TENURE AND GLOBAL 
CLIMATE CHANGE (TGCC)
Evaluation Report
This publication was produced at the request of the United States Agency for International 
Development. It was prepared independently by The Cloudburst Group.
This report was developed by Heather Huntington, Aleta Starosta, Ben Ewing and Nicole Walter, with 
research support from Aidan Schneider and Carola Cerda-Zamudio. The research team thanks Dan 
Mattingly, Ajay Shenoy, M. Mercedes Stickler and Lauren Persha for invaluable design and research 
inputs. The evaluation would not have been possible without the support and collaboration of the 
TGCC program team, led by Matt Sommerville, or the tireless data collection efforts of Rural Net Inc. 
Our biggest acknowledgement is to the thousands of Zambian respondents who took the time to 
participate in the study.
Photo Credit: Sandra Coburn—The Cloudburst Group
Prepared for the United States Agency for International Development, USAID Contract Number AID￾OAA-TO-13-00019, Evaluation, Research and Communication (ERC) Task Order under Strengthening 
Tenure and Resource Rights (STARR) IQC No. AID-OAA-I-12-00030.
Implemented by: 
The Cloudburst Group
8400 Corporate Drive, Suite 550
Landover, MD 20785-2238
Tenure and Global Climate 
Change (TGCC)
Evaluation Report
JULY 2018
DISCLAIMER
The authors' views expressed in this publication do not necessarily reflect the views of the United States 
Agency for International Development or the United States Government.
CONTENTS
1.0 EXECUTIVE SUMMARY ........................................................................................................ 5
2.0 BACKGROUND AND CONTEXT...................................................................................... 15
DEVELOPMENT CHALLENGE ............................................................................................................................................. 15
LAND TENURE AND ADMINISTRATION....................................................................................................................... 15
AGROFORESTRY..................................................................................................................................................................... 16
LINKED LAND TENURE AND AGROFORESTRY INVESTMENT ............................................................................. 17
3.0 TGCC OVERVIEW................................................................................................................ 20
TENURE SECURITY STRENGTHENING INTERVENTION ......................................................................................... 21
AGROFORESTRY EXTENSION INTERVENTION.......................................................................................................... 24
4.0 EVALUATION METHODS .................................................................................................. 27
EVALUATION DESIGN .......................................................................................................................................................... 27
OUTCOME FAMILIES, HYPOTHESES, AND INDICATORS ........................................................................................ 27
SAMPLING DETAILS AND DATA COLLECTION ......................................................................................................... 29
ANALYTICAL APPROACH ................................................................................................................................................... 34
5.0 OUTCOME FAMILY I, TENURE SECURITY..................................................................... 37
SUMMARY OF KEY FINDINGS ............................................................................................................................................ 37
RESULTS...................................................................................................................................................................................... 38
DISCUSSION.............................................................................................................................................................................. 41
6.0 OUTCOME FAMILY II, LAND GOVERNANCE................................................................ 46
SUMMARY OF KEY FINDINGS ............................................................................................................................................ 46
RESULTS...................................................................................................................................................................................... 47
DISCUSSION.............................................................................................................................................................................. 50
7.0 OUTCOME FAMILY III, UPTAKE OF AGROFORESTRY PRACTICES & SEEDLING 
SURVIVAL .................................................................................................................................... 52
SUMMARY OF KEY FINDINGS ............................................................................................................................................ 52
RESULTS...................................................................................................................................................................................... 53
DISCUSSION.............................................................................................................................................................................. 57
8.0 OUTCOME FAMILY IV, FIELD INVESTMENT................................................................. 65
SUMMARY OF KEY FINDINGS ............................................................................................................................................ 65
RESULTS...................................................................................................................................................................................... 67
9.0 OUTCOME FAMILY V, LONG TERM OUTCOMES: AGRICULTURAL 
PRODUCTIVITY & LIVELIHOODS.......................................................................................... 68
SUMMARY OF KEY FINDINGS ............................................................................................................................................ 68
RESULTS...................................................................................................................................................................................... 68
ANNEX 1. GEOSPATIAL CONTEXT...................................................................................... 69
ANNEX 2. WITH CERTIFICATE VERSUS WITHOUT CERTIFICATE.............................. 75
ANNEX 3. AGROFORESTRY .................................................................................................... 78
ANNEX 4. MULTIPLE TEST CORRECTION .......................................................................... 89
ANNEX 5. SUPPLEMENTAL SUMMARY STATISTICS AND REGRESSION TABLES .... 90
ANNEX 6. SUPPLEMENTAL EVALUATION DOCUMENTATION.................................... 91
ANNEX 7. ATTRITION TABLE ................................................................................................ 92
ANNEX 8. DISCLOSURE OF ANY CONFLICTS OF INTEREST......................................... 93
REFERENCES............................................................................................................................... 94
ACRONYMS AND 
ABBREVIATIONS
Ag Agroforestry
Ag+LT Agroforestry + Land Tenure
CDLA Chipata District Land Alliance
CLC Customary Land Certificate
COMACO Community Markets for Conservation 
CSA Climate-Smart Agriculture 
ERC Evaluation, Research and Communication
IE Impact Evaluation
ITT Intent to Treat
M&E Monitoring and Evaluation 
LT Land Tenure
RCT Randomized control trial
STARR Strengthening Tenure and Resource Rights
TGCC Tenure and Global Climate Change
TOT Treatment on the Treated
USAID United States Agency for International Development 
VLC Village Land Committee
YGL Yield Group Leaders
5
1.0 EXECUTIVE SUMMARY
Agroforestry is widely perceived as a long-term sustainable land use practice that can help meet a range 
of rural development objectives in sub-Saharan Africa. Expected benefits include increased crop yields 
and more effective adaptation and mitigation responses to climate-change impacts (Mbow et al., 2014). 
Favorable Zambian agricultural policy has encouraged Climate-Smart Agriculture (CSA), and a number 
of organizations have actively promoted conservation agriculture and agroforestry to encourage food 
security, especially in Eastern Province. However, uptake of CSA practices, in particular agroforestry, 
remains limited and dis-adoption is high, despite the expected benefits to Zambia’s small holder farmers 
who struggle with low yields, unreliable access to fertilizer and limited resilience.
Insecure resource tenure is hypothesized to constrain smallholder investment in the long-term 
productivity of their fields. Many studies provide empirical evidence of the benefits of tenure security for 
promoting greater field investment outcomes (Deininger et al., 2011; Deininger and Chamorro, 2004; 
Feder et al., 1988; Holden et al., 2009; Jacoby et al., 2002; Rozelle and Swinnen, 2004). A basic premise 
of stronger and more secure land tenure is that the enforcement of these rights lessens the risk of 
forcible displacement and allows for a level of long-term security and a sense of permanence that 
encourages land-related investment (Besley, 1995). 
Consequently, land tenure security and property rights governance issues represent a central focus in 
Zambia for a range of rural development initiatives to address agricultural livelihoods and poverty 
reduction. In Zambia, where the majority of the land is under customary management by traditional 
chiefs, smallholders commonly have no documentation of their land rights, which can result in complex 
land disputes over boundaries or defense of rights in the event of divorce, death of a family member or 
arbitrary reallocation of land by chiefs or headmen. Uncertainties over land allocation processes within 
villages also contribute to ongoing land conflicts. This is an especially pressing issue in the rural areas of 
Zambia, where insufficient access to arable land is a recognized driver of continued impoverishment. 
Prior research indicates large variation in the size of farmer landholdings among village households, 
significant numbers of land constrained households even in villages where unallocated land is present and 
widely varying perceptions about land availability and ease of acquisition for farmland expansion (Jayne et 
al., 2009).
Nevertheless, many questions remain around the efficacy of activities that are hypothesized to 
strengthen farmer perceptions of the tenure security of their farm holdings, as well as the extent to 
which strengthened land tenure security incentivizes farmers to undertake longer-term sustainable land 
investments such as agroforestry. Although some studies have found strong evidence of positive impacts 
for land formalization (Deininger et al., 2011; Goldstein et al., 2018), no clear consensus has emerged 
from empirical studies across varying sub-Saharan Africa contexts on whether and how stronger tenure 
security may, as a whole, incentivize farmer decision-making and pursuit of different land investment 
strategies on their farms (Place, 2009). 
Recent literature has paid particular attention to the role of customary land registration in sub-Saharan 
Africa as a means of strengthening smallholder perceptions of tenure security and altering their land use 
decision-making strategies to undertake longer-term land investments (Place and Otsuka, 2001; Smith, 
6
2004). However, there remains a dearth of rigorous empirical evidence on the efficacy of interventions 
to strengthen customary land tenure. 
With a view towards addressing these limitations, USAID’s E3/Land and Urban Office piloted the 
Tenure and Global Climate Change Program (TGCC) in Zambia to explore the relationship between 
secure resource tenure and the adoption of agroforestry practices. TGCC was implemented in Chipata 
District, one of nine districts of Zambia’s Eastern Province, from 2014–2017. The intervention was 
designed to strengthen customary tenure security at the chiefdom, village, and household levels, while 
also supporting agroforestry extension services, primarily at the village level. 
TGCC piloted two treatments: 
1. Land Tenure (LT): centered on a village-level land tenure intervention consisting of participatory 
mapping, village headperson land administration support and the facilitation of informal customary 
land use certificates for households1;
2. Agroforestry (Ag): centered on agroforestry extension in villages to facilitate tree planting adoption 
and survivorship on smallholder farms; and 
This report presents the results of an impact evaluation of TGCC. The primary objective of the impact 
evaluation (IE) was to determine whether TGCC’s village and household land tenure interventions
strengthen the security of land tenure and resource rights for smallholders, thereby increasing farmer 
investment in sustainable agroforestry and uptake of other CSA practices. 
The overarching research question that motivated this evaluation was:
“How do changes in property rights that strengthen a farmer’s perception of long term security 
over farmland affect a farmer’s decision to practice climate-smart agriculture, 
including agroforestry, on their own farms?”
To answer this central question, TGCC was implemented as a four-arm village-level randomized control 
trial (RCT) with three key sub questions for hypothesis testing: 
1. Whether an agroforestry extension program increased farmer investment in sustainable agroforestry; 
2. Whether a village and household informal land registration and governance intervention strengthened
perceptions of land tenure security and resource rights for smallholders in a customary context; and, 
3. Assuming farmers’ perception of tenure security were strengthened, whether stronger perceptions 
of tenure security increased farmer investment in sustainable agroforestry. 
Villages in the evaluation sample were randomized into four treatments (Agroforestry, Land Tenure, 
Agroforestry and Land Tenure, and Pure Control) across four chiefdoms. The implementation of TGCC 
as an RCT enabled a rigorous assessment of the direct and joint impacts of the agroforestry extension 
and tenure security strengthening interventions. 
The analysis for this report relies primarily on a household panel dataset consisting of approximately 
3,200 respondent observations across 246 villages. Baseline data was collected from June to August 
2014, prior to the start of the TGCC program, and follow-up data was collected from June to August 
2017, following the completion of the program. To supplement the primary outcome analysis, we 
analyze several additional sources of data, including a quantitative village level survey conducted with the 
 1 Note that TGCC program did not have sufficient scope and capacity to undertake institutional capacity building for a land administration 
system.
7
village headman or headwoman and quantitative key informant interviews with a lead farmer and a village 
land committee member in each village. In addition, 62 focus group discussions were conducted in 28 
villages, with special attention paid to women and youth. Additional focus groups were held with TGCC 
program staff to collect information about variation in implementation and other village-level data. 
To investigate each research question, we developed a series of primary and secondary indicators across 
five outcome families: tenure security, land governance, agroforestry uptake and survivorship, 
investment in agriculture, and agricultural productivity and livelihoods. For each indicator, we present 
the average treatment effects, in addition to heterogeneous impacts for female-headed, youth, elderly, 
poor and land constrained households. This is our primary impact analysis for the evaluation and relies 
on the household panel dataset. 
Additionally, we support this primary analysis with descriptive statistics based on the panel results, as 
well as full follow-up survey sample results for the section that discusses the agroforestry results. We 
also include a secondary “within-treatment” analysis for subgroups to assess whether there were 
differential effects within the treatment groups to gauge the tendency for elite capture. Finally, we 
incorporate additional qualitative findings from key informant interviews and focus group discussions to 
further ground the analysis. 
To our knowledge there is no experimental evidence studying the impact of strengthening customary 
land tenure on perceived tenure security, and little or no evidence of any sort on whether granting 
customary documentation to a producer makes her or him more likely to adopt new technology. As 
such, the TGCC evaluation has generated new knowledge around the impacts of informal customary 
land registration on household-level development outcomes. This contributes towards enhanced policy 
and programming and provides insights on the role of land rights clarification and enhancement in 
customary contexts for meeting broader development objectives.
KEY HOUSEHOLD FINDINGS
The evaluation results for the household panel analysis suggest:
• significant and positive impacts of informal customary land registration on short and long term 
perceived tenure security and decreased perceived expropriation risk by internal and external actors; 
• a positive impact of the agroforestry extension program on uptake of agroforestry tree planting, 
although the actual rate of tree planting and seedling survival remain low; 
• no evidence that greater perceived tenure security motivated higher rates of agroforestry 
investment;
• no evidence of treatment impacts for indicators measuring land governance, agricultural productivity,
or livelihoods. Given the short timeline between the close of the intervention and follow-up data 
collection, the absence of a treatment impact for long-term indicators, such as crop yields and 
welfare impacts, was expected. 
The findings indicate that the TGCC land registration and governance intervention successfully increased 
perceptions of tenure security. The process of boundary demarcation and the expectation (if not the 
actual delivery2) of receiving customary certificates for fields did indeed make households feel more 
secure in their property rights. Treatment households perceive that their fields are more secure from 
reallocation or unauthorized appropriation from both internal and external threats in the next three 
 2 Delivery of customary land certificates was not uniform across the treatment areas, as detailed in Section 3. 
8
years and beyond four years. Looking across the binary measures for the six sources of dispossession, 
we find that the land registration and governance intervention raised the probability that households 
think encroachment or unauthorized appropriation is “impossible” by 5 to 6 percentage points for 
headman, neighboring villages and other households within their home village. 
Almost by definition, marginalized groups—particularly women, youth and poor households—are most 
vulnerable to unauthorized appropriation. These households often lack the social power to resist 
expropriation by elites or encroachment by other households. In addition to strong household effects, 
the analysis finds evidence of positive subgroup treatment impacts for primary and secondary indicators 
in villages that received the Land Tenure intervention. These are found across tenure security, land 
governance, agroforestry uptake and investment outcomes. The regression results, in combination with 
a secondary analysis that shows an absence of systematic “within-treatment” differences between 
subgroups and the average household results, point to important equity benefits to the TGCC program 
and an absence of elite capture. 
Nevertheless, the Land Tenure intervention had no effect on agroforestry adoption. Households that 
received both the Land Tenure and Agroforestry interventions were 15 percentage points more likely 
to use agroforestry on their plots—an effect almost identical to that on households that received only
the Agroforestry intervention. 
These results suggest that either tenure insecurity does not actually deter households from making 
agroforestry investments, or only after households have felt secure for some time do they begin 
adopting technology and making investments.
The presence of significant tenure security results, despite no changes in the governance indicators 
examined for this study, suggest that changes for these governance indicators do not necessarily 
represent a pre-condition for improved perceptions of tenure security. This has important implications 
for the Theories of Change that inform development programming related to customary land protection. 
Tenure Security
The TGCC Land Tenure intervention had a positive impact on perceived tenure security for all primary 
perception indicators, although there is no evidence that TGCC had an impact on the actual prevalence 
of land disputes.3 In addition, we find positive subgroup treatment impacts for female-headed and elder￾headed households receiving the Land Tenure intervention. 
The planting of agroforestry trees is not linked to greater perceived secure tenure for the full household 
sample or for the subgroup results for primary indicators. However, we do find some evidence of 
marginal improvements for the joint Agroforestry + Land Tenure treatment group among secondary
indicators for female-headed households. This means that women may experience additional tenure 
security benefits from involvement in the agroforestry extension program, compared to only 
participating in the Land Tenure intervention. However, this is the only evidence that we find of a 
positive marginal effect for tenure security on agroforestry investment.
Furthermore, the regression analysis indicates a potential dampening effect on perceptions of tenure 
security for elderly and land-constrained involved in the Agroforestry program. Despite the strong Land 
 3 The number of disputes is low at both baseline and follow-up, which may explain why the tenure security intervention found no impact on 
prevalence of land disputes 
9
Tenure effects across primary and secondary indicators for elderly respondents, there is no evidence of 
a positive treatment effect in the Agroforestry + Land Tenure group for this subgroup.
Although the analysis is descriptive, we do not find evidence of elite capture or uneven program impacts 
for women and marginalized groups receiving the Land Tenure intervention. Women and other 
vulnerable subgroups experience similar positive treatment impacts to their counterparts within the 
treatment sample. This provides some positive evidence of equity benefits for the program participants 
receiving the Land Tenure intervention. 
Land Governance
There is no evidence of a treatment impact on primary land governance outcomes, which measure 
households’ perceptions of the transparency of land allocation and accountability of leaders, as well as 
satisfaction with the resolution of land disputes. 
For secondary outcomes, on average, we see increased participation in land related meetings for
households in villages receiving the Land Tenure treatment and greater trustworthiness of leaders. The
participation result most likely reflects program output versus outcomes, but should be tested through a 
longer-term study. The subgroup results are somewhat mixed for the Land Tenure intervention and 
there does not appear to be a clear narrative to explain the findings. We find consistently positive 
results for secondary indicators for elderly- and youth-headed households but several negative findings 
spread across female-headed, poor- and land-constrained households. 
There are also unexpected results for this outcome family for some subgroups that received the 
Agroforestry treatment. Households in villages that received the agroforestry extension perceive less 
fairness in leaders’ decision-making regarding land management issues and less equity in land distribution 
for vulnerable groups. 
For the within-treatment analysis, we find some evidence of positive subgroup benefits for the Land 
Tenure intervention for primary indicators compared to the average treatment effect. These results—in 
combination with the Tenure Security Outcome Family findings—lend further support to an argument 
that the program was not subject to elite capture and provided important benefits to vulnerable groups. 
Agroforestry Uptake
The agroforestry uptake findings speak to the fundamental research question driving the study—
whether stronger property rights affect a farmer’s decision to practice climate-smart agriculture, 
including agroforestry. The results show a positive aggregate TGCC Agroforestry treatment impact for 
our two primary agroforestry uptake measures. Correspondingly, all subgroups—female-headed, youth, 
elders, poor, and land constrained—show significant uptake in agroforestry and increased agroforestry 
planting across fields for the Agroforestry intervention.
