when the information on composting is spread using existing social networks in Mali 
(where networks are among lead male farmers) less influential farmers and female farmers, in 
particular, lose out on receiving valuable information. 
3.4. Hypotheses
The paper considers member-mandated functions as specified in the WUA Law (Republic of Tajikistan, 
2006) , and introduces a number of hypotheses regarding the effects of longer training on these 
indicators. A list of indicators is provided in Table 1.
a) Irrigation fees: Longer training is expected to increase the probability of a farm paying the 
seasonal irrigation fees.
b) WUA membership fees: Longer training is expected to increase the probability of a farm paying 
its annual membership fees
c) Participation in pre-irrigation cleaning of canals: Longer training is expected to increase the 
number of person-days of labor that the member-farm contributes towards cleaning. 
d) Legal relations: Longer training is expected to increase the probability of a farm signing a 
contract with a WUA; and is expected to increase the probability of the manager or (listed) 
worker of a farm attending a WUA meeting. 
Training in Tajikistan was imparted to managers (lead farmers), most of whom are male. Since the 
networks among male farmers are likely to be strong, we expect participation to not to be affected 
when non-trained males operate the farm, but to be lower when females operate the farm. 
48
4. Methodology
Consider the following equation:
𝑌𝑗𝑡 = 𝜇 + 𝛾𝑆𝑗 + 𝜃𝑡 + 𝜔(𝑆𝑗 × 𝑡) + 𝛽𝑋𝑗𝑡 + 𝜗𝑗𝑡 (1),
where 𝑌𝑗𝑡 refers to a participation indicator for farm j at time 𝑡. 𝑆𝑗
is a categorical variable that denotes 
the treatment status of farm𝑗, with 𝑆𝑗 = 1 if the farm manager received longer training, and 𝑆𝑗 = 0 if 
the manager received shorter training. The variable 𝜔 reports the causal effect of longer training on 
participation. 𝑋𝑗𝑡 refers to a set of farm-specific covariates at time 𝑡 that might also influence 𝑌𝑗𝑡. 𝜗𝑗𝑡 is 
the error term.8
The difference-in-difference (DID) technique identifies the causal effect of longer training by comparing 
the average change in participation over a time period for the treatment group to that for the control 
group, while controlling for differences at the starting points and common time trends. Typically, the 
DID technique would be executed by collecting data on participation and other covariates from farms, 
first at the start of the training (i.e., when 𝑡 = 0), and again at some time 𝑡 > 0 after training was 
completed, thus creating a panel dataset. Therefore, the standard DID technique assumes that both 
treatment and control groups were participating even before the training began (that is 𝑌𝑗𝑡 ≠ 0 when
𝑡 ≤ 0), allowing for a pre-training comparison of trends in outcomes between the two groups. Under 
this condition (called the standard condition), the DID technique eliminates time-invariant unobservable 
selection effects (𝜇 in Equation 1)—such as any farm-specific, area-specific, or agency-specific fixed 
effects that are constant over time but may drive differences in level of participation—and provides an 
unbiased estimate of 𝜔. 
However, if participation indicators take values of zero in the pre-intervention period (𝑌𝑗𝑡 = 0 ∀ 𝑡 ≤ 0), 
then there are no pre-intervention trends to compare. In the case of Tajikistan, there were no 
participatory institutions for farms to participate in, before training began in the area under study. If the 
DID technique were implemented by collecting the first set of data on participation and covariates at the 
start of the training (when 𝑌𝑗0 = 0 for all farms), and by collecting the next set of data at some time 𝑡 >
0 for all farms (where 𝑌𝑗𝑡 ≥ 0 for all farms); then this would mathematically be equivalent to using 
 
8 The error structure is assumed to follow: 𝐸 (𝜗𝑗𝑡| 𝑆𝑗
,𝑡) = 0 ∀ 𝑆𝑗 ∀𝑡). This is because the identifying assumption is 
that, by explicitly accounting for , , the errors are uncorrelated with 𝑆𝑡 and 𝑡 . 
49
cross-sectional data, rather than panel data. Under this condition, the DID-technique would not be able 
to eliminate the time-invariant unobservable selection effects (𝜇), and would provide a biased estimate 
of 𝜔. 
Since the case of Tajikistan imposes the condition, where for all farms, 𝑌𝑗𝑡 = 0 ∀ 𝑡 ≤ 0; the difference￾in-difference estimator is used in a modified setting, where both the first and the second sets of data are 
collected in time periods after the training was completed. This modification of standard practice 
eliminates bias in the estimation of 𝜔 due to time-invariant unobservable selection effects (𝜇), but 
introduces a potential bias due to measurement error. However, the magnitude of any bias due to 
measurement error is reduced to zero as the first set of data is collected closer to the time when 
training commenced (i.e. closer to 𝑡 = 0). Therefore, collecting the first set of data as close as possible 
to 𝑡 = 0 would minimize bias due to this measurement error, while also eliminating bias due to time￾invariant unobservable selection effects (𝜇). A mathematical proof of the elimination of bias due to 
these unobservable selection effects (𝜇) and a minimization of the bias due to measurement error as the 
first set of data is collected closer to the time when training commences can be found in 
Balasubramanya et al. (2018).
Apart from time-invariant unobservable selection effects, there may be time-varying unobservable 
selection effects that also bias the estimation of 𝜔. These are controlled for by including a host of farm￾specific time-varying covariates (𝑋𝑗𝑡) on the right-hand-side of the modified DID equation, as 
demonstrated in Balasubramanya et al. (2018). 
5. Study Design
5.1. Determining sample size, number of clusters and number of observations per cluster
A WUA usually provides water to member-farms in one or two subdistricts9
, enabling each subdistrict to 
be classified as either a treatment subdistrict (where farm managers received longer training) or a 
control subdistrict (where farm managers received shorter training). Power calculations were conducted 
to determine the number of treatment and control subdistricts, the number of observations within a 
 
9
It is unlikely that a dehkan farm is not a member of a WUA. WUAs have been created across all gravity schemes in 
Southern Tajikistan, and data collected from WUAs in a separate study demonstrate that all farms in the command 
area are member-farms (Balasubramanya et al., 2016)
50
subdistrict, and the sample size. The minimum detectable effect (MDE) size was calculated using the 
formula:
MDE = 2.487 σy√(1 − R2) [(1 − ρ) (
1
atbt
+
1
acbc
) + ρ (
1
at
+
1
ac
) ] (2)
In equation (2), ac and at
represent the number of control and treatment subdistricts; and bc and bt
represent the number of member-farms per subdistrict. The variable σy is the standard deviation of the 
outcome variable; and the variable ρ refers to the intra-subdistrict correlation associated with that 
outcome variable. R
2
is the coefficient of determination. Outcome variables and their means, standard 
deviations (σy), and intra-subdistrict correlations (ρ) were taken from the Tajikistan Living Standards 
Measurement Survey (T-LSMS) (World Bank, 2003).10 The coefficient of determination and the level of 
confidence were set at (conventional) rates of 0.8 and 0.95 respectively. Two outcome variables from 
the T-LSMS database were used as proxies for the range of indicators to be assessed. These were: 
proportion of farms that irrigated their agricultural plot; and proportion of farms that believed their plot 
received adequate water supply. 
A sensitivity analysis was subsequently carried out by varying the number of control and treatment 
subdistricts (respectively ac and at
), the number of member-farms per subdistrict (bc and bt
) and the 
sample size. Results using the proportion of farms that irrigated their agricultural plot are reported in 
Figures 1-2. For a given sample size, the MDE is smaller as the number of clusters in the sample 
increases (Figure 2), and as the number of observations per cluster in the sample decreases (Figure 2). A 
study design of 40 treatment and 40 control subdistricts, with 25 farms per subdistrict, emerged 
suitable; the MDE for this study design is an increase of 10% in the proportion of households irrigating 
their plots; and an increase of 6% in the proportion of households that felt that their plots had adequate 
water. For the design of 40 treatment and 40 control clusters, the MDE falls as the sample size increases 
(i.e., the number of observations per cluster increases), but the change in the MDE is rather small 
(Figure 3).11
5.2. Selecting the clusters
 
10 The T-LSMS (2003) was preferred to the T-LSMS from 2007 and 2009 due to a larger sample of rural households 
in the agricultural provinces.
11 Results from using the proportion of households that believed their plot received adequate water supply are 
also similar.
51
Since assignment to longer training was not random, treatment and control subdistricts to be sampled in 
this assessment were selected by constructing propensity scores, and then using a 1:1 matching process 
without replacement to select matched pairs of treatment and control subdistricts. A pre-sampling 
survey of all subdistricts where irrigated cultivation of wheat and cotton was practiced (164 of the 406 
subdistricts) was conducted. Of the164 subdistricts, 116 were in Khatlon Province, 21 in Sughd Province 
and 27 in DRS Province. Information on demographic attributes, agricultural practices, land use and farm
attributes, and irrigation infrastructure was collected. Propensity scores were constructed to calculate 
the probability of each subdistrict being treated (i.e., where farm managers received longer training). A 
complete list of attributes that were used to construct the propensity scores and the model of 
treatment can be found in Table 2.12
Using the propensity scores, subdistricts with farms whose managers received longer training were 
matched (using a caliper size of 0.12) to subdistricts with farms whose managers received shorter 
training, without replacement to their nearest neighbor (1:1 match), to select 40 subdistricts of each 
type. The differences between the subdistrict attributes for unmatched and matched subdistricts is 
displayed in Table 3. When unmatched, treatment and control subdistricts displayed significant 
differences on a number of attributes; these differences did not emerge for the matched pairs. 
5.3. Selecting Dehkan Farms
Records of the population of dehkan farms were not available in any government office at the national 
level. Therefore, a census of all dehkan farms in the 80 selected subdistricts was undertaken by the 
research team. Information on the name of the farm, and the name of the manager of the farm was 
collected. In addition, farms were categorized on two key variables: the type of canal from which the 
farm was irrigated (primary, secondary or tertiary); and the farm’s location on that canal (head, middle, 
or tail). These two variables affect water availability at the farm level and may influence participation. 
For example, managers of farms located on the tail of a tertiary canal may be more inclined to attend 
WUA meetings because their access to water is deeply dependent on the actions of farmers at the head. 
On the other hand, these managers may be less likely to pay the irrigation fees if they perceive that they 
 
12 The propensity score also takes into account ethnic composition of subdistricts, the number of rural health 
centers and schools, and the number of agricultural markets in the subdistrict. It also takes into account whether 
land reforms have been completed, and the number of years of tenure of the current subdistrict leader. These 
could affect selection into treatment, and hence were accounted for while selecting the treated and control 
groups. 
52
are unlikely to receive water anyway. A stratified random sampling method using these two 
characteristics was used to select 25 dehkan farms from each of the 80 selected subdistricts, totaling 
2,000 farms. This process randomly selects the nine types of farms in proportion to their numbers in the 
population, producing a representative sample of farms in each subdistrict. 
5.4. Clustering
The sample was selected by first selecting clusters (subdistricts) and then selecting farms within each 
cluster. Consequently, the econometric analysis of the data in the paper has been conducted by 
clustering results at the sub-district level, to account of the fact that two farms within the same cluster 
are likely to be more similar that two farms in different clusters. 
6. Data
A panel data set was collected through surveys conducted with the 2,000 farms. The first survey was 
conducted in 2015 to collect information on the 2014 calendar year. The second survey was conducted 
in 2017 to collect information about the 2016 calendar year.
Respondent: Both surveys were targeted at the farm-managers, since they are the operators of the 
farms and had received the training. However, this was often not possible, due to overseas or rural-to￾urban male migration. For the first survey, if the manager had not migrated and was in residence during 
2014-2015, the manager was interviewed13. If the manager was not in residence during 2014-2015, the 
(listed) worker of the farm who had taken on the operations of the farm was interviewed. For the 
second survey, if the respondent of the first survey was still in residence, they were interviewed.14 If the 
respondent of the first survey had migrated, then the person who had taken on the operations of the 
farm was interviewed. Interviews were scheduled in advance to check whether the person who had 
answered the first survey was available, and to coordinate with the new operator when needed. Data 
were also collected on the gender of the respondent (who was the operator of the farm for that 
calendar year). 
 
13 We did not find any case where the manager who was trained was in residence but was not the operator of the 
farm. 
14 We did not find any cases where the respondent of the first survey was still in residence and was not the prime 
operator of the farm.
53
Attrition between first and second survey: In the first survey, respondents from 1, 957 of the 2,000 
member-farms agreed to participate in the study and were consequently interviewed. The second 
survey was answered by 1,855 of the 1,957 member-farms. Using data collected from the first survey, 
no statistically significant differences were observed between farms in the treatment and control group 
within the subsample of 102 member-farms that did not answer the second survey. The primary reason 
that these member-farms could not be surveyed again was because production on these dehkan farms 
had ceased after the first survey and before the implementation of the second survey most often due to 
male migration. 
Final sample: Data pertaining to 1,855 member-farms are used in the analysis, with 933 farms whose 
managers received longer training (treatment group) and 922 farms whose managers received shorter 
training (control group).
Left-hand-side variables: Indicators pertaining to member-farm participation were constructed to reflect 
the roles as delineated in the WUA Law of 2006. Respondents were asked if the farm has paid its 
irrigation fees and its WUA membership dues for the calendar year. Respondents were also asked to 
report on the number of (listed) workers of the farm who had participated in canal cleaning and the 
number of days of labor each of those workers had contributed, in order to calculate person-days of 
labor that the farm had contributed towards repair and maintenance. Finally, respondents were asked if 
the farm had a signed a contract with the WUA for the calendar year, and if the farm had been 
represented at the WUA planning meetings in the calendar year. Data on these indicators were collected 
in both surveys. 
Right-hand-side variables: Data on farm membership size and demographics, cotton acreage, and 
cultivation of other crops were collected. These data were also elicited in both surveys. Table 4 provides 
summary statistics for these key variables. 
In 2014, a male who was not directly trained operated 30% of farms in the treatment group and 35% of 
farms in the control group. In 2016, males who were not directly trained operated 29% of farms in the 
treatment group and 31% of farms in the control group. Regarding the gender of the farm operator, 
females operated 10% of the treatment group farms and 11% of the control group farms in 2014. For 
2016, these values changed to 16% and 19% for treatment group and control group respectively. The 
number of (listed) workers of the farm did not change significantly within and between groups across 
54
years; they fell from ~ 7 workers to ~6 workers for the treatment group and from ~8 workers to ~7 
workers for the control group. 
The gender composition of the workers of the farm listed on the farm-title changed within and between 
groups; while 46% and 48% of listed farm-workers in the treatment and control groups during 2014 
were female, 53% and 51% of listed farm-workers in the treatment and control groups during 2016 were 
female. Their difference-in-difference ((53-46) – (51-48) ) is 4%, and is significant at 5%. The share of 
workers of the farm listed on the farm-title spending the majority of their working time on the farm 
increased within both groups but not between groups. While 79% of listed farm-workers in the 
treatment group and 72% of listed farm-workers in the control group worked a majority of their time on 
the farm during 2014, these increased later to 87% and 82% for the treatment and control groups, 
respectively. The number of households associated with the member-farms did not change, staying at 
around three households for both groups during both surveys. The area of land covered by the title of 
the member-farms also did not change much between the two surveys for both groups. This is because 
rental markets for land are not well-developed in Tajikistan.15
The cultivated area of the member-farms did not significantly change within and between the two 
groups over the two surveys. In 2014, the treatment group cultivated 4.19 ha per member-farm on 
average, and the control group cultivated 4.43 ha; these changed to 4.01 ha and 4.48 ha in 2016 for the 
treatment and control group respectively. Irrigated area of the member-farms also did not significantly 
change within and between the two groups. In 2014, the treated group irrigated 4.15 ha and the control 
group irrigated 1.05 ha on average; these changed to 3.95 ha and 4.18 ha in 2016. The share of member￾farms cultivating cotton remained the same within each group; in 2014, 67% of treatment member￾farms and 47% of control member-farms cultivated cotton, and in 2016, these changed to 70% of 
treatment member-farms and 49% of control member-farms. The area under cotton cultivation also 
remained the same between both within and between the groups. In 2014 the treatment member￾farms cultivated 3.27 ha of cotton on average and the control member-farms cultivated 3.95 ha of 
cotton on average. In 2016, the treatment member-farms were cultivating 3.18 ha of cotton while the 
control member-farms were cultivating 3.81 ha of cotton, on average.
 