However, there is no evidence of marginal improvements in agroforestry uptake for households that 
also received the land registration and governance intervention.4 Thus, for the overall household sample, 
there is no evidence to support a link between perceived tenure security and agroforestry uptake. In 
contrast, the subgroup findings for female-headed households, poor and elderly respondents indicate 
 4 This finding holds across the village-wide analysis of the program’s intent to treat (ITT) and for the treatment on the treated, where we focus 
on the results for only those households that actively participated in COMACO training (Treatment on the Treated (TOT)). TOT results 
are presented in the Annex 3, whereas the body of the report presents the ITT findings. In fact, unexpectedly, the ITT results are greater in 
magnitude and present for both Land Tenure and Agroforestry + Land Tenure, in contrast to the TOT results. This could be due to 
inaccurate reporting by respondents about their involvement in COMACO. 
10
marginal benefits to linking land tenure and agroforestry. This lends some limited support to the 
argument that, at least for more marginalized groups, stronger property rights affect a farmer’s decision 
to practice CSA including agroforestry. For the long term, it will be valuable to track whether there is a 
relationship between tree survival and perceived tenure security.
Finally, there is no strong or systematic evidence of within-treatment differences between subgroups 
and the aggregate treatment estimates. Similar to other outcome families, this bodes well for equity in 
the distribution of program benefits and absence of elite capture.
In comparison to a 50–75 % survival rate found in other studies (Fink et al., 2014), the seedling and tree 
survival rates for the program are low. Across all years of the intervention, over a third of households 
who engaged in agroforestry report that less than 25% of their Musangu and Gliricidia seedlings have 
survived. For Musangu, the rate of survival is even lower. Only 22% (N=23) of treatment households 
have a survival rate above 75% for trees planted in 2014, and the percentage is even lower for trees 
planted in 2015 (17%, N=14). These low survival rates do not appear to correlate with the geographical 
location of the villages, or with the location of wells or water points. However, it is important to note 
that the entire Chipata District was impacted by a drought in both the 2014–2015 and 2015–2016
agricultural seasons, which impacted seedling survival. Moreover, the seedling survival declines over time 
and is the lowest in 2016. This suggests that the support provided by Community Markets for 
Conservation (COMACO) during the agroforestry program was critical to the survival of seedlings, and 
withdrawing this support has had negative impacts on seedling survival rates.
The difficulty and demands of agroforestry—combined with perceptions of relatively small and distal 
benefits—suggest a need to reconsider agroforestry as a key CSA development intervention in this 
context. Other interventions, such as fruit trees or woodlots, or other CSA methods like minimal 
tillage, may prove more worthwhile. 
Field Investment
We do not find significant treatment effects for field investments measured at the household level.5 Our 
data suggest fallowing may increase tenure insecurity (or at least its perception). This is driven by a fear 
that unproductive land will be reallocated by the headperson, given increasing land scarcity and a 
growing population. For example, households that have recently left land fallow are 10 percentage points 
more likely to report a fear of expropriation. Fear of expropriation is especially high for the largest plots 
(2 to 3 hectares) as compared to smaller plots (less than 1 hectare), by a difference of 10 percentage 
points. The plots that may most benefit from agroforestry or fallowing are also those perceived to be 
the least secure.
The results also show a mix of positive and negative heterogeneous treatment impacts on subgroups. 
Positive results for investment indexes are found among youth and land constrained in the Land Tenure 
group. However, we also find that the poor are significantly less likely to fallow in the Land Tenure 
group, whereas there is weak evidence that youth are less likely to have fallowed in the Agroforestry + 
Land Tenure treatment. There are no differential impacts for female or elder-headed households.
In line with the other outcome families, the findings do not reveal systematic evidence of a difference 
between the average treatment effects and subgroup results for female-headed, elderly, poor or land 
constrained households. The exception to this is the results for youth. We find that compared to the 
 5 Note that subsequent research conducted at the field level reveals positive Land Tenure treatment impacts on field investments (See 
Huntington et al., 2018). 
11
average treatment effect for ‘non-youth,’ youth receiving the treatment are less likely to have fallowed in 
the past 3 years. 
CONSTRAINTS AND LIMITATIONS
Full certification versus boundary demarcation
The distribution of certificates to households at the time of the follow-up survey represents an
important deviation from the original TGCC IE research design. In particular, although all chiefs and 
villages completed the customary land certification process, several chiefs had not signed and distributed 
certificates to households in their villages at the time of data collection. As a result, certificates were not 
distributed to all households before the follow-up survey took place. As of June 2017, based on program 
Monitoring and Evaluation (M&E) data, the breakdown of certificate distribution across chiefdoms 
expected by the evaluation team was as follows: Mkanda (complete); Maguya (complete); Mshawa 
(distribution ongoing but not complete by follow-up); and Mnukwa (distribution did not take place by 
follow-up).
Just over half of households receiving the Land Tenure treatment (57%, N=709) have a paper document 
for at least one of their fields, and these documents are overwhelmingly identified as customary land 
certificates. However, there is some confusion about which households took part in the land 
registration process within villages. We know from M&E data that nearly every household in a treatment 
village took part in the land registration process. The low number of households reporting that they 
have a paper document within villages for their field is likely related to the fact that physical certificates 
were not distributed to every chiefdom at the time of the follow-up survey. 
Unfortunately, the household data regarding certificate distribution does not match the headperson 
data—and there are some discrepancies between the headperson and program M&E data regarding 
certificate distribution.6 Moreover, in the 33 villages where the headperson says 75% to 100% of 
households have received certificates, only 45% (on average) of households say they received a 
certificate.
Furthermore, certificate distribution is not random, which introduces selection bias into the comparison 
of chiefdoms/villages with and without certificate distribution. For example, chiefdoms where certificate 
distribution took place quickly may have customary leaders who are more supportive of the Land 
Tenure program. Villages certified first may have more effective leaders and better governance, fewer 
foreigners7 and be easier to access or less remote than villages that have not yet received paper 
documentation.
As a result, the evaluation relies on the intent-to-treat estimates of household engagement in the 
process of customary land mapping and the expectation of receiving a paper certificate. This is the 
primary analysis for this study and preserves the rigor of the RCT design. We use qualitative focus 
group data to highlight the importance of respondents’ expectation of receiving documentation for their 
responses to survey indicators. The household sample is used to conduct an exploratory analysis of 
estimates for households that received actual paper copies of customary land certificates for their fields, 
 6 Breaking it down by chiefdom for the headperson data, 18 of the villages who received certificates by the launch of the follow-up survey are 
in the Maguya chiefdom, with seven land tenure villages remaining in Maguya. There are 18 land tenure villages in Mkanda with certificates, 
with nine remaining, zero in Mnukwa, with 25 remaining, and finally two in Mshawa, with 40 remaining.
7 Some chiefs refused to present customary land certificates to households from Malawi, and required households to show their National 
Registration Card (NRC) prior to receiving their customary land certificate. 
12
versus those that participated in the certification process but did not receive a physical land certificate. 
This regression output and write-up can be found in Annex 2. Given the selection bias and discrepancy 
in M&E, headperson and household data regarding certificate distribution, these results should be 
treated with caution.
Detecting long term effects
There is strong reason to expect that TGCC effects in the long run may differ from those in the short 
run. It may take time for households to trust that the guarantees of land tenure will be honored. 
Households that adopt agroforestry may subsequently abandon it. We recommend a third round of data 
collection in 2–3 years that revisits the same households who took part in the baseline and follow-up
surveys in order to investigate the longer term effects of the TGCC program. A third round of data 
collection will provide additional evidence about the impact of the land registration and governance 
intervention on field investments, agricultural productivity and livelihoods, as well as the impact of
strengthening perceived tenure security on agroforestry seedling survival. This will promote a better 
understanding of the TGCC program’s full policy potential and value for money, and inform other 
stakeholders’ decisions to take the program to scale in Zambia and other African countries with similar 
customary land systems. 
Following the completion of the follow-up data collection, the TGCC program expanded the Land 
Tenure intervention to cover control households. This will reduce the rigor of the evaluation design, 
depending on how the certificates are distributed in control areas. The optimal scenario for the long 
term study is for TGCC to distribute certificates across all control villages. This would create a lagged 
design where the control areas still serve as a valid comparison, however, inferences about the 
treatment impact would be based on the maturity of the treatment. We would still expect to see better 
outcomes in the treatment areas compared to the control, although the treatment effect would most 
likely be dampened. 
More problematic would be if the Land Tenure intervention was implemented in only one or two 
chiefdoms instead of uniformly across all of them. This would result in selection bias and the inability to 
distinguish whether changes in villages with certificates are driven by receiving land certificates or by 
something inherently different about the chiefdom. For example, chiefs who opted to expand the 
treatment in their chiefdom may be more invested in protecting their land from investors, or they may 
have better governance than chiefs who did not expand the program. In the latter case, the study will 
need to compensate with additional statistical corrections. 
PROGRAMMING & POLICY RECOMMENDATIONS
Overall, the impact evaluation findings provide a basis for the following policy recommendations: 
Scaling-up the Land Tenure intervention
As the Government of Zambia prepares to revise the Land Act with anticipated technical support from 
USAID, the results from the evaluation support the scale-up of TGCC in Zambia and program piloting in 
other customary land systems in Africa. Supplemental research analysis focused on the intervention’s 
positive effect on field investment indicators at the field level, which provides evidence that a policy of 
customary land formalization may support a viable pathway for smallholder-led agricultural 
transformation. A land policy to support smallholders through customary land certification may address 
Zambia’s central development challenge of achieving greater agricultural productivity and balanced, 
inclusive economic growth (Huntington et al., 2018). 
13
Reconsidering the benefits of Agroforestry extension
Findings about the impact of the Agroforestry extension intervention are important, as donors and the 
government continue to promote climate-smart agricultural practices while seeing relatively little 
increase in uptake. Although the TGCC agroforestry extension program did promote an increase in 
agroforestry adoption, seedling survival rates are on average less than 40% and declined over time. The 
qualitative data highlights a number of challenges to keeping seedlings alive, including water shortages, 
the need for significant labor investments during the harvest season and constant threats from fire, pests 
and livestock. 
Given these contextual and environmental challenges to keeping seedlings alive, agroforestry may not be 
the best climate-smart agricultural investment to offer. Households may be more responsive to a 
program that promotes reforestation or growing community woodlots, as many respondents mentioned 
the need for trees for fuelwood and climate resilience. Other CSA possibilities with greater benefits to 
farmers include minimal tillage and crop rotation. Minimal tillage requires greater labor than traditional 
tillage, but can be done in the farming offseason where labor is less constrained. Crop rotation requires 
no additional labor, but does require a change in farming inputs that may need to be provided. 
Adjust agroforestry extension to address low seedling survival 
If programs choose to continue with agroforestry, several issues should be addressed. First, seedling 
survival will continue to be low if traditional village rules regarding grazing and setting fires are not 
addressed. Future programs should incorporate improved land management rules in communities with 
the aim of increasing seeding survival rates. One potential rule in the Zambian context is to prohibit 
livestock from grazing on fields at any point in the harvest cycle. Other areas for improvement include 
addressing inequalities regarding seedling distribution and who works in the nursery. 
Another possible option for future agroforestry programs is to incorporate incentives to motivate 
better survival rates. In another RCT of an agroforestry program in Eastern Province, Jack et al. (2016) 
found that paying cash incentives to farmers who kept at least 70% of their trees alive after one year 
increased the seedling survival rate in the short term, as well as increasing program participation and the 
number of trees planted. The same study also found that weekly monitoring visits from program officers 
improved seedling survival, which, though costly, could be incorporated into future programs. 
In Zambia, given a long history of radio broadcasts and agroforestry activities, improving farmer’s 
knowledge of agroforestry might not need to remain an intervention focus. The ecological benefits of 
agroforestry are widely known, even at baseline and across control groups, and the evaluation findings 
do not support the claim that lack of knowledge is a barrier to uptake. To help inform the design of 
future programs, baseline data can be used to assess farmers’ knowledge of agroforestry or other CSA 
practices in areas with a long history of development interventions.
Increase participation in programs by increasing the number of sensitization visits 
Some households wanted to participate in the Agroforestry intervention, but were reportedly refused. 
Other households signed up expecting a different type of intervention, or were not clear about who in 
their household was allowed to attend trainings.
To allow for greater participation and reduce confusion, future programs could increase the number of 
sensitization visits they make to each village and permit households to join the program until the end of 
the final sensitization visit. Additional sensitization will spread information about the program and give 
households extra time to decide if they would like to participate. The meeting should clearly lay out the 
14
expected time commitments, number of household members who are permitted to join and/or attend 
trainings and the types of inputs the households would receive, all things which generated confusion 
during the Agroforestry intervention and may explain some of the negative effects discussed in the 
sections below.
Increase village land committee (VLC) training on systems for updating customary land certificates
The follow-up survey shows that many VLC members did not understand the system that the Chipata 
District Land Alliance (CDLA) put in place for updating customary land certificates or applying for new 
ones. This is worrisome for the sustainability of the land certification program. Future iterations of the 
program would do well to increase training on these steps, and perhaps place greater emphasis on 
recruiting VLC members who are literate. This of course, has its own tradeoffs that would need to be 
weighed carefully before choosing a path forward. In-line with recommendations from TGCC, it may be 
more effective and sustainable to create VLCs that operate at the sub-chiefdom level as opposed to the 
village level. This would decrease the number of people who would need to be trained in these systems. 
Increased and continued sensitization about land certification process 
Several years into the program, there is still some skepticism about the true purpose of the land 
registration and governance program. Households are leery the program could lead to taxation, or even 
land grabbing, and still more households, particularly women, were confused about the certification 
process. Additional sensitization by CDLA and the VLCs will help alleviate these concerns and increase 
household buy-in for registration. 
Focus group discussions reveal that households do not know how to use their certificates to resolve a 
conflict in the future, or how to ask for the certificates to be updated. An additional village-wide meeting 
conducted by CDLA after the distribution of certificates would help to explain these processes and 
increase the sustainability of the certificates.
Involve the entire village in the boundary demarcation process
Qualitative findings showed a clear link between high levels of household participation in the boundary 
demarcation process and fewer conflicts over boundaries. We recommend that every effort be made to 
include the entire village in the demarcation process, including women. This is in-line with the best 
practices during boundary demarcation recommended by other organizations that specialize in 
interventions to protect customary lands and natural resources. 
Improve dispute data collection and organization
For future parcel mapping and dispute objections and correction efforts, TGCC should improve its 
dispute data collection and organization methods. This will allow the program to better track where, 
when and how disputes were identified and resolved over time. As part of this process it will be 
important to have dispute attributes tied to each parcel such as the date the dispute was identified, the 
type of dispute, if (and when) the dispute was resolved, as well as what method was used to resolve the 
dispute. Assuming this data is tracked by parcel, this would allow for analysis of disputes by village or 
help identify patterns in the types of parcels or types of disputes that occur. Additionally, it will be 
important to ensure that this data collection effort is streamlined to minimize any additional workload 
for the TGCC program. Overall, this level of dispute analysis will identify lessons learned as well as help 
pinpoint how future programming efforts in Zambia or elsewhere could be modified to more efficiently 
identify and resolve land parcel disputes.
15
2.0 BACKGROUND AND 
CONTEXT
DEVELOPMENT CHALLENGE
Land tenure security and property rights governance issues have long been a central focus for a range of 
rural development initiatives in sub-Saharan Africa that address poverty reduction, agricultural 
livelihoods, natural resource management and gender disparities, among other issues. To motivate the 
evaluation, this section provides brief background on land tenure and administration in the Zambian 
context. It discusses agroforestry as a risk-smoothing activity and form of climate-smart agriculture. It 
considers several known barriers to agroforestry adoption in sub-Saharan Africa—including insecure 
resource tenure—and examines research on programs and policies in tenure security and agricultural 
investment.8
LAND TENURE AND ADMINISTRATION
Smallholder farmers, particularly in Zambia’s Eastern Province, grow subsistence crops of maize, as well 
as cash crops of cotton and tobacco, on customary lands controlled by local chiefs. USAID investments 
have long focused on improving agricultural production and increasing access to markets. While there 
has been a great deal of USAID and other donor research on constraints facing smallholder farmers, we 
do not fully understand the influence of resource tenure and the effects of tenure security on 
smallholder investment in long-term land productivity in the country. A number of Zambian legal and 
customary practices related to resource rights could be acting as disincentives to smallholder 
investment.
The 1995 Land Act of Zambia vests all land in the Zambian President and recognizes only two types of 
land: customary and state land. State land includes all land occupied by the national government, as well 
as land held by individuals who lease the land from the state, including those lands that previously were 
freehold estates. Customary land, which under the law is administered by chiefs, represents the 
remainder of land in Zambia, estimated at between 66 and 95% of land. 
While it does not specifically define property rights in land, the Zambian Constitution of 1991 does 
recognize individual property rights and protects those rights against deprivation by the government, 
except in cases authorized by law. Customary lands, which are not registered with the government, are 
largely regulated outside the statutory and official realm of Zambian government. Local chiefs have the 
authority to administer customary land within their chiefdoms. The traditional leaders grant use and 
occupancy rights, regulate transfers of land, control use of communal land and hear disputes (Tetra 
Tech, 2014).
 8 For an in-depth discussion of the research motivation and supporting literature, please refer to Persha and Huntington (2016) and Persha et 
al. (2016). 
16
Customary lands in Zambia therefore fall under the complete control of the chiefs. The chiefs exercise 
this authority through their headmen/women (often more than 300 per chiefdom) and are advised by a 
council of indunas consisting of a dozen to a few dozen individuals. At the local level, the headmen who 
have direct authority over the villages within their domain make decisions about local land management 
and allocations. While villages do often have some small amount of community land, these community 
resources are not farmland—they may be forested land, woodlots or community grazing areas. 
Individual smallholders commonly have no documentation of their rights to land, which can result in 
complex land disputes over boundaries or defense of rights in the event of divorce, death of a family 
member or arbitrary reallocation of land by chiefs or headman. Chiefs may make such allocations, for 
instance, to other villagers or to outside investors via the conversion of customary lands to title deeds. 
Both traditional leaders and subjects are increasingly attuned to the use of documentation, such as 
customary land certificates, as a mechanism to increase household security over occupancy rights to 
land and to help resolve conflicts. 
Uncertainties over land allocation processes within villages also contribute to ongoing land conflicts, and 
insufficient access to arable land is a recognized driver of continued impoverishment in rural areas (Jayne
et al., 2009). Prior research points to large variation in farmer landholdings among village households, 
significant numbers of land-constrained households even in villages where unallocated land is present and 
widely varying perceptions around land availability and ease of acquisition of land for farm expansion. 
Although many factors are likely to feed into such variations, key characteristics include relations to 
local headmen; distance to markets, roads and district administrative centers; and whether a household 
is female headed (Jayne et al., 2009).
AGROFORESTRY
Agricultural production supports the livelihood of over 70% of Zambia’s population, including 78% of 
women. Relative to other countries in the region, Zambia, and in particular Eastern Province, has an 
abundance of fertile land, water and a favorable climate for agricultural production. Yet, despite these 
favorable conditions, 80% of rural Zambians live in extreme poverty, and stunting and malnutrition 
impact rural communities at much higher rates than their urban counterparts. Individual land holdings 
are, on average, small and a quarter of the rural population nationwide farms on only one hectare of 
land. The primary crop grown is corn, and for most farmers it is the only crop they grow, which 
increases vulnerability to weather conditions or pests that damage the crop. Yields for crops in Zambia 
are well below global averages, and despite efforts by the Zambian government and NGOs, adoption 
rates for chemical fertilizer, hybrid maize seeds, herbicide and other agricultural investments remain low 
(Sitko et al., 2011).
Agroforestry is widely perceived as a longer-term sustainable land use practice that can help to meet a 
range of rural development objectives related to improved land use and farmer livelihoods throughout 
sub-Saharan Africa. Agroforestry can act as a mechanism for diversification of a farming portfolio. 
Farmers may use trees to complement rather than replace their crop-planted acreage, for example, 
through nitrogen-fixing legumes that provide additional nutrients to the soil. They may use trees to 
improve soil management in order to improve yields or reduce risk. Indeed, decades of existing research 
points to a range of realized or expected farmer benefits from agroforestry investment, including 
increased crop productivity and reduced variability in yields through such outcomes as increased soil 
fertility; improved livelihoods from higher and more reliable farm income; risk smoothing through crop 
17
diversification; and additional direct benefits provided by trees on farms, such as fuel, wood or fodder 
(Franzel et al., 2001; Mbow et al., 2014; Mercer, 2004).
In more recent years, heightened awareness over projected negative effects of climate change across the 
region has promoted enhanced interest and effort for expanding agroforestry efforts in sub-Saharan 
Africa. Advocates of this approach cite it as a type of longer-term sustainable land use that can not only 
improve farmers’ livelihoods but also enable more effective adaptation and mitigation responses to 
climate-change impacts in already food-insecure regions of the continent (Mbow et al., 2014).
However, despite decades of advancements in agroforestry research, low adoption rates across the 
tropics continue to serve as a substantial barrier to wider realization of agroforestry benefits, as well as 
to the theorized improvement in rural development outcomes (Franzel et al., 2001; Mercer, 2004).
Indeed, uptake of CSA practices, in particular agroforestry, remains very limited in Zambia,
notwithstanding favorable Zambian agricultural policy encouraging CSA and agroforestry promotion 
among a number of organizations. Statistics analyzed from the 2012 Rural Agricultural Livelihoods 
Survey from Chipata District show agroforestry species were planted on 6% of fields (N=84) and in 8%
of households (n=31) surveyed.9
Several barriers to widespread agroforestry adoption persist. Prior research has tended to emphasize 
identifying the biophysical properties of agroforestry systems rather than examining cultural, 
demographic and socioeconomic factors that might impede wider adoption (Ajayi, 2007; Sirrine et al., 
2010). Barriers to adoption include the financial outlay required, explicit and implicit investments in 
labor and the extent to which farmers have the necessary technical knowledge and skills to establish 
trees on farms and effectively engage in agroforestry. Wealth-based and gendered aspects of 
agroforestry uptake are also reported in a number of studies across the sub-Saharan region (Kiptot et 
al., 2014; Phiri et al., 2004). For example, additional disadvantages emerge around women’s frequently 
greater insecurity over land and tree resources, as well as their access to labor, capital and knowledge 
services (Kiptot et al., 2014).
LINKED LAND TENURE AND AGROFORESTRY INVESTMENT
In their 2003 meta-analysis of barriers to agroforestry adoption across 32 empirical case studies, 
Pattanayak et al. (2003) identified tenure security and extension support as two of the most important 
determinates of increased agroforestry uptake (finding tenure security significant in 72% and extension 
support significant in 90% of cases that included these factors in their analyses) (Pattanayak et al., 2003). 
Compared with annual crops, trees require longer periods to produce mature crops—five to fifteen 
years for Musangu trees and three to five years for the Gliricidia trees that are the focus of the TGCC 
intervention. These longer time frames may influence the decisions of farmers to engage in such planting, 
especially given that decisions to plant trees may be influenced by perceived tenure security and 
expectations about access to and control of the land over longer planning horizons. Uncertainty over 
land security has implications for undertaking investments with future payoffs and may limit incentives 
that rely on a long time horizon. The lack of security may serve as a disincentive for farmers to engage 
in agroforestry, given the required upfront financial, labor and other investments; as well as delays of 
several years to realize expected soil fertility, yield, livelihoods and other benefits (Mbow et al., 2014).
 9 If missing responses are included at the field level, the Rural Agricultural Livelihoods Survey data shows agroforestry planted on only 5% of 
fields. 
18
Secure land tenure alone is widely hypothesized to be a necessary condition for individuals to undertake 
productivity-enhancing investments on their land. Numerous studies have suggested positive impacts of 
greater land tenure security on agricultural outcomes and investment in rural land (Deininger et al., 
2011; Deininger and Chamorro, 2004; Feder et al., 1988; Holden et al., 2009; Jacoby et al., 2002; Rozelle 
and Swinnen, 2004). Across a longer timeframe, investment in agroforestry is hypothesized to increase 
famer’s agricultural productivity as the mature trees improve the quality of the soil and reduce the need 
for chemical fertilizer. Consequently, increased agricultural productivity improves household wealth 
through higher yields.
In Zambia and other countries where most rural residents depend on land for their livelihoods, insecure 
tenure and conflicts that can result from uncertainty over ownership or boundaries undermines 
household productivity. Although empirical evidence demonstrating a strong link between strengthened 
land rights and reduced land conflict is relatively scarce, some studies do indicate that land registration 
programs can have the ability to reduce boundary disputes and litigation arising from such conflicts. A 
basic premise of stronger and more secure land tenure is that the enforcement of these rights lessens 
the risk of being forcibly displaced and allows for a level of long-term security and a sense of 
permanence that encourages land-related investment (Besley, 1995). Increased tenure security is also 
thought to reduce the need for smallholders to expend resources to defend their land claims, which can 
be particularly important for women and other vulnerable groups whose rights may not be sufficiently 
protected under traditional practices (Joireman, 2008).
In sub-Saharan Africa, many questions remain around the efficacy of interventions designed to 
strengthen perceptions of tenure security and subsequently promote longer-term sustainable land use 
investments. Some studies have found very strong evidence of positive impacts (Deininger et al., 2011; 
Deininger and Jin, 2006). However, to date no clear consensus has emerged from empirical studies 
across varying sub-Saharan African contexts on whether and how stronger land tenure security 
incentivizes farmer decision-making and pursuit of different land investment strategies on their farms 
(Lawry et al., 2014; Place, 2009). More recent literature has paid particular attention to the role of 
customary land titles (Goldstein et al., 2018; Place and Otsuka, 2001; Smith, 2004). 
The TGCC IE was designed to address these evidence and knowledge gaps and to inform future 
programs and policy formulation by measuring the direct and joint impacts of the TGCC land 
registration and governance and agroforestry extension interventions on three main types of outcomes. 
These categories are: 
• Changes in household perceptions of tenure security over their smallholdings;
• Planned and actual agricultural investment resulting from perceived tenure security, including 
improved adoption of agroforestry and related CSA activities; and 
• Distal outcomes around agricultural productivity and livelihood improvements, which are expected 
to flow from the interventions over longer time horizons and are of broader interest to USAID.
The overarching question that underlies and motivates this evaluation is:
“How do changes in property rights that strengthen a farmer’s perception of long term security 
over farmland affect a farmer’s decision to practice climate-smart agriculture, 
including agroforestry, on their own farms?”
19
The evaluation also advances understanding of the following secondary questions.
1. Do chief—and village—level tenure strengthening activities around participatory village mapping and 
clarified land allocation processes reduce land disputes within villages?10
2. To what extent do improvements in village land management, allocation and adjudication processes 
contribute to more positive perceptions of tenure security over farmland, as well as encourage the 
adoption of longer-term CSA land uses, including agroforestry and soil and water conservation?
3. Are land tenure strengthening activities alone sufficient to change farmer behavior towards greater 
agroforestry uptake, or is it necessary to couple land tenure strengthening with agroforestry 
extension in order to see significant change in agroforestry uptake rates? 
4. Given existing relevant information, and technical and institutional barriers to agroforestry, how does 
improved farmer access to agroforestry knowledge, inputs and related extension resources alone 
(absent any land tenure intervention) affect farmer decisions to engage in agroforestry?
5. How does improved farmer access to agroforestry extension resources additionally impact a farmer’s 
decision to engage in agro-forestry?
 10 Note that there is a discrepancy from the original design report and pre-analysis plan in this question. As ‘sustainable land use planning’ was 
planned but not actually included in the TGCC intervention plan. We have removed that reference from this secondary research question. 
20
3.0 TGCC OVERVIEW
This section describes the TGCC interventions under evaluation. We use M&E data collected by the 
evaluation team and implementing partner, as well as key informant interview data, to detail the planned 
versus actual activities for each intervention and discuss obstacles to implementation. Note that the 
M&E data reported by the implementing partner extends through February 2018, in contrast to 
evaluation data sources collected July–August 2017. The key informants include Village Land Committee 
(VLC) members for villages in the Land Tenure intervention and yield group leaders or lead farmers for 
the Agroforestry intervention. 
The evaluation focused on two 
interventions:
• Land Tenure—A village-level land 
registration and governance
intervention consisting of 
participatory mapping, village 
headperson land administration 
support and provision of land 
information and dispute resolution 
training, including the facilitation of 
informal customary land use 
certificates for households. This 
intervention was implemented by
Chipata District Land Alliance 
(CDLA), a community based 
organization.11
• Agroforestry—Agroforestry 
extension in villages to facilitate tree 
planting adoption and survivorship 
on smallholder farms, implemented 
by the NGO COMACO.
Both interventions took place in 
Chipata District in Eastern Province 
from late summer 2014 through the 
end of 2016.
12 The villages participating 
in this evaluation and their treatment 
status are shown in Figure 3-1.
 11 District Land Alliances (DLAs), such as the CDLA, are community-based organizations, founded under the broader umbrella consortium of 
the national Zambia Land Alliance. They promote greater security of land access and ownership via advocacy activities and community 
outreach in their respective districts of operation. 
12 The Land Tenure program finished by 2016, but the distribution of certificates continued through early 2018. 
FIGURE 3-1: TGCC TREATMENT STATUS BY 
VILLAGE
21
TENURE SECURITY STRENGTHENING INTERVENTION
The land-tenure intervention was implemented by the CDLA in four chiefdoms: Mnukwa, Mkanda, 
Mshawa and Maguya. The intervention aimed to strengthen customary tenure through a set of activities 
that took place at the chiefdom, village and household levels. The IE focuses on identifying the effects of 
the village- and household-level interventions. A description of the land tenure intervention at each level 
is provided below: 
CHIEFDOM LEVEL: 
The chiefdom-level activities aimed to increase transparency of land allocation, administration and 
decision processes and to strengthen smallholder rights to land and trees by: 
• Facilitating a dialogue on land use management and improved tenure governance with chiefs and their 
indunas (advisory councils);
• Identifying and developing opportunities to make decisions on land allocation and land disputes more 
transparent; 
• Chiefdom-level mapping of customary resources within the Chiefdom, particularly communal 
resources;
• Supporting the Chief through facilitating the delivery of household/family level customary land 
certificates, following a boundary clarification process; 
• Documenting customary rules around land administration and management; and
• Providing basic training in administrative support, such as record keeping and map reading, where 
relevant.
VILLAGE AND HOUSEHOLD LEVEL:
The primary activities under the land tenure intervention consisted of establishing Village Land 
Committees (VLCs), conducting participatory mapping and facilitating the issuance of informal customary 
land certificates. In particular, the Land Tenure intervention included: 
• Holding community workshops to establish VLCs;
• Providing training to VLCs about land management, conflict resolution, customary land certificates 
and the customary land registration process;
• Conducting participatory mapping through the development of a common village map that can be 
used as a tool by the headperson when allocating land;
• Providing households with information on land law and rights (CDLA and VLCs), as well as 
information about customary land certificates; and,
• Facilitating the process for households to obtain informal customary land use certificates (CDLA and 
VLCs). The certification process included: 
– Demarcation of land through boundary walks;
– Adjudication and support for land dispute resolution; 
– Facilitating an objections and corrections process for field and village boundary review; 
– Distribution of land certificates that confirm the right to use the land, but not to sell it; and,
– Ongoing support to post-certification activities in Chipata. 
VLCs are responsible for a variety of land related activities, detailed in Table 3-1 below. The most 
common activities conducted by VLCs during the intervention include the resolution of household and 
village level conflicts and clarifying household and village level boundaries. The resolution of conflicts 
22
between households in the village was ranked as a primary VLC activity by key informants in the Land 
Tenure villages. M&E data from CDLA shows that VLCs report resolving 574 conflicts (or an average of 
5 per Land Tenure village) between September 2015 and February 2018, roughly 30% of all conflicts that 
VLCs report occurring.
Several governance activities—originally planned to be conducted through the VLCs—were less 
common or were eliminated from intervention activities. Less common activities included holding village 
meetings and informing the community about their land rights; key informants reported that these occur 
in slightly more than a third of treatment villages.13 Only 20% of villages opened the VLC meetings up to 
the entire community. Moreover, creating and enforcing village rules about land use were not currently 
widespread and are not planned for expansion in the future. Only eight VLC KIIs (7%) report allocating 
land as an activity they have ever or are currently doing, though M&E data from CDLA reports that 287 
new parcels of land have been allocated between September 2015 and February 2018. Land use planning 
activities, documenting land use rules and regulations and the use of a paralegal to support adjudication 
and land dispute resolution were not included in the intervention.
TABLE 3-1: OCCURRENCE OF GOVERNANCE ACTIVITIES BY VLCS
Ever did 
Activity 
Currently doing
Activity
Will do Activity 
in Future
LT + (Ag+LT) LT + (Ag+LT) LT + (Ag+LT)
Resolve HH conflicts 71% (81) 58% (54) 38% (42)
Clarify HH boundaries 47% (54) 31% (34) 28% (31)
Resolve village conflicts 46% (52) 36% (40) 28% (31)
Clarify village boundaries 42% (48) 38% (41) 21% (23)
Hold meetings about land management issues 36% (40) 32% (35) 27% (30)
Inform members of the village about land rights 33% (37) 25% (27) 34% (37)
Create rules about land management 16% (18) 8% (9) 12% (13)
Enforce village rules about land management 10% (11) 5% (5) 15% (16)
Allocate land 7% (8) 7% (8) 17% (19)
Award land certificates 7% (8) 8% (9) 14% (15)
Update land certificates 2% (2) 1% (1) 6% (7)
PARTICIPATION 
According to key informants, the Land Tenure intervention was widely supported by chiefs, headpersons
and the village as a whole. Almost 90% (N=73) of key informants responded that their chief “strongly 
supports” the land registration and governance program, and only one key informant said the chief was 
unsupportive. There was no notable difference between chiefdoms. 
Participation by households was high in villages where the program was offered. Those who did not 
participate most frequently did so because they were out of the village or unavailable during the process. 
Only six key informants cited households not wanting a certificate as a reason that households did not 
participate. Focus group discussions indicate that households who did not want a certificate likely did 
 13 M&E data from CDLA report an average of 24 meetings per treatment village between September 2015 and February 2018, with 27,462 
non-unique participants overall.
23
not want them because of a fear of being taxed or belief that the customary system is strong enough to 
protect their land without additional documentation. 
IMPLEMENTATION HETEROGENEITY
There were some important differences in how the CDLA program was implemented across chiefdoms, 
particularly in Maguya. Households in the Maguya chiefdom were not originally offered the option to 
register a parcel as “family land”; instead, all land had to be registered as household land. Households in 
Maguya were also originally only permitted to register one person as a joint landholder. CDLA returned
to these households and gave them the option to register multiple joint landholders and register land as 
family land, however, few households opted to make these changes. Also different in Maguya chiefdom, 
villages boundary walks took place without input from neighboring villages. This was modified in other 
chiefdoms to include these stakeholders. In another deviation, households in Mshawa chiefdom received 
the intervention several months later than the other chiefdoms due to a change in chiefdom leadership. 
Randomization occurred within versus across chiefdoms and these differences do not threaten the 
validity of the overall treatment impacts, however, they do have important implications for the 
interpretation of what constitutes the ‘treatment’ under investigation. 
There are also village level differences in the establishment and composition of the VLCs. According to 
the program design, VLC members were meant to be elected. However, key informant interviews 
indicated that in approximately 60% of villages VLC members were appointed by the village as a whole, 
whereas 20% of villages elected VLC members. Data from focus groups with the implementing partners 
indicates that villages were first asked to elect members, but if no one volunteered, people would be 
appointed. Half of the members of the VLC were required by CDLA to be women. However, the 
recruitment of women proved to be more difficult, as women did not readily volunteer. Instead, village 
leadership would appoint women and inform them that the role was mandatory. These recruiting 
methods led to VLCs that are, on average, 46% women with 86% of villages (N=110) having at least one 
female member.14
SUSTAINABILITY 
The M&E findings suggest possible sustainability problems for the land certificates as CDLA hands over 
the process of awarding and updating these certificates to the VLCs. Most key informants do not report 
the VLCs being involved in the awarding or the updating of customary land certificates, either in the 
past, present or future. While awarding the certificates falls to the chief during the period of the 
intervention, future awards and the updating of certificates are meant to be done by the VLC. Only 15 
key informants believe that awarding land certificates would be an activity in the future, and seven stated 
that VLCs would play a role in updating land certificates in the future. Correspondingly, there appears to 
be a lack of familiarity with the DHIS2 reporting system that CDLA uses for updating existing or issuing 
new land certificates. Only 35% (N=30) of VLC members from the KII sample know how to use the 
DHIS2 reporting system.15 Given the current state of land policy and administration in Zambia, 
maintaining the TGCC’s Land Tenure outputs, outcomes and achievements will require a broader and 
more systematic policy and programming approach. 
 14 CDLA did not focus on youth while recruiting VLC members, but 36% of VLC members are under the age of 35, and 79% (N=100) of KII 
respondents report at least 1 youth member on their VLC.
15 It is possible that other members of the VLC than the member selected for the KII interview know how to the DHIS2 reporting system, as 
M&E data from CDLA shows that an average of 70% of villages have submitted monthly reports using the DHIS2 system.
24
AGROFORESTRY EXTENSION INTERVENTION
The Agroforestry intervention was implemented by COMACO in five chiefdoms: Mnukwa, Mkanda, 
Mshawa, Maguya, and Saili. The fifth chiefdom, Saili, was not a part of the randomized control trial for 
the evaluation.16 Thus, the results for Saili are not analyzed and described in this follow-up report that
focuses on the RCT. 
The Agroforestry intervention included activities at the village level, as well as basic interactions with 
chiefs. Through the intervention, an extension agent provided support related to planting and 
establishment of Musangu (Faidherbia albida) trees and/or Gliricidia (Gliricidia Sepium) on cropland. 