15 In the first survey, the per-farm land rented was 0.14 ha for the treatment group and 0.13 ha for the control 
group. In the second survey, these numbers were 0.15 ha and 0.09 ha for the treatment and control group 
respectively. 
55
The following variables were only measured during the first survey, because they either changed at the 
same rate over time for both groups, or were time-invariant. Treatment and control member-farms 
were of the similar age, with treatment member-farms ~4.83 years old (standard deviation (sd) = 0.32) 
in 2015 and control member-farms around 5.63 years old in 2015 (sd = 0.47). The average treatment 
member-farm was ~1.75 km away from the nearest road (sd = 0.27), while the average control member￾farm was 2.03 km away (sd = 0.33). The age of the member-farm manager was similar for both groups, 
with managers around 48 years of age in 2015 (standard deviation of 0.43 and 0.48 for the treatment 
and control group respectively). The treatment group had a slightly higher share of member-farm 
managers who had completed secondary education, with 29% as compared to 24% for the control 
group. These variables were also included on the right-hand-side of the difference-in-difference 
estimating equations but the estimated coefficients are not reported in the paper because, being time￾invariant, they are eliminated by the difference-in-difference technique during estimation.
7. Results
7.1. Causal effects of longer training on participation and cooperation
Member-farms whose managers were provided with longer training were 8% more likely to pay their 
membership fees than member-farms whose managers were provided with shorter training (p <0.10) 
(Table 5). These member-farms contributed seven more person-days of labor per member-farm (p 
<0.01) towards pre-irrigation season cleaning of canals, were 19% more likely to have signed a contract 
with their WUAs (p <0.01) and 9% more likely to attend the WUA meetings (p<0.05). 
The coefficients of determination (R-squared) in these regressions are low, because these are not 
ordinary least squares regressions with cross-sectional data. Since these regressions are difference-in￾differences (fixed effects) using panel data which is clustered at the subdistrict, the F-statistic is a more 
reliable indicator of explanatory power. These values are below 0.1 for the regressions where the effect 
of training is positive and significant at the usual levels. 
7.2. Effect on participation when farm was operated by non-trained male worker
When the farm was operated by a male workers who had not been directly trained, that farm 
contributed two fewer man-days of labor than when the farm was operated by the manager who had 
been directly trained (p <0.1; Table 5). However, the likelihood of a farm paying its irrigation fees, WUA 
56
membership fees, signing a contract with the WUA and being represented at WUA planning meetings 
was not affected when the farm was operated by a non-trained male worker (Table 5). 
7.3. Effect on participation when farm was operated by female worker
The gender of the operator of the farm was significant in determining participation. As seen in Table 5, a 
farm operated by a female worker was 9% less likely to pay its membership fees (p<0.01); 11% less likely 
to sign a contract (p<0.05); and 3% less likely to attend WUA meetings (p<0.1). 
8. Discussion
In countries such as Tajikistan, formal participatory institutions are being newly developed after the 
state-control era of the Soviet Union. At the same time, larger macroeconomic and labor-market forces 
are changing the gender-based roles in agriculture. Understanding the role of training in enhancing 
participation in WUAs, and examining diffusion of information on participation from trained farmers to 
other male and female workers is important to provide evidence for program design and policymaking.
Quantitative assessment of participation in newly created community management institutions is 
challenging because control groups and baselines are often difficult to identify. Selection into 
participatory organizations is not random, complicating the isolation of causal effects from observable 
and unobservable confounders. Further, when participatory institutions are created for the first time, all 
participation variables take values of zero at baseline because the intervention also created the notion 
of membership, further complicating the elimination of unobservable confounders. 
In this paper, quasi-experimental methods are used to construct a control group by calculating 
propensity scores at the cluster level, and then selecting matched pairs of treatment and control 
clusters without replacement, to control against bias due to selection effects at the cluster level. The 
first measure of participation is made after the newly created organizations started functioning, so that 
participation could be measured and did not have to take a default value of zero. Using the difference￾in-difference technique with right-hand side indicators where both sets of data are collected after 
intervention began introduces a potential bias due to measurement error, but controls against bias due 
to selection effects at the farm-level. Despite the measurement bias, the estimates provide empirical 
evidence for the effects of longer training on participation and the diffusion of information from trained 
managers to non-trained workers in a context where almost all trained managers were male.
57
In this paper, the results demonstrate that farms that had longer training had a higher probability of 
paying their membership fees, signing a water contract, and attending WUA meetings. Information on 
participation likely diffused from trained male managers to untrained male workers, but not from 
trained male managers to female workers. This is demonstrated by the absence of a significant effect on 
participation when farms were operated by male workers who were not directly trained, and by the 
presence of a significantly lower effect on participation when farms were operated by female workers. 
Such evidence is useful for programming purposes. Since longer training produces greater participation, 
supplemental or refresher-training modules can be designed to provide targeted training in areas where 
shorter training was provided. If more female workers are likely to operate farms, then investing in the 
human capital of female workers by directly training them in participatory management may be needed, 
rather than relying on traditional methods of training (male) lead farmers and expecting diffusion to 
other farmers across gender lines. In addition, in the current context where most migrants are males, 
the functioning of young participatory organizations such as WUAs may be enhanced if knowledge is 
housed with female workers. These trainees are economically less mobile, and have to continue farming 
to feed their families even while males are absent. 
Despite the possibility of bias in results from delaying the inception of data collection, evidence that is 
less than first-best can still be useful for adaptive sectoral development when institutions are young.
Acknowledgments
This research is funded through a project titled ‘Impact of Water Users Associations on Water and Land 
Productivity, Equity and Food Security in Tajikistan’ funded by the United States Agency for International 
Development under USAID Grant Number AID-BFS-G-11-00002. The opinions expressed herein are those 
of the author and do not necessarily reflect the views of the U.S. Agency for International Development. 
I would like to thank Ted Horbulyk for comments and suggestions; Joseph P.G. Price for research 
assistance; and Sohrob Aslamy and Katherine MacDonald for supervising data collection. This study was 
co-designed with Marie-Charlotte Buisson. I would also like to thank participants at a World Bank 
Brown Bag Seminar organized by the Inclusion in Water Program of the Water Global Practice, for 
comments on this work. All errors are mine.
Bibliography
58
Abadie, A. and Gardeazabal, J. (2003). ‘The economic costs of conflict: a case study of the Basque 
country’. American Economic Review, volume 93, pp.113-132.
Alaverdyan, A. and Houston, J.E. (2004). ‘Water user association in Armenia: current situation and future 
trends’. Selected Paper prepared for presentation at the Western Agricultural Economics Association 
Annual Meeting. Honolulu, Hawaii, 30th June-2
nd July, pp. 1-32.
Aydogdu, M. H., Yenigun, K., and Aydogdu. M. (2015). ‘Factors affecting farmers’ satisfaction from 
Water Users Association in the Harran Plain-GAP region, Turkey’. Journal of Agricultural Science and 
Technology, volume 17, pp. 1669-1684.
Balasubramanya, S, M-C Buisson, P Saikia, K MacDonald, S Aslamy, T Horbulyk, C Hannah, M Yakubov 
and A Platonov (2016). Impact of water user associations on water and land productivity, equity and 
food security in Tajikistan. Baseline report, prepared for the United States Agency for International 
Development (USAID), Colombo: International Water Management Institute, pp. 1–127. 
https://agrilinks.org/sites/default/files/resource/files/Feed%20the%20Future%20Tajikistan%20Water%
20User%20Associations%20Impact%20Evaluation%20-%20B.pdf (last accessed15 June 2017).
Balasubramanya, S., Price, J.P., and Horbulyk, T. (2018). Impacts assessments without true baselines: 
assessing the relative effects of training on the performance of water user associations in Southern 
Tajikistan. Water Economics and Policy, online ready: doi: 10.1142/S2382624X18500078. 
Beaman, L. Dillon, A. (2018). ‘Diffusion of agricultural information within social networks: evidence on 
gender inequalities from Mali’. Journal of Development Economics, online ready: doi: 
10.1016/j.jdeveco.2018.01.009. 
Beresford, P. (2010). ‘Public partnerships, governance and user involvement: a service user perspective’. 
International Journal of Consumer Studies, volume 34, issue 5, pp. 495-502.
Bovaird, T., Van Ryzin, G.G., Loeffler, E., Parrado, S. (2015). ‘Activating citizens to participate in collective 
co-production of public services’. Journal of Social Policy, volume 44, issue 1, pp. 1-23.
Brady, M. K. and Cronin Jr., J.J. (2001). ‘Some new thoughts on conceptualizing perceived service quality: 
a hierarchical approach’. The Journal of Marketing, volume 65, number 3, pp. 34-49.
59
Cieslewska, A. (2010). ‘Tradition and poverty reduction – mahalla and its significance in development 
process in Tajikistan’. International Conference on Eurasian Economies 2010, University of Warsaw, pp. 
258-263.
Clausen, A.S., Jors, E., Athuhaire, A., Thomsen, J.F. (2017). ‘Effect of integrated pest management 
training on Ugandan small-scale farmers’. Environmental Health Insights, volume 11, pp.1-10.
Cleaver, F. (1999). ‘Paradoxes of participation: questioning participatory approaches to development’. 
Journal of International Development, volume 11, issue 4, pp. 597-612.
Feder, G., Murgai, R., Quizon, J.B. (2004). ‘The acquisition and diffusion of knowledge: the case of pest 
management training in farmer field schools, Indonesia’. Journal of Agricultural Economics, volume 55, 
number 2, pp. 221-243.
Fernández-Pacheco, D.G., Ferrández-Villena, M., Molina-Martínez, V., Ruiz-Canales, A. (2015). 
‘Performance indicators to assess the implementation of automation in water user associations: A case 
study in southeast Spain. Agricultural Water Management, volume 151, pp. 87-92,
Food and Agriculture Organization of the United Nations (FAO). (2012). ‘AQUASTAT: Tajikistan’. 
http://www.fao.org/nr/water/aquastat/countries_regions/TJK/. (last accessed 9th January 2017).
Food and Agriculture Organization of the United Nations (FAO). (2018). ‘Gender and Land Rights Data 
Base: Tajikistan’. http://www.fao.org/gender-landrights-database/country-profiles/countries-list/land￾tenure-and-related-institutions/en/?country_iso3=TJK (last accessed 23rd January 2018).
Fuglie, O., Kascak, C.A.(2001). ‘Adoption and diffusion of natural-resource conserving agricultural 
technology’. Review of Agricultural Economics, volume 23, number 2, pp. 386-403.
Gronroos C. (1982). ‘An applied service marketing theory’. European Journal of Marketing, volume 16, 
number 7, pp. 30–41.
Gunchinmaa, T. and Yakubov, M. (2009). ‘Institutions and transition: does a better institutional 
environment make water users associations more effective in Central Asia?’ Water Policy, volume 12, 
number 2, pp. 165-185.
60
Gurung, M.B. (1992). ‘Concept, procedure and strategy of participatory management program for large 
irrigation projects under ILC financing, Western Regional Irrigation Directorate’, in Raj Pant, S., Valera, A. 
and Pradham, U. (editors). Proceedings of the National Workshop on Participatory Management in 
Agency-managed Irrigation Systems in Nepal. Nepal: Department of Irrigation and International 
Irrigation Management Institute, pp. 27-34.
Hu, X-J., Xiong, Y-C., Li, Y-J., Wang, J-X., Li, F-M., Wang, H-Y. and Li, L-L. (2014). ‘Integrated water 
resources management and water users’ associations in the arid region of northwest China: a case study 
of farmers’ perceptions’. Journal of Environmental Management, volume 145, pp. 162-169.
Johnson, S.H. and Stoutjesdijk, J. (2008). ‘WUA training and support in the Kyrgyz Republic’. Irrigation 
and Drainage, volume 57, pp. 311-321.
Kazbekov, J., Abdullaev, I., Manthrithilake, H., Qureshi, A., Jumaboev, K.(2009). ‘Evaluating planning and 
delivery performance of Water User Associations (WUAs) in Osh Province, Kyrgyzstan’. Agricultural 
Water Management, volume 96, issue 8, pp.1259-1267,
Martin, G.P. (2009). ‘Public and user participation in public service delivery: tensions in policy and 
practice’. Sociology Compass, volume 3, number 2, pp. 310-326.
Mukhamedova, N., Wegerich, K. (2018). ‘The feminization of agriculture in post-Soviet Tajikistan’. 
Journal of Rural Studies, volume 57, pp. 128-139.
Nagrah, A., Chaudhry, A.M. and Giordano, M. (2016). ‘Collective action in decentralized irrigation 
systems: evidence from Pakistan’. World Development, volume 84, pp. 282-298.
O’Hara, S.L. (2000). ‘Lessons from the past: water management in Central Asia’. Water Policy, volume 2, 
pp. 365-384.
Ostrom, E. (1990). Governing the commons: the evolution of institutions for collective action. Cambridge, 
United Kingdom: Cambridge University Press, pp. 1-294.
Parasuraman A., Zeithaml V. and Berry L. (1985). ‘A conceptual model of service quality and its 
implications for future research’. Journal of Marketing, volume 49, pp. 41–50.
61
Pemsl, D.E., Waibel, H., Witt, R. (2006). ‘Diffusion of information among small-scale farmers in Senegal: 
the concept of Farmer Field Schools’. Proceedings of the German Development Economics Conference, 
Berlin 2006 / Verein fürSocialpolitik, Research Committee Development Economics, No. 30. Available at: 
http://hdl.handle.net/10419/19855
Pratchett, L. (1999) ‘New fashions in public participation: towards greater democracy?’ Parliamentary 
Affairs, volume 52, number 4, pp. 616-633.
Qiao, G., Zhao, L. and Klein, K.K. (2009). ‘Water user associations in Inner Mongolia: factors that 
influence farmers to join’. Agricultural Water Management, volume 96, issue 5, pp. 822-830.
Rola, A.C., Jamias, S.J., Quizon, J.B.(2002). ‘Do Farmer Field School Graduates Retain and Share What 
They Learn? An Investigation in Iloilo, Philippines’. Journal of International Agricultural and Extension 
Education, volume 9, number 1, pp. 65-76.
Shahriari, H. (2009). Activities, water and gender in Tajikistan’s rural sector: a social assessment of 
Konibodom, Bobojon Ghafurov, and Yovon. Social Development Department, Europe and Central Asia 
series. Washington, DC: World Bank. Available at: 
http://documents.worldbank.org/curated/en/364511468117242593/pdf/514450WP0TJ0So10Box34202
8B01PUBLIC1.pdf (last accessed June 15, 2017). 
Tellez Foster, E., Rapoport, A. and Dinar, A. (2017). ‘Groundwater and electricity consumption under 
alternative subsidies: evidence from laboratory experiments’. Journal of Behavioral and Experimental 
Economics, volume 68, pp. 41-52. 
T-LSMS. (2003). Tajikistan Living Standards Survey. Available at: 
http://microdata.worldbank.org/index.php/catalog/278. 
Tripp, R., Wijeratne, M., Piyadasa, V.H. (2005). ‘What should we expect from farmer field schools? A Sri 
Lanka case study’. World Development, volume 33, number 10, pp. 1705-1720. 
Wegerich, K. (2000). ‘Water user associations in Uzbekistan and Kyrgyzstan: study on conditions for 
sustainable development’. Occasional paper – Water Issues Study Group, London: School of Oriental and 
African Studies, University of London, number 32, pp. 1-23.
62
Yakubov, M. (2011). ‘Assessing irrigation performance from the farmers’ perspective: a qualitative 
study’. Irrigation and Drainage, volume 61, issue 3, pp. 316-329.
Yap-Salinas, L.H. (1994). ‘Converging factors in the successful transfer of irrigation management 
responsibilities to water users’ associations in the Dominican Republic’. Paper presented at the 
International Conference on Irrigation Management Transfer, 20th-24th September, Wuhan, China, pp. 
117-127.
Zeithaml, V., Parasuraman A. and Berry L. (1990). Delivering Quality Service Balancing Customer 
Perceptions and Expectations. New York: The Free Press, pp. 1-226.
63
Figure 6: Relationship between Minimum Detectable Effect and number of clusters for a given sample 
size
Figure 7: Relationship between MDE and number of observations per cluster for a given sample size
Figure 8: Relationship between MDE and sample size (n>=2,000)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0 20 40 60 80 100
MDE
Total # clusters
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0 50 100 150 200 250
MDE
# observations / clusters
0.05
0.10
0.15
0.20
0.25
0 1000 2000 3000 4000 5000 6000 7000 8000
MDE
Sample size
64
Table 1: Participation and satisfaction indicators
Irrigation Fees
Fees were paid for both irrigation seasons in the year
Farm owed arrears in irrigation fees in the calendar year
WUA membership fees
Membership fees were paid for the calendar year
Participation in pre-irrigation cleaning of canals
# of days donated by farm towards cleaning
# of people from the farm who participated
Legal relations
Farm signed a contract with the WUA
Farm member(s) attended WUA meetings
65
Table 2: Constructing propensity scores 
Logit
Treatment 
Subdistrict
Population of the subdistrict in 2014 0.00 (0.00)
Number of villages 0.033 (0.05)
Total area of the subdistrict 0.00 (0.00)***
Majority of population Tajik (dummy) -0.69 (0.54)
Number of secondary schools in subdistrict 0.07 (0.08)
Number of rural health centers in the subdistrict -0.54 (0.19)***
Number of agricultural markets in the subdistrict 0.40 (0.29)
Chairman born in the subdistrict (dummy) 0.16 (0.64)
Number of years of election of the chairman -0.07 (0.07)
Elevation of the subdistrict (m ASL) 0.00(0.00)
Sandy soil (dummy) 0.00 (0.55)
Deep groundwater level (dummy) -0.81 (0.52)
Land reform completed (dummy) 2.54 (0.69)***
Cotton main crop of the subdistrict (dummy) 2.84 (0.66)***
Subdistrict irrigated by gravity system (dummy) 0.61 (0.67)
Constant -3.89 (1.49)***
Pseudo R2 0.402
Sample size 164
66
Table 3: Differences between treatment and control group for unmatched and matched sample
Treatment - Control
Unmatched Matched
Mean (Std Dev) Mean (Std Dev)
Population of the subdistrict in 2014 797.18 (1446.5) -845.48 (2000.07)
Number of villages -1.34 (1.08) -0.775 (1.45)
Total area of the subdistrict 2770.21 (2888.16) 1537.04 (3992.63)
Majority of population Tajik (dummy) -0.13 (0.07)* -0.1 (0.09)
Number of secondary schools in subdistrict 0.51 (0.69) -0.125 (0.98)
Number of rural health centers in the 
subdistrict
-1.03 (0.33)*** -0.125 (0.28)
Number of agricultural markets in the 
subdistrict
0.09 (0.15) -0.125 (0.22)
Chairman born in the subdistrict (dummy) 0.05 (0.07) -0.1 (0.09)
Number of years of election of the chairman 0.12 (0.62) 0.4 (0.81)
Elevation of the subdistrict -178.22 (56.08)*** -56.48 (70.73)
Sandy soil (dummy) 0.04 (0.07) -0.025 (0.1)
Deep groundwater level (dummy) -0.23 (0.08)*** -0.11 (0.11)
Land reform completed (dummy) 0.34 (0.08)*** 0.125 (0.09)
Cotton main crop of the subdistrict (dummy) 0.48 (0.08)*** 0.17 (0.11)
Subdistrict irrigated by gravity system (dummy) 0.06 (0.07) -0.025 (0.09)
Number of observations 164 80
67
Table 4: Summary statistics for first and second survey data
First Survey 2014 Second Survey 2016
Treatment Control Treatment Control Difference-in 
Difference
N Mean (Std Err) N
Mean (Std 
Err) N
Mean (Std 
Err) N
Mean (Std 
Err)
Mean (Std 
Err)
Farm operated by non-trained male 933 0.30 (0.02) 922 0.35 (0.02) 933 0.29 (0.02) 922 0.31 (0.02) -0.02 (0.02)
Farm operated by female 933 0.10 (0.02) 922 0.11 (0.01) 933 0.16 (0.02) 922 0.19 (0.02) 0.00 (0.02)
Number of members 928 6.71 (0.47) 908 7.84 (0.70) 928 6.07 (0.39) 908 6.96 (0.46) 0.24 (0.60)
Share of members that were female 910 0.46 (0.01) 875 0.48 (0.01) 910 0.53 (0.01) 875 0.51 (0.01) 0.04 (0.01)**
Share of members spend majority of 
time 911 0.79 (0.02) 833 0.72 (0.02) 911 0.87 (0.01) 883 0.82 (0.02) -0.03 (0.03)
Number of households 931 3.18 (0.26) 911 3.20 (0.31) 931 3.06 (0.27) 911 3.17 (0.22) -0.09 (0.29)
Area with official title (ha) 933 5.15 (0.44) 921 5.33 (0.71) 933 4.53 (0.29) 911 5.38 (0.90) -0.65 (0.54)
Cultivated area (ha) 932 4.19 (0.25) 918 4.43 (0.54) 932 4.01 (0.26) 918 4.48 (0.55) -0.23 (0.29)
Irrigated area (ha) 933 4.15 (0.24) 917 4.05 (0.51) 933 3.95 (0.25) 917 4.18 (0.47) -0.32 (0.29)
% farms cultivating cotton 929 0.67 (0.03) 919 0.47 (0.06) 929 0.70 (0.03) 919 0.49 (0.05) 0.01 (0.02)
Area under cotton cultivation (ha) 545 3.27 (0.29) 382 3.95 (0.73) 545 3.18 (0.29) 382 3.81 (0.84) 0.05 (0.27)
68
Table 5: Effect of longer training on farmer participation
Irrigation fees Membership fees # man-days Farm signed Farm attended
paid paid labor
a water 
contract WUA meetings
Longer training -0.06(0.05) 0.08 (0.05)* 7.10 (2.40)*** 0.20 (0.05)*** 0.09 (0.04)**
Farm operated by non-trained male -0.02 (0.04) -0.02 (0.02) -2.43 (1.85)* -0.02 (0.03) -0.01(0.02)
Farm operated by female 0.03 (0.05) -0.09 (0.03)*** 3.21 (1.94) -0.11 (0.04)** -0.03 (0.01)*
Number of members -0.00 (0.00) -0.00 (0.00) -0.09 (0.11) 0.00 (0.00)** 0.00 (0.00)
Share of members that were female -0.05 (0.05) -0.01 (0.03) 2.88 (2.29) -0.00 (0.05) 0.01 (0.03)
Share of members that work 
permanently -0.03 (0.03) -0.00 (0.03) -2.96 (2.47) 0.01 (0.05) -0.03 (0.03)
Number of households 0.01 (0.00)* 0.00 (0.000 0.03 (0.28) -0.00 (0.00) -0.00 (0.00)
Area with official title 0.00 (0.00) 0.00 (0.00)* 0.03 (0.08) 0.00 (0.00) -0.00 (0.00)
Cultivated area 0.01(0.01) 0.00 (0.00) 0.52 (1.03) 0.01 (0.01)* 0.01 (0.01)
Irrigated area -0.02(0.01)** -0.00 (0.00) -0.39 (0.92) 0.01 (0.01)* -0.01 (0.00)*
Area under cotton cultivation -0.00 (0.01) -0.00 (0.00) -0.20 (0.37) -0.00 (0.00) 0.00 (0.00)
Number of observations 1753 1753 1561 1753 1753
F-statistic F (20, 60) = 1.21 F(20, 60)= 0.98 F(20, 60)=2.34 F(20,60)=3.48 F(20,60)=1.51
Prob > F 0.28 0.57 0.01 0.00 0.09
R-squared 0.02 0.04 0.02 0.09 0.03
Updated June 2018 69
Appendix 3 
Buisson, M.C.; Balasubramanya, S. 2018. The effect of irrigation service delivery and training in 
agronomy on crop choice in Tajikistan. In review at Land Use Policy.
Updated June 2018 70
Title: The effect of irrigation service delivery and training in agronomy on crop choice in Tajikistan
Keywords: irrigation delivery, agricultural extension, water user associations, agricultural production,
Tajikistan
Abstract: The aim of this paper is to analyze the effect of irrigation delivery services and agricultural 
extension services on crop choice in southern Tajikistan. This analysis is motivated by the government’s 
recent efforts to address the country’s severe malnutrition problem by supporting changes in irrigation 
service delivery and agronomy to increase diversity in agricultural production and consumption, in an 
environment where the cultivation of cotton had, until recently, been mandatory. Water management 
in Tajikistan has largely been transferred to the community through the creation of water users’ 
associations (WUAs) between 2011 and 2013. While all WUAs received training to improve irrigation 
delivery services, some also received training in cropping alternatives and improving cultivation 
practices through agricultural extension services. Through specific empirical analysis conducted on a 
primary panel dataset of 1,855 farms in southern Tajikistan, we identify the extent to which 
improvements in irrigation services and agronomy training through extension services affect decisions 
pertaining to cultivated areas of cotton and wheat (the traditional crops) and the cultivated area and 
number of (newer) high-value crops. We also examine the effect of water delivery and agricultural 
extension services on crop diversity and cropping intensity (how often land is used in a calendar year). 
We find that improvements in irrigation delivery services affect cultivated areas of cotton and wheat. 
Cultivation of high value crops is significantly influenced by agricultural extension services. While 
cropping intensity depends on water delivery services, crop diversity depends on extension services. 
From a policy perspective, these results highlight the importance of agricultural programs for stimulating 
agricultural value added in landscapes historically characterized by limited crop choice and a collapse of 
the agricultural sector. 
Updated June 2018 71
1. Introduction
In this article, we analyze the effect of irrigation service delivery and training in agronomy on crop 
choice made on Tajik dehkan farms. Through an empirical analysis, we aim to identify the effect on the 
acreage of cotton and wheat cultivation (the traditional crops); the acreage and number of high-value 
crops (alternatives to traditional crops); crop diversity on the farm; and the cropping intensity of the 
farm. 
An examination of land use is motivated by the government’s recent efforts to address the country’s 
severe malnutrition problem by supporting changes in irrigation service delivery and agronomy, and by 
increasing awareness, to increase diversity in agricultural production and consumption (WFP, 2017; FAO, 
2018), in an environment where the cultivation of cotton had, until rather recently, been mandated. The 
2017 Global Hunger Index suggests that ~30% of the country’s population in undernourished (WFP, 
2017). In 2013, around 26% of children under the age of five were stunted, while 10% suffered from 
wasting (Statistical Agency under the President of the Republic of Tajikistan (SA), Ministry of Health 
(MOH), and ICF International, 2013). The Tajik diet, which is poor in mineral-rich vegetables and fruits 
has been identified as an important factor for persistent malnutrition (FAO, 2018). The Government has 
made efforts to tackle this problem, through programs such as ‘Scaling Up Nutrition’, which aim to 
improve diversity in agricultural production and consumption (FAO, 2018). However, agricultural 
markets continue to support the cultivation of cotton even today (Boboyorov, 2016). An understanding 
of how improvements in irrigation and training in alternative crops influences crop choice and land use 
would be important for designing and managing interventions to support the government’s 
development goals.35,36
When the Soviet Union disintegrated, the government of Tajikistan introduced land reforms in the early 
1990s to confer cultivation rights on households by decollectivizing the Soviet collective farms that 
specialized in cotton cultivation. However, large stretches of agricultural land were left fallow during the 
1990s and early 2000s due to the lack of irrigation services that could manage the needs of thousands of 
smaller farms, rather than one large collective farm (Gunchinma & Yakubov, 2010). With support from 
 