Activities consisted of establishing Farmer Groups in treatment villages, establishing nurseries, 
distributing seedlings, and providing training and agricultural extension support services about 
agroforestry to farmer groups. Specific activities included:
• Conducting awareness meetings with chiefs and headpersons;
• Forming village-level Farmer Groups open to any household in the village;
• Facilitating the use of Lead Farmers to disseminate information; 
• Leading trainings on agroforestry with Farmer Groups;
• Distributing high-quality Gliricidia and Musangu seedlings and supplies for nurseries; 
• Leading additional trainings on nursery management and field establishment, including information on 
best planting times, sites, and intercropping; 
• Assisting with field-planting crops and trees and promote best practices to promote the protection of 
seedlings during establishment; 
• Providing additional resources to remove constraints to investment, such as groundnut seeds17 or 
wells,18 as necessary; and,
• Collecting monitoring data on seed/seedling planting, survival, and threats to tree survival through 
TGCC mobile data collection units.
There were limits to the number of household members who could officially participate in Farmers 
Groups due to limitations in the amount of inputs provided to each Group. For each one lima19 of land a 
household owns, one member was able to join. Although not required, this member was typically the 
household head. Household members who were not official members were welcome to attend meetings 
and trainings, and frequently, a household head would send his wife to attend training. Married women 
were unlikely to join independently from their husbands, though they were allowed to if they wished. 
Female headed households were as likely as male headed households to join, and when women did join, 
they were more likely to participate in meetings and trainings then men.
According to the program design, every participating village should have participated in all four of the 
major program activities listed in Table 3-2 below. However, key informants indicate that—to the 
respondents’ knowledge—not every activity was offered. In treatment communities, KIIs were most 
likely to report that households received agroforestry seedlings (82%, N=129), and least likely to report 
 16 Saili was added to serve as a separate “Agroforestry control” for evaluating the impacts of the agroforestry intervention because no 
chiefdom level interventions took place in Saili. Therefore, unlike the other four chiefdoms, which contain a mix of control villages as well as 
villages that received the land and agroforestry interventions, Saili only contains villages that received the agroforestry treatment.
17 In year two, every village was given access to a groundnut “seed fund” to provide groundnut seeds to households who wished to intercrop 
their trees with groundnuts. This was not part of the original intervention design, but developed organically from community needs.
18 As a result of severe water shortages that threatened seedling survival, 47 communities were provided with a well as part of the 
agroforestry intervention. 
19 A lima of land is the equivalent of one-quarter of a hectare 
25
households worked in the nursery (54%, N=86). Forty-one percent (N=50) of key informants report 
their village participated in all four activities. This could suggest either some deviation in the planned 
agroforestry extension program, or that KII respondents do not accurately recall which activities took 
place. Nurseries, for example, were only established in 77 villages and shared between nearby villages, 
so work in the nursery may not have been shared equally by all participants.
TABLE 3-2: AGROFORESTRY ACTIVITIES
Ag + (Ag+LT)
Attended trainings on agroforestry 79% (96)
Received agroforestry seedlings 85% (104)
Planted agroforestry seeds 79% (96)
Worked in a nursery 57% (69)
Village participated in all four of the activities 41% (50)
There are unexpected differences in the types of trees reportedly planted as part of the program, as 
noted below in Table 3-3. Overall, only 51 KIIs in the 107 treatment communities said that all three of 
the species of interest to TGCC were planted in their village as part of the Agroforestry intervention. 
Nearly all key informants in treatment villages said COMACO distributed Musangu seedlings, which are 
the most well-known agroforestry species in the region. A similarly high number report the planting of 
Gliricidia seedlings. However, key informants are less likely to say households planted pigeon peas as part 
of the Agroforestry intervention. Additional information from the implementing partner revealed that 
there were not enough pigeon peas available to distribute to each village. 
TABLE 3-3: TREE PLANTING ACTIVITIES IN 
AGROFORESTRY VILLAGES
Ag + (Ag+LT) (N=122)
Households planted Musangu trees 93% (113)
Households planted Gliricidia trees 91% (111)
Households planted pigeon peas 32% (39)
Households planted all 3 COMACO species 31% (38)
SUSTAINABILITY
There are several implementation issues that suggest sustainability problems for the agroforestry 
extension intervention. We examine these in significant detail in Section 6.0 (Agroforestry findings
section). First, a number of seedling nurseries are no longer operational, as illustrated in Table 3–4. 
Approximately 42% of villages reported having an operating nursery, and an additional 18% used to have 
a nursery, but it is no longer operational. Second, a large amount of household labor is required in the 
nursery during the growing season. Amongst treatment villages that have had or currently have 
nurseries, 73% of KIIs (N=71) reported that households are required to work in the nurseries to 
receive seedlings. Households work an average of 70 hours per growing season in the nursery. Third, 
according to key informants, most households expected other programming or inputs from COMACO 
beyond agroforestry, including improved crop seeds (70% (85)), guaranteed market access for crops 
(37% (N=45)), higher prices for crops (34% (N=42) and fuel-efficient cook stoves (41%(N=50). Without 
the follow-on benefits, household may not have sufficient incentives to continue the labor and time 
investment necessary to ensure seedling survival. 
26
TABLE 3-4: NURSERIES
Ag + (Ag+LT)
Have a nursery 42% (53)
Do not have a nursery 40% (50)
Had a nursery, but is no longer operational 18% (22)
27
4.0 EVALUATION METHODS
EVALUATION DESIGN
FACTORIAL DESIGN 
The evaluation utilizes a factorial design to test the impacts of the TGCC intervention. To assess the 
individual and joint effect of each of the Land Tenure and Agroforestry interventions, the evaluation 
design is a four-arm village-level RCT. Figure 4-1 illustrates the four treatment arms of the IE, consisting 
of the different interventions or combination of interventions that treatment villages received (pure 
control, agroforestry extension, tenure-security strengthening activities or both). Villages were 
randomized into these four treatments across four chiefdoms: Mnukwa, Mkanda, Mshawa and Maguya. A 
comparison of findings in villages receiving the Agroforestry or Land Tenure condition versus control
villages provides the average program impact on each of these interventions. The comparison of the 
average outcomes in the group receiving both the Land Tenure and Agroforestry treatment provides 
evidence about the additional effect of land tenure certification in promoting agroforestry uptake and 
CSA. To ensure that the four treatment groups are similar enough to provide a valid counterfactual for 
comparison, balance tests were conducted in the Baseline Report and the Pre-Analysis Plan,
20 and do 
not reveal any meaningful differences between the treatment and control groups that would threaten 
the validity of the study. 
In Saili chiefdom, no chiefdom level land tenure activities were completed, nor did any village receive the 
land tenure intervention. Saili is not included in the RCT, as no interventions were randomized across 
this chiefdom. Instead, the villages in the Saili chiefdom only received an agroforestry intervention. Note 
that the results for Saili are not analyzed and described in this first follow-up report, which focuses on 
the RCT.21
OUTCOME FAMILIES, HYPOTHESES, AND INDICATORS
Following from the evaluation questions, program theory and hypotheses, we developed a set of five 
‘outcome families’ with thematic groupings of hypothesized program effects: Tenure Security, Land 
Governance, Uptake of Agroforestry Practices, Investment in Agriculture and the long-term effects of 
Agricultural Productivity and Livelihoods. We test a series of hypotheses that link the expected effect of 
the program to outcomes in each family. To understand how and why the program may have led to 
changes, we developed and pre-specified a set of primary and secondary indicators. Data on these 
indicators measure and track changes at the household level across baseline and follow-up data 
collection. The hypotheses and primary and key secondary household indicators for each outcome 
family are listed below in Table 4-1. Please refer to Annex 6 for a full list of all evaluation indicators, and 
to the Pre-Analysis Plan for additional details about each of the evaluation indicators. 
 20 Please refer to Annex 6 for links to all TGCC Impact Evaluation reports and supplemental documentation.
21 The main objective of studying the impact of the agroforestry program in Saili will be to better understand the impact of the chiefdom-level 
land-tenure intervention by comparing the “Agroforestry” group in the RCT with the “Agroforestry Control” group in Saili. Also, by
comparing Saili with the main RCT sample we can measure spillovers or what economists sometimes call "general equilibrium effects."
28
TABLE 4-1: HYPOTHESES AND PRIMARY HOUSEHOLD INDICATORS
Outcome 
Family Hypothesis Primary Indicators 
Tenure 
Security
H1: Households receiving the TGCC 
intervention have different levels of village￾wide incidence of land conflicts.
H2: Households in villages receiving the 
TGCC intervention perceive different 
levels of tenure security.
TS-1: Overall PCA index of HH perception of 
tenure security
TS-2: Long-term PCA index of HH perception 
of tenure security
TS-3: Short-term PCA index of HH perception 
of tenure security
TS-4: PCA index of HHs perceived 
expropriation risk from internal actors
TS-5: PCA index of HHs perceived 
expropriation risk from external actors
TS-6: Experienced at least one dispute on field
Land 
Governance
H3: Households receiving the TGCC 
intervention perceive different levels of 
transparency regarding the land allocation 
process and accountability of land 
allocation decision makers.
H4: Households in villages receiving the 
TGCC intervention have different levels of 
satisfaction regarding the resolution of land 
disputes.
G-1: Overall land governance indicator (Index)
G-3: Household perception of land leaders 
(Index)
Uptake of 
Agroforestry 
Practices
H5: Households in villages receiving the 
TGCC intervention have greater uptake of 
agroforestry. 
H6: Households in villages receiving the 
TGCC intervention have greater rates of 
agroforestry seedling survival. 
A-1: Household engages in agro-forestry 
A-2: Extent of field(s) planted with 
agroforestry trees or shrubs
A-3: Agroforestry seedling survival rates
Field 
Investment 
H7: Households in villages receiving the 
TGCC intervention have greater uptake of 
climate-smart agriculture (excluding 
agroforestry).
H8: Households in villages receiving the 
TGCC intervention have greater uptake of 
short and long term field investments. 
TS-2: Length of fallowing period
Long term: 
Agricultural 
Productivity 
and Livelihoods
H9: Households in villages receiving the 
TGCC intervention have higher 
agricultural productivity and crop yields. 
H10: Households in villages receiving the 
TGCC intervention have different 
livelihood and welfare outcomes.
AP-1: Household reported experiencing 
improved crop growth 
AP-2: Household reported experiencing 
improved crop yields 
L-1: Asset-based wealth index: Assets (counts), 
Livestock (counts), land area owned, roof 
construction/material
29
SAMPLING DETAILS AND DATA COLLECTION
DATA COLLECTION 
ENDLINE DATA COLLECTION
Follow-up data collection took place between June 2017 and August 2017. The household, headperson, 
and key informant surveys were collected through a cloud-based mobile data collection effort. Data was 
entered directly into Android phones using a mobile data collection platform, SurveyCTO. Rural Net 
Associates (Rural Net), a Lusaka-based data collection firm, conducted the follow-up data collection in 
close collaboration with Evaluation, Research and Communication (ERC). Rural Net also conducted the 
TGCC baseline data collection in 2014. The research team did not experience unexpected 
circumstances in the field during data collection.
Enumerator training took place from June 5 through June 16, 2017, including a pilot exercise in 
Chongwe, a rural community just outside of Lusaka. The ERC Project Manager and an ERC analyst led a 
four-day training of the trainers for seven Supervisors and seven Deputy Supervisors, as well as the Field 
Manager and Deputy Field Manager. Training focused on the five TGCC instruments (household, 
headperson, Yield Group Leader key informant, Village Land Committee key informant and Focus 
Group Discussion), SurveyCTO, electronic data collection, supervisor responsibilities, as well as
information about sampling, tracking and spotchecks. Training contained lectures, role plays and group 
exercises. All team members were fluent in English and Chinyanja, and the majority had at least some 
post-secondary education. Roughly half were female.22
Training of the full enumerator team took place over five days, and was led by the Field Manager and 
Deputy Field Manager with ERC support. Fifty-seven enumerators were trained on best practices for 
interviewing, the ethics of research with human subjects and the household survey instrument. Survey 
instruments were practiced in Chinyanja. The training provided two days for enumerators to practice 
the survey in small groups, share their questions and advice and practice using SurveyCTO, the survey 
platform selected for this project. Feedback from this training allowed ERC to improve the instruments 
and further adapt them to the local context before data collection. 
The four-person qualitative team received an additional day of training by ERC. Qualitative enumerators 
were trained on best practices for qualitative data collection, the ethics of research with human subjects, 
the FGD instrument and objectives, respondent selection and recruiting and qualitative data 
management. A majority (~75%) of the qualitative enumerators were female in an attempt to ensure 
that women’s FGDs would be led by a female enumerator. The qualitative team also received a second 
day of training to serve as auditors. Auditor training consisted of reviewing protocols for audits as well 
as practicing each of the three versions of the audit survey. A pilot was conducted in Chongwe on June 
16, 2017 to give all team members direct experience using SurveyCTO and practicing the household 
survey. The data collected from this pilot also led to several minor improvements in the survey 
instrument.
 22 In total, the survey team consisted of one Field Manager, one Deputy Field Manager, seven Supervisors, seven Deputy Supervisors, 35 
enumerators, four qualitative enumerators/auditors and one mobilizer, who traveled ahead of the team to make survey appointments in 
villages. The team was subdivided into seven teams consisting of one supervisor, one deputy supervisor and five enumerators. Each small 
team was responsible for surveying one village per day (15 household surveys, 1 headman survey, 1 VLC KII and 1 YGL KII). Supervisors 
conducted the headman survey, Deputy Supervisors conducted the VLC and YGL KII surveys, and enumerators conducted the household 
survey. 
30
In line with the requirements for human subjects’ protection, approval was received from the Clark 
University Institutional Review Board in May 2016 and from the ERES Institutional Review Board in 
Lusaka in June 2017. Written informed consent was received from each participant after reading a 
statement about the purpose of the research, the content of the survey, any risks or benefits and the 
time commitment. Participants were assured their participation was voluntary and could be withdrawn 
at any point and their answers would be kept confidential. They were also informed that their responses 
would be shared through public posting and publication in a way that protected their identities. 
Participants who agreed to participate in the research signed or stamped their fingerprint on a paper 
consent form. In addition, consent was recorded in the electronic survey device.
DATA QUALITY
The TGCC follow-up data collection effort utilized the following quality control measures: observation 
of enumerators by supervisors and the Field Manager and Deputy Field Manager; site presence by the 
Field Manager and Deputy Field Manager; daily quality control checks by ERC staff; and auditing of 
respondents. Each enumerator was observed, or “spotchecked,” by their Supervisor and the Field 
Manager a minimum of two times each week, and the Supervisor or Field Manager was present for the 
entire interview. This observation exercise had an accompanying checklist through which the Supervisor
or Field Manager scored the enumerator on a scale from 1–5 on their surveying technique, including the 
informed consent process, probing ability and relationship with the respondent. The checklist was 
designed by ERC, and the scores could be used for positive incentives (bonuses) or for reprimanding 
(verbal warning, or in extreme cases, dismissal), as the firm saw fit. Feedback from the Field Manager 
and Supervisors was continuously used to improve enumerator performance and discourage data 
falsification. 
In addition to the spotchecks, 15% of surveys were audited by the audit team. These audits were 
stratified by enumerator, and each enumerator had surveys audited every week. The audit data was 
compared to the original data by the ERC team and the number of discrepancies were recorded. If a 
large number of discrepancies were found, additional training was offered to the enumerator. If 
necessary, additional targeted audits were used to investigate unusual patterns that could indicate data 
falsification. 
Finally, the most thorough checks were back-checks conducted by the ERC evaluation team. These 
checks were conducted on 100% of all household, spotchecks, and leader surveys using SurveyCTO, and 
results were compiled and shared with the survey firm. The back-checks compared survey responses by 
each enumerator to search for patterns indicating data falsification or systematic errors that should be 
corrected, including short survey times, missing responses, a low average number of “other, specify” 
responses or multiple selections, low average number of rows completed on each roster and any other 
significant irregularities by day, village, survey team or enumerator. 
31
SAMPLE CHARACTERISTICS
At follow-up, household and headman surveys were conducted as panel surveys, and as many baseline 
respondents as possible were re-interviewed at follow-up.
23 If the original respondent could not be 
interviewed, they were replaced with another adult member of the household. If the entire household 
could not be tracked, another household in the village was selected.24
ATTRITION 
Due to the time between baseline data collection and follow-up data collection (three years), the survey 
team encountered some difficulty tracking all baseline respondents. Whenever possible, mobilizers and 
supervisors were instructed to first try to replace missing respondents from someone within the 
household, and then to replace the respondent using the randomized protocol. In some villages, the 
small village population made it impossible to replace attrited households. Before a household was 
considered untrackable, the survey team conducted three visits. 
In total, 514 households were added to the follow-up data collection. These households were selected 
randomly to either replace households who were interviewed at baseline but could not be tracked, or in 
cases where fewer than fifteen households were interviewed from that village at baseline, to supplement 
the sample size. These respondents were not included in the panel analysis, but were included in 
descriptive statistics for the follow-up survey. 
Altogether, 3,522 households were interviewed at baseline, and 3,403 households were interviewed for 
the follow-up survey. The panel dataset consists of 2,937 households that were surveyed at baseline and 
then again for the follow-up. The study had an attrition rate of 15% for the sample that includes Saili and 
13% for the randomized sample. Table A7-1 in Annex 7 shows that attrition appears to be random, and 
there is no selection bias due to attrition. The primary causes of attrition are households leaving the 
study area. Because our indicators of interest are closely tied to geography, the survey firm did not 
follow households who moved outside of the study area, though they did attempt to follow households 
who moved to another community that was a part of the sample. 
HOUSEHOLD SURVEY
TGCC treatment impacts discussed in this report are based on the panel dataset of 2,937 households. 
Table 4-2 shows the breakdown of panel household observations by chiefdom and treatment group.
TABLE 4-2: HOUSEHOLD PANEL OBSERVATIONS
Mnukwa Mkanda Mshawa Maguya Saili Overall
Agroforestry 132 142 164 147 0 585
Land Tenure 125 111 223 147 0 606
Agroforestry & Land 
Tenure
121 112 259 154 0 646
Control 141 121 208 110 0 580
Agroforestry Control 0 0 0 0 520 520
Total 519 486 854 558 520 2,937
 23 Please refer to Annex 6 TGCC Baseline Report for detailed sampling information on baseline data collection.
24 In Agroforestry, Agroforestry Control, and Land+Agroforestry villages, households who participated in the agroforestry intervention were 
given preference for replacement. In all other villages, replacement households were selected at random using SurveyCTO. 
32
The dataset for the follow-up household survey includes 3,391 respondents and 293 village observations. 
Table 4-3 shows the breakdown of household observations for the follow-up survey by chiefdom and 
treatment group. 
TABLE 4-3: FOLLOW-UP HOUSEHOLD OBSERVATIONS
Mnukwa Mkanda Mshawa Maguya Saili Overall
Agroforestry 178 162 166 168 0 674
Land Tenure 168 131 271 178 0 720
Agroforestry & 
Land Tenure
146 125 271 199 0 741
Control 175 150 220 142 0 674
Agroforestry 
Control
0 0 0 0 582 582
Total 667 568 887 687 582 3,391
For the full household panel sample, the average respondent age is 45 years old (sd=16). Sixty-six
percent of respondents are monogamously married (N=1,595), and 5% are polygamous (N=121). An 
additional 17% are widowed (N=404). Nearly a fifth of the sample has no formal education (19%, 
N=526), but 40% (N=1,120) have reached at least grade seven. Households own an average of two
hectares of land. 
Fifty-six percent (N=1,593) of respondents identify as members of the patrilineal N’goni tribe, and 
another third identify as members of the matrilineal Chewa tribe (35%, N=994), the two major tribes in 
Eastern province. Respondents who identify as Chewa are most common in Mkanda chiefdom. Ninety￾three percent (N=2,610) of respondents are Christian. Sixty percent (N=1,702) were born in the village 
where they currently live, and those who were not born in the village have lived there for an average of 
22 years (sd=15). 
HEADPERSON SURVEY 
TABLE 4-4: HEADMAN OBSERVATIONS
Mnukwa Mkanda Mshawa Maguya Saili Overall
Agroforestry 13 12 19 11 0 55
Land Tenure 12 13 20 11 0 56
Agroforestry & 
Land Tenure
13 14 22 14 0 63
Control 15 14 19 10 0 58
Agroforestry 
Control
0 0 0 0 39 39
Total 53 53 80 46 39 271
Table 4-4 shows headman observations by chiefdom and treatment group. In total, 271 headmen were 
interviewed and 90% (N=243) of respondents are men. Mirroring the household sample, the most 
common tribes are N’goni (67%, N=183) and Chewa (30%, N=82). Chewa respondents are most 
33
common in Mkanda chiefdom (N=49). Women are more likely to serve as the headperson in Mkanda 
chiefdom, where Chewas, a matrilineal tribe, are dominate. 
The headperson is, on average, older than household survey respondents, with an average age of 57 
(sd=14). Eighty-seven percent (N=238) of headperson respondents have attended some formal 
schooling, and 44% (N=121) have reached grade seven or higher, a higher rate than for the average 
household. Eighty-nine percent (N=243) are married, and 11% are polygamous (N=29). Two-thirds of 
headmen were born in the village they currently lead (67%, N=184), and those who were not born in 
the village have lived there for 33 years on average (sd=18). 
Though all the villages in the sample have a relatively small population (approximately 200 people), there 
is a great deal of variance in population size (from 10 to 850). On average, villages contain 41 
households (sd=48); the smallest village has only two households, while the largest has 300. Sixty-eight 
percent (N=186) of villages have mobile phone service within the village, and 53% of villages have a 
borehole (N=146). Seven percent (N=19) have a primary school in the village, two villages have a 
secondary school (<1%), and six villages (2%) have a health clinic. 
KEY INFORMANT INTERVIEWS
Key informant respondents were selected with assistance from the headperson. In villages without 
VLCs, key informants include a village member who was knowledgeable about land issues, often the 
Induna. In villages without agroforestry Yield Group Leaders (YGL), the YGL key informant interview
was conducted with the COMACO lead farmer in agroforestry villages, or a village member who was 
knowledgeable about agriculture in villages without the COMACO program. Table 4-5 shows key 
informant observations by chiefdom and treatment group. 
TABLE 4-5: KEY INFORMANT OBSERVATIONS
Mnukwa Mkanda Mshawa Maguya Saili Overall
YGL VLC YGL VLC YGL VLC YGL VLC YGL VLC YGL VLC
Agroforestry 14 13 20 19 14 15 11 11 0 0 59 58
Land Tenure 12 14 20 21 10 12 14 14 0 0 56 61
Agroforestry & 
Land Tenure
12 14 25 26 14 11 15 15 0 0 66 66
Control 19 16 20 20 17 13 10 9 0 0 66 58
Agroforestry 
Control
0 0 0 0 0 0 0 0 40 38 40 38
Total 57 57 85 86 55 51 50 49 40 38 287 281
FOCUS GROUP DISCUSSIONS
In addition to the four quantitative surveys, the evaluation conducted 62 focus group discussions in 27 
purposefully selected villages across all five treatment arms, shown in Table 4-6. This sampling plan is 
described in detail in the Pre-Analysis Plan (see Annex 6). Respondents were selected to capture a 
variety of experiences within the Agroforestry and Land Tenure treatment arms, including women, 
youth and individuals within treatment villages who did not participate in the interventions (“non￾participants”). Focus group participants were recruited by enumerators with assistance from the 
headperson, taking care to avoid selection bias. When possible, respondents were selected to be 
34
homogenous in age, gender and socioeconomic status, however, given the small sizes of the villages, the 
majority of village residents would often participate in the discussions. 
TABLE 4-6: FGDS PER TREATMENT ARM
Treatment Number of villages Number of FGDs
Agroforestry 7 14 
Land 8 16
Agroforestry + Land 6 20
Control 4 8
Agroforestry Control (Saili) 2 4
TOTAL 27 62
ANALYTICAL APPROACH
The benefit of an RCT is that individuals offered the treatment are, on average, no different from those 
held as controls. Any difference between the two groups must be caused by the treatment, assuming 
correct implementation and a sufficient sample size to ensure baseline balance. However, the individuals 
who actually take-up or adhere to the treatment are not random. Those who expect to benefit most, or 
are most well-informed, or have more trust in institutions may be the ones who take up treatment. 
Comparing that group to the control group would be invalid. Thus, the evaluation examines the effect of 
being offered treatment, which is randomly allocated, rather than the effect of taking up treatment, 
which is not. From a policy perspective, Intent to Treat (ITT) is more relevant than Treatment on the 
Treated (TOT), since when a program is implemented in an area, not all potential participants will take 
part in the program. 
MAIN EFFECTS: INTENT TO TREAT (ITT) 
The IE is designed to rigorously assess the direct and joint impacts of the agroforestry extension 
intervention and tenure security strengthening interventions on the five outcome families described 
above. In this follow-up report, the analysis tests the impact of TGCC on the primary and secondary 
household indicators described in Table 4.1 above. Our estimation compares the following three groups: 
Group 1—Each treatment arm (Agroforestry, Land Tenure, Agroforestry + Land Tenure) to the pure 
control. 
Group 2—The Agroforestry + Land Tenure arm to the Agroforestry intervention. This estimation of 
the joint effect of strengthening land tenure in areas receiving the agroforestry extension represents the 
primary objective of the IE. It enables the research to determine whether or not improving perceived 
tenure security increases farmer investment in sustainable agroforestry and uptake of other CSA
practices.
Group 3—The Agroforestry + Land Tenure arm to the Land Tenure intervention. This estimation of 
joint effects is designed to assess whether tree planting and the promotion of other field level 
investments motivates higher levels of perceived tenure security than the standalone land intervention. 
35
HETEROGENEOUS TREATMENT EFFECTS
Based on the program theory and literature, we expect to find variation in the treatment effect across a 
number of subgroups. Where applicable, outcomes are tested for heterogeneous treatment effects 
across a number of household subgroups. The study may not be well-powered enough to detect every 
treatment effect; however, the results that are significant show that the program had differential effects. 
This represents a secondary versus primary analysis, and is applied across the regression analysis for 
Groups 1–3 described above. 
Our primary household subgroups of interest include: 
• Household head gender (male-headed households versus female-headed households);
• Household baseline wealth status (continuous asset-based wealth index, and lowest quartile vs. 
others);
• Household baseline landholding (land-constrained vs. others); and,
• Age of household head at baseline (youth (under 35 vs. others) and elders (above 55 vs. others).
The main effect of the land tenure treatment is estimated with the following specification for household 
panel outcomes: 
[1a] Yij=β0+β1TijAGRO+β2 TijTENURE+β3 TijTENURE* AGRO +Yij0 +uij
where Yij is the outcome measure of household i in village j. Tij is the treatment dummy for each of the 
three treatment arms of interest. Yij0 is the baseline vectors for the outcome measure and uij are robust 
standard errors clustered at the village level, using Huber-White sandwiched standard errors (Lin et al., 
2013).
To test for heterogeneous treatment effects across the population subgroups, we estimate the following 
equation:
[1b] Yij=β0+β1TijAGRO+β2 TijTENURE+β3 TijTENURE* AGRO +Yij0 +Hetij+ β4TijAGRO*
Hetij +β5TijTENURE* Hetij +β6 Tij(TENURE* AGRO)* Hetij + uij
Please refer to Annex 6 (Pre-Analysis Plan) for more details on the empirical approach.
MULTIPLE TESTING
A ‘false discovery rate’ adjustment was used, to correct p-values25 from each test for the fact that 
several tests were run within each outcome family (Benjamini and Hochberg, 2000). Given the number 
of tests that were run, some portion of the significant results obtained may be simply due to chance. Put 
differently, the more tests that are run, the higher the likelihood that some of them will come back 
significant, but some of these are likely to be false positives. Results that maintain their significance even 
after the p-values were adjusted via the ‘false discovery rate’ correction are considered highly robust.
Given the number of outcomes tested in the evaluation, there may be false positives in our results. 
Thus, we report both uncorrected p-values and corrected p-values using the Benjamini and Hochberg 
(1995) False Discovery Rate Correction. Our main report findings and summary sections rely on the 
uncorrected values, because we are analyzing a number of closely related interdependent outcomes and, 
 25 The p-value is the level of significance within a statistical hypothesis test that represents the probability of the occurrence of a given event. 
36
therefore, the standard corrections for the false discovery rate are likely too conservative (Gelman et 
al., 2012). 26 The corrected values and summary write-up are included in Annex 5. The statistical 
significance of the primary outcome indicators does not change when the p-values are corrected. 
EXPLORATORY, DESCRIPTIVE AND QUALITATIVE ANALYSIS 
Additionally, we support this primary analysis with descriptive statistics based on the panel results, as 
well as full follow-up survey sample results for the section that discusses the agroforestry results. We 
also include a secondary “within-treatment” analysis for subgroups to assess whether there were 
differential effects within the treatment groups to gauge the tendency for elite capture. Finally, we 
incorporate additional qualitative findings from key informant interviews and focus group discussions to 
further ground the analysis. These findings are integrated throughout the body of the report. 
As a supplemental analysis for the Land Tenure intervention, we also examine treatment effects for 
households that received physical land certificates versus households who only completed the process of 
land registration at the time of the follow-up survey. As described in the introduction, these results are 
not experimental; they are exploratory and should be treated with caution. The distribution of land 
certificates was not random, and has the potential to introduce selection bias, since chiefs who were 
quicker to distribute the certificates may be more supportive of the Land Tenure intervention and more 
likely to uphold the rights granted by the certificate. Correspondingly, for the Agroforestry intervention, 
we also estimate the impact of the program on those individuals in the treatment arms who actually 
participated in the agroforestry extension intervention. Since not all household chose to participate, 
there was not comprehensive uptake across villages. Summary write-ups for these results are included in 
Annex 3 and should be treated as investigative and not measures of a causal program impact. 
 26 Gelman and his co-authors note here that for most social science studies, where the effects may be small but are unlikely to be exactly zero, 
the corrections are likely too conservative.
37
5.0 OUTCOME FAMILY I, 
TENURE SECURITY
Hypothesis 1: Households in villages receiving the TGCC intervention perceive different levels of tenure 
security.
Hypothesis 2: Households in villages receiving the TGCC intervention have different levels of village-wide 
incidence of land conflicts.
SUMMARY OF KEY FINDINGS 
• There is strong empirical evidence that the TGCC land registration and governance intervention had 
a positive impact on perceived tenure security for primary and secondary indicators.
• According to the regression results, there is no quantitative evidence that the Land Tenure 
intervention had an impact on the prevalence of land disputes.
• We find positive Land Tenure subgroup treatment impacts for female- and elder-headed households.
• For the overall household panel sample, there is no evidence of marginal improvements to perceived 
land tenure security for the joint Agroforestry + Land Tenure treatment group. However, we do find 
evidence of marginal improvements for the joint Agroforestry + Land Tenure treatment group
among female-headed households. 
• There is some evidence of a negative or dampening treatment effect on perceived land tenure 
security for households receiving the Agroforestry intervention.
• According to the exploratory analysis for the “within-treatment” effects, we do not find evidence of 
elite capture or uneven program impacts for marginalized groups. Women and other vulnerable 
subgroups experience similar positive treatment impacts to their counterparts within the treatment 
area.
Tables 5-1 through 5-3 illustrate the statistically significant regression findings for the household and 
subgroup panel samples. The coefficient on the treatment estimate is included in the table; a (P) 
designates a primary indicator and an (S) designates a secondary indicator. For indexes, we report the 
standard deviation above the control mean and for binary indicators, we report the marginal effect on 
the probability the household reports the chance of encroachment across all of their fields for a 
particular source of dispossession is “impossible.” A discussion of the magnitude of effects is included
below in the Results section. The level of statistical significance is indicated by the color of the table cell 
where: p < 0.1; p < 0.05; p < 0.01.
38
RESULTS
Tables 5-1 through 5-3 below display the overall and subgroup treatment effects for primary and 
secondary indicators. These tables, along with Figures 5-1 and 5-2, show that all primary household 
indicators of perceived tenure security are statistically significant for households receiving the land 
registration and governance intervention. For all Figures in the report, positive (or negative) impacts are 
represented by the vertical confidence bars that do not overlap with an effect size of ‘0’; the magnitude 
of effects is indicated by a greater distance from the ‘0’ effect size line. 
TABLE 5-1: TENURE SECURITY OVERALL SUMMARY TABLE
(BY TREATMENT GROUP)
AG LT AG+LT
Overall PCA index of HH perception of tenure security (P) 0.16 0.11
Long-term PCA index of HH perception of tenure security (P) 0.15 0.10
Short-term PCA index of HH perception of tenure security (P) 0.15 0.10
PCA index of HHs perceived expropriation risk from internal actors (P) 0.15 0.11
PCA index of HHs perceived expropriation risk from external actors (P) 0.15 0.09
Experienced at least one dispute on field (P)
HHs perceived expropriation risk from other HHs (S) 6%
HHs perceived expropriation risk from elites (S)
HHs perceived expropriation risk from neighboring villages (S) 6%
HHs perceived reallocation risk from chief (S)
HHs perceived reallocation risk from headperson (S) 5%
HHs perceived expropriation risk from extended family (S)
TABLE 5-2: TENURE SECURITY SUBGROUPS SUMMARY TABLE
(LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
Overall PCA index of HH perception of tenure security (P) 0.21 0.25
Long-term PCA index of HH perception of tenure security (P) 0.18 0.25
Short-term PCA index of HH perception of tenure security (P) 0.21 0.25
PCA index of HHs perceived internal actor expropriation risk (P) 0.21 0.26
PCA index of HHs perceived external actor expropriation risk (P) 0.16 0.22
Experienced at least one dispute on field (P)
HHs perceived expropriation risk from other HHs (S) 10% 10%
HHs perceived expropriation risk from elites (S)
HHs perceived expropriation risk from neighboring villages (S) 14%
HHs perceived reallocation risk from chief (S) 10%
HHs perceived reallocation risk from headperson (S) 12% 10%
HHs perceived expropriation risk from extended family (S) 11%
39
TABLE 5-3: TENURE SECURITY SUBGROUPS SUMMARY TABLE
(AGROFORESTRY + LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
Overall PCA index of HH perception of tenure security (P) 0.19
Long-term PCA index of HH perception of tenure security (P) 0.19
Short-term PCA index of HH perception of tenure security (P) 0.19
PCA index of HHs perceived internal actor expropriation risk (P) 0.19
PCA index of HHs perceived external actor expropriation risk (P) 0.18
Experienced at least one dispute on field (P)
HHs perceived expropriation risk from other HHs (S)
HHs perceived expropriation risk from elites (S)
HHs perceived expropriation risk from neighboring villages (S)
HHs perceived reallocation risk from chief (S)
HHs perceived reallocation risk from headperson (S)
HHs perceived expropriation risk from extended family (S)
FIGURE 5-2: TREATMENT EFFECT ON TENURE SECURITY—
DISPUTES
FIGURE 5-1: TREATMENT EFFECT ON TENURE SECURITY—PCA 
INDECES
40
The results for this outcome family provide strong evidence that the TGCC Land Tenure intervention 
had a positive impact on perceived tenure security. In contrast, the regression findings do not indicate a 
Land Tenure impact on the prevalence of land disputes. The number of disputes reported by both 
headmen and households did decrease, from an average of 2.42 conflicts per village to 1.54 conflicts per 
village at the time of the follow-up survey. However, this decrease in conflicts occurs across both 
treatment and control groups, and the treatment results are not statistically different from the control 
group. It is unlikely that this decrease across all groups is caused by spillover from villages that received 
the land tenure treatment, as a similar decrease in conflict is found in the Saili chiefdom which did not 
receive the land tenure treatment. 
The regression results indicate a positive and statistically significant treatment impact on perceived 
tenure security for all indexes, with the overall index having the largest effect at 0.16 (p < 0.01) standard 
deviations above the control mean. Treatment households perceive that their fields are more secure 
from reallocation or unauthorized appropriation from both internal (0.15 (p < 0.01)) and external (0.14
(p < 0.01)) threats in the next three years (0.15 (p <0 .01)) and beyond four years (0.15 (p < 0.01)). 
Looking across the binary measures for the six sources of dispossession, we find that the land 
registration and governance intervention raised the probability households think encroachment or 
unauthorized appropriation is “impossible” by 5 to 6 percentage points for headman, neighboring villages 
and other households within their home village. However, we do not find a marginal impact for the 
treatment for chiefs, elites or family members. Moreover, there is no evidence that the intervention had 
any short-term impact on the already low prevalence of reported land disputes in the study area. 
In addition, we find a number of positive treatment impacts for elder- and female-headed households. 
The magnitude of these impacts is greater than those of the overall household sample. Mirroring the 
average household results, female- and elder-headed households in the treatment perceive that their 
fields are more secure from reallocation or unauthorized appropriation overall (0.21 (p < 0.05); 0.25 (p 
< 0.01)) from both internal (0.21 (p < 0.05); 0.26 (p < 0.01)) and external (0.16 (p < 0.05); 0.22 (p < 
0.01)) threats in the next three years (0.20 (p < 0.05); 0.25 (p < 0.01)) and beyond four years (0.18 (p < 
0.01); 0.25 (p < 0.01)), respectively. 