35 Understanding how perceptions and dietary choices can be influenced at the point of food consumption is also 
vital, but is beyond the scope of this paper. 
36 Tajikistan is characterized by migration of rural males, with 48% of rural households in Khatlon Province in 
southern Tajikistan having a male migrant. The World Bank estimates that around 55% of individuals in agriculture 
are now female. An understanding of how this feminization affects labor availability, that in turn influences crop 
choice would also be important for designing and managing future interventions that meet the government’s 
development goals, but is also beyond the scope of this paper.
Updated June 2018 72
the United States Agency for International Development (USAID), participatory irrigation management 
was introduced in 2006, and enabled by legislation defining water user associations (WUAs). WUAs were 
mandated to restore irrigation services by designing irrigation schedules, coordinating water use and 
collecting payments, in order to improve the timeliness of water delivery, the distribution of water, the 
quantity of water delivered, and the condition of canals; and consequently to expand cropping area. 
WUAs were created between 2010 and 2014, both by USAID and by the government in southern 
Tajikistan, and at present, around 400 WUAs are registered and functional. The USAID WUAs differ from 
other WUAs in that they provided longer trainings on irrigation service delivery and water governance 
along with training in improved agricultural practices through agricultural extension services. Other 
WUAs were provided shorter training in irrigation service delivery without any extension information 
(though information on changing and improving cultivation practices may have diffused from USAID 
WUAs to other WUAs). 
Understanding how land use changes are influenced by improvements in irrigation service delivery and 
by training in agronomy can provide important evidence for policy and planning. This paper contributes 
to the literature by providing empirical evidence based on a study design that controls for other factors 
that may influence crop acreage, crop type, diversity and intensity of land use. Propensity scores using 
historical agronomic and socioeconomic data were constructed for sub-districts served by USAID and 
non-USAID WUAs and then these subdistricts were selected for detailed examination in matched pairs, 
to control for observable factors in the past that may drive current production decisions, besides 
irrigation and extension services. In addition, this examination of the effects of irrigation services and 
extension services is based on a panel dataset of farms, rather than relying on a cross-section. A total of 
1,956 dehkan farms, with a equal number served by USAID WUAs and other WUAs were surveyed 
twice—once for the 2014 calendar year and again for the 2016 calendar year—to examine changes in 
the acreage and cropping behavior. The use of panel data allows for identification of more robust 
correlations between the irrigation, extension and decisions pertaining to cultivated areas, crop diversity 
and cropping intensity by controlling for unobserved time-invariant effects that may also affect 
agricultural decisions. Finally, a representative sample of farms in southern Tajikistan within the selected 
subdistricts was selected through a stratified random sampling process.
In spite of this contribution, one limitation of the paper is that the results of the analysis cannot be 
interpreted as causal estimates of the effects or impacts of USAID WUA intervention, which combined 
water management with extension services. This is because of the absence of data on production 
choices before the interventions began. 
Updated June 2018 73
The results suggest that decisions pertaining to cultivated areas of cotton and wheat, and the cropping 
intensity of the farm are dependent on water delivery services, while decisions pertaining to the 
cultivated areas of high-value crops and crop diversity are dependent on agricultural extension services. 
From a policy perspective, the results provide evidence for designing programs that coordinate water 
and agricultural extension interventions for stimulating the agricultural sector.
This article is organized as follows. In the second section, details of the context of the agriculture sector 
in Tajikistan and the WUA interventions are provided. The third section provides a conceptual 
framework that motivates the analysis in the paper. Section 5 provides a description of the econometric 
methods, while Section 5 provides a description of sampling strategy and the data collected. Section 6 
provides results from descriptive and econometric analysis. Finally, the policy implications of the key 
findings are discussed in Section 7.
2. Background
In Tajikistan, agriculture is the main sector of the economy, providing around 50% of the employment 
and contributing around 25% of the country’s gross domestic product (ADB, 2016). After the steady and 
intense agricultural growth during the Soviet period up to 1980, the agriculture sector stagnated in the 
1980s. The post-independence transition phase from 1991 to 1997 was characterized by a decline of 
55% in the gross agricultural output. Land reforms, which began in 1992, dissolved the Soviet-era 
collectives that specialized in the production of cotton, first into smaller collectives and then further into 
individual or family farms, called dehkan farms. A typical dehkan farm is around 3-5 ha in size; and is 
usually farmed by around 3 households37. By 2008, the reforms undertaken after the civil war brought 
agricultural production back to its level at the time of independence (Lerman & Sedik, 2008). Land 
continues to remains state property, but farmers have the right to cultivate dehkan farms and to 
transfer their allotted farms through inheritance (Abbott, 2016). 
As compared to the pre-independence system, these land reforms reduced control by the government 
on production choices. Yet, cotton fiber continues to be the country’s second highest export earner, 
accounting for 15% of the total export revenues in 2014. The Government of Tajikistan continues to 
regulate cotton prices, and all cotton harvests are purchased by a few companies that export to the 
global market. At the country level, areas under cotton cultivation declined in the 1990s and 2000s, but 
production has remained stable after 2008, suggesting an increase in yields (Government of Tajikistan, 
 
37 A household is defined as a set of individuals who live in the same dwelling and share food cooked in the same 
kitchen. 
Updated June 2018 74
2007). At the farm level, cotton is still the most commonly cultivated cash crop, in part because 
cultivating cotton provides inputs such as stalks for heating and crop residue for fodder. The well￾regulated structure of the cotton market; relations between farmers and cotton purchasers that provide 
inputs and loans towards cotton cultivation; and community norms contribute to the prevalence of 
cotton cultivation (Boboyorov, 2012; Boboyorov, 2016; Hofman, 2017) .
Bread, mostly made from wheat flour, is the staple food in Tajikistan and the main source of nutrition, 
as it provides 52% of the daily calorific intake in Tajikistan (Muminjanov et al., 2016). Around 40% of the 
wheat consumed in Tajikistan is imported from Russia (Muminjanov et al., 2016). Tajikistan is the 
leading consumer of wheat per capita in the world (Husenov et al. 2015), with an average consumption 
of 166.3 kilograms (kg) per person per year in 2009. These dietary customs combined with fluctuating 
prices of imported wheat provide strong incentives for farmers to choose to cultivate wheat for self￾consumption. The area under wheat cultivation has more than doubled since independence in 1991. 
This pattern of cotton and wheat cultivation is rather predominant, and is colloquially termed a ‘cotton 
for cash, wheat for food’ model. The entire cotton harvest is sold for cash, while a significant share of 
the wheat harvest is retained for self-consumption.
While water delivery was centrally administered by state irrigation departments for collective farms in 
the Soviet Union, the creation of dehkan farms was accompanied by the introduction of decentralized 
and participatory management of water (Gunchinma & Yakubov, 2010). Tajikistan passed the ‘Law on 
WUAs’ in November 2006 (with support from USAID), under which WUAs were recognized as the 
primary organization responsible for delivering water to dehkan farms. The principal functions of WUAs, 
as mandated by the law, are the operation of local water infrastructure, design and implementation of 
an irrigation schedule to deliver water to farms, repair and maintenance of irrigation canals, collection 
of fees (membership fees and water delivery service fees) and water-related conflict resolution (Sehring, 
2009). Training materials for establishing and supporting WUAs were developed under the Family 
Farming Program (FFP), launched by USAID in 2010, which was incorporated into the Feed the Future 
(FTF) initiative in 2011. USAID set up 70 WUAs in Khatlon Province in southern Tajikistan during 2011-
2013. Local governments (with funding from other international organizations) also used USAID￾developed training materials to establish WUAs in Khatlon during 2011-2013 (and across the country 
thereafter). As compared to other WUAs, USAID WUAs were provided longer training in irrigation 
delivery and management. 
Beyond the primary training related to water delivery, and in the absence of formal agricultural 
extension services in the country, USAID-established WUAS were also imparted training on agricultural 
Updated June 2018 75
technologies both formally through training and informally through farmer gatherings. Information on 
improved cultivation practices and inputs for cotton and wheat (the traditional crops) was provided. In 
addition, demonstration plots for cultivating high value crops such as fruits and vegetables were 
undertaken, and farmers in USAID WUAs were trained in cultivating fruits and vegetables at larger scales 
(such commodities largely made their way into Tajikistan from other republics under the central 
planning economic system). In WUAs set up by the government, such information and trainings were 
not provided; however farms served by non-USAID WUAs may have acquired such information through 
diffusion of information and through WUA exchange visits, which were organized by USAID and the 
government. 
3. Conceptual framework 
Cotton and wheat production, the traditional crops in the irrigated agricultural areas, of Tajikistan faces 
several risks. The productivity of wheat and cotton is likely to fall due to the impacts of climate change in 
the Central Asian region (Mannig et al., 2013; Bobojonov & Aw-Hassan, 2014; Sommer et al., 2013).
38
Cotton is a highly water-dependent cash crop; deterioration of irrigation infrastructure and irrigation 
management decreases the quantities of water available, and the periods during which water is 
available for cultivation, thus affecting farmers’ choices of cultivated area. The production of wheat, 
which is less dependent on water, depends on the types of cultivation practices (Wang et al., 2009; 
Wang et al., 2018); which could be influenced through the provision of extension services. In the 
context of Tajikistan, improvements in water delivery services and introduction of better technologies to 
cultivate cotton and wheat may increase agricultural production by expanding cultivated areas or 
improving yields (Birthal et al., 2015; Kasem and Thapa, 2011).
Improvements in irrigation infrastructure and management may improve water availability and increase 
the cultivation period, allowing crops of different maturation periods to be cultivated on the farm, thus 
increasing the number of crops cultivated and their acreage (Zimmerer, 2014; McCord et al., 2015). In 
addition, agricultural extension can play an important role in affecting and expanding crop choice 
especially in contexts where knowledge about cultivation of alternative crops may be limited (Calub et 
al., 2005; Oladele, 2005). In the Soviet Era, Tajikistan specialized in cultivating cotton and wheat, relying 
on other republics for other commodities under the USSR’s central economic planning system (Pomfret 
2008; Bobojanov et al., 2013). With the mandatory cultivation of cotton relaxed, engagement with 
 
38 Risk management mechanisms such as forward markets and crop insurance are hitherto limited in their 
availability in Tajikistan.
Updated June 2018 76
agricultural extension services may likely expand the variety of crops cultivated (Winters, 2006). 
Extension services have been found to play an important role helping farmers in Thailand diversify crop 
cultivation, when national policy moved away from promoting cultivation of rice (Kasem and Thapa, 
2011). Similarly, in Kenya, exposure to extension officers was found to increase the variety of crops 
cultivated especially for smallholder farmers (McCord et al., 2015). 
Increases in cultivation of number and acreage of other crops in Tajikistan may take place along with 
increases or decreases in the area under cotton and wheat cultivation on the farm. For example, 
Pradhan and Ranjan (2016) demonstrated that improvements in irrigation services, which increased 
water availability, also increased the cultivation of dry season rice in Bangladesh (the traditional crop) 
along with other seasonal crops. In the context of Tajikistan, an increase in the variety of crops 
cultivated can reduce environmental damages generated due to historical mono-cropping of cotton by 
reducing: nutrient loss in soils (Hooper and Vitousek, 1997; Reich et al., 2001); soil salinity (Bobojonov et 
al., 2013); and risks from pests and climate change (Winters et al., 2006). The incentives to increase 
variety of crops are likely to arise more from the perspective of improving quality of recently 
decollectivized land that farmers now bear user rights to, and to cultivate consumption crops that were 
historically imported. Income generating incentives that encourage expansion of the number of crops 
cultivated may also gain importance as the (nascent) agricultural processing sector and markets develop 
(Bobojonov and Lamers, 2008). 
4. Econometric methods
In this section, we analyze the determinants, at the farm level, of the following: cultivated areas of 
cotton, wheat, and other crops; number of crops cultivated (excluding cotton and wheat); crop diversity; 
and cropping intensity. Specifically, we examine the extent to which water delivery services and 
agricultural extension services influence crop production decisions on the farm.
The empirical model estimated is the following:
𝐷𝑖𝑗𝑡 = 𝛽1 𝐹𝐶𝑖𝑗𝑡 + 𝛽2 𝑀𝐶𝑖𝑗𝑡 + 𝛿D1 𝑊𝐷𝑖𝑗𝑡 + 𝛿D2 𝐴𝐸𝑖𝑗𝑡 + η𝑖𝑗𝑡 (1)
𝐷𝑖𝑗𝑡 represents the following (in different regressions): cultivated areas of cotton, of wheat and of other 
crops on farm 𝑖 in subdistrict39 𝑗 in time 𝑡; the number of high-value crops cultivated (beyond wheat and 
cotton) on farm 𝑖 in subdistrict 𝑗 in time 𝑡 ; the cropping intensity of farm 𝑖 in subdistrict 𝑗 in time 𝑡; the 
Margalef’s index of diversity for farm 𝑖 in subdistrict 𝑗 in time 𝑡. 
 
39 Tajikistan is divided into 4 provinces. Each province is divided into districts. Each district is divided into 
subdistricts. WUAs typically provide water to dehkan farms in 1-2 subdistricts. 
Updated June 2018 77
𝑊𝐷𝑖𝑗𝑡 represents a vector of variables that contain water delivery characteristics for farm 𝑖 in subdistrict 
𝑗 in time 𝑡. In an ideal world, the presence of meters and gauges would have enabled the inclusion of 
quantity of water delivered during peak irrigation time. In the absence of such meters and gauges, 
perceptions regarding fairness of water sharing, timeliness of irrigation services, quantity of water 
delivered, condition of the watercourse canal, and condition of the distributary canal are included, 
assuming that farmers’ practices are correlated with their perceptions. 
𝐴𝐸𝑖𝑗𝑡 is a vector capturing the agricultural extension services received by farm 𝑖 in subdistrict 𝑗 in time 𝑡. 
While such training was imparted to farms that are served by USAID WUAs, such information may have 
spread to other areas, with farmers in other areas visiting demonstration plots in USAID WUAs. 
Therefore, an indicator variable that denotes whether farm members were directly trained and the 
frequency of interaction with other members of an agricultural or water group is included in this vector.
𝛿D1 and 𝛿D2 are the coefficients of interest to be estimated. η𝑖𝑗𝑡 is an error term which is assumed to 
have the following structure
𝜂𝑖𝑗𝑡 = 𝜗𝑖𝑡 + 𝛼𝑗 + 𝜏𝑡 (2)
η𝑖𝑗𝑡 is the term of error which is structured to have a time component (𝜏𝑡), a subdistrict component 
(𝛼𝑗)and a farm component (𝜗𝑖𝑡).
Other variables that also affect production decisions are included. 𝐹𝐶𝑖𝑗𝑡 refers to a vector of the 
characteristics of farm 𝑖 in subdistrict 𝑗 in time 𝑡, with the area of the farm, the distance to the road and 
the age of the farm included in this vector.40 𝑀𝐶𝑖𝑗𝑡 represents a vector with the characteristics of the 
shareholders of farm 𝑖 in subdistrict 𝑗 in time 𝑡, which include the number of farm shareholders, the 
proportion of female shareholders, and the age, education and sex of the manager of the farm (who is 
the farm head). 
In equation (1) the variables related to the selling price of the commodity, the consumer price of the 
product and the input prices have been voluntary omitted, because prices of cotton and wheat do not 
vary much across different locations. For non-regulated crops, the domestic price of high-value crops is 
influenced by world prices, which implies that prices are unknown in advance and unlikely to be used to 
make cultivation decisions (Bobokhonov et al., 2017).
The empirical model equation (1) is estimated using a random-effects tobit model when the 
determinants of cultivated areas of cotton, wheat and high-value crops; and the Margalef’s index of 
 
40 Farmers were not able to influence or decide the area of their farm when de-collectivization occurred. Also, land 
acquisition through rental or purchase is not common. Therefore, the area of the farm is not likely to be 
endogenous. 
Updated June 2018 78
diversity41 are examined, to take into account truncated values. A generalized least squares random￾effects estimator is used when determinants of the number of high-value crops cultivated, and the 
cropping intensity42 are examine. In all cases, standard errors are clustered at jamoat level to allow for 
intragroup correlation of the standard errors. 
5. Study design and data
This study is based in subdistricts of southern Tajikistan that are irrigated by gravity schemes. While 
farms in some subdistricts are served by USAID WUAs, farms in other subdistricts are served by WUAs 
established by the government. A pre-sampling survey was conducted in all subdistricts served by WUAs 
in gravity schemes in Tajikistan in 2014. Data from 164 subdistricts in Khatlon (116), Sughd (21) and 
Districts of Republican Subordination (27) provinces were collected. Information on land use and 
agricultural practices; irrigation infrastructure and schemes; the presence and characteristics of WUAs; 
and demographic characteristics was collected from the administrative office of each of the subdistricts. 
Based on this data, propensity scores were constructed to calculate the probability of each subdistrict 
being treated by USAID WUAs. A complete list of attributes used to construct these propensity scores 
can be found in Table 1. Using the propensity scores, subdistricts served by USAID WUAs were matched 
(using a caliper of 0.12) to those that are served by government WUAs, without replacement to their 
nearest neighbor, to select 80 subdistricts—40 served by USAID and 40 served by government WUAs. 
Matching was conducted to control for selection bias due to observable characteristics, since 
assignment of farms to USAID and non-USAID WUAs was not random. The propensity score matching 
improves the comparability of farms served by USAID and government WUAs.
Next, dehkan farms were selected within each sampled subdistrict. In the absence of consolidated lists 
of irrigated farms at the local or national level, a census of farms was conducted by the study team, to 
collect information on the name of the farm, name of the manager of the farm, type of canal serving the 
farm (primary, secondary, tertiary), and the farm’s location along the canal (head, middle or tail). These 
characteristics together make up nine types of farms (three canal types by three canal locations). A 
stratified random sampling method was used to select 25 dehkan farms from each of the selected 
subdistricts (a sample size of 2,000 farms in all). This sampling method randomly selects the nine types 
 