For the sources of dispossession, we again find significant and more substantial treatment effects for 
female and elderly-headed households. The intervention raised the probability that female- and elder￾headed households think encroachment or unauthorized appropriation is “impossible” by 9 to 12 
percentage points for headman (12% (p < 0.05); 9% (p < 0.05)) and other households within their home 
village (10% (p < 0.10), 10% (p < 0.10)). Moreover, for elder-headed households, we also find a positive 
program effect for neighboring villages (14% (p < 0.01)), as well as for sources of dispossession where 
there is no average household effect, including chief (10% (p < 0.05)), and family (11% (p < 0.05)). These 
are the largest treatment impacts related to the Tenure Security Outcome Family that we find in the 
analysis. There is no evidence of differential effects between treatment and control groups for poor, 
youth or land-constrained respondents. 
Thus, although we see strong treatment impacts for the aggregate-, female- and elder-headed 
households across all primary indicators the impact of TGCC is not consistent across sources of 
dispossession and is null for elite appropriation. 
The planting of agroforestry trees is not linked to greater perceived secure tenure for the full household 
sample. There is no evidence of marginal improvements to perceived land tenure security for 
41
households in the Agroforestry + Land Tenure treatment group for the aggregate household sample. 
Furthermore, the regression analysis indicates a potential dampening effect on perceived tenure security 
for elder-headed households involved in the agroforestry extension program. Despite the strong Land 
Tenure effects across primary and secondary indicators for elder-headed household respondents, there 
is no evidence of a positive treatment effect in the Agroforestry + Land Tenure group for this subgroup. 
Finally, our exploratory analysis of a within treatment differential does not reveal any differences 
between the average treatment effect and the results for female-headed and other vulnerable subgroups 
that received the Land Tenure treatment. Put differently, female and vulnerable group effects are not 
statistically different from the average treatment effect. This provides some positive evidence of equity 
benefits for the program participants receiving the Land Tenure intervention. 
DISCUSSION
The regression analysis provides strong evidence that the TGCC Land Tenure intervention had a 
positive impact on perceived tenure security. In this section, we explore the pathways or mechanisms 
linking the TGCC intervention to the positive impacts identified above. This section investigates 
household and key informant descriptive and qualitative data to understand why we see treatment 
effects.
BENEFITS FROM CUSTOMARY LAND CERTIFICATES, NOW AND IN THE FUTURE
As described above in Section 3, the TGCC Land Tenure intervention focused on boundary 
demarcation, land governance and customary land documentation through informal customary land 
registration. Table 5-4 below displays the benefits of customary land certificates (CLC) that are 
expected by respondents. Roughly 80% of treatment households believe that having a customary land
certificate will make it less likely for their land to be taken, both now and in the future. The rate is 
similar between male- and female-headed households. Qualitative findings illuminate why households feel 
that customary land certificates make it less likely that land will be taken. Focus group discussions (FGD)
reveal that households believe that having well-known and clearly defined boundaries through the 
demarcation process in combination with paper “evidence” of their ownership are the primary drivers 
of the increased sense of tenure security. The children are the primary recipients of these benefits and 
will be protected from other family members grabbing land after the death of their father. This series of 
FGD quotes illustrate this sense of security. 
“When [the land] was merely referred to as “land of our parents,” I did [not] have the sense of 
personal ownership of the land. The land belonged to us as a family. But now, what I have is much 
more. It is my land. The certificate has my name and that of my wife and children. When I die, the 
certificate will be issued in the name of my children. I feel that the future of my children is much 
more secure. No one will take the land away from them" (Mkanda).
"My brothers and sisters no longer have a say over my land as they did when we called it the land of 
our parents. My wife and children have much more security over my land, because their names are 
on the certificate. I have secured their future by having a certificate for my land" (Mkanda).
Households also believe that the certificates will stop the headman or chief from reallocating 
unproductive land, which is increasingly in demand as the population grows. A male FGD participant in 
Mnukwa chiefdom described the benefit this way: “…after the field has been mapped the headman 
cannot add another person because the boundary has been measured already and the land is now [the 
42
owners’], unlike the way it was without certificates… If a person was given 20 hectares, then he had to 
cultivate the entire 20 hectares, but if he failed and only managed 3 or 4 it used to be like a punishment 
to people. If the certificates come, then maybe this year [if] things won’t go well such that I won’t 
manage to cultivate the whole field, they cannot add someone else.” 
Focus group discussions also highlight citizens’ hope that certificates can protect households from 
powerful elites by providing them some amount of power, as long as the government and customary 
leaders respect their certificate. "To add on what is agreed by maybe the headman, or the chairman, or 
maybe the chief, there are people who come from town, from town they come to see the chairman or 
the headman, those people come with money how can we find a place to cultivate here? so when those 
people come, they have their own money, so they will use their money ‘If you give us such an amount 
we will give you that area’…They come in through the headman, the chief and the chairman, so they 
come to provoke this one [current land owner] and he has no power, where will he get that?... What 
needs to protect the person is the certificate and then the government should defend this person even 
the headman and the chief should support that person who has the certificate, so this is what is needed”
(Mnukwa).
TABLE 5-4: EXPECTED BENEFITS OF CUSTOMARY LAND CERTIFICATES
LT (N=208) Ag+LT (N=230)
Benefit Now Future Now Future
Land is less likely to be taken from the household 81% (168) 81% (168) 83% (190) 80% (183)
Land is less likely to be taken from the village 40% (83) 33% (69) 37% (84) 37% (86)
Fewer disputes about boundaries 45% (93) 50% (105) 47% (108) 52% (119)
Fewer disputes about inheritance 42% (87) 52% (108) 39% (89) 54% (125)
Fewer disputes about land allocation 25% (51) 27% (57) 17% (39) 28% (64)
Gain access to credit 3% (6) 3% (7) 2% (4) 3% (7)
Gain ability to buy or sell land 2% (4) 2% (4) 1% (3) 1% (2)
More control over agricultural decision making 13% (27) 14% (29) 10% (24) 10% (22)
There are no benefits 9% (18) 1% (2) 9% (20) 3% (6)
The qualitative analysis indicates that a subset of households report that they have already experienced 
benefits from the land governance and registration intervention, citing a decrease in disputes about 
boundaries, inheritance, and land allocation. In contrast to the regression results, the qualitative data 
highlight households’ belief that CLCs have served to reduce disputes about boundaries. Qualitative 
findings also show that the demarcation process itself strengthened perceptions of tenure security by 
making boundaries clear and well-known throughout the village. A FGD participant in Mnukwa 
commented “There were land disputes before Land Alliance came because the field boundaries were 
not clear but now they are clear, there is no one who can say ‘you have taken part of my field.’ No. 
Now you have to be in agreement for someone to use part of your field."
FGDs indicate that conflicts were less likely to occur in villages with widespread participation in the 
demarcation process, as opposed to just a few households, or only men.27 According to a woman in 
Mshawa describing the demarcation process— “they had chosen 10 people, we were the ones who 
were going round the fields... I was the only woman the rest were men, they were many…women 
stayed behind they [men] told us to stay behind and cook relish."
 27Note that female participation was a part of the original design of the boundary walk for the Land Tenure intervention.
43
The descriptive statistics show that households believe that CLCs have reduced disputes about 
inheritance (40%, N=176), and even more believe they will do so in the future (53%, N=233).
A male FGD participant in Mshawa elaborated on the certificate’s benefit for women after the death of 
their husband. “Now I know, once I get my certificate, even my wife and children will be secure after I 
am dead. Because what happens is that the woman is told to go back to her parent’s village and the 
relatives get the land for themselves… If on the certificate you are there registered, you will have the 
right because there is nowhere they can take you, the evidence is the certificate.”
However, qualitative findings are mixed about whether this protection will extend to women without 
children. In most cases, it appears that customary land certificates protect women’s land rights through 
their children. Women are the stewards of the land until their children are grown, then the land passes 
to the adult children, who in turn provide for their mother. If a woman without children is widowed or 
divorced, in most patrilineal tribes she must still return to her village and get land from her parents or 
brothers. As one woman from Mshawa chiefdom described in a FGD, “Us as women cannot be involved 
in land issues, it is the men’s duty to do so. The men are the ones who are supposed to sign for the land 
because they own the land, it is them who brought us here. We left our villages and came here, can we 
say that we have land? We have no land.”
Across all treatment and control areas, the majority of households without documentation for their land 
would like some form of documentation to reduce the likelihood of losing their land (84%, N=830), and 
to strengthen the ability of their children to inherit the land (69%, N=688). Very few households in 
either treatment group believe that CLCs will lead to access to credit, either now or in the future.28 Just 
over a quarter (25%, N=63) of households in the Land Tenure group without land certificates believe 
that CLCs will protect their field investments. Table 5-5, below, shows the most common benefits 
associated with a CLC. 
TABLE 5-5: WHY WOULD HOUSEHOLDS WANT A CUSTOMARY LAND 
CERTIFICATE?
LT Ag+LT Control 
Reduce likelihood of losing my land 77% (202) 71% (217) 75% (411)
Protects investments I have made on my 
land 25% (63) 29% (86) 16% (85)
Strengthens the ability for my children to 
inherit the land 63% (163) 62% (189) 62% (336)
Helps me to obtain access to credit 3% (7) 4% (13) 2% (13)
Of the households who do not have CLC in the Land Tenure group, nearly all of them (87%, N=501) 
would like to have documentation, and would pay an average of 75 ZMW ($7.50) for documentation, as
shown in Table 5-6. Interestingly, households in the treatment area who do not have certificates are 
willing to pay 5% less of what households in control areas are willing to pay. This could represent the 
strong demand in control communities for certification. An alternative explanation is that since 
treatment communities have been told by CDLA they will receive certificates for free, they are willing
to pay less. Another explanation is that treatment communities believe it may be a possibility that they 
 28 CDLA programming specifically stated that customary land certificates could not be used as collateral.
44
will be required to pay to receive certificates that have not been distributed and subsequently choose a
more affordable amount.
Overall, the results for village key informants (KI) regarding benefits from the land certification process 
mirror those found in the household sample. This is illustrated in Table 5-7. Consistent with the 
household results, the main benefits cited by KI in Land Tenure villages is that land is less likely to be 
taken from households, along with decreases in disputes about inheritance boundaries. One exception is 
that key informants highlight that the TGCC Land Tenure intervention also provided greater protection 
for community land. No KI respondents reported increased access to credit or the ability to purchase 
and sell land; the number of KI respondents that expect to see these benefits in the future are similarly 
low. As the CDLA program specifically told participants that certificates could not be used for these 
purposes, it is promising that these limitations are well-understood. 
TABLE 5-7: KII BENEFIT FROM LAND CERTIFICATION PROCESS
Benefit
Present Future
LT + (Ag+LT) LT+ (Ag+LT)
Fewer disputes about inheritance 62% (52) 71% (60)
Land is less likely to be taken from households 59% (62) 78% (66)
Fewer disputes about boundaries 57% (48) 62% (52)
Land is less likely to be taken from the village 52% (44) 60% (51)
Fewer disputes about land allocation 27% (23) 36% (30)
Households gain access to credit 0 (0%) 0 (0%)
Households gain ability to buy or sell land 0 (0%) 5% (4)
Households have more control over agricultural 
decision making 6% (5) 12% (10)
Women have more equal access to land 14% (11) 15% (12)
Youth have more equal access to land 11% (9) 13% (11)
Poor households have more equal access to land 6% (5) 10% (8)
No benefits 7% (6) 2% (2)
NEGATIVE EFFECTS OF CUSTOMARY LAND CERTIFICATES, NOW AND IN THE 
FUTURE 
In general, the number of households and village key informants who expect any negative impacts as a 
result of their CLC is low, shown below in Table 5-8. The biggest negative impact reported by 
households and KI is fear of paying taxes or fees in the future. As a female FGD participant said, "We 
might just hear in the future that you should be paying every month since you already signed on the 
certificates.” Some households also believe that CLCs have made land harder to acquire, and will 
continue to do so in the future. There are a handful of households who believe that CLCs have 
increased disputes about inheritance (4%, N=16). Similarly, only a few key informants believe that CLCs 
TABLE 5-6: DEMAND FOR CUSTOMARY LAND CERTIFICATES 
LT + (Ag+LT) Control 
Percentage of households without 
documentation who would like documentation 87% (501) 89% (490)
Average amount households would pay to 
receive documentation for their fields (in 
kwacha)
76 (204) 80 (230)
45
have increased conflicts, increased corruption or increased the likelihood of land being taken from 
households of the community. 
The qualitative findings reveal that the initial skepticism some felt about the program, primarily that the 
certification process was a way for the government to grab their land, is decreasing. A Mkanda FGD 
respondent stated “That is why you can see that even those who were reluctant to have their fields 
demarcated by CDLA, are now changing their minds and now also requesting for their fields to be 
demarcated. Those fears and misconceptions they earlier had on this program, are now slowly 
dissipating." Still, ongoing and additional sensitization by CDLA about the program would be beneficial. 
There are also a fair number of FGD participants that point out that regardless of the land certificates, 
their tenure security is derived from customary systems in the chiefdom. They feel secure in their land 
tenure because the chief gave them their land, not because of the certificates. As a youth from Mshawa 
commented "Our headman and chief will always protect our interests when it comes to securing our 
fields."
TABLE 5-8: NEGATIVE IMPACTS OF CUSTOMARY LAND CERTIFICATES
Households
LT + (Ag+LT) (N=493) LT + (Ag+LT) (N=493)
Present Future
There are no negative effects 90% (396) 87% (379)
Key informants 
LT + (Ag+LT) LT + (Ag+LT)
Present Future
There are no negative effects 86% (73) 69% (59)
46
6.0 OUTCOME FAMILY II, 
LAND GOVERNANCE
Hypothesis 3: Households in villages receiving the TGCC intervention perceive different levels of transparency 
regarding the land allocation process and accountability of land allocation decision makers.
Hypothesis 4: Households in villages receiving the TGCC intervention have different levels of satisfaction 
regarding the resolution of land disputes.
SUMMARY OF KEY FINDINGS 
• There is no quantitative evidence of aggregate household or subgroup impacts for the primary 
governance outcomes analyzed in the regression analysis. 
• For secondary outcomes for the overall household sample, we see a 4% higher perception of leaders 
as trustworthy in villages receiving the Land Tenure treatment. There is a 5% greater participation in 
land related meetings in villages receiving the Agroforestry treatment. 
• For secondary subgroup outcomes, the findings are mixed. We find some positive results in the Land 
Tenure group. In particular, there are some positive treatment effects for elderly-headed households; 
they are more likely to believe that leaders are trustworthy, accountable, make transparent decisions 
and protect natural resources, in comparison to their control counterparts. Moreover, poor 
households are 10% less likely to report that households have been disadvantaged by land allocation 
decisions. However, in the Agroforestry and Agroforestry + Land Tenure groups, we find negative 
treatment impacts across all subgroups. There is scattered evidence that treatment subgroups 
perceive equity or accountability problems with local leaders and additional disadvantages for 
vulnerable groups related to land management and allocation.
• For the within-treatment analysis, we find some evidence of positive subgroup benefits for primary 
indicators compared to the average treatment effect. These results—in combination with the Tenure 
Security Outcome Family findings—lend further support to an argument that the program was not 
subject to elite capture.
Tables 6-1 through 6-4 illustrate the statistically significant regression findings for the household and 
subgroup panel samples. The coefficient on the treatment estimate is included in the table; a (P) 
designates a primary indicator and an (S) designates a secondary indicator. For indexes, we report the 
standard deviation above the control mean and for binary indicators, we report the marginal effect on 
the probability the household reports that they ‘agree’ or ‘strongly agree’ with a question. A discussion 
of the magnitude of effects is included below in the Results section. The level of statistical significance is 
indicated by the color of the table cell where: p < 0.1; p < 0.05; p < 0.01.
47
RESULTS
Although the Land Tenure intervention was inherently a governance program, there is no evidence of an 
impact on primary governance outcomes for the household sample. Tables 6-1 through 6-4 below display 
the overall and subgroup treatment effects for primary and secondary indicators. These tables, along 
with Figure 6-1, show that the primary household indicators of governance outcomes are not 
statistically significant for households and subgroups receiving the land registration and governance 
intervention. For all Figures in the report, positive (or negative) impacts are represented by the vertical 
confidence bars that do not overlap with an effect size of ‘0’; the magnitude of effects is indicated by a 
greater distance from the ‘0’ effect size line. 
TABLE 6-1: GOVERNANCE OVERALL SUMMARY TABLE 
(BY TREATMENT GROUP)
AG LT AG+LT
Overall PCA index of governance (P)
PCA index of HH perception of land leaders (P)
HH perception of land allocation, 0 = worse, 4 = better (S)
HH believes leaders are trustworthy (S) 4%
HH believes leaders protect natural resources (S)
HH believes decisions are fair (S)
HH believes decisions are transparent (S)
HH believes rules are clear (S)
HH believes leaders are accountable (S)
PCA index of equity in land allocation across vulnerable groups (S) -.16
HH participation in land management-related meetings (S) 0.05
HH feels women disadvantaged in land allocation decisions (S) -7%
HH feels the elderly disadvantaged in land decisions (S) -6%
HH feels the poor disadvantaged in land decisions (S) -5%
HH feels minority tribe members disadvantaged in land decisions (S) -6%
HH feels vulnerable groups disadvantaged in land decisions (S) -9%
TABLE 6-2: GOVERNANCE SUBGROUPS SUMMARY TABLE
(LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
Overall PCA index of governance (P)
PCA index of HH perception of land leaders (P)
HH perception of land allocation, 0 = worse, 4 = better (S)
HH believes leaders are trustworthy (S) 6%
HH believes leaders protect natural resources (S) 5%
HH believes decisions are fair (S)
HH believes decisions are transparent (S) 7%
HH believes rules are clear (S)
HH believes leaders are accountable (S) 7%
PCA index of equity in land allocation across vulnerable groups (S)
HH participation in land management-related meetings (S)
HH feels women disadvantaged in land decisions (S)
HH feels elderly disadvantaged in land decisions (S)
HH feels poor disadvantaged in land decisions (S)
HH feels minority tribe members disadvantaged in land decisions (S)
HH feels vulnerable groups disadvantaged in land decisions (S) 10%
48
TABLE 6-3: GOVERNANCE SUBGROUPS SUMMARY TABLE
(AGROFORESTRY HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
Overall PCA index of governance (P)
PCA index of HH perception of land leaders (P)
HH perception of land allocation, 0 = worse, 4 = better (S)
HH believes leaders are trustworthy (S) -6%
HH believes leaders protect natural resources (S) -6%
HH believes decisions are fair (S) -12% -8%
HH believes decisions are transparent (S)
HH believes rules are clear (S)
HH believes leaders are accountable (S)
PCA index of equity in land allocation across vulnerable groups (S) -0.22
HH participation in land management-related meetings (S) 8% 7%
HH feels women disadvantaged in land decisions (S) -10%
HH feels elderly disadvantaged in land decisions (S) -8%
HH feels poor disadvantaged in land decisions (S) -7% -12%
HH feels minority tribe members disadvantaged in land decisions (S)
HH feels vulnerable groups disadvantaged in land decisions (S) -10%
TABLE 6-4: GOVERNANCE SUBGROUPS SUMMARY TABLE
(AGROFORESTRY + LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
Overall PCA index of governance (P)
PCA index of HH perception of land leaders (P)
HH perception of land allocation, 0 = worse, 4 = better (S) -0.26
HH believes leaders are trustworthy (S) -16%
HH believes leaders protect natural resources (S)
HH believes decisions are fair (S) -13%
HH believes decisions are transparent (S)
HH believes rules are clear (S)
HH believes leaders are accountable (S) -12%
PCA index of equity in land allocation across vulnerable groups (S)
HH participation in land management-related meetings (S)
HH feels women disadvantaged in land decisions (S)
HH feels elderly disadvantaged in land decisions (S)
HH feels poor disadvantaged in land decisions (S) -8% 7%
HH feels minority tribe members disadvantaged in land decisions (S)
HH feels vulnerable groups disadvantaged in land decisions (S)
49
The analysis of secondary indicators presents a more complicated and mixed story. We find some 
positive results for the Land Tenure group, but negative or unexpected results for households in villages 
receiving the Agroforestry treatment. The analysis of secondary indicators shows that households 
receiving the Land Tenure treatment are 4% (p<0.05) more likely to rate their leaders as trustworthy 
above the control mean. We find a slight increase in participation in community land meetings for 
households receiving the Agroforestry treatment (5% (p<0.10)); this includes an 8% (p<0.05) increase 
for female-headed households and a 7% (p<0.05) increase for youth-headed households. 