41 The Margalef’s Index of Diversity (MID), commonly used for measuring biodiversity, is defined in this context as 
the number of crops cultivated per hectare of land area, and is calculated as: 
MID=(N-1)/(ln(A+1)), where N is the number of crops cultivated and A is the total area cultivated by the farm.
42 Cropping intensity of a plot on the farm was calculated as the gross cropped area divided by the net sown area, 
multiplied by 100 per plot. This indicates the number of times the same plot on the farm was used during one 
agricultural year.
Updated June 2018 79
of farms in proportion to their numbers in the population, and allows for a spatially representative 
sample and for robust econometric identification of correlations between water delivery and 
agricultural extension interventions, and crop choice. Power calculations, taking into account intra￾subdistrict correlation and non-response rates, conducted before the sample was selected, were 
undertaken to determine the appropriate sample size, the number of USAID and non-USAID subdistricts 
(clusters), and the number of farms per USAID and non-USAID subdistrict.
Panel data were collected from this sample. The person leading operation on the farm was interviewed 
a first time in 2015 and a second time in 2017. The 2015 survey collected information for the 2014 
calendar year, while the 2017 survey collected information for the 2016 calendar year. The 2015 survey 
was answered by 1,956 farms. The 2017 survey was answered by 1,855 farms. Information on number 
and cultivated areas of crops, farm characteristics, and characteristics of the shareholders of the farm; 
perceptions regarding improvements in water delivery and the condition of infrastructure, training 
received from extension services, and frequency of interactions with an agricultural or water community 
group were collected in both surveys. Table 2 contains a summary of these variables, which are also 
used during the econometric analysis in section 7 to control for other factors that affect choices 
pertaining to cultivated areas, number of crops cultivated, crop diversity, and cropping intensity.
6. Results
6.1. Descriptive statistics
The average farm area was 4.40 ha (standard deviation(s.d) of 8.70) in 2014, and 4.24 ha (s.d 8.15) in 
2016. In 2014, farms had an average of ~7.4 members (s.d 11.10), which slightly fell to 6.55 in 2016 (s.d 
9.77). Around 47% of farm members were female in 2014; this increased to 52% in 2016. In 2015, 13% 
of farms were headed by women, this increased to 38% in 2018. 
Cotton and wheat are the most commonly cultivated crops by farms. In the sample, 56.1% of farms 
cultivated cotton and 58.5% cultivated wheat in 2014 (Figure 1). This scenario did not change much 
between 2014 and 2016, with 59.5% farms cultivating cotton and 63.7% farms cultivating wheat in 2016. 
High-value crops were not cultivated by more than 15% of the farms in 2014. There was an increase in 
the number of farmers cultivating tomatoes, onions, potatoes, melons and clover between 2014 and 
2016. Potatoes were cultivated by 10.9% of the farms in 2014 and increased to 17.7% in 2016 (p <0.01).
A significant share of the net cultivated area constituted cotton cultivation. In 2014, cotton was 
cultivated on 3.5 ha, which was equivalent to 67.1% of the net cultivated area of the farm (Figure 2). The 
cultivated area of wheat was lower, at 1.6 ha in 2014, which was equivalent to 44% of the net cultivated 
Updated June 2018 80
area. Except for maize and clover, which require larger areas, the cultivated areas of high-value crops 
did not exceed 1 ha, on average. Between 2014 and 2016, there was a small increase in the cultivated 
areas of tomatoes, onions, potatoes and melons. There was a decrease in the areas under cotton and 
wheat by 7.1% and 15.2%, respectively, between 2014 and 2016.
The number of crops cultivated by a dehkan farm, on average, was 2.9 in 2014 and this increased to 3.4 
in 2016 (Table 3). The difference between the two years is significant at 1%. The Margalef’s Index of 
Diversity (MID), commonly used for measuring biodiversity and defined in this context as the number of 
crops cultivated per hectare of land area, was calculated as: 
𝑀𝐼𝐷 =
𝑁−1
ln(𝐴+1)
(3)
where 𝑁 is the number of crops cultivated and 𝐴 is the total area cultivated by the farm. On average, 
this indicator was 1.67 in 2014 and increased to 2.03 in 2016, with the difference significant at 1%.
Finally, the cropping intensity of a plot on the farm was calculated as the gross cropped area divided by 
the net sown area, multiplied by 100 per plot. This indicates the number of times the same plot on the 
farm was used during one agricultural year. In the sample, cropping intensity was limited, centered 
around 100, and generally, only one crop was cultivated per plot per year even if the cropping intensity 
significantly increased from 2014 to 2016. Wheat is usually sown in September/October and harvested 
in June/July; and cotton is sown in April and harvested from September to November (FAO, 2017). With 
these two traditional crops, the possibility of increasing cropping intensity on a plot is limited. A shift 
towards vegetables requiring a shorter cultivation duration could increase cropping intensity in the 
future. 
6.2. Determinants of cultivated area, number of crops, crop diversity and cropping intensity
6.2.1.Cultivated areas
Table 4 presents the determinants of the cultivated areas of cotton, wheat and high-value crops. Water 
delivery services had a modest impact on the cultivated area of cotton, wheat and high-value crops. 
Farmers who thought that fairness in water sharing had improved increased cultivated areas of cotton 
by 0.23 ha (p < 0.1); and farmers who perceived an improvement in the condition of the watercourse 
increase cultivated areas of high-value crops by 0.23 ha (p< 0.05). In contrast, farmers who perceived an 
improvement in the quantity of water delivered reduced cultivated areas of wheat by 0.49 ha (p < 0.05). 
Wheat is not a water-intensive crop, and any improvement in the quantity of water may provide an 
opportunity to cultivate high-value crop, which are more water intensive.
Updated June 2018 81
Receiving training did not have any impact on the cultivated areas of cotton, wheat, and high-value 
crops. However, additional interactions with an agricultural or water community group increased the 
cultivated area of high-value crops by 0.14 ha (p < 0.05). 
These results suggest that cultivated areas of cotton and wheat depend on water delivery, while those 
of high-value crops depend on extension services and farmer exchanges. 
6.2.2.Number of crops
Table 5, column 1 presents results pertaining to the determinants of the number of crops cultivated, 
other than cotton and wheat. Farmers who perceived an improvement in the physical state of the 
watercourse cultivated 0.16 additional crops (p < 0.05); while farmers who perceived an improvement in 
the condition of the distributary canal cultivated 0.12 additional crops (p < 0.05). Farmers who received 
formal training in cultivation of high-value crops (which was only provided to members of USAID 
established WUAs) cultivated 0.28 additional crops. 
These results suggest that infrastructure maintenance and improvements, and formal training in 
cultivating high-value crops are important for increasing the number of crops cultivated. 
6.2.3.Crop diversity
Table 5, column 2 contains results pertaining to the determinants of crop diversity. The condition of the 
watercourse canal was again a significant determinant of the MID; farmers who perceived that the 
condition of the watercourse canal had improved had a MDI of 0.13 more that those who thought that 
there were no such improvements (p<0.05). The MDI has also higher by 0.29 (p<0.01) when a 
shareholder had received formal training in agricultural extension. 
6.2.4.Cropping intensity
Column 3 of Table 5 reports determinants of cropping intensity. A perceived improvement in the 
condition of the watercourse increased cropping intensity by 3.06%. Formal training in extension and 
the frequency of interaction with agricultural and water groups did not have any effect on cropping 
intensity. 
7. Conclusion
The government of Tajikistan has adopted a policy of diversification in agricultural production and 
consumption to tackle the problem of malnourishment. This policy is being implemented by supporting 
improvements in irrigation delivery, encouraging cultivation of high-value crops, and by increasing 
awareness about the benefits of diversified production and diets. The result in this paper suggest that 
improvements in irrigation delivery services affect the cultivated areas of cotton and wheat, as well as 
Updated June 2018 82
that of high-value crops. Irrigation services also determine the number of high-value crops cultivated 
and the cropping intensity. Formal training in agricultural extension significantly influences crop 
diversity. 
From a policy perspective, this analysis highlights the role of a multifactorial approach in increasing 
cultivated areas, the number of crops cultivated, crop diversity, and cropping intensity. In a landscape 
where agricultural production was historically characterized by limited crop selection and where 
production in the agricultural sector had collapsed, better coordination between water management 
and agricultural extension programs may stimulate the agricultural sector. Moving forward, developing 
national agricultural extension services, along with expanding agricultural markets that supply quality 
inputs for cultivating fruits and vegetables will be important for easing supply-side constraints on 
diversification in production. 
Acknowledgments: 
This research is funded through a project titled ‘Impact of Water Users Associations on Water and Land 
Productivity, Equity and Food Security in Tajikistan’ funded by the United States Agency for International 
Development under USAID Grant Number AID-BFS-G-11-00002. The opinions expressed herein are those 
of the authors and do not necessarily reflect the views of the U.S. Agency for International 
Development. We would like to thank Ted Horbulyk for comments and suggestions; and Sohrob Aslamy 
and Katherine MacDonald for supervising data collection. Both authors contributed equally to this work. 
All errors are ours.
Updated June 2018 83
References
ADB (2016). Sector Assessment (Summary): Agriculture and Natural Resources. Available at: 
https://www.adb.org/sites/default/files/linked-documents/cps-taj-2016-2020-ssa-03.pdf (accessed June 
25, 2018). 
Abbott, E. A. (2016). Assessing Increased Awareness and Use of Land Restructuring and Land Use Rights 
and Their Impacts on Farming , Crop Diversification and Household Food , and Gender in 12 Project 
Rayons of Khatlon Region , Tajikistan. Dushanbe.
Birthal, P. S., Roy, D., & Negi, D. S. (2015). Assessing the Impact of Crop Diversification on Farm Poverty 
in India. World Development, 72, 70–92. https://doi.org/10.1016/J.WORLDDEV.2015.02.015
Bobojonov, I., & Aw-Hassan, A. (2014). Impacts of climate change on farm income security in Central 
Asia: An integrated modeling approach. Agriculture, Ecosystems & Environment, 188, 245–255. 
https://doi.org/10.1016/J.AGEE.2014.02.033
Bobojonov, I., Lamers, J.P.A., Bekchanov, M., Djanibekov, N., Franz-Vasdeki, J., Ruzimov, J., & Martius, C. 
(2013). Options and constraints for crop diversification: a case study in sustainable agriculture in 
Uzbekistan. Agroecology and Sustainable Food Systems, 37, 788-811. 
Bobojonov, I., & Lamers, J. P. A. (2008). Analysis of agricultural markets in Khorezm, Uzbekistan. In 
Continuity and change: Land and water use reforms in rural Uzbekistan. Socio-economic and legal 
analyses for the region Khorezm, eds. P. Wehrheim, P., A. Schoeller-Schletter, and C. Martius, 165–182. 
Halle/Saale, Germany: IAMO.
Bobokhonov, A., Pokrivcak, J., & Rajcaniova, M. (2017). The impact of agricultural and trade policies on 
price transmission: The case of Tajikistan and Uzbekistan. The Journal of International Trade & Economic 
Development, 26(6), 677–692. https://doi.org/10.1080/09638199.2017.1287212
Boboyorov, H. (2012). Personal Networks of Agricultural Knowledge in the Cotton Growing Communities 
of Southern Tajikistan. Demokratizatsiya, 20(4), 409.
Boboyorov, H. (2016). Constraints and Concerns for Farming and Cropping Types and Strategies in 
Southern Tajikistan. In A.-K. Hornidge, A. Shtaltovna, & C. Schetter (Eds.), Agricultural Knowledge and 
Knowledge Systems in Post-Soviet Societies (p. 301). Peter Lang.
Calub, B.M., Tanje, D. & Phouyyavong, K. (2005). Promotion of soybean as a cash crop: determinants 
and possible impacts. LSUAFRP Field Report No. 05/09, NAFRI Vientiane, Lao PDR.
FAO (2017). GIEWS - Global Information and Early Warning System, Country Briefs Tajikistan. 
www.fao.org/giews/countrybrief/country.jsp?code=TJK. Accessed the 15 May 2018. 
Government of Tajikistan. (2007). The Coton Sector of Tajikistan New Opportunities for the International 
Updated June 2018 84
Cotton Trade, (April).
Gunchinma, T., & Yakubov, M. (2010). Institutions and transition: Does a better institutional 
environment make water users associations more effective in central Asia? Water Policy, 12(2), 165–
185. https://doi.org/10.2166/wp.2009.047
Hofman, I. (2017). Soft budgets and elastic debt: farm liabilities in the agrarian political economy of 
post-Soviet Tajikistan. Journal of Peasant Studies, 0(0), 1–21. 
https://doi.org/10.1080/03066150.2017.1293047
Hooper, D., & Vitousek, P.M. (1997(. The effect of plant composition and diversity on ecosystem 
processes. Science, 277, 1302–1305.
Husenov, B., Makhkamov, M., Garkava-Gustavsson, L., Muminjanov, H., & Johansson, E. (2015). 
Breeding for wheat quality to assure food security of a staple crop: the case study of Tajikistan. 
Agriculture & Food Security, 4(1), 8. https://doi.org/10.1186/s40066-015-0029-1
Kasem S. & Thapa, G.B. (2011). Crop diversification in Tahiland: status, determinants, and effects on 
income and use of inputs. Land Use Policy, 28, 618-628. 
Lerman, Z., & Sedik, D. (2008). The Economic Effects of Land Reform in Tajikistan. Policy Studies on Rural 
Transition.
Mannig, B., Müller, M., Starke, E., Merkenschlager, C., Mao, W., Zhi, X., … Paeth, H. (2013). Dynamical 
downscaling of climate change in Central Asia. Global and Planetary Change, 110(Part A), 26–39. 
https://doi.org/https://doi.org/10.1016/j.gloplacha.2013.05.008
McCord, P.F., Cox. M., Schmitt-Harsh, M., & Evans, T. (2015). Crop diversification as a smallholder 
livelihood strategy within semi-arid agricultural systems near Mount Kenya. Land Use Policy, 42, 738-
750.
Muminjanov, H., Otambekova, M., & Morgounov, A. (2016). The history of wheat breeding in Tajikistan. 
In The world wheat book: a history of wheat breeding, volume 3 (Lavoisier, pp. 283–305). Hampshire, 
United Kingdom.
Sehring, J. (2009). Path dependencies and institutional bricolage in post- soviet rural water governance. 
Water Alternatives, 2(1), 53–60.
Oladele, O.I., 2005. A profit analysis of propensity to discontinue adoption of agriculture technology 
among farmers in Southwestern Nigeria. Journal of Central European Agriculture, 6, 249–254.
Pradhan, D. & Ranjan, R. (2016). Achieving Sustainability and Development through Collective Action? 
An Empirical Analysis of the Impact of the Bore Pool Sharing Program on Farm Incomes and Crop 
Choices. World Development, 88, 152–174. 
Updated June 2018 85
Pomfret, R. (2008). Tajikistan, Turkmenistan and Uzbekistan. In Distortions to agricultural incentives in 
Europe’s transition economies, eds. K. Anderson and J. Swinnen, 297–338. Washington, DC: World Bank.
Reich, P. B., J. Knops, D. Tilman, J. Craine, D. Ellsworth, M. Tjoelker, T. Lee, D. Wedin, et al. (2001). Plant 
diversity enhances ecosystem responses to elevated CO2 and nitrogen deposition. Nature, 410, 809–
810.
Sommer, R., Glazirina, M., Yuldashev, T., Otarov, A., Ibraeva, M., Martynova, L., … de Pauw, E. (2013). 
Impact of climate change on wheat productivity in Central Asia. Agriculture, Ecosystems & Environment, 
178(Supplement C), 78–99. https://doi.org/https://doi.org/10.1016/j.agee.2013.06.011
Statistical Agency under the President of the Republic of Tajikistan (SA), Ministry of Health (MOH), and 
ICF International (2013). Tajikistan Demographic and Health Survey 2012. Dushanbe, Tajikistan and 
Calverton, Maryland, USA: SA, MOH, and ICF International.
Wang, F., He, Z., Sayre, K., Li, S., Si, J., Feng, B., King, L. (2009). Wheat cropping systems and technologies 
in China. Field Crops Research, 111, 181-188. 
Wang, L., Sun, J., Zhang, Z., Xu, P., Shangguan, Z. (2018). Winter wheat grain yield in response to 
different production practices and soil fertility in China. Soil & Tillage Research, 10-17. 
Winters, P., Cavatassi,R., & Lipper, L. (2006). Sowing the seeds of social relations: The role of Social 
capital in crop diversity. ESA Working Paper No. 06-16. Rome, Italy: FAO.
World Food Programme (WFP) (2017). “Tajikistan Overview.” Available at 
http://www.wfp.org/countries/tajikistan/overview .
Zimmerer, K.S., 2014. Conserving agrobiodiversity amid global change, migration,and nontraditional 
livelihood networks: the dynamic uses of cultural landscapeknowledge. Ecology and Society, 19 (2), 1.
Updated June 2018 86
Table 1: Constructing propensity scores
Logit
Treatment 
Subdistrict
Population of the subdistrict in 2014 0.00 (0.00)
Number of villages 0.033 (0.05)
Total area of the subdistrict 0.00 (0.00)***
Majority of population Tajik (dummy) -0.69 (0.54)
Number of secondary schools in subdistrict 0.07 (0.08)
Number of rural health centers in the subdistrict -0.54 (0.19)***
Number of agricultural markets in the subdistrict 0.40 (0.29)
Chairman born in the subdistrict (dummy) 0.16 (0.64)
Number of years of election of the chairman -0.07 (0.07)
Elevation of the subdistrict (m ASL) 0.00(0.00)
Sandy soil (dummy) 0.00 (0.55)
Deep groundwater level (dummy) -0.81 (0.52)
Land reform completed (dummy) 2.54 (0.69)***
Cotton main crop of the subdistrict (dummy) 2.84 (0.66)***
Subdistrict irrigated by gravity system (dummy) 0.61 (0.67)
Constant -3.89 (1.49)***
Pseudo R2 0.402
Sample size 164
Notes: Standard errors included in parentheses, next to the coefficients. 
***implies p<0.01, ** implies p<0.05, and * implies p<0.1
Updated June 2018 87
Table 2: Variables for which data was collected for the 2014 and 2016 cropping year
2014 2014
# Mean Std. 
Dev # Mean Std. 
Dev
DEPENDANT VARIABLES
Cultivated area of cotton (ha) 1956 1.95 5.10 1956 1.82 5.53
Cultivated area of wheat (ha) 1956 0.96 2.32 1956 0.89 2.04
Cultivated area of high-value crops (ha) 1956 1.22 4.22 1956 1.45 3.95
Number of high-value crops (#) 1956 1.79 1.59 1848 2.10 1.74
Margalef's index 1917 1.67 1.71 1845 2.03 1.87
Cropping intensity 1902 96.90 34.75 1825 107.87 39.47
FARM CHARACTERISTICS
Area of the farm (ha) 1950 4.40 8.70 1854 4.24 8.15
Distance of farm to road (kms.) 1914 1.43 2.19 1914 1.43 2.19
Age of farm (years since farm allotted) (#) 1899 6.26 4.87 1899 8.26 4.87
SHAREHOLDER 
CHARACTERISTICS
Number of shareholders (#) 1936 7.42 11.10 1853 6.55 9.77
Proportion of female shareholders 1882 0.47 0.25 1848 0.52 0.24
Age of the manager of the farm (#) 1854 49.39 12.81 1854 50.39 12.81
Education of farm manager(categorical) 1855 3.89 1.23 1855 3.89 1.23
Female farm manager (dummy) 1956 0.13 0.33 1956 0.38 0.48
WATER DELIVERY
Perception of fairness of water sharing (categorical) 1327 2.90 0.59 1848 2.81 0.59
Perceived timeliness of water distribution improved 
(dummy) 1956 0.42 0.49 1956 0.42 0.49
Perceived quantity of water received improved(dummy) 1956 0.43 0.49 1956 0.43 0.49
Perceived condition of water course canal (categorical) 1893 2.45 0.85 1822 2.58 0.80
Perceived condition of distributary canal (categorical) 1876 2.39 0.84 1812 2.48 0.81
AGR EXTENSION 
SERVICES
Shareholder received training in extension (dummy) 1899 0.29 0.45 1855 0.38 0.48
Frequency interaction with agriculture/water group 
(categorical) 1937 0.91 1.42 1956 1.19 1.19
Updated June 2018 88
Table 3: Number of crops, crop diversity and cropping intensity of the farm
2014 2016
Difference
Mean (2016 – 2014)
Standard 
Deviation
Mean
Standard 
Deviation
Number of crops cultivated 2.93 1.59 3.37 1.74 7.984***
Margalef’s Index of Diversity 1.67 1.71 2.03 1.87 6.239***
Cropping intensity 96.90 0.79 107.87 0.92 9.018***
Notes: Standard errors included in parentheses, next to the coefficients. 
***implies p<0.01, ** implies p<0.05, and * implies p<0.1.
Updated June 2018 89
Table 4: Determinants of cultivated areas of cotton, wheat and high-value crops
Area of 
cotton
Area of 
wheat
Area of 
high-value
cultivated cultivated crops cultivated
(tobit) (tobit) (tobit)
Area of the farm (ha) 0.51 (0.01)*** 0.15(0.01)** 0.31 (0.01)***
Distance of farm to road (km) -0.09 (0.06)* 0.05 (0.03)* (-0.02 (0.04)
Age of the farm (#) -0.02 (0.02) 0.03 (0.01)** 0.03 (0.02)**
# shareholders 0.07 (0.01)*** -0.01 (0.01)* -0.03 (0.01)***
% female shareholders 1.37 (0.3)*** -0.07 (0.19) -0.59 (0.27)**
Age of manager (#) -0.01 (0.01) 0.00 (0.00) 0.00 (0.01)
Education of the manager (categorical) -0.09 (0.09) -0.02 (0.04) 0.08 (0.06)
Female manager (dummy) 0.04 (0.14) -0.00 (0.09) 0.49 (0.13)***
Fairness in water distribution (categorical) 0.23 (0.12)* -0.11 (0.08) -0.12 (0.12)
Timeliness of water distribution improved (dummy) 0.03 (0.48) 0.28 (0.21) 0.18 (0.30)
Quantity of water received improved(dummy) 0.77 (0.48) -0.49 (0.21)** -0.08 (0.30)
Condition of watercourse (categorical) -0.11 (0.11) 0.07 (0.07) 0.23 (0.10)**
Condition of dist. canal (categorical) 0.11 (0.11) -0.11 (0.07) 0.10 (0.10)
Shareholder received training in extension (dummy) 0.09 (0.13) 0.02 (0.09) -0.02 (0.13)
Frequency interaction with agriculture/water group 
(categorical) -0.04 (0.05) 0.06 (0.03) 0.14 (0.05)***
Number of observations 2,874 2,874 2,874
Number of farms 1,730 1,730 1,730
Notes: Robust standard errors are included in parentheses. ***implies p<0.01, ** implies p<0.05, and * 
implies p<0.1. In addition to the variables included in this table, the following variables were also 
included: a constant, dummies variables for the type of canal serving the farm (primary, secondary, 
tertiary); dummy variables for the location of the farm along the canal (head, middle, tail); and dummy 
variables for the rivers that were the source of water for the canals (Vakhsh, Pyanj, Kafirnigan, 
Sukhandarya).
Updated June 2018 90
Table 5: Determinants of number of crops, crop diversity and cropping intensity
# of high-value Margalef's index
Cropping 
intensity
crops of Crop Diversity of plots on the
cultivated 
(GLS) (tobit) farm (GLS)
Area of the farm 0.02 (0.01)** -0.03 (0.01)*** -0.53 (0.190***
Distance to road -0.03 (0.02) -0.05 (0.02)*** 0.01 (0.37)
Age of the farm 0.03 (0.01)*** -0.01 (0.01) -0.17 (0.21)
# shareholders 0.00 (0.01) -0.00 (0.00) 0.07 (0.08)
% female shareholders -0.21 (0.14) -0.29 (0.16)* 0.21 (3.27)
Age of manager 0.00 (0.00) 0.00 (0.000 -0.00 (0.07)
Education of the manager 0.00 (0.03) -0.00 (0.03) 0.45 (0.65)
Female manager (dummy) -0.08 (0.07) 0.07 (0.08) 4.12 (2.06)**
Fairness in water distribution (categorical) 0.08 (0.07) 0.04 (0.07) 0.48 (1.68)
Timeliness of water distribution improved (dummy) 0.06 (0.14) 0.12 (0.16) 1.15 (2.84)
Quantity of water received improved (dummy) -0.04 (0.150 -0.04 (0.16) -0.23 (2.83)
Condition of watercourse (categorical) 0.16 (0.06)** 0.13 (0.07)** 3.07 (1.38)**
Condition of dist. canal (categorical) -0.12 (0.06)** 0.00 (0.06) -1.58 (1.52)
Shareholder received formal training in extension 0.28 (0.09)*** 0.29 (0.08)*** 1.48 (1.84)
Frequency of interaction with agriculture/water group 
(categorical) -0.02 (0.03) -0.03 (0.03) 0.42 (0.71)
Observations 2,874 2,854 2,823
Number of farms 1,730 1,728 1,717
Notes: Robust standard errors are included in parentheses. ***implies p<0.01, ** implies p<0.05, and * 
implies p<0.1. In addition to the variables included in this table, the following variables were also 
included: a constant, dummies variables for the type of canal serving the farm (primary, secondary, 
tertiary); dummy variables for the location of the farm along the canal (head, middle, tail); and dummy 
variables for the rivers that were the source of water for the canals (Vakhsh, Pyanj, Kafirnigan, 
Sukhandarya).
Updated June 2018 91
Figure 9: Share of farms cultivating different crops
Source: Authors’ calculation based on survey data collected 
Figure 10: Cultivated area of different crops (ha)
Source: Authors’ calculation based on survey data collected
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
100.00
Cotton Wheat Any
other
crops
Maize Tomato Onion Potato Melon Clover
2014 2016
0
0.5
1
1.5
2
2.5
3
3.5
4
Cotton Wheat Maize Tomato Onion Potato Melon Clover
2014 2016
Updated June 2018 92
Appendix 4 
Price, J.; Balasubramanya, S. 2018. The role of the mahalla in local water dispute resolution 
in Tajikistan. Revise and resubmit at Central Asian Survey.
Updated June 2018 93
Title: The role of the mahalla in local water dispute resolution in Tajikistan
Keywords: dispute resolution, mahalla, water user associations, Tajikistan
Abstract: This article examines how mahalla committees interact with newly introduced water 
user associations (WUAs) in Tajikistan, following institutional reforms after the disintegration of 
the Soviet Union and the civil war. The study adopts a qualitative approach using data from elite 
interviews in Dushanbe, and a case study in Khatlon Province. While not well-recognised in formal 
legislation, the mahalla committee is found here to play a significant role in local water 
governance and dispute resolution in practice. Since they regulate water for ‘kitchen gardens’, 
mahalla leaders often coordinate with WUA leaders to resolve prevailing disputes between water 
users from different plot types. In some cases, however, mahalla leaders dominate new water￾related institutions and services. These findings are particularly significant due to the increase in 
kitchen gardens in the context of increased male out-migration to Russia, and might suggest a 
need for revisiting the status of the mahalla in policy.
1. Introduction
Within the theory and practice of institutional reform, there has been a growing interest in the importance 
of pre-existing forms of governance (World Bank, 2017). In developing countries, these forms can often 
be informal – not legally recognised – but still wield significant and persistent power. Local water 
governance and dispute resolution in Tajikistan presents one such case. Newly introduced water user 
associations (WUAs)1 are mandated by law to provide irrigation services, including a dispute resolution 
mechanism, but water disputes prevail and are also found to be resolved by alternative means 
 
1 WUAs have been introduced in Tajikistan since the earlier 2000s, as part of water sector reforms after the 
disintegration of the Soviet Union in 1991 and the ensuing civil war.
Updated June 2018 94
unidentified in recent studies (Balasubramanya, Price and Horbulyk, 2018). This presents an empirical 
puzzle. 
Local governance in Tajikistan and wider Central Asia historically has been characterised by the 
existence of mahalla committees, which have endured pre-Soviet, Soviet and civil war eras (Sievers, 2002). 
The mahalla committee can be defined as ‘elected representatives of community members’, including a 
leader, ‘who mediate to stabilise contested political and economic relations’ (Asian Development Bank 
(ADB), 2011: 1; Mandler, 2010: 7). The word ‘mahalla’ translates as ‘neighbourhood’, or the unit of the 
village in rural contexts. Mahalla committees are known traditionally to be an important institution in 
community affairs, from water allocation to resolving general conflict. This article therefore seeks to 
answer: what is the role of the mahalla committee in local water dispute resolution in Tajikistan, and how 
does the mahalla committee interact with WUAs?
Mahalla committees are often over-looked among policy interventions and donor programmes, 
as well as in the literature (United States Agency for International Development (USAID), 2013; Republic 
of Tajikistan, 2009). As suggested by Freizer (2004), ‘there is room for greater cooperation and division of 
competencies between jamoats2
, mahallas, higher government authorities, local NGOs and international 
donor agencies’.
Tajikistan has considerable potential for, and experience with, water disputes at both local and 
transboundary levels. Localised conflict has punctuated the contested border regions in the Ferghana 
Valley in the north of the country (Bichsel, 2009), while the densely populated cotton-growing Khatlon 
Province experiences disagreements over in-demand land and water resources. This research therefore 
addresses a salient issue, identifying the de facto collection of governance arrangements – formal and 
informal. 
 
2 A jamoat is an administrative unit within a district. In other countries, the equivalent of a jamoat is a sub-district. 
In Tajikistan, a collection of jamoats makes up a district.
Updated June 2018 95
The article employs qualitative methods to investigate governance forms and phenomena that 
are not easily visible from quantitative or formal legal analysis. It advances the proposition that ‘policy 
making and policy implementation do not occur in a vacuum. Rather, they take place in complex political 
and social settings, in which individuals and groups with unequal power interact within changing rules as 
they pursue conflicting interests’ (World Bank, 2017: 283). The article examines Tajikistan as one such 
setting, using data from interviews conducted in Dushanbe and rural Khatlon in 2017 to form the basis of 
the analysis. This is supplemented by consulting formal legal documents, against which to contrast de 
facto governance dynamics. Interviews were conducted at multiple levels of governance, to uncover the 
causal mechanisms of disputes and better define resolution processes in practice. There is a lack of up-to￾date accounts of the role and influence of mahalla committees in general in Tajikistan, but particularly 
regarding how mahalla committees relate to institutions and institutional reforms introduced following 
the disintegration of the Soviet Union and the civil war in the 1990s. This article addresses this broader 
gap in the literature, as well as the specifics of water dispute resolution in practice.
The research finds that the mahalla committee does indeed play a significant role in local water 
governance in Tajikistan. The mahalla leader regulates water for household ‘kitchen gardens’, while WUAs 
govern matters concerning private dehkan farms. There are many disputes between different water users 
from different plot types, due to kitchen garden water users diverting canal irrigation water to their plots 
(before the water reaches dehkan farms). In these instances, the mahalla leader often coordinates with 
the WUA leader to resolve the dispute. The form of dispute outlined also relates to fee payment disputes 
between dehkan farms and WUAs, since water users from dehkan farms resent paying fees when kitchen 
garden water users are perceived as not paying for water-related services in the same way. Although 
effective coordination between the mahalla committee and the WUA exists, there are also cases of 
mahalla leaders seeking to dominate new water-related services. 
Updated June 2018 96
Section 2 provides a context for examining water disputes and their resolution in Tajikistan. 
Section 3 reviews the literature; first constructing a conceptual framework, then comparing empirical 
cases where these concepts apply. Section 4 outlines the methodology, while Section 5 proceeds with the 
analysis of qualitative data. Finally, Section 6 aggregates the results and discusses them in relation to the 
existing literature and the context, before Section 7 offers policy implications and avenues for future 
research.
2. Context
The mahalla committee as an institution can be traced back to medieval Central Asia, but is neither 
temporally static nor spatially uniform. Originally, the mahalla was a central unit of socio-political 
organisation based around kinship ties and the geographic area, and it often superseded other potentially 
more divisive features such as religion, language or class (Sievers, 2002). The mahalla committee provided 
services for its residents and a framework for general ‘social securing functions’ (Poos, 2011: 5), outside 
of the structures of the state. More recently, during the Soviet period, mahalla committees were 
subverted but not entirely abolished3
. 
In post-war Tajikistan, Mandler (2010: 2) advises that ‘it is not the state that controls local affairs’. 
This is especially the case in rural communities, where local networks and kinship ties shape daily life. The 
mahalla committee remains a mediator of these affairs. It is known to play a role in general relations 
between neighbours and within families (Freizer, 2004). Less clear, is the specific role that it plays in local 
water governance, in relation to WUAs. 
Several authors maintain how, in some villages, the mahalla committee may be more trusted as 
a reliable form of governance than the state and development agencies (Boboyorov, 2013; Sehring, 2009). 
 
3 The reasons for this are subject to debate, but are suggested to be either due to risks of revolt or that the mahalla
committee could complement Soviet societal goals, if moderated (Sievers, 2002).
Updated June 2018 97
Boboyorov (2013: 26) goes as far as to assert that ‘people do not trust the state and development 
institutions due to the short-term, unstable and unpredictable protection they offer in Tajikistan. Instead, 
mahalla serves to transform these institutions into long-term reciprocal networks which then reduce 
political and economic uncertainties and threats’. For example, formal courts may be distrusted or 
deemed ineffective by citizens. To give a sense of the potential local power held by mahalla committees, 
Noori (2006: 538) explains that ‘neither state courts nor the local government are authorised to intervene 
in disputes involving residents, unless one of the parties obtains written permission from his or her 
mahalla committee documenting the committee’s intervention or attempt at reconciliation.’
In the agricultural communities in southern Tajikistan, the mahalla committee often consists of 
an elected mahalla leader, village elders, respected Islamic clerics, and other locally powerful individuals 
(Boboyorov, 2013; Cieslewska, 2010). These figures form a committee and typically meet in a designated 
chaikhana (teahouse) to discuss and mediate affairs. They base their activities and decisions on traditional 
unwritten rules. Sievers (2002) notes how social harmony is often emphasised over absolute justice as a 
goal of dispute resolution. Some scholars and Western policymakers view this as one of the negative 
aspects of mahalla committees, alongside perceptions that they are defined by patriarchal or unequal 
power structures (Cieslewska, 2010). As such, the mahalla committee is viewed as a negative institution 
by some actors, and as a positive one by others. 
Tajikistan is home to a predominantly agrarian population, many of whom inhabit the densely 
populated arable land in cotton-growing Khatlon in the south-west (Abdullaev and Barnes, 2001). There 
is a strong agricultural tradition, advanced by the Soviet Union which designated the southern part of the 
Tajik Soviet Socialist Republic as Central Asia’s main cotton producing hub. Since the disintegration of the 
Soviet Union in 1991, cotton cultivation – which requires substantial water – remains the main source of 
Gross Domestic Product in Tajikistan (Food and Agriculture Organization of the United Nations, 2012). 
Updated June 2018 98
However, while agricultural activities have continued, the land tenure underpinning them has 
shifted significantly. Collective farms from the Soviet Union were broken down into private dehkan farms, 
with their irrigation services provided by community-based participatory WUAs, as part of a logic of 
increased decentralisation and democratisation. Notably, WUA membership is for dehkan farms and other 
entrepreneurial entities (Republic of Tajikistan, 2006). Recently, land use has been influenced by an 
increase in male out-migration to Russia for labour. As a result, women are increasingly left to cultivate 
crops at home on ‘kitchen gardens’, which are small plots attached to houses (or in the vicinity of a 
residence) used to grow crops for subsistence. The increase in male migration has thus led to a greater 
prominence of kitchen gardens. Kitchen gardens cannot become members of WUAs by law, since they are 
not considered entrepreneurial entities (Republic of Tajikistan, 2006).
3. Literature review
Conceptual framework
While one may associate the term ‘governance’ with official organs of the state, practitioners and scholars 
have recently acknowledged that governance may be provided by a variety of actors (Skarbek, 2011; 
Denyer Willis, 2014; World Bank, 2017). For example, Skarbek (2011: 702) underlines that ‘centralised 
governments and competing, overlapping, governance organizations can both provide governance 
institutions that resolve disputes, secure property rights, and limit negative externalities’. A broad 
conception of governance is thus: ‘the process through which state and non-state actors interact to design 
and implement policies within a given set of formal and informal rules that shape and are shaped by 
power’ (World Bank, 2017: 3). 
In light of this definition, there is a distinction between ‘formal’ and ‘informal’ governance (i.e. 
official governance provided for, or regulated, by the state, and governance provided by non-state actors).
The mahalla council is an example of a mode of informal governance. Especially in contexts of political 
Updated June 2018 99
and economic transition, some scholars reflect on the reality of ‘hybrid arrangements’ of formal and 
informal governance (Meagher, De Herdt and Titeca, 2014; Koehler, 2004). Rather than newly created 
institutions discretely replacing old ones, the reality on the ground is often that a bricolage of institutions 
interact, and contest or coordinate in a more disorderly fashion (Sehring, 2009). In particular, this may be 
the case when the institutions are externally devised (as is the case with WUAs). As a result, one can view 
institutions in terms of their ‘form versus function’ (World Bank, 2017: 5). Different forms of institution 
(formal or informal) may carry out the same functions in practice (or may be capable of doing so); or may 
combine to carry out a function. Some scholars argue further that governance functions must be 
accompanied by specific accountabilities and oversight in order to be complete or effective (Jepson, 
2005).
Water dispute resolution is an example of a governance function (Joffé, 2015); it may be provided 
by a wide range of institutions - the state, non-governmental organisations (NGOs), foreign governments, 
development agencies, or traditional informal institutions. If a function such as dispute resolution is 
provided by a traditional institution, then the power dynamics and historical continuity associated with 
the institution mean that it may remain the outlet of preference for citizens, and may be firmly established 
as the provider of the governance function due its influence in local politics. These considerations arise 
during and after institutional reform, where carefully-designed policies can seem optimal in principle, but 
may meet a complex political environment on the ground. In such cases, interaction and compromise may 
be unavoidable (Knaus and Stewart, 2011).
Empirical cases
Globally, there are numerous cases of informal modes of governance playing an important role in local 
water dispute resolution, where one can examine the institutional interactions born from modern policy 
interventions. Joffé (2015) outlines water governance mechanisms in North Africa, highlighting the 
durable nature of traditional water allocation and dispute resolution practices within the Berber 
Updated June 2018 100
population. He demonstrates how these practices have, in fact, informed recent democratic 
developments in the region, rather than colliding with formal and newer forms of governance. Stewart 
(2004) provides an account of the patterns of local governance in Afghanistan, observing the local power 
held by informal village-level structures. By contrast, he describes the challenges arising when external 
powers (such as foreign governments and development agencies) sought to establish formal democratic 
and decentralised institutions, and how these collided with traditional informal institutions when there 
was insufficient recognition of their pre-existence.
In Central Asia, Sehring (2009) notes that path dependencies4
can limit the effectiveness of 
institutional reform. She explains how ‘rules and organisations established formally by the state and or 
international donor organisations are undermined by informal institutions. Yet informal institutions are 
not only an obstacle to reform, but can also support it’. Related to this, Koehler (2004) describes how, in 
practice, such hybrid arrangements combine to impact the occurrence and processing of conflict in Central 
Asia. 
Stevens (2005: 282) reflects on interactions between mahalla committees and modern NGOs in 
Uzbekistan, citing the ‘Mahalla Initiative Program’ – ‘one of the earliest attempts at incorporating the 
mahalla into donor programming’. He concludes that there is significant complementarity between NGOs 
and mahalla committees, and cautions that, without accommodation, mahalla leaders may ‘seek to 
consolidate their constituencies and may be best placed to take advantage of political liberalisation’ 
(Stevens, 2005: 293). In Kyrgyzstan, WUAs were introduced in the 1990s -slightly earlier than in Tajikistan. 
Farmers were still found to turn towards traditional institutions in the case of conflict, but, again, Sehring 
(2005) indicates how these institutions ‘can support WUAs’. These findings from elsewhere in the region 
frame the potential issues surrounding the role of the mahalla committee in Tajikistan that are examined 
in this empirical enquiry.
 