However, the analysis highlights several negative results for equity indicators among participants 
receiving the Agroforestry condition. Among the secondary indicators, we find negative impacts for 
households in the Agroforestry intervention for an index (-0.16 (p<0.05)) and questions that ask 
respondents to assess whether vulnerable groups (women (7% (p<0.05)), elderly (6% (p<0.05)), poor
(5% (p<0.05)) and minority tribes (6% (p<0.05)) have been disadvantaged in decisions about land 
reallocation. Respondents in the Agroforestry group are 9% (p<0.01) more likely to say that any 
vulnerable group is disadvantaged in land decisions, compared to the control mean. 
For our analysis of subgroup effects for secondary indicators, we find evidence of positive effects in the 
Land Tenure group. There are some positive treatment effects for elderly-headed households; they are 
more likely to believe that leaders are trustworthy (6% (p<0.05)), accountable (7% (p<0.10)), make 
transparent decisions (7% (p<0.10)) and protect natural resources (5% (p<0.05)), in comparison to their 
control counterparts. Moreover, poor households are 10% (p<0.10) less likely to report that 
households have been disadvantaged by land allocation decisions. 
However, in the Agroforestry group, we find negative treatment impacts across all subgroups. There is 
scattered evidence that treatment subgroups perceive equity or accountability problems with local 
leaders and additional disadvantages for vulnerable groups related to land management and allocation. In 
the Agroforestry group, female-headed households are less likely to report that leaders are trustworthy 
FIGURE 6-1: TREATMENT EFFECT ON LAND GOVERNANCE—PCA INDECES
50
(6% (p<0.05)), protect natural resource (6% p<0.10)), and make fair decisions (12% (p<0.01)). They are 
more likely to report that women (10% (p<0.05)) and the poor (7% (p<0.10)) are more likely to be 
disadvantaged in land decisions. Youth-headed households are more likely to say that land decision 
making is not fair (8% (p<0.10)), and elderly-headed households are more likely to say that vulnerable 
groups overall (10% (p<0.10)) have been disadvantaged in land decisions, as well as elderly (8% (p<0.10))
and poor (12% (p<0.01)). 
Similarly, in the Agroforestry + Land Tenure group, we find that poor households are less likely to say 
that land was allocated fairly (-0.26 (p<0.05)) and women are (8% (p<0.10)) more likely to report that 
the poor are disadvantaged in land related decisions. Land constrained households are less likely to 
report that leaders are trustworthy (16% (p<0.05)) and accountable (12% (p<0.10)). Youth are more 
likely to say that decision making is fair (13% (p<0.01)) and the poor are disadvantaged in land related 
decision making (7% (p<0.10)). 
The within-treatment analysis shows no significant difference between the average treatment effect and 
the subgroups for primary indicators. This exploratory evidence suggests that certain subgroups within 
the treatment—such as men or those that are relatively well-off—did not experience differential 
governance impacts. However, the within-treatment analysis for secondary indicators highlights positive 
impacts, along with further evidence of problematic results for households receiving the agroforestry 
extension program. On the positive side, we find several examples where subgroup impacts are higher 
(or more positive) than the average household effect. In particular, poor respondents receiving the land 
registration and governance program are more likely to participate in land meetings compared to their 
treatment counterparts. An, youth in the Agroforestry group are more likely to report greater equity in 
land allocation. On the negative side, we find significant lower results of decision-making fairness for 
female-headed households, youth and poor households, as well as a perception of additional 
disadvantages for vulnerable groups for female-headed households and elders. Finally, for the 
Agroforestry + Land Tenure treatment group, land constrained households are less likely to participate 
in land related meetings compared to the average treatment respondent and poor households are less 
likely to feel like land decisions are fair, transparent and accountable. 
DISCUSSION
As demonstrated in Section 5, the land registration and governance program had a strong and positive 
impact on tenure security, however, the regression results presented in this section indicate that this 
effect may not have occurred through the expected theory of change and governance indicators tested 
by the evaluation. The Land Tenure intervention was comprised of several governance activities—as 
defined by the implementing partner and USAID—including establishing Village Land Committees 
(VLCs), conducting participatory mapping, and facilitating the issuance of customary land certificates. 
VLCs also provided households with information on land law and rights (CDLA and VLCs), as well as 
information about customary land certificates. 
One explanation for the strong tenure results but underwhelming governance results could be that the 
tenure results are being primarily driven by the land registration and boundary demarcation activities—
and that the governance components have little effect. Another explanation may be that, although an 
extensive survey module was included, the governance indicators measured by the evaluation do not 
capture the types of governance changes that are occurring in village land administration due to the 
program. Another explanation may be the strong baseline state of governance in the study area.
51
In terms of descriptive findings, survey respondents were asked to rank a variety of community actors 
on a “ladder of power,” which indicates how much power various people have over land management 
decisions in the community. This is shown in see Figure 6-2. Unsurprisingly, the chief has the most 
decision-making power, followed by the headperson and village elders. This is consistent across all
households in the Land Tenure, Agroforestry and Agroforestry + Land Tenure group, and there is no 
substantive difference between these rankings for traditional authorities between treatment and control 
areas. Similarly, the treatment versus control rankings for the village as a whole, women and youth are 
essentially equivalent across treatment and control villages. Households in villages receiving the land 
registration and governance treatment rank the VLC as having the least amount of decision-making 
power for all groups.
FIGURE 6-2: LADDER OF POWER
52
7.0 OUTCOME FAMILY III, 
UPTAKE OF AGROFORESTRY 
PRACTICES & SEEDLING 
SURVIVAL
Hypothesis 5: Households in villages receiving the TGCC intervention have greater uptake of agroforestry. 
Hypothesis 6: Households in villages receiving the TGCC intervention have greater rates of agroforestry 
seedling survival.
SUMMARY OF KEY FINDINGS 
• The regression results show that the Agroforestry treatment was successful in motivating greater 
uptake of agroforestry.
• There is no quantitative evidence from the aggregate household findings to support the primary 
hypothesis that strengthening tenure security leads to greater agroforestry uptake. 
• All subgroups show significant uptake in agroforestry and increased agroforestry planting across fields 
for treatment villages receiving agroforestry extension. Descriptively, we see a small upward trend 
for subgroups in the Agroforestry + Land Tenure group of marginal benefits to linking agroforestry 
extension with land registration and governance.
• Within treatment areas, the seedling and tree survival rates are low. For the long term, it will be 
valuable to track whether there is a relationship between tree survival and tenure security 
perceptions.
Tables 7-1 through 7-3 illustrate the statistically significant regression findings for the household and 
subgroup panel samples. The coefficient on the treatment estimate is included in the table; a (P) 
designates a primary indicator. For agroforestry extent, we report the percent of agroforestry coverage
above the control mean, and for the binary indicator of agroforestry adoption, we report the marginal 
effect on the probability the household reports that they engaged in agroforestry. A discussion of the 
magnitude of effects is included below in the Results section. The level of statistical significance is 
indicated by the color of the table cell where: p < 0.1; p < 0.05; p < 0.01.
TABLE 7-1: AGROFORESTRY UPTAKE SUMMARY TABLE 
(BY TREATMENT GROUP)
AG LT AG+LT
% of field(s) planted with agroforestry trees or shrubs (S) 14% 15%
Field is planted with agroforestry trees or shrubs (Y/N) (S) 30% 30%
53
RESULTS
The findings in this section speak to the fundamental research question driving the study—whether 
stronger property rights affect a farmer’s decision to practice climate-smart agriculture, including 
agroforestry. The results show a positive aggregate Agroforestry treatment impact for agroforestry 
adoption (14% (p<0.01)) and the extent of agroforestry planted (30% (p<0.01)), illustrated in Figure 7-1. 
Correspondingly, all subgroups show significant uptake in agroforestry and increased agroforestry 
planting across fields—female (14% (p<0.01), 28% (p<0.01)), youth (13% (p<0.01), 28% (p<0.01)), and 
elder-headed (16% (p<0.01), 32% (p<0.01)) households, as well as poor (13% (p<0.01), 20%(p<0.01))
and land constrained households (15% (p<0.01), 22% (p<0.01)). 
TABLE 7-2: AGROFORESTRY UPTAKE SUBGROUPS SUMMARY TABLE
(AGROFORESTRY HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
% of field(s) planted with agroforestry trees or shrubs (S) 14% 13% 15% 13% 16%
Field is planted with agroforestry trees or shrubs (Y/N) (S) 28% 20% 22% 28% 32%
TABLE 7-3: AGROFORESTRY UPTAKE SUBGROUPS SUMMARY TABLE
(AGROFORESTRY + LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
% of field(s) planted with agroforestry trees or shrubs (S) 18% 14% 10% 14% 16%
Field is planted with agroforestry trees or shrubs (Y/N) (S) 33% 25% 19% 26% 35%
FIGURE 7-1: TREATMENT EFFECT ON AGROFORESTRY UPTAKE
54
For the overall household sample, there is no evidence to support a link between perceived tenure 
security and agroforestry uptake. Specifically, there is no evidence of marginal improvements in 
households that received the Agroforestry + Land Tenure treatment versus those that only received the 
Agroforestry treatment. This finding is consistent across the primary household panel analysis, as well as 
for the exploratory analysis where we focus on the results for only those households that actively 
participated in the COMACO training.
In contrast, the subgroup findings for female-headed and poor households indicate a slight uptick (5%)
from linking land tenure and agroforestry, although not statistically significant. This lends some 
qualitative support to the argument that, at least for more marginalized groups, stronger property rights 
may affect a farmer’s decision to practice climate-smart agriculture (CSA) including agroforestry. Finally, 
there is no strong/systematic evidence of within treatment differences between subgroups and the 
aggregate treatment estimates. This bodes well for equity in the distribution of program benefits and 
absence of elite capture. 
Within treatment areas, the seedling and tree survival rates are low. Figure 7-2 shows the survival rates 
of each village in the sample. Figure 7-3 displays seedling survival rates across all tree species from 2014–
2016 for treatment and control households. These low survival rates do not appear to correlate with 
the geographical location of the village, or with the location of wells or water points. It is important to 
highlight that the entire Chipata district experienced droughts during the 2014–2015 and 2015–2016 
growing seasons.
FIGURE 7-3: SEEDLING SURVIVAL RATES BY 
TREATMENT GROUP
FIGURE 7-2: SEEDLING SURVIVORSHIP BY 
VILLAGE
55
Across all years of the intervention, the majority of households who engaged in agroforestry report that 
less than 50% of their Musangu seedlings on their fields survived, as shown in Figure 7-4. However, 
treatment households have slightly higher survival rates than control households. Since the overall
sample size of control households who planted seedlings is low, the percentages may be misleading. In 
2014, the first year of the COMACO program, about a quarter of fields planted with Musangu in 
treatment communities (24%, N=25) have between 76–100% of seedlings alive today. Six fields in 
treatment communities have a zero percent survival rate, compared to one in control communities. 
Survival rates for Musangu trees planted in 2015, the second year of the program, are slightly lower than 
they were for the first year of the program. In the second year of the program, COMACO discontinued 
extension services and only provided seeds which may explain the lower survival rate. Only 17% (N=14) 
of fields planted by treatment households have a survival rate above 75%. Still, this is higher than control 
households, where only one household has a survival rate that high. It is also important to note that the 
2014–2015 and 2015–2016 growing seasons were years with historically bad rainfalls and low crop 
yields. 
In 2016, after COMACO ended agroforestry support in treatment villages, survival rates for Musangu 
trees fell below the survival rates of trees in control villages. Only four treatment household fields 
report a survival rate higher than 75%, which is equal to the number of fields as in the control group. 
Table 7-4 shows the average household survival rate for Musangu trees in 2014–2016 by treatment 
group. 
FIGURE 7-4: MUSANGU SURVIVORSHIP BY YEAR
56
TABLE 7-4: MUSANGU TREE SURVIVAL RATES
Musangu Trees Ag Ag+LT Control
Percentage of trees planted in 2014 that are alive now
0% 11% (4) 3% (2) 1
1–25 % 34% (12) 29% (20) 1
26–50% 26% (9) 26% (18) 5
51–75% 9% (3) 19% (13) 0
76–100% 20% (7) 23% (16) 2
Percentage of trees planted in 2015 that are alive now
0% 13% (4) 12% (6) 2
1–25% 34% (11) 25% (13) 3
26–50% 13% (4) 29% (15) 4
51–75% 22% (7) 13% (7) 2
76–100% 19% (6) 15% (8) 1
Percentage of trees planted in 2016 that are alive now
0% 2 0
1–25% 5 5 2
26–50% 4 4 0
51–75% 8 5 4
76–100% 3 1 4
Survival rates for Gliricidia trees are slightly more optimistic, though the majority of households still 
report survival rates under 50%, as shown in Figure 7-5. Of the seedlings planted in 2014, 32% (N=48) 
of fields planted by treatment households have more than 75% are alive today, compared to zero
control household fields. The figures are similar for seedlings in fields that treatment households planted 
in 2015 (33%, N=35) and 2016 (28%, N=20). The number of fields with a survival rate higher than 75% 
planted by control households increased from eight in 2015 to nine fields in 2016. Table 7-5 shows the 
average household survival rate for Gliricidia trees in 2014–2016 by treatment group. 
FIGURE 7-5: GLIRICIDIA SURVIVORSHIP BY YEAR
57
TABLE 7-5: GLIRICIDIA TREE SURVIVAL RATES
Gliricidia Trees Ag Ag+LT Control
Percentage of trees planted in 2014 that are alive now
0% 11% (9) 14% (10) 5
1–25% 22% (18) 21% (15) 2
26–50% 20% (17) 17% (12) 2
51–75% 14% (12) 10% (7) 1
76–100% 28% (23) 35% (25) 0
Percentage of trees planted in 2015 that are alive now
0% 11% (7) 12% (7) 3
1–25% 26% (16) 23% (13) 3
26–50% 15% (9) 18% (10) 6
51–75% 13% (8) 16% (9) 9
76–100% 31% (19) 28% (16) 8
Percentage of trees planted in 2016 that are alive now
0% 3% (1) 9% (4) 2
1–25% 23% (7) 22% (10) 3
26–50% 19% (6) 13% (6) 2
51–75% 16% (5) 27% (12) 5
76–100% 29% (9) 24% (11) 9
Overall, seedling survival for Musangu and Gliricidia was the lowest in 2016. This suggests that the 
support provided by COMACO during the agroforestry program was critical to the survival of seedlings, 
and withdrawing this support has had negative impacts on seedling survival rates.
DISCUSSION 
This section explores the factors driving program participation, expected benefits and the main 
challenges to agroforestry, in order to better understand uptake and seedling survival. 
PARTICIPATION
In communities where the agroforestry extension was offered 57% (N=711) of households had at least 
one household member participate in the program. The benefits of agroforestry are well-known to 
households in Chipata district. Focus group participants easily articulated the benefits, even if they did 
not participate in the agroforestry extension intervention.
An Agroforestry participant in Mnukwa chiefdom described the benefits this way: “Growing agroforestry 
in our fields will help us save tremendously on the purchases of fertilizers. At least we will have money 
to spend on other necessities, like our children’s education… It will also help us maintain the fertility in 
our fields, rather than depending on these fertilizers that even end up destroying the fertility of the soils 
in our fields.”
A female FGD participant in Maguya chiefdom described how she expects to benefit from agroforestry. 
“Growing little food has troubled us here in this village. If you don’t have fertilizer and at home you have 
a big family including grandchildren and the field is not productive so to hear that there are trees that 
can make soil fertile… so when fertility returns to the soil, people can grow enough food for their 
58
families. That is why we want agroforestry trees. Most of us can’t afford to buy fertilizer. So we want 
fertility to return to the soil so that we can reduce hunger in our homes”
Table 7-6 illustrates the many reasons households chose to participate in the Agroforestry intervention.
The primary reason households participate is to reap the agricultural benefits from agroforestry (78%, 
N=554), such as improving soil fertility and reducing the need for fertilizer. Other common reasons 
included wanting to learn new farming techniques (36%, N=261) and wanting advice about agroforestry 
(37%, N=263). Other draws to the program were receiving free inputs (8%, N=58), and wanting
fuelwood (2%, N=14). 
TABLE 7-6: WHY PARTICIPATE IN COMACO PROGRAM
Reasons for participating Ag Ag+LT
Want to higher soil fertility and other benefits from 
agroforestry 
77% (250) 78% (304)
Want to learn about new farming techniques 37% (120) 36% (141)
Want advice about agroforestry 31% (101) 42% (162)
Want free inputs 6% (20) 10% (38)
Want fuelwood 1% (3) 3% (11)
My friend/neighbor/relative urged me to participate 1% (2) 1% (5)
The headman urged me to participate 1% (2) 2% (6)
Want a guaranteed buyer/better price for my crops 1% (4) 1% (2)
Want to benefit from fuel-efficient cookstoves 1% (3) 1% (2)
Want fodder for livestock 0% (0) 0% (0)
Table 7-7 illustrates the reasons that households did not participate in the agroforestry extension
program. The biggest reason for nonparticipation was inability to attend the meetings (40%, N=141). 
This suggests that modifications to recruitment protocols, such as greater outreach before signups 
began, or allowing for multiple rounds of program signups as news of the program permeates the 
villages, might engender higher participation rates. 
Other reasons households did not participate include an overall lack of interest (7%, N=29), a desire to 
see if other households were successful before trying agroforestry (7%, N=26), and because 
agroforestry requires too much labor for the household (6%, N=21). Not being allowed to plant trees is 
almost never a reason households do not participate (1%, N=5), suggesting that for most households, 
there are no rules prohibiting them from planting trees.