4
i.e. current practices based on historical preference
Updated June 2018 101
4. Methodology
The research adopts a qualitative approach using data from interviews, and from a case study of a village 
in southern Tajikistan to provide supportive evidence. Combined with elite interviews of broader 
geographic scope, the case study allows for a deeper analysis to uncover and define the causal 
mechanisms of disputes, and the detail of their resolution in practice. Interviews were conducted, ranging 
from the highest relevant central government authorities, through local government, and the WUA 
leader, mahalla leader and water users in the case study village (see Table 1). Consulting all levels allows 
for comparison of issues and themes to determine whether different actors are highlighting the same 
points. The case study selection was derived from quantitative results in a previous study on WUAs in 
Tajikistan (Balasubramanya, Price and Horbulyk, 2018.). The WUA jurisdiction with the greatest number 
of water disputes in that study’s sample, and where disputes were not always resolved by the WUA alone, 
was selected (a village in Vakhsh District, Khatlon Province – ‘Obishirin 2016’ WUA). 
A purposive sampling method was used in the selection of formal pre-arranged interviews, while
a snowballing sample method allowed for informal discussions with water users in the case study village 
(Table 1). Interviews were semi-structured, with mostly open-ended questions. A voice recorder was used 
to record interviews (if endorsed by the interviewee), and anonymity was guaranteed for government 
officials. Interviews were later analysed using framework analysis (Ritchie and Lewis, 2003). Finally, after 
conducting the interviews, further information regarding the legal status of the mahalla was collected.
Findings from a single case study cannot claim to provide generalisable results that are 
representative of the region or country at large. In addition, fieldwork time was limited to ten days, 
meaning it was not possible to interview a very large number of stakeholders, or spend an extended 
period of time observing in great detail water governance and dispute resolution practice in the case study
village. Nevertheless, the fieldwork still produced rich and nuanced perspectives from key actors at all 
Updated June 2018 102
levels. It uncovered some of the mechanisms of dispute resolution that are not always visible through 
quantitative survey data or formal legal analysis.
Table 1: Interviews conducted during fieldwork in Tajikistan, August 2017
1. Tax Administration Authority interviewee
2. Ministry of Land Reclamation and Water Resources interviewee
3. Local government (district) interviewee
4. WUA leader (case study village, Vakhsh)
5. Mahalla leader (case study village, Vakhsh)
6. Water users (case study village, Vakhsh)
5. Analysis
The mahalla, de jure
The mahalla is almost completely absent in Tajik legislation today, and had no clear legal status and 
definition during the Soviet period. After the collapse of Soviet Union in 1991, Tajikistan launched a series 
of public administration reforms, which defined the main public authorities. New laws were articulated 
and established accordingly. Part of this re-structuring involved the designation of WUAs as irrigation 
water providers, as stipulated in the WUA Law (Republic of Tajikistan, 2006). WUAs are mandated to 
perform functions including the ‘provision of fair, effective and timely allocation of water resources’, 
‘collection of payments for water supply services’; and ‘solving disputes’ (Kabilov, 2017). 
In 2008, a new law on ‘self-governance organisations’ was enforced (Republic of Tajikistan, 2009). 
This formalised the generic concept of self-governance organisations, which may be established by 
members of local communities. Mahalla committees can only register officially as this type of 
organisation. Any group of local people may register a broad variety of organisations as self-governance 
organisations, which means that there are no clear formal jurisdictions or authorities for mahalla 
committees per se. As a result, mahallas do not have an explicit legal status. Although some legal 
documents indirectly refer to the mahalla, there are no separate regulations defining them and their 
Updated June 2018 103
responsibilities. Indeed, the leader of the mahalla committee in the case study village, Vakhsh District, 
was unsure of his exact status in Tajik legislation. He was of the conviction that the activities of the mahalla 
committees are somehow related to the government and public affairs, but was not aware of any explicit 
written statements about the mahalla in Tajik law.
The mahalla, de facto
The role of the mahalla committee in local water governance
Most interviewees voluntarily shared information on the de facto role of the mahalla committee, 
including officials from government. They outlined the role of the mahalla committee in water governance 
across Tajikistan, describing how the mahalla leader typically regulates water for drinking and for kitchen 
gardens. The interviewee from the Ministry of Land Reclamation and Water Resources expressed that, 
currently, in many districts outside of Dushanbe, ‘the mahalla committee plays an important role in 
solving this or that problem or improving the quality of life’. He described the form and functions of the 
institution, stating that the mahalla committee has a leader that is elected by people living there, and 
performs tasks such as ‘creating schedules for using water’ as a regulatory mechanism, and ensuring 
uniform use and cleanliness.
The mahalla leader in the case study village described a role in which he is in charge of water 
management for drinking water and for kitchen gardens for the people living within his mahalla (informal) 
jurisdiction. For drinking water, he is responsible for keeping the water clean. When there is a shortage of 
water, he creates the schedule for both drinking water and kitchen garden water supplies. The water for 
kitchen gardens comes primarily from taps which were fitted in the Soviet period. When kitchen garden 
water users have a problem, they visit the mahalla leader for support or advice, and the mahalla 
committee assembles to discuss these various community issues. For straightforward disputes, the 
Updated June 2018 104
mahalla leader works as a mediator, bringing people together and ‘kindly presenting both sides of an 
argument to each party’ to stabilise community relations (mahalla leader, case study village).
Water users in the case study village recognised the importance of both the mahalla leader and 
the WUA leader. The users viewed the leaders’ roles as complementary, seeing the mahalla leader as a 
coordinator, while the WUA leader implements more technical plans. Residents in the area seem 
comfortable approaching either leader, and do so depending on the issue in question. In some villages, 
the mahalla leader may be viewed as being more important than the WUA leader. Notably, the local 
government interviewee gave the strongest expression of the mahalla leader’s local power in water 
governance which will be elaborated upon further.
The role of the mahalla committee in water dispute resolution
Because of the mahalla committee’s local influence and importance – due to its role in regulating kitchen 
gardens and stabilising community relations - when a dispute arises, the mahalla leader often plays a 
direct role in dispute resolution. It was observed by the interviewee from the Tax Administration Authority 
that mahalla leaders’ decisions may carry a lot of weight in communities and are often honoured in 
practice. One prevalent form of dispute concerns kitchen gardens and dehkan farms, and was detailed by 
both the local government interviewee and the mahalla leader in the case study village. When the taps 
for kitchen gardens are not functioning or there are water shortages, kitchen garden farmers seek 
alternative sources of irrigation water. In many villages, farmers from kitchen gardens set their pipes in 
canals and irrigate their plots. These water users divert water from canals that are administered by the 
WUA and are thus formally for use by dehkan farms. Especially when water is scarcer, this may mean that 
insufficient irrigation water reaches some of the dehkan farms, and they, too, do not receive enough water 
for cultivation. This problem is the cause of ‘huge disputes’, as formal dehkan water users visit the WUA 
leader and complain that kitchen gardens are taking their water. The dehkan farmers ask: ‘Why is there a 
way for others to receive water for free?’ (local government interviewee).
Updated June 2018 105
The mahalla leader is often central to the resolution of this form of dispute, due to his 
involvement with kitchen garden water users. The mahalla leader in the case study village explained how 
he is called upon by the WUA leader to resolve such disputes and the two leaders must coordinate 
together. The mahalla leader and the WUA leader thus meet with the parties concerned to resolve the 
dispute jointly in a conciliatory fashion. This form of dispute also compounds another form of dispute, 
whereby dehkan farm water users are reluctant to pay WUA fees for irrigation services. In part, this 
dispute is driven by lingering Soviet attitudes of non-payment for services deemed communal, and a lack 
of clarity in the law regarding what precisely is being paid for5
. However, there is also a degree of 
resentment and reluctance to pay fees when dehkan farm water users observe kitchen garden water users 
receiving water for free.
Further mahalla committee-WUA interactions
The interactions examined, between the mahalla committee and the WUA appear inevitable in many 
settings in Tajikistan, such as the case study village in Vakhsh, Khatlon. In addition to dispute resolution, 
the mahalla leader in in the case study defined other more general interactions with the WUA leader. The 
mahalla leader is sometimes called upon by the WUA leader to provide assistance, and information shared 
in the interview suggested that water management-related responsibilities are well divided between 
these two important community figures. The mahalla leader must also interact and coordinate with other 
stakeholders involved in local governance. He negotiates with mahalla committees from neighbouring 
areas when necessary (and may be required to resolve disputes jointly with them in a similar manner as 
with the WUA leader). The mahalla leader must also interact with local government to gain permission 
for allocating drinking water to homes. This is despite the fact that the mahalla committee is not a formally 
recognised institution in the law. However, information from interviews shows that, in practice, the 
 
5
In the Water Code (Republic of Tajikistan, 2000), water itself is free; water users only pay for water-related services 
such as water delivery or related electricity fees. However, due to miscommunication of the laws, there is some 
confusion, such that some water users believe that they are paying for the water resources.
Updated June 2018 106
mahalla committee is accepted by other official local actors as playing an important role in local 
governance. District state authorities are known to authorise informally the mahalla committee as the 
institution to mediate and resolve some disputes.
According to the interviewee from the Tax Administration Authority, who offered an overview of 
mahalla-WUA interactions nationally, the mahalla committee does not always work closely with the WUA 
in every case. This can be the cause of some tensions. Since the WUA ‘s mandate is to serve dehkan farms 
alone, the WUA’s conduct may not align with the mahalla leader’s activities in regulating water for kitchen 
gardens. The local government interviewee shared stories of resource and service misappropriation in his 
district and elsewhere. In some cases, foreign governments and NGOs have built pumps and/or 
established WUAs in villages, but after the project was concluded, the mahalla leader dominated and 
regulated the water-related services according to his own policy; anyone who wanted water then had to 
consult the mahalla leader. The interviewee reported that this 'happens in many places' when projects 
are not established with a long-term vision. These findings suggest that the scope and dominance of the 
mahalla committee can vary in different contexts; its role may be more pervasive in some than others, 
and can depend on personalities and other circumstances.
6. Discussion
Analysis of de facto water governance and disputes in Tajikistan in this study indicates that the mahalla 
committee does indeed play a significant role, despite being absent in Tajik legislation. Involved in a range 
of water-related community affairs, the mahalla leader regulates water for household kitchen garden 
water users. This reality was found to have implications for water dispute resolution involving kitchen 
gardens and dehkan farms. Interviewees expressed that there are frequent disputes between different 
water users from different plot types, due to kitchen garden water users diverting canal irrigation water -
formally for use by dehkan farm water users. In these instances, in the article’s case study, the mahalla
Updated June 2018 107
leader coordinates with the WUA leader to resolve the dispute. The form of dispute outlined also relates 
to fee payment disputes between dehkan farms and WUAs, since water users from dehkan farms resent 
paying fees when they observe kitchen gardens obtaining water for free. Although coordination between 
the mahalla committee and the WUA was found to exist in the case study village in Vakhsh, Khatlon (and 
elsewhere), there are also cases of mahalla leader dominance. This demonstrates the potential local 
power held by mahalla leaders, and that they may involve themselves in certain activities without 
accountability or oversight.
Overall, these mixed findings showed both coordination and dominance as possible outcomes of 
interactions between the mahalla committee and a WUA. Two scenarios were observed: that either newly 
established WUAs function cooperatively alongside pre-existing mahalla committees (i.e. coordination), 
or that the mahalla committee asserts its local influence or dominates a newly established WUA (i.e. 
dominance). From this particular research, it does not appear that WUAs discretely replace pre-existing 
forms of local governance in Tajikistan.
The role of the mahalla in local disputes and their resolution in Tajikistan is therefore one of influence, 
regulation, coordination and mediation. It also has the potential to be one of misappropriation and 
dominance of local water governance. In practice, the mahalla committee appears to undertake specific 
tasks related to water governance. These tasks are recognised by officials in government, even if they are 
not formally recognised in the law. From this research, the main emphasis of mahalla committees’ 
governance seems to be stabilising community relations to maintain social order.
Similar to Boboyorov (2013) and Noori’s (2006) discussions, this research finds that some water 
users continue to recognise and turn to a traditional, informal institution (the mahalla committee) in 
Tajikistan. In addition, the findings are particularly significant in the context of the changes in the 
movement of labour and the associated increase in importance of kitchen gardens. In turn, this has 
implications for the role of the mahalla committee and its involvement in dispute resolution.
Updated June 2018 108
Findings here demonstrate continuity with, but also add to, existing concepts and accounts in the 
literature. The World Bank (2017: 3) definition of governance seems highly relevant in this case of rural 
Tajikistan. In villages, a function of governance – water dispute resolution - is provided by formal modes 
of governance (WUAs), but also informal modes of governance (the mahalla). The case study offers one 
example of ‘hybrid arrangements’ of governance. Here, different institutions interact to produce functions 
in a complex socio-political setting. This research resonates with Stewart (2004) and Sehring’s (2009) 
analysis, highlighting the potential for traditional practices and informal institutions to undermine 
institutional reform in the region. In the accounts of mahalla leader dominance and misappropriation, 
there is evidence of mahallas seeking to ‘consolidate their constituencies’ as Stevens (2005: 293) warned. 
However, as Sehring (2009) suggests, such institutions may also coordinate with and support new policy 
endeavours, as was more the case in the village in Vakhsh, Khatlon. In resolving disputes and stabilising 
political and economic relations, the mahalla may play a significant role governing water resources and 
influencing community cohesion. Whether this influence is always positive or negative in terms of 
governance outcomes such as justice or equality is not conclusive from this research.
7. Policy implications and future research
This research has three implications for policy in Tajikistan, and may also provide broader lessons. Firstly, 
since the findings indicate that the mahalla committee is active in local water governance de facto, and is 
recognised by local citizens, there may be a need for greater formal recognition in the law, and when 
developing, implementing and monitoring institutional reforms. Secondly, the findings suggest that 
encouraging coordination between the mahalla committee and the WUA (particularly the leaders) may 
result in more functional governance and the stabilisation of disputes. It appears important to recognise 
the realities on the ground, and the role of mahalla committees, since failure to do so may facilitate 
dominance and misappropriation. Finally, the potential for misappropriation of newly introduced water-
Updated June 2018 109
related services underlines the importance of long-term monitoring and evaluation of institutional reform 
projects.
Mahalla committees fulfil certain roles without formal accountability or oversight. However, it 
was beyond the scope of the article to assess the effectiveness or desirability of the mahalla committee’s 
role in water governance, and to trace what happens after water disputes are resolved in the manner 
documented in this article. There was no opportunity to examine scenarios of non-compliance or the 
possible sanctions that a mahalla committee might impose. Further research into water disputes and 
institutional interactions addressing these issues could build on the findings from this research, and, in so 
doing, develop a greater understanding of the scope of the mahalla committee in Tajikistan.
Acknowledgments: The research for this article was funded by the United States Agency for 
International Development (USAID), under USAID Grant Number AID-BFS-G-11-00002. The opinions 
expressed herein are those of the authors and do not necessarily reflect the views of USAID. We thank 
Ted Horbulyk for providing comments on a draft version of this article. We also thank: Bakhadur 
Khabibov for clarifying legal information about the mahalla, Sino Kadyrov for translating interviews, and 
Sohrob Aslamy and Katherine MacDonald for their assistance with fieldwork logistics.
Bibliography
Abdullaev, K. and C. Barnes. 2001. ‘The politics of compromise: the Tajikistan peace process’. Accord, 
London: Conciliation Resources. http://www.c-r.org/downloads/10_Tajikistan.pdf.
Asian Development Bank (ADB). 2011. ‘Civil society briefs: Tajikistan’. Dushanbe: ADB Tajikistan Resident 
Mission. https://www.adb.org/sites/default/files/publication/29302/csb-taj.pdf. 
Updated June 2018 110
Balasubramanya, S., J. P.G. Price and T. M. Horbulyk. 2018. ‘Impacts assessments without true baselines: 
assessing the relative effects of training on the performance of water user associations in southern 
Tajikistan’. Water Economics and Policy (online first). https://doi.org/10.1142/S2382624X18500078. 
Bichsel, C. 2009. ‘It’s about more water: natural resource conflicts in Central Asia’. In: Péclard, D. 
Environmental peacebuilding: managing natural resource conflicts in a changing world. 
http://www.zora.uzh.ch/id/eprint/35018/3/bichsel_It%E2%80%99s_about_More_Water_2009.pdf.
Boboyorov, H. 2013. ‘The ontological sources of political stability and economy: mahalla mediation in 
the rural communities of southern Tajikistan’. Crossroads Asia: working paper series, 13: 1-33.
Cieslewska, A. 2010. ‘Tradition and poverty reduction – mahalla and its significance in development 
process in Tajikistan’. Paper presented at the International Conference on Eurasian Economies 2010, 
Warsaw.
Denyer Willis, G. 2014. ‘The gun library: an ethic of crime in Sao Paulo’. Boston Review, 8th April 2014. 
http://bostonreview.net/world/graham-denyer-willis-pcc-gun-library-sao-paulo-prisons-crime.
Food and Agriculture Organization of the United Nations (FAO). 2012. ‘AQUASTAT: Tajikistan’. 
http://www.fao.org/nr/water/aquastat/countries_regions/TJK/. (last accessed 9th January 2017).
Freizer, S. 2004. ‘Tajikistan local self-governance: a potential bridge between government and civil 
society?’ In: De Martino, L. Tajikistan at a crossroad: the politics of decentralization. Geneva: Cimera 
Publications.
Jepson, P. 2005. ‘Governance and accountability of environmental NGOs’. Environmental Science & 
Policy, no. 8: 515-254.
Joffé, G. 2015. ‘Traditions of governance in North Africa’. Journal of North African Studies, 20 (5): 1-15.
Kabilov, F. 2017. ‘Local water management in Tajikistan: legal framework’. Central Asian Journal of 
Water Research, 3 (2): 73-88.
Updated June 2018 111
Knaus, G. and R. Stewart. 2011. Can intervention work? Amnesty International Global Ethics Series, New 
York: W. W. Norton and Company.
Koehler, J. 2004. ‘Assessing peace and conflict potentials in the target region of the GTZ Central Asia and 
Northern Afghanistan programme to foster food security, regional cooperation and stability’, Berlin: 
Analysis and Research Consulting (ARC).
Mandler, A. 2010. ‘The influence of local governance on agricultural advisory services in Tajikistan’. 
Presented at the Leibniz Institute of Agricultural Development in Transition Economies (IAMO) Forum 
2010: Institutions in Transition - Challenges for New Modes of Governance, Halle: IAMO.
Meagher, K., T. De Herdt and K. Titeca. 2014. ‘Unravelling public authority: paths of hybrid governance 
in Africa’. IS Academy – Human Security in Fragile States, research brief no. 10: 1-8.
Noori, N. 2006. ‘Expanding state authority, cutting back local services: decentralization and its 
contradictions in Uzbekistan’. Central Asian Survey, 25 (4): 533-549.
Poos, H. 2011. ‘The local governance of social security in rural Surkhondarya, Uzbekistan: post-Soviet 
community, state and social order. Zentrum für Entwicklungsforschung (ZEF) Working Paper Series, no. 
87, Bonn: University of Bonn. 
Republic of Tajikistan. 2000. ‘Water code’. Law of the Republic of Tajikistan, no. 34. Available at: 
http://extwprlegs1.fao.org/docs/pdf/taj34375E.pdf. (last accessed 7th December, 2017).
Republic of Tajikistan. 2006. ‘Water user association law’. Law of the Republic of Tajikistan, pp. 1-12. 
Republic of Tajikistan. 2009. ‘About self-government institutions of settlements and villages’. Law of the 
Republic of Tajikistan, no. 590. Available at: http://cis-legislation.com/document.fwx?rgn=28938. (last 
accessed 20th October 2017).
Ritchie, J. and J. Lewis. 2003. Qualitative research practice: a guide for social science students and 
researchers. London: SAGE Publications.
Updated June 2018 112
Sehring, J. 2005. ‘Water user associations (WUAs) in Kyrgyzstan: a case study on institutional reform in 
local irrigation management’. Zentrum für internationale Entwicklungs- und Umweltforschung der 
Justus-Liebig-Universität Gießen – Discussion paper, no. 24: 1-49. 
Sehring, J. 2009. ‘Path dependencies and institutional bricolage in post-Soviet water governance’. Water 
Alternatives, 2 (1): 61-81.
Sievers, E. W. 2002. ‘Uzbekistan’s mahalla: from Soviet to absolutist residential community 
associations’. Chicago-Kent Journal of International and Comparative Law, 2 (1): 91-157.
Skarbek, D. 2011. ‘Governance and prison gangs’. American Political Science Review, 105 (4): 702-716.
Stevens, D. 2005. ‘NGO-mahalla partnerships: exploring the potential for state-society synergy in 
Uzbekistan’. Central Asian Survey, 24 (3): 281-296.
Stewart, R. 2004. The places in between. London: Picador.
United States Agency for International Development (USAID). 2013. ‘USAID Family Farming Program, 
Tajikistan’. Quarterly report: year three – third quarter, Dushanbe: USAID.
World Bank. 2017. ‘Governance and the law’. World Development Report, Washington: World Bank.
 
Updated June 2018 113
Appendix 5 
Kabilov, F. 2017. Local water management in Tajikistan: legal framework. Central Asian Journal of 
Water Research 3(2): 73-88. Special Issue on Water Use Management Challenges in Central Asia and 
Afghanistan. Available online at: http://www.water-ca.org/article/3214-local-water-management￾in-tajikistan-legal-framework
Updated June 2018 114
Title: Local water management in Tajikistan: legal framework
Keywords: Tajikistan, WUA, water law, water management, contract
Abstract: Water management plays an important role in the economy of Central Asian states. With the 
transition into post-Soviet era, the States implemented several phases of agricultural reforms. The 
establishment of the Water User Associations (WUAs) in Tajikistan started in the 1990s following the 
implementation of the first phase of Land Reforms in 1998-2000. Main purposes of creating such WUAs, 
often initiated and supported by international donors, is to operate, maintain and use on-farm irrigation 
system with the purpose of on-time, adequate and reliable water supply to its water users. No reforms 
take place in a vacuum. Effective legislative framework is needed to support new initiatives and 
institutional settings. While a WUA being a separate legal entity, it has certain obligations both provided 
in law and in the contracts they sign. These obligations then transformed into liabilities when non￾performance or mal-performance occur. Therefore, it is important to know what the law says about the 
chain of water management and the relationships involved, where WUA stands as an intermediate 
institution between State water organizations and farmers, the ultimate water users. This paper will look 
into these legal settings and endeavours to explain the complex nature of local water management in 
Tajikistan from the perspective of the law. 
 
1. Introduction
Tajikistan is a civil law country with laws being codified in the codes and legal-normative acts, and the 
judicial decisions as legal precedent, unlike in common law countries, are not the sources of law. Since 
its independence in 1991, Tajikistan has undertaken an effort to build its legal system entirely anew 
and towards the end of 1990s has seen most of the laws and other normative acts being adopted by 
independent Tajikistan. 
However, it should be noted that many laws in Tajikistan, especially the ones that are believed to play 
important role in the legal system and in the economic development, have been developed with the 
assistance of foreign legal experts. Therefore, in the context of water, the presumption is that when it 
comes to the players in the local level, no clarity exists on the prescription of the law regarding the 
rights and obligations of WUAs and farmers, who are the ultimate water users. The law which regulates 
WUA, its status in the society, and its relationship with other water players in the chain of water 
management is the Law on Water Users Associations (hereinafter, the WUA Law) adopted in 2006. 
Although WUA creation began in the country several years’ prior adoption of the law, the WUA Law 
was expected to fill the gap in the law. It is supposed to provide, alongside other laws such as Civil 
Code of the Republic of Tajikistan, a legal framework for the establishment and operation of WUAs 
throughout the country. 
This paper will explore the relevant provisions in the law that define the relationships among the 
stakeholders in the chain of water management in Tajikistan with the view that it will contribute to fill 
the knowledge gap in this important sector of the economy. After all, effectiveness and long-lasting 
Updated June 2018 115
positive impact of any reform depends on the laws, the clarity of the rights and obligations, and the 
legal awareness of the subjects of that reform. When things go wrong you better know your point of 
redress. 
Introduction section is followed by a general background text on the relationship among different 
types of legal regulatory acts and those subject to discussions in this paper. Then, relevant provisions 
are discussed in terms of institutional framework: main actors involved, definitions and who has what 
competence in the chain of water management, including WUAs, their operations, finances and etc. 
Remaining sections will cover dispute resolution, issues of water quality and a small discussion on 
gender as covered in the law, with a brief conclusion ending the paper. 
Disclaimer should be made at this stage that this is a purely legal paper, which looks into the law and 
gives a picture of what the law says. Direct references are made to the relevant provisions of the law 
with some analysis from the author on implications for practice. Real picture on the ground can be 
substantially different, provided this is a general tendency in post-soviet world: laws are adopted for 
the sake of adoption but not altering the practice.
 
2. General Background
According to the Constitution, Tajikistan is a state based on the rule of law1
. The legal system in 
Tajikistan is based on the hierarchy of legal norms, which are also divided based on the area of 
regulation. This hierarchy is provided by the Law of the Republic of Tajikistan on Legal and Normative 
Acts, a principal document which sets up the relationship of different legal acts (laws, by-laws, decrees 
and regulations and etc.). 
Article 7 provides the following system of laws: 
 Constitution of the Republic of Tajikistan;
 Laws of the Republic of Tajikistan, adopted through referenda;
 International legal acts (treaties), recognised by Tajikistan; 
 Constitutional laws;
 Codes, Laws (legislative acts adopted by the Parliament); 
 Joint resolutions of the Majlisi milli (upper house) and Majlisi namoyandagon (lower house) of 
Majlisi Oli (Supreme Council, the Parliament); 
 Resolutions of the Majlisi milli and Majlisi namoyandagon of Majlisi Oli (the Parliament) of the 
Republic of Tajikistan;
 Orders of the President of the Republic of Tajikistan;
 Decrees of the Government of the Republic of Tajikistan; 
 Directives of the ministries and state committees (of the Government); 
For the purpose of this paper following legal mechanisms will be subject to analysis: 
 
1 Article 1, the Constitution of the Republic Tajikistan, adopted on 6 November 1994. 
Updated June 2018 116
1. The Constitution of the Republic of Tajikistan – Article 13 provides that ‘the land, its resources, 
water, airspace, fauna and flora, and other natural resources are exclusively the property of the 
State, and the State guarantees their effective use’; 
2. The Water Code of the Republic of Tajikistan, adopted in 2000 with the latest amendment being 
made in 2012. A code is ‘a unified and ordered law through which a particular area of public 
relations is in full, directly and systematically regulated’. 2
It has to be borne in mind that the text 
of the Code is approved and adopted by a separate law of the Parliament. In other words, the 
Parliaments adopts the law to approve the Code. And because of this fact, it is suggested that the 
normative status of the Code is equal to any other laws (legislative acts), similarly adopted by the 
Parliament, such as the next instrument; 
3. The Law on Water Users Associations (the WUA Law) of the Republic of Tajikistan, adopted in 2006. 
Law (in a narrower sense) is ‘a normative legal act adopted by the legislative authority in 
accordance with the established procedures, which regulates the most important public 
relations’3
; 
4. The Law on Environmental Protection of the Republic of Tajikistan, adopted in 1994. The document is 
relevant when it comes to issues of environmental protection of water resources and water quality 
issues; 
5. Decree No. 281 of the Government of the Republic of Tajikistan on approval of the Regulations on 
the procedure of charging consumers of public irrigation systems for water supply services, 
adopted in 1996. Decree (or Resolution), in this context, is defined as ‘a normative legal act of the 
Government of the Republic of Tajikistan, adopted in the form of Decree of the Government of the 
Republic of Tajikistan’.4
6. Decree No. 39 of the Government of the Republic of Tajikistan on approval of the Regulation on the 
delineation of the powers of specially authorized state bodies for regulating the use and protection 
of water resources, adopted in 2002. The document established clearly defined responsibilities on the 
relevant ministries and State committees in the sphere of water management and protection. 
The above mentioned instruments are provided in hierarchal order. As mentioned, the normative statuses 
of the Water Code and the WUA Law are equal and referring to the Article 3 of the Water Code, both are 
constituent parts of the water legislation of the Republic of Tajikistan.5
What if conflict of laws? The Constitution is the supreme law in the country and no national law may 
contradict its provisions. However, Article 10 of the Constitution provides that in case of a conflict 
between the laws and the recognized international legal treaties, the international mechanism shall 
prevail.
Article 70 (1) of the Law on Legal-Normative Acts provides that if there is conflict between two 
different instruments applicable on one particular issue, the privilege is given either to (1) the 
 
2 Article 17 of the Law of the Republic of Tajikistan on Normative Legal Acts, adopted in 2009 with the latest 
update being made in 2011.
3 Article 18, Ibid.
4 Article 21.2, Ibid. 
5 Article 3 reads as follow: “Water legislation of RT is based on the Constitution of RT and consists of the present 
Code, laws, and legal-normative acts of the RT and the international acts recognized by the RT”.
Updated June 2018 117
instrument that is adopted later than the one it is in conflict with or if (2) the instrument is considered 
to be more specifically tailored to deal with the situation or question under consideration (lex 
specialis).
3. Institutional Framework
3.1. Main Actors and Definitions 
The primary source of water legislation in Tajikistan is the Constitution and the Water Code of the Republic 
of Tajikistan. The Constitution guarantees that the water is the exclusive property of the State6
, while the 
Water Code elaborates that principle setting detailed guidance for all kind of sectors dealing with water. 
Article 1 of the Code in its provision 1 provides that the Code aims for ‘strengthening of the lawfulness 
and protecting the rights of individuals and legal entities in the field of water relations’. The next section 
of this paper shall discuss the water relations as performed by various actors, bodies responsible for water 
management and use in the local level – namely (i) local executive bodies, (ii) water management 
organizations (vodkhozes or basin authorities), (iii) water user associations (WUAs) and (iv) water users 
(individual farming entities). The complex nature of the water relations played by the above mentioned 
actors shall make up the institutional framework for the local level water management in Tajikistan. 
For the purpose of clarity and intellectual curiosity, it is important that legal definitions for certain 
terminology and concepts are brought at this stage of the study. The Code in its Article 2 defines following 
terms, some of which will be referred throughout this study: 
 General water use – water use without use of water infrastructure or engineering tools; 
 Special water use – water use with the means of water infrastructure and engineering tools; 
 Primary water users – natural and legal entities, who are provided with water bodies for use; 
 Secondary water users – natural and legal entities, who are given permission by the primary water 
users to use water bodies on a contractual basis and with the authorisation of the State regulatory 
bodies in the field of use and protection of water resources; 
 Water User Association – non-commercial organization, established by legal entities who have 
the right to agricultural land use and commercial organizations; 
 Basin Water Management Organization (BISA) – Water management organization serving water 
users within a river or main canal basin7
. 
 
6 Article 13 of the Constitution of the RT provides that ‘the land, its resources, water, airspace, fauna and flora, and 
other natural resources are exclusively the property of the State, and the State guarantees their effective use’. 
Article 5 of the Water Code reiterates this provision. 
7 BISAs are created to replace provincial water management organizations called “OblVodkhozes”, while 
RayVodkhozes refer to previous district water authorities. Newly created organizations are gradually being 
replaced as the local executive branches of the Agency for Land Reclamation and Irrigation, central executive 
authority in the area of land reclamation and irrigation.
Updated June 2018 118
3.2. The Scope of Competence of the Various Actors 
Due to the fact that the focus area of this study is the local level actors in the water management 
hierarchy, the study will concentrate primarily on the relationship between various actors at local level as 
provided by the relevant laws. Reflections are the result of legal opinion based on the statutory law as it 
stands but where necessary reflections are also drawn on the implications of the law for practice. 
The principles and the hierarchy of the water management is provided in Article 9 of the Water Code 
which states that: 
‘The State management in the field of use and protection of water resources is based on the combination 
of basin and border and administrative-territorial management principles and implemented by the 
Government of the Republic of Tajikistan, the executive authorities, local authorities, as well as by the 
authorized State agencies on water management and protection’. 
The sequence of implementers in the provision can be interpreted as the intention of the legislators in 
terms of setting up the institutional hierarchy in water relations, which provides for the following order in 
the local water management context: 
1. The executive authorities (provincial);
2. Local authorities (district); 
3. Authorised state water management agencies (vodkhozes and/or basin authorities); 
Thus to the full picture of water relations on the ground we can conclude that beneficiaries would be: 
4. WUAs (as primary water users); and 
5. Individual water users (as secondary water users).
3.3. Competence of the Local Executive Branches of the State (Khukumats) 
Article 7 of the Water Code defines jurisdictional scope of the local state authorities within their respective 
territories. It includes but not limited to: 
- Determining main direction of water management and protection in the territory of their 
respective jurisdiction. In practice it may mean that local executive authority dictates or should 
dictate the trend of water management in their respective territories. 
- Guaranteeing lawfulness and legal order in the field of regulating water management and 
protection. The practical implication of this provision is that, in case of breach of law and 
established norms, a party at loss may bring a claim against a party at fault to the attention of 
local executive branches of the state. In addition, this provision should be read together with 
Article 12 of the Water Code, which provides that State control on water use and protection shall 
be vested upon authorised state water management agencies (vodkhozes or basin authorities). 
The implication of the combined reading of these provisions might suggest that local executive 
bodies of the State should maintain more of role of a watchdog while authorised state water 
management agencies are the ones responsible to take an action. However, it has to be noted, 
Updated June 2018 119
the claim mentioned above is an administrative claim and there is nothing in law that prohibits 
parties to sue each other for an alleged wrongdoing. 
- Registration and assessment of the state of water and water bodies, control of water use and 
protection, observance of the established water use limits; 
- Carrying out activities for the conservation and improvement of the status of water bodies, 
prevention and liquidation of harmful effects, as well as water pollution, restore facilities damaged 
by accidents, floods, landslides and other natural disasters. This has to be read together with other 
provisions of this Code and other relevant laws on rehabilitation and construction of works. It can 
be suggested that the responsibility of the local executive authorities in rehabilitating and 
restoring water infrastructure primarily covers overseeing overall status of water and water 
infrastructure in the territory it administers and that responsibility particularly becomes 
important in emergency situations; 
- Cooperation with the specially authorised state water management and protection agencies 
(vodkhozes); 
3.4. Competence of the State agencies on water management and protection (vodkhozes)
The State agencies on water management and protection on provincial and district level, namely 
Vodkhozes or/and Basin water management agencies, are in the position of entering into contractual 
relations with the primary water users for special water use, whereas primary water users can establish
contractual relations with the secondary water users for this purpose (Article 25 of the Water Code). The 
following mechanisms define the contractual rights and obligations of the parties: 
1. Contract itself
2. The Water Code 
3. 2006 Law on Water Users Associations (WUA Law). 
The rights of water users for the special water use can be restricted by the authority who vested those 
rights upon them, that is Vodkhozes upon primary water users and primary users upon secondary users 
(Article 44(4), Water Code). This provision is also reflected in Article 49, which provides for the 
circumstances when the right to water use can be terminated. This includes but not limited to expiration 
of the permission for a special water use or systematic violation of a water supply contract by a water 
user. Vodkhozes as water supply agencies maintain direct right to terminate water use right of a primary 
water user (e.g. WUA) and supervisory role in case of a primary water user terminates the right of a 
secondary water user (e.g. WUA terminating the right of its member) (Article 50). However, as stated 
above, water users maintain the right to challenge the legality of the decision on termination of their 
respective water use rights and bring it to the attention of a respective Vodkhoz and/or local executive 
body of the State. 
Water users can establish a water user association (WUA). WUAs then become a primary water user for 
the purpose of acquiring special water use rights under Article 25 above, and its respective members are 
classified as secondary users. Water users have the right to demand from Vodkhozes to take necessary 
measures to prevent exhaustion and contamination of the water fund and providing water resources in 
appropriate quantity and quality (Article 43, Rights of Water Users). However, the latter should be read 
Updated June 2018 120
in the light of obligations of the water users to make timely payment for water delivery services; 
maintaining irrigation and drainage systems in an appropriate functioning state and comply with 
established requirements and conditions of a water supply contract with Vodkhozes (Article 45). 
3.5. Responsibility of Vodkhozes in relations to the Water Users
Article 48 of the Code defines the responsibilities of Vodkhozes and it includes but not limited to: 
- Providing appropriate maintenance of irrigation systems in order to deliver the necessary quantity 
of water to water users as indicated in water supply contracts. The implication of this provision 
for practice would be that the law as such does not define what irrigation systems are meant by 
this provision. However, as the delivery of water takes place at the point of water distribution –
the WUA gate e.g. – it can be suggested that it is up to that point of irrigation system that Vodkhoz 
has to undertake maintenance work; 
- Preparing drains, water catchment-discharge systems, water ways, and water supply
- Supplying water to water users according to norms and at fixed times8
; 
- With the agreement of water users, installing water level indicators at the point of water 
distribution in order to identify the quantity of delivered water in accordance with a water supply 
contract. Implications for practice would be that it is the responsibility of Vodkhozes to install 
water measurement units at a WUA water in-take gates. 
The above mentioned responsibilities are positive obligations held in relation to the water users and if 
breached becomes liability serving as a ground for claim by the water users. 
3.6. Competence of Water Users 
According to Article 27 of the Water Code, ‘natural and legal persons, no matter of the form of ownership, 
whose activities are based in the territory of the Republic of Tajikistan and on the laws of the Republic of 
Tajikistan can become a water user’. Water users of general water use benefits from the use free of 
charge, while water users benefiting from special water use shall make payment for the use of water and 
the delivery service (Article 35, Water Code). Any analysis on the competence of the water users should 
also be undertaken in the light of the responsibilities of the players in the higher hierarchy of water 
relations, as those responsibilities are translated into the rights for water users. 
In the following sub-section, the competence of WUAs is discussed in the context of a water user 
(primary), while the next sub-section will look at WUA competence in terms of a water supplier (to 
individual members/farmers). 
3.7. Responsibilities of the Water Users 
Responsibilities of water users can be defined as contractual obligations and those provided in the Water 
Code and other normative acts, such as WUA Law. If looked from the perspective of vodkhozes, those 
responsibilities laid down in the Water Code are applicable both on WUAs and its members. However, the 
WUA Law makes the responsibilities of WUAs and individual water users clearly distinctive from each 
 
8
Subject to Article 44 of the Water Code as stated above 
Updated June 2018 121
other. Therefore, assuming individual water-users are members of a WUA their responsibilities are held 
in relation to that WUA. Meanwhile, the responsibilities of the WUA are held in relation to the respective 
vodkhoz in the capacity of a water user and in relation to its members in the capacity of water supplier. 
Responsibilities of the water users as provided in the Water Code include but not limited to (applicable 
both on individual water users and WUAs as collective water users)9
: 
- Rational use of water, saving water resources, and rehabilitating and improving water quality; 
- Making timely payment for water use and water use services. Implication of this provision would 
be that individual water users shall make payment to WUAs and WUAs to vodkhozes;
- Maintain the irrigation, drainage, wastewater treatment and other facilities in a repaired 
condition and improve their performance. One can suggest that individual water users are 
responsible to maintain the infrastructure in their own territory, including tertiary canals and in 
some cases secondary canals, and WUAs are responsible for main (the part located in WUA’s 
territory) and secondary canals. However, the provision itself provides no clarification in terms of 
territorial limits of this responsibility. 
- Comply with the established requirements and conditions of water supply contracts; 
Furthermore, the WUA Law places particular responsibilities to WUAs in its capacity as a water user. These
responsibilities basically reiterate those mentioned above, namely making timely payment to vodkhozes 
for the water supply services and maintaining water infrastructure in a working condition, although the 
latter is more of a responsibility in the capacity of a water supplier (Article 11, WUA Law). 
3.8. Competence of Water User Associations as a Water Supplier
The Water Code provides fundamental aims of establishing a WUA, which are: 
(i) Maintenance and exploitation of inter-farm amelioration and irrigation systems, which are in 
collective and individual use;
(ii) Provision of fair, effective and timely allocation of water resources among farming entities;
(iii) Collection of payments for water supply services; and 
(iv) Solving disputes among its members on matters related to water allocation (Article 43(2)). 
The WUA Law reiterates the above mentioned aims in a more elaborative manner in Article 3 (aims and 
tasks of WUA) and in Article 11 (responsibilities of WUA). In addition, the Law gives right to WUA to enter 
into contractual relations with a respective vodkhoz and individual water users, members and non￾members alike. Under the contract, WUA shall collect payment from its members and non-members and 
make payment to vodkhozes for the amount of delivered water. 
In addition to the aims, there are also principles upon which WUAs shall function. Interestingly, from the 
list of principles, the first comes ‘involvement of the WUA members in the management, repair and 
rehabilitation of the irrigation systems’ (Article 4). If the provision is read together with the first aim, which 
is on maintenance too, one can draw a conclusion that the law clearly places responsibility on WUAs to
undertake necessary repair and reconstruction works and the WUA members are under a positive duty to 
avail themselves when their involvement is required. Accordingly, the role of the WUA are more of a 
 
9 Article 45, Water Code. 
Updated June 2018 122
mobilizer of collective efforts. However, this statement may also be a misguiding because it is impossible 
to separate a member from WUA and vice-versa. WUA is association established by members, with 
collegial managerial body, composed also from the members. Then if WUA is responsible according law, 
then this responsibility is the equal subsidiary responsibility of each member of WUA. 
Operation and functioning of the WUA shall be organized in the following manner/order (Article 21 and 
22): 
1. The amount of water to be delivered to each water user and the schedule under which the delivery 
takes place is provided in the water supply contract – amount and schedule can be referred as the 
plan. 
2. Then, this plan is reflected in the internal timetable (plans) of the WUA. 
3. Then, these plans are brought to the attention of water supply organization to seek their 
confirmation. 
4. Then, the approval is sought from local executive bodies for these plans.
As the law stands, it is not quite clear how this works: whether contractual arrangements are made first 
and reflected within internal plans of WUA or contractual calculations are the reflection of the WUA 
internal plans. The whole process is referred as State regulation over the activities of WUAs. 
4. Payment for Water Supply Services
The 2006 WUA Law defines two types of payment the members shall make to WUA (Article 2): 
1. Membership fee – payment to be collected from the WUA members according to the decision of the 
general meeting (the highest management body) and to be utilized on management and 
maintenance of irrigation system and the functioning of the WUA.10; 
2. Payment for the water supply services – payment to be collected from the members and other water 
users in order to make payment to vodkhozes for the provision of water to the zone of WUA 
coverage. 
The membership and thus the above mentioned payments give right to water users (members and non￾members alike) to claim the water as stated in their contract and claim for damages from WUA in case of 
the breach of contract by WUA (Article 8). 
In addition to collecting membership fees, there is nothing in law that prohibits WUA to obtain extra 
funding from various sources to maintain its activities. A WUA also has a right to acquire a property, 
property and non-property rights11
. 
WUA can establish its own membership fee and any other fees (this may include punitive fees for non￾compliers), which has to be incorporated in its Charter. The law provides that WUA has a right ‘to demand 
fee for water supply services and any other established fees’12. Thus, the scope of any other fees can be 
wide enough, to the extent not prohibited in law. 
 
10 Article 19 of the Law asserts that the repair and reconstruction works to be covered from the WUA budget. 
11 Article 10, WUA Law 
12 Ibid
Updated June 2018 123
Furthermore, the law does not prohibit WUAs to build up a capital surplus. 
Chapter 6 of the Water Code establishes that water supply for irrigation purposes is subject to payment, 
which has to be performed on contractual basis. The contractual relations between a water supplier and 
a user creates certain rights and obligations, which are specified in Chapter 7 of the Code. Making payment 
for water supply is one of the obligations of the water users, breach of which may result in partial or full 
termination of the water supply. The same Chapter also provides for implicit risk assignments in case of 
emergency or low water availability, where the user’s right to a particular amount of water shall be limited 
in order to accommodate State interests and the interests of other water users. Drinking and domestic 
water uses shall be given priority in situations mentioned above.13
 
Detailed methods of charging for water supply is provided in the regulatory document Decree No. 281 of 
the Government of the Republic of Tajikistan on approval of the Regulations on the Procedure of Charging 
Consumers of Public Irrigation Systems for Water Supply Services, adopted in 1996. 
Accordingly, Ministry of Melioration and Water Resources and other sub-organizations, such as Vodkhozes
and Irrigation System Authorities are defined as water supplier and agricultural water consumers (and 
others) are defined as water consumer. 
Article 1.2 provides that water supply is to be conducted on the basis of a contract between supplier and 
consumer, whereas Article 1.3 provides that non-legal water consumers/users can unite within an 
association that is a single water consumer who will be entitled to enter into a water supply contract with 
the Water supplier (the case with WUA). 
Article 1.5 provides for dealing with emergency situations, according to which, ‘both parties shall 
participate with their respective technical, material and other means to eliminate the consequences of 
emergency situation and each side bears the cost of the works undertaken in its own respective side’. The 
same principle applies when the provision is interpreted in the WUA context, that is to say expenses for 
accident’s repair and liquidation of its consequences must be covered by respective institution, who owns 
this part of irrigation system. This provision if read together with the responsibilities of WUAs, once again 
clarifies that each side is responsible to bear its own cost in its respective territories.
Measurement unit for water is m3 and for irrigated land is hectare (ha). The amount of water delivered to 
the consumer is measured with special gauges. Where there is no a gauge, delivered water is measured 
with instrumental measurement techniques and the processes is supervised by authorised persons from 
each side and duly recorded. On the basis of these records at the end of each month an act of receiving a 
water supply is made. The act serves as a document for payment collection (Article 2). 
Article 3 and its provisions define the order of payment. Payment is calculated according to the tariff for 
each m3 delivered to the consumer. Tariffs for water supply services are determined based on the total 
regulatory costs: for maintenance and repair of public irrigation and drainage systems and their 
structures, full restoration of depreciated assets, mandatory payments, the insurance fund to cover 
expenses during events such as extremely low availability of water or flooding, expansion of production, 
scientific, technological and social development of the Supplier. Salary costs of the staff of water supply 
 
13 Article 44, Water Code
Updated June 2018 124
organization are not mentioned in the provision, as the water supply organization is not a self-sustaining 
commercial entity and thus it is under the state budged. 
40% advance payment is applicable for water supply services for irrigation purposes as provided in Article 
3.7. Payment for water supply services is paid according to established tariff as long as water withdrawal 
remains within established agreed limits. Over the limit water withdrawal will be charged with multiplying 
factor equal to 1/2 and an unauthorized withdrawal will be charged three times more. 
Within associations united as a single consumer (e.g. WUA), tariffs (how much water is requested) will be 
determined by consumers, subject to agreement by the Supplier.
5. Conflict Resolution 
The hierarchy of the institutions with the capacity to resolve disputes related to water management is as 
follow: 
1. The Government of the Republic of Tajikistan; 
2. Local State Authorities; 
3. Authorised Water Management Bodies; and 
4. The Court. 
WUA has a positive duty to resolve disputes among its members and non-members14 through establishing 
an ad hoc commission comprising three people. Commission adopts an act with the decision and the act 
is given to an appropriate body of WUA for implementation15
. 
Resolving inter-WUA disputes is the responsibility of the authorised water management body (Vodkhoz) 
who gave them the right for water use.16 Any dispute can be brought to the attention of the Court if the 
above mentioned institutions are not able to settle dispute for the parties. 
6. Water Management and Water Quality – Legal Framework 
A special chapter of the Water Code deals with the water protection issues (Chapter 21 Water Protection). 
According to it, ‘all waters (and water infrastructure) are subject to protection from pollution, clogging 
and depletion’. But some harm is permissible, as long as it does not undermine the natural capacity of 
water to self-rehabilitate. 
The Law obliges all legal entities, whose activities have impact on the quality of water, ‘to undertake 
technological, forest-ameliorative, agro-technical, hydro-technical, sanitary and other measures to 
provide protection against water pollution, clogging and depletion and facilitates improvement of water 
quality and water regime’17. These rules are one of a preventive character. 
Two practical implications can be drawn from the above provision: 
 
14 Article 11, WUA Law (2006) 
15 Article 
16 Article 116, 117, Water Code
17 Article 120, Water Code 
Updated June 2018 125
1) all legal entities include water user associations, as well as WUA members alike with a legal entity 
status; 
2) the law does not actually prohibit all the harm but it obliges entities causing harm to undertake 
measures to mitigate the harm and thus improve quality of the water. In other, words the rule is one 
of due diligence18. Thus, a conclusion would be an entity who is causing harm and not addressing it 
shall be considered as the one who is causing harm to the quality of water. This conclusion will be 
relevant for the purposes of remedy. 
Water pollution and clogging is considered to be a violation of the law 19 and subject to damages in the 
amount and order established by law20, However, payment in the form of remedy does not discharge the 
violator from undertaking the measures to mitigate and eliminate the harm21
. 
Territorial planning of water protection measures is the responsibility of the local executive authorities, 
while implementing costs of water quality measures shall be borne by the State and the water users like.22
Water users (both legal and individuals) are obliged to undertake measures to warn and liquidate adverse 
impact on water caused by accidents and emergency situations. These measures are subject to 
coordination and agreement with the state water management agencies23
. 
Coordination of urgent matters related to warning and liquidation of adverse impacts on water will be 
undertaken by a commission established by the Cabinet of Ministries and the state local authorities24
. 
Overall control and monitoring the changing quality of water resources is the responsibility of the State 
Committee on Nature Protection25
. 
6.1. Right to a Healthy Environment
The foundation of the right to a healthy environment is to be found in the Constitution of the Republic of 
Tajikistan. Article 36 provides that the State guarantees each citizen’s right to a healthy environment. This 
right is further reiterated in the Law on Environmental Protection (1994), whose scope is extended to 
cover the surface as well as groundwater water resources26. The Law gives right to both individuals and 
public and non-governmental associations (this includes WUAs) to bring a complaint to the authorised 
state agencies in the field of environmental protection, who then may take an administrative and legal 
action against a party at fault27
. 
 
18 Due diligence is an extremely important concept in environmental regulation and compliance. When properly 
done, it provides a defence to regulatory charges or, at least, can mitigate the amount of fines imposed.
19 Article 142, Water Code 
20 Article 144, Water Code and Article 20 of the 1994 Law on Environmental Protection 
21 Article 20 (4), the 1994 Law on Environmental Protection 
22 Article 122, Water Code
23 Article 129, Water Code 
24 Article 130, Water Code 
25 Decree No. 39 of the Government of the Republic of Tajikistan on approval of the Regulation on the delineation 
of the powers of specially authorized state bodies for regulating the use and protection of water resources, 
adopted in 2002
26 Article 6, Law on Environmental Protection 
27 Article 9 on the Competence of the Authorised State Agencies in the Field of Environmental Protection 
Updated June 2018 126
Individuals and public and non-governmental associations may also bring a direct legal claim to the court 
in case their right to a healthy environment is infringed28. This means WUAs fall within the latter category 
and thus having the right to act as a claimant in legal proceedings. 
However, it has to be noted that if an environmental harm is caused due the fault of one of the members 
of the WUA, the respondent (defendant) shall be the member itself but not the WUA, because, according 
to the legislation, the WUA is not responsible for the activities of its members29
.
Remedies under the law includes damages, specific performance (that is rehabilitation and maintenance 
works and etc.) and potential losses30. The Water Code also provides that the right to use water may be 
taken away from the entity in case environmental balance is substantially violated31. It can comfortably 
be suggested that ecosystems and environmental flows can be included in this provision, meaning if the 
share of water of the environment is taken away from the environment this may mean that environmental 
balance is infringed. These rules mentioned above are of a remedial character.
7. Law, Gender and Equality
The Constitution of the Republic of Tajikistan recognizes international law as a component part of the 
national legal system, and Tajikistan is a Party to the Convention on the Elimination of All Forms of 
Discrimination Against Women (CEDAW) and to other fundamental human rights treaties. In 2014, the 
parliament ratified the Optional Protocol to CEDAW, which allows individual women in Tajikistan to 
submit complaints to the CEDAW Committee and gives them an additional remedy for violations of the 
convention. Important steps have also been taken to implement UN Security Council resolutions on 
women, peace, and security (1325 and 2122) with the drafting of a national action plan. 
The Constitution guarantees equal rights on the basis of sex (Article 17), and principles of non￾discrimination are enshrined in basic legislation, for example, the Family Code, the Labour Code, the Land 
Code, the Criminal Code, the Law on Education, and the Law on Public Health.
No laws that have been discussed above in the context of water management make reference to gender 
issues. But if read in the light of national and international mechanisms in the field of protection of 
women’s right, the laws in the field of water management have equal application on both man and 
woman.
8. Concluding remarks 
The paper has reviewed legislative framework in Tajikistan in order to understand the functioning of water 
user associations and whether the law supports smooth operation of local water management if 
interpreted and applied correctly. Reforms and transition become sustainable if they are supported by 
law and order. Effective functioning of the law and order depends to the extent people can rely on them. 
 
28 Article 12 and 13 respectively 
29 Article 11, WUA Law (2006) 
30 Article 77-81 on dispute settlement and remedies, Law on Environmental Protection 
31 Article 49 (2), Water Code
Updated June 2018 127
Reliance become stronger if people know what the law says in terms of their rights and obligations, making 
legal capacity building and raising awareness important component of any developmental endeavours. 
The analysis shows that the frame is good enough to provide for fair and equitable water management 
system. However, due to the nature of the legal system of the country (this is true for almost entire post￾soviet world), making sense out of the law is not easy as documents make endless reference to existing 
legislation. This becomes particularly erroneous for organizations when resources are limited. But then, 
even one takes an effort for detailed analysis there are still lack of clarity in the law, e.g. in terms of 
responsibilities of the stakeholders for rehabilitation works, water measuring and competences of Local 
Executive Branches of the Government and Vodkhozes.
Moreover, water is still a political phenomenon in Tajikistan and in neighbouring states. This makes power 
politics in resources management as an integral part of the system. But if we talk about sustainability, 
efforts should be made to abandon old habits and make the way for clear, effective and just laws to do 
the job for us. 
Acknowledgement
This research is funded through a project titled ‘Impact of Water User Associations on Water and Land 
Productivity, Equity and Food Security in Tajikistan’ funded by the United States Agency for International 
Development under USAID Grant Number AID-BFS-G-11-00002, awarded to the International Water 
Management Institute, Colombo. The author would like to thank Ted Horbulyk and Soumya 
Balasubramanya for comments on an earlier version of this paper. 
Updated June 2018 128
Updated June 2018 129
Appendix 6 
Wiberg, D. 2018. Remotely-sensed cropping and productivity differences between areas 
with water user associations and areas without. Technical note. 
Updated June 2018 130
Title: Remotely-sensed cropping and productivity differences between areas with water user 
associations and areas without
Introduction and methodology
One of the major activities of USAID’s Family Farming Program (FFP) was the creation of WUAs to 
improve timely, regular, and equitablee access to on-farm water for improving yields, crop diversity, and 
enhancing livelihoods. Since 2011, USAID set up WUAs in 12 districts in southern Khatlon Province, to 
improve on-farm water delivery, which could impact yields, crop choice and activities of agricultural 
workers.
To support the evaluation of the impact of these water user associations on productivity in cultivated 
areas, a remote sensing analysis was performed to assess difference in productivity between areas 
where WUAs have been established and those where they have not been established. This analysis 
focused on cotton and wheat cultivation and used satellite images during the growing seasons from 
April to Nov. and Dec. to April, respectively, for USAID project and non-project areas between 2010 and 
2017. The year 2010 has been taken as a base year, since private farms did not exist before 2010. 
Agricultural land was still collectivized at that time, and farmers couldn’t decide what to cultivate.
Landsat satellite images, having suitable temporal and spatial, 30 m, resolution were chosen to carry out 
the analysis. Landsat scenes with minimum cloud and snow cover, for an area of 2.3 million ha of land 
spread over 12 districts of southwest Khatlon province, between the years 2010 and 2017, were 
selected and downloaded. The downloaded scenes were acquired by Landsat 5, Landsat 7 ETM+ and 
Landsat 8 sensors. 
For each year two sets of images were used for the analysis. One set for the cotton growing season, and 
the other for the wheat growing season. The study area covers multiple satellite scenes. Five satellite 
images are required to cover the study area for one season. Suitable cloud free images were not 
available for the years 2012 and 2013. Sixty images were required for the analysis of both the seasons in 
the remaining six years. The downloaded satellite images were pre-processed including layer stacking 
and resolution merge techniques for the data obtained from Landsat 7 and Landsat 8, taking advantage 
of presence of panchromatic band with higher resolution in both Landsat 7 & 8.
The selected satellite images were classified into 11 classes following the supervised classification 
technique, using the maximum likelihood algorithm. The signatures for various land use types were 
generated/gathered based on the ground truth data provided and high-resolution images from Google 
Earth. The classes the satellite images were classified into included Cotton, Wheat, Rice, Other 
Crops/Vegetables, Fallow land, Bare land/Shrub/Fodder, Fruit Orchards/tree cover, Urban/Built-up, 
Wetland, Water, and Snow with an extra category to indicate the area covered by Clouds.
Cotton and wheat coverage have been assessed during their growing seasons, August through October 
for Cotton, and November through March for Wheat. The satellite images for the months of September 
and October were preferred for the identification of cotton as during these months cotton crops start 
bearing flowers resulting in a unique reflection recorded by the satellite sensors and the specific spectral 
reflectance pattern of cotton is more distinct from other crops during this period. Similarly, the satellite 
images for month of March were preferred for the identification of wheat crops as most of the wheat 
fields were fully grown giving a strong vegetative reflection.
Updated June 2018 131
Results
Time series analysis
Analysis of the outputs do not indicate major changes between 2010 and 2017 in the overall area under 
cotton cultivation (Table 1, Fig. 1). However, wheat production appears to be increasing dramatically 
between 2014 and 2017 (Table 2, Fig 2). 
Fig. 1. Area under cotton cultivation in the study area
No. Land Cover
Area in hectare during the cotton season of the year (ha)
2010 2011 2014 2015 2016 2017
1 Cotton 231591 237227 231062 163700 220707 224977
2 Rice 2964 1540 9859 3602 797 13412
3 Other Crops 17073 80653 52058 152628 89299 113691
4 Fallow land 474850 436171 460894 430172 421756 396524
5 Bare land/Shrub/ 
Fodder 1322510 1252510 1252850 1321930 1322780 1251200
6 Fruit orchards /Tree 
Cover 119266 119983 120356 120568 119920 120288
7 Built-up 69172 69184 72895 72975 72991 73679
8 Wetland 20768 20896 27124 4313 24995 20284
9 Water 49012 51321 39317 32806 42165 50725
10 Snow 10502 1381 2772 6121 2256 293
11 Clouds 46764 48574 8938 85 52691
Table 1: Area of different land use types during cotton cultivation season of the year in the entire study 
area
0
50000
100000
150000
200000
250000
2010 2011 2014 2015 2016 2017
Hectare
Year
Area of Cotton Cultivation
Updated June 2018 132
Fig 2. Area under wheat cultivation in the study area
No. Land Cover
Area in hectare during the wheat season of the year (ha)
2010 2011 2014 2015 2016 2017
1 Wheat 57618 93145 74581 109051 125809 135310
2 Rice 14043 286.537 0 1108 3092 12351
3 Other Crops 119469 93020 58185 136207 139897 130991
4 Fallow land 548194 517224 549350 419103 449711 451212
5 Bare land/Shrub/ 
Fodder 1322560 1249850 1252710 1251370 1322740 1250910
6 Fruit orchards /Tree 
Cover 118083 119402 120248 117923 119994 116856
7 Built-up 68205 68711 72908 72103 72992 71715
8 Wetland 16701 5830 7611 15422 27157 18431
9 Water 52830 32330 43340 41389 41181 46896
10 Snow 82550 86472 143443 11642 83089
11 Clouds 55291 52355 10608 3533
Table 2: Area of different land use types during wheat cultivation season of the year in the entire study 
area
Analysis of Water User Association areas
A comparison of areas where water user associations are (WUA) present and where those institutions 
are not present indicates significant differences. A comparison of different land cover types between 
WUA areas and remaining areas can be found in Table 3 & 4. The percentage of cotton and wheat 
cultivated areas within agriculture lands is higher where WUAs are present compared to other areas. 
(Fig 3 & 4). While extent of cotton cultivation has not changed much over time, the proportion of cotton 
cultivation in croplands in WUA areas is much higher than remaining areas. The proportion of wheat 
0
20000
40000
60000
80000
100000
120000
140000
160000
2010 2011 2014 2015 2016 2017
Hectare
Year
Area of Wheat Cultivation
Updated June 2018 133
cultivation in croplands in WUA areas are not only higher compared to the remaining areas, but these 
areas also show significant increases in wheat cultivation over time. The cultivated area of wheat where 
WUAs are not present is stagnant.
Figure 3. Comparison of percentage of cotton cultivated areas in agriculture lands, in areas where water 
user associations are present and the remaining areas.
0%
10%
20%
30%
40%
50%
60%
2010 2011 2014 2015 2016 2017
Area (%)
Year
Percentage area under cotton cultivation
Water use Association present No Water use association present
Updated June 2018 134
Figure 4. Comparison of percentage of wheat cultivated areas in agriculture lands, in areas where water 
user associations are present and the remaining areas.
0%
5%
10%
15%
20%
25%
2010 2011 2014 2015 2016 2017
Area (%)
Year
Percentage area under wheat cultivation
Water use Association present No Water use association present
Updated June 2018 135
Water User Association present Water User Association not present
No. Land Cover 2010 2011 2014 2015 2016 2017 2010 2011 2014 2015 2016 2017
1 Cotton 174141 182649 178557 117777 167975 167956 57450 54578 52505 45923 52732 57021
2 Rice 698 192 6219 1975 46 4260 2266 1348 3640 1626 751 9152
3 Other Crops 10157 49850 27024 97200 53097 59151 6916 30803 25034 55428 36202 54540
4 Fallow land 129770 108973 90217 105657 94167 102905 345080 327198 370677 324515 327589 293619
5 Bare land/ 
Shrub/Fodder 17602 9636 17927 13689 15183 9080 1304908 1242874 1234923 1308241 1307597 1242120
6 Fruit orchards/ 
Tree Cover 40612 22620 52737 38481 44200 29159 78654 97363 67619 82087 75720 91129
7 Built-up 41241 40939 42178 41412 40043 42107 27931 28245 30716 31563 32948 31572
8 Wetland 3331 2467 3617 268 2522 2731 17438 18429 23507 4046 22473 17553
9 Water 2525 2732 1077 1808 2846 2685 46487 48589 38241 30998 39319 48040
10 Snow 0 16 518 0 0 0 10502 1365 2254 6121 2256 293
11 Clouds 0 0 8 1815 0 48 0 46764 48565 7123 85 52643
Table 3: Area of land cover types during the cotton growing season in locations where Water User Associations are present and where not 
present
Updated June 2018 136
Water User Association present Water User Association not present
No. Land Cover 2010 2011 2014 2015 2016 2017 2010 2011 2014 2015 2016 2017
1 Wheat 32499 40705 32890 58033 80339 75803 25118 52440 41691 51018 45470 59507
2 Rice 8245 186 0 104 2374 7610 5798 101 0 1004 718 4741
3 Other Crops 69965 53084 26220 62743 94239 64356 49504 39936 31965 73464 45658 66635
4 Fallow land 211853 225093 262618 207990 130567 177551 336341 292131 286732 211113 319144 273661
5 Bare land/ 
Shrub/Fodder 13677 18146 24313 21482 18832 13338 1308883 1231704 1228397 1229888 1303908 1237572
6 Fruit orchards/ 
Tree Cover 37914 40330 26623 21303 45942 35150 80169 79072 93625 96620 74052 81706
7 Built-up 40148 40406 41885 41794 40415 41419 28057 28305 31023 30309 32577 30296
8 Wetland 3051 727 329 1527 3772 2577 13650 5103 7283 13894 23385 15854
9 Water 2725 626 1752 3192 2594 2275 50104 31704 41588 38197 38587 44621
10 Snow 0 0 7 61 281 0 0 82550 86465 143382 11361 83088
11 Clouds 0 772 3443 1853 726 0 0 54519 48911 8756 2807 0
Table 4: Area of land cover types during the wheat growing season in locations where Water User Associations are present and where not 
present
Updated June 2018 137
Discussion
The results appear to indicate that the development of Water User Associations have benefited crop 
production. In the case of cotton, the cultivated area is much higher where WUAs are present. However, 
since the cultivated area was higher throughout the entire time period, it may simply be that land where 
WUAs are present is more suitable for cotton anyway. A stronger case can be made for wheat 
cultivation of the beneficial impact of WUAs, since wheat cultivated areas are both larger and growing 
where WUAs are present throughout the period, but not growing where WUAs are not present.
As already alluded to, the results presented here come with a few caveats. The results come from 
remotely-sensed images, and therefore provide only an indication of the area of land covered by the 
crops assessed. Image classification is also not perfect. Results can always benefit from further 
verification, both in terms of checking and verifying the methodology, and comparing to any ground 
data and statistical information that is available. Furthermore, production changes have multiple causes 
and influences. The remote sensing analysis presented here only provides an indication of changes in 
land use and land cover. It does not by itself establish the causality of the patterns presented.
By assessing only land cover, this was also only an initial step in the types of analysis that could be 
performed using remote sensing. Additional remote sensing analysis could further aid the assessment of 
the effectiveness of water user associations through estimations of water productivity on these lands, 
for example. 
The classified land cover maps for each year are provided on the following pages.
Updated June 2018 138
Updated June 2018 139
Updated June 2018 140
Updated June 2018 141
Updated June 2018 142
Updated June 2018 143