In the FGDs, a lack of incentive to counteract the high labor and time costs is a consistent theme for 
households who chose not to participate. They would prefer a program that provided inputs that 
materialized in the short term, such as seeds, to the long-term benefits associated with agroforestry. 
This is especially true in areas that currently have virgin land or relatively fertile soil. The two quotes 
below illustrate this sentiment:
“The reason why I did not take part is that doesn’t this project go with other additional crops such as 
beans and cowpeas? So you find that you as a farmer can’t start with planting trees without first planting 
crops because as a farmer you first need a seed that will lift you out of your problems” (Mshawa, 
Agroforestry non-respondent).
“How can a program be only about trees! It takes a lot of years before you can cut a ‘beam’ out of 
it?...We would be glad if this program will go with seeds that we plant in the fields. These same seeds 
59
will also be planted in the same field but now if we have only planted trees, what benefit will we achieve 
if there is no seed to plant in it? Then poverty will continue but trees and seeds ought to go together so 
that we can also sell” (Mshawa, male Agroforestry non-participant).
TABLE 7-7: WHY NOT PARTICIPATE
Reasons for not participating Ag
(N=174)
Ag+LT
(N=177)
Wanted to participate but failed to attend the 
meetings 40% (69) 41% (72)
Not interested in trying a new farming method 7% (13) 7% (13)
Wanted to see if other households had success 
before trying 
5% (9) 7% (12)
Required too much labor 6% (10) 6% (11)
No benefits to agroforestry 5% (9) 3% (5)
Not allowed to plant trees on my land 1% (2) 2% (3)
AGROFORESTRY UPTAKE 
Households in Agroforestry treatment villages (54%, N=771) are more likely to have engaged in 
agroforestry than control households (24%, N=163). Agroforestry + Land treatment households have 
the highest participation rate (56%, N=412). Agroforestry households have a participation rate of 53%
(N=359).
The most common tree species planted are the Musangu tree and Gliricidia, each planted by roughly 40% 
of all households in the treatment group.
29 It can be grown amidst any crop, but COMACO encouraged 
farmers to plant their fields where Musangu seedlings were being grown with low-growing crops such as 
groundnuts, to ensure that the seedlings would get enough sunlight. Households were provided with 25 
Musangu seedlings. Just over a third of households (35%, N=205) receiving the Agroforestry treatment 
and (41%, N=263) of households receiving the Agroforestry + Land Tenure treatment planted Musangu 
trees. Uptake for the control group is substantially lower (14%, N=79). 
Gliricidia is also grown from a seedling and intercropped. Households struggled to keep the seedlings 
alive in the nursery, as Gliricidia is particularly sensitive to water shortages. This may explain the slightly 
lower rates of Gliricidia adoption, though households who participated in the extension program 
received both species. Households in the treatment group were provided with 100 Gliricidia seedlings
and have a rate of adoption slightly under 40% (39%, N=552)) The rate for control villages is significantly 
lower (15%, N=102). Female headed households in treatment communities are slightly less likely to have 
planted Gliricidia trees than male headed households (FHH: 23%, MHH: 28%, significant at the 1% level). 
Based on Monitoring and Evaluation (M&E) conversations with the COMACO team, female-headed 
households were more likely to struggle to transport their seedlings from the nursery to their fields, and 
the large number of Gliricidia seedlings (100 Gliricidia vs. 25 Musangu) may have been too much of an 
obstacle for female-headed households. 
 29 Musangu trees are grown from seedlings. It is intercropped with the field’s main crop, and is best suited to being planted in a 5–by–5 meter 
grid.
60
The third species of tree provided by the program was pigeon peas, though the provision of these seeds 
was not uniform. The original design called for 500g of pigeon pea seeds to be distributed to 
households, which would be planted directly in their fields. Pigeon peas grow into a bush that produces 
an edible fruit (the pigeon pea), and are highly desirable for consumption. Despite their desirability, less 
than 5% of treatment households planted pigeon peas (4%, N=48).
30
TABLE 7-8: HH TREE SPECIES PLANTED
Species Ag (N=585) Ag+LT (N=646) Control (N=580)
Musangu 35% (205) 41% (263) 14% (79)
Gliricidia 40% (236) 38% (243) 15% (88)
Pigeon Peas 5% (31) 3% (17) 1% (5)
EXPECTED BENEFITS FROM AGROFORESTRY UPTAKE
Households have a clear understanding of the expected benefits for agroforestry trees. The most 
common expected benefit cited by treatment groups both now and in the future is improved soil 
fertility (18%, N=82 for Musangu; 26%, N=136 for Gliricidia). The number of households who expect to 
see that benefit in the future is double the number of households who currently see the benefit (35%, 
N=164 for Musangu; 48%, N=231 for Gliricidia), which seems to indicate households understand that the 
benefits of agroforestry accrue in the future. 
Similarly, improved crop growth around trees is another benefit cited for both species, particularly in 
the future. Ten percent of households say they expect to see improved crop growth in the future for 
Musangu trees (N=48), as do 13% of households for Gliricidia (N=60). This is a lower percentage than 
we would expect, since higher yields and/or improved crop growth should be the ultimate benefit to 
households. 
A benefit emphasized in focus group discussions is the increased availability of fuel wood. It appears that 
at least some households view their agroforestry trees as an opportunity for a woodlot as opposed to 
caring for their trees to improve their field’s agricultural productivity. This is further supported by the 
number of households who do not believe that there are any yield-related benefits to the trees. The 
household statistics are less striking than the qualitative analysis, but are worth noting—5% (N=22) of 
households believe their Musangu trees will increase the availability of fuelwood in the future. This is 
highest in the Agroforestry group (5%, N=11). The percent of households who believe that Gliricidia
trees will increase the availability of fuelwood in the future is slightly higher (6%, N=31).
As Table 7-9 below illustrates, agroforestry adoption is not associated with greater perceived tenure 
security. Almost no households believe that planting agroforestry trees reduce the fear of their land 
being taken, or raise the value of their land for collateral, either now or in the future. This coincides 
with the regression results presented in Section 5 that show no significant difference between the Land 
Tenure and Agroforestry + Land Tenure findings for perceived tenure security.
TABLE 7-9: MUSANGU AND GLIRICIDIA TREE BENEFITS 
Benefits Present Future Present Future
Musangu trees Ag (N=205) Ag+LT (N=263)
 30 COMACO did not distributed pigeon peas in all villages due to a lack of seeds.
61
No benefits 23% (47) 8% (16) 35% (91) 11% (29)
Improved soil fertility 18% (36) 31% (63) 18% (46) 38% (101)
Improved crop growth around trees 9% (19) 9% (18) 4% (10) 11% (30)
Higher overall crop yield 3% (7) 9% (18) 3% (9) 13% (33)
Increased fuel wood availability <1% (1) 5% (11) 2% (4) 4% (11)
Reduced labor time on weeding 0% (0) 0% (0) 0% (0) <1% (1)
Reduced weeds 0% (0) 1% (2) 0% (0) 2% (4)
Reduced fear of land being taken 0% (0) <1% (1) 0% (0) <1% (1)
Raised value of the land for collateral 0% (0) 0% (0) 0% (0) 0% (0)
Gliricidia trees Ag (N=236) Ag+LT (N=243)
Improved soil fertility 32% (75) 49% (115) 25% (61) 48% (116)
No benefits 38% (90) 15% (35) 40% (98) 15% (37)
Improved crop growth around trees 9% (21) 11% (27) 9% (21) 14% (33)
Higher overall crop yield 6% (14) 18% (42) 7% (16) 19% (45)
Increased fuel wood availability 1% (3) 7% (16) 2% (5) 6% (15)
Reduced labor time on weeding 0% (0) <1% (2) 1% (3) 3% (7)
Reduced weeds <1% (1) 1% (3) <1% (1) 4% (9)
Reduced fear of land being taken <1% (1) <1% (1) <1% (1) 1% (2)
Raised value of the land for collateral 0% (0) 0% (0) 0% (0) 0% (0)
Taken together, these findings suggest that the program helped households understand the major 
ecological benefits of agroforestry trees, as well as the fact that these benefits will likely manifest in the 
future as opposed to the present.
Despite understanding the benefits, households seem skeptical that the benefits will actually materialize 
for them, either now or in the next 3–5 years. At present, between 30–40% of treatment households 
see no benefits to their agroforestry trees (29%, N=138 for Musangu; 39%, N=188 for Gliricidia), though 
as expected, this number drops substantially when households are asked about benefits they expect in 
the next 3–5 years (10%, N=45 for Musangu; 15%, N=72 for Gliricidia). Still, the number of people who 
do not expect any benefits from their agroforestry trees seems high, considering the amount of time 
and labor households put into their trees, and raises the question of why these households are engaging 
in agroforestry at all. There is nothing definitive in the data to answer this question, but focus group 
responses do allude to households joining the program from a sense of obligation.
CHALLENGES TO SURVIVAL 
Table 7-10 shows the types of challenges to agroforestry tree survival that households encountered. 
The most common challenges to agroforestry seedling survival, for both Gliricidia and Musangu, include a 
lack of water for seedlings, fires burning trees, pests killing the trees and animals grazing in the field.31
 31 The greatest challenge facing households planting pigeon peas is not lack of water, but animals grazing in the field, followed by pests killing or 
damaging the plants. Lack of water is the third most common challenge. 
62
Given the drought conditions during implementation, lack of water for seedlings was the number one 
challenge for both Musangu and Gliricidia seedlings, identified by 20% (N=95) of treatment households 
for Musangu trees, and 27% (N=130) of treatment households for Gliricidia trees. Fourteen percent of 
treatment households (N=113) say that their nursery never had enough water, and 12% (N=97) say 
their nursery only had enough water part of the year.32 COMACO attempted to address this challenge 
by installing wells in villages with significant water scarcity, however, it does not appear that the wells 
improved seedling survival in subsequent years. 
FGDs tell a similar story. A participant in the agroforestry extension from Mnukwa chiefdom 
commented “The challenge we are facing is lack of water. Even if they brought more seed for us to 
plant, once the rains stop, we work in vain. So we don’t know how you are going to help us, once you 
help us with water, aah we will have trees all over here." 
Traditional land management practices also present a challenge to the survival of agroforestry trees. In
particular, animals grazing trees, which occurs when animals are allowed to graze anywhere in the fields 
after crops have been harvested, was cited as a challenge to growing Musangu by 6% of treatment 
households (N=27). Another traditional practice, setting fires to fields to clear crops after harvest or to 
harvest mice to eat, resulted in 8% of treatment households (N=38) reporting that their Musangu 
seedlings were burned by fires. 
Pests killing and damaging trees was also a challenge for households growing both Musangu and Gliricidia, 
especially in the control areas. The fact that treatment communities were slightly less likely to have their 
trees killed by pests may speak to the success of COMACO’s training about how to protect their trees. 
Finally, it is worth noting that lack of knowledge was not a prevalent challenge for households. 
Additional agroforestry trainings in the future would likely have limited impact on uptake or 
survivorship. 
 32 Thirty-one percent (N=456) of household respondents did not know if their nurseries had water during the year.
63
TABLE 7-10: AGROFORESTRY CHALLENGES
Musangu Gliricidia
Ag Ag+LT Control Ag Ag+LT Control
Lack of water 19% (38) 22% (57) 22% (17) 28% (67) 26% (63) 24% (21)
Pests killed/damaged 
trees
10% (20) 13% (34) 21% (27) 18% (43) 18% (44) 20% (17)
Animals grazed trees 6% (12) 6% (15) 5% (4) 10% (24) 9% (22) 10% (9)
Fires burned the 
trees
3% (6) 12% (32) 9% (7) 12% (29) 9% (21) 13% (11)
Trees were chopped 
down by mistake 4% (8) 5% (12) 13% (10) 6% (13) 5% (11) 6% (5)
Lack of knowledge 4% (9) 2% (6) 6% (5) 6% (15) 3% (8) 8% (7)
Difficulty transporting 
seedlings to field 0% (0) 1% (2) 0% (0) 3% (6) 3% (7) 2% (2)
Not enough labor to 
care for trees <1% (1) 2% (5) 1% (1) 1% (2) 2% (5) 0% (0)
Could not access 
seeds
<1% (1) <1% (1) 1% (1) 3% (6) 1% (3) 1% (1)
Did not see benefits 
so stopped caring for 
trees 
0% (0) 0% (0) 0% (0) 0% (0) 1% (2) 1% (1)
VILLAGE LAND MANAGEMENT PRACTICES 
Land management rules in a village play an important role in the success or failure of agroforestry. Based 
on the key informant YGL responses, the agroforestry program had a limited effect on the existence or 
adoption of village rules related to land management. The descriptive results do not indicate a 
substantive difference in rule adoption or the types of rules between treatment and control 
communities. 
Rules about setting fires and grazing livestock are particularly important, since seedlings are susceptible 
to being grazed or trampled by livestock or burned by fires, as discussed above. Half of villages have a 
rule that forbids lighting fires in fields at any point in the year (53%, N=146). The other half of villages 
either have no rule about lighting fires, or allow fires after the crops are harvested. Figure 7-6 shows 
seedling survival in villages that have a rule that forbids fires on fields at any time in the year. The results 
show a potential association between the presence of rules and higher survival rates for agroforestry 
trees.
Grazing rules are another possible means for increasing seedling survival. Traditionally, animals are 
allowed to graze in fields after crops are harvested, which leaves seedlings vulnerable to being grazed or 
trampled by cattle, goats and other livestock. We examined the seedling survival rates for villages that 
forbid livestock grazing on fields, or allow it only if the livestock are accompanied (68%, N=187), and 
compared to villages that either have no rule about grazing or allow livestock to graze unaccompanied 
on fields (32%, N=86). Figure 7-7 shows the results of the comparison. There does not appear to be as 
strong of an association between grazing rules and seedling survival, compared to the rules about 
starting fires. One reason that grazing rules may not have the same effect on survival as fires is that even 
if supervised, it is not always possible to control livestock, and seedlings may still be grazed or trampled 
even if the rule is being obeyed. Another possible explanation is a lower rate of adherence to rules 
about grazing than rules about fire, as reported by households. Five percent of households in treatment 
64
areas (N=46) say that households do not follow rules about grazing, compared to 10% of households in 
treatment area (N=99) who say households do not follow rules about fires.
Focus group discussions further highlight the importance of good grazing practices. A participant from 
Mnukwa described the challenge this way: “We will first need to have proper prevention plan from 
livestock. Sometimes we are not around, we may be attending a funeral at a neighboring village, so if we 
do not secure these trees from livestock then there is nothing that will happen. The chief need to teach 
the headmen how to look after livestock and the headmen should in turn teach their subjects, because if 
these livestock are left like this destroying plants… we need strict rules on livestock, if not it will 
continue to be a serious problem."
FIGURE 7-6: SURVIVORSHIP RATE & FIRE 
RULES BY VILLAGE
FIGURE 7-7: SURVIVORSHIP RATE & GRAZING 
RULES BY VILLAGE
65
8.0 OUTCOME FAMILY IV, 
FIELD INVESTMENT
Hypothesis 7: Households in villages receiving the TGCC intervention have greater uptake of climate-smart 
agriculture (excluding agroforestry).
Hypothesis 8: Households in villages receiving the TGCC intervention have greater uptake of short and long 
term field investments.
SUMMARY OF KEY FINDINGS 
• For the household level panel analysis, there is no evidence of greater labor or cost intensive field 
investments for households in the Land Tenure treatment group. 
• Across the three treatment groups, we find evidence of negative investment impacts for all subgroups 
except female-headed households. 
• For the within-treatment analysis, we generally find that the average treatment impacts are 
equivalent to the subgroup impacts, although there is evidence that youth in treatment areas are less 
likely to fallow or engage in labor or cost intensive field investments. 
Tables 8-1 through 8-4 illustrate the statistically significant regression findings for the household and 
subgroup panel samples. The coefficient on the treatment estimate is included in the table; a (P) 
designates a primary indicator and an (S) designates a secondary indicator. For binary indicators, we 
report the marginal effect on the probability the household reports that they have undertaken the field 
investment in question. A discussion of the magnitude of effects is included below in the Results section. 
The level of statistical significance is indicated by the color of the table cell where: p < 0.1; p < 0.05; p < 
0.01.
TABLE 8-1: FIELD INVESTMENT OVERALL SUMMARY TABLE 
(BY TREATMENT GROUP)
AG LT AG+LT
HH engages in fallowing (Y/N) (S) -4%
Minimum number of seasons HH left a field fallow (P) 
HH practiced zero tillage on field (S)
HH applied manure or compost on field (S)
HH used fertilizer on field (S)
HH constructed planting basins in field (S)
HH used crop rotation on field (S)
TABLE 8-2: FIELD INVESTMENT SUBGROUPS SUMMARY TABLE
66
(LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
HH engages in fallowing (Y/N) (S)
Minimum number of seasons HH left a field fallow (P) 
HH practiced zero tillage on field (S) -7%
HH applied manure or compost on field (S) -21%
HH used fertilizer on field (S)
HH constructed planting basins in field (S) -15%
HH used crop rotation on field (S)
TABLE 8-3: FIELD INVESTMENT SUBGROUPS SUMMARY TABLE
(AGROFORESTRY HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
HH engages in fallowing (Y/N) (S) -6%
Minimum number of seasons HH left a field fallow (P) 
HH practiced zero tillage on field (S)
HH applied manure or compost on field (S)
HH used fertilizer on field (S) 14%
HH constructed planting basins in field (S)
HH used crop rotation on field (S)
TABLE 8-4: FIELD INVESTMENT SUBGROUPS SUMMARY TABLE
(AGROFORESTRY + LAND TENURE HOUSEHOLDS ONLY)
FHH POOR LC YOUTH ELDERLY
HH engages in fallowing (Y/N) (S) -6%
Minimum number of seasons HH left a field fallow (P) 
HH practiced zero tillage on field (S) -8%
HH applied manure or compost on field (S)
HH used fertilizer on field (S)
HH constructed planting basins in field (S)
HH used crop rotation on field (S)
67
RESULTS
There is no evidence of improved long term or labor/cost intensive field investments (planting basins, 
rotating crops fallowing and fertilizer application) for primary or secondary indicators measured at the 
household level for the Land Tenure intervention. However, supplemental research conducted at the 
field level does uncover some empirical support to the hypothesis that greater perceived tenure security 
is associated with increased field investments (See Huntington et al., 2018). 
Although land constrained households are more likely to apply chemical fertilizer, the subgroup results 
show several negative heterogeneous treatment impacts for youth- and elder-headed households, as 
well as poor and land constrained. There are no differential impacts for female-headed households. For 
Land Tenure, the results show that elderly households are 7% (p<0.05) less likely to practice zero tillage 
methods and that land constrained households are less likely to engage in manuring or composting (21%
(p<0.05)), or planning basins (15% (p<0.05)). Poor households in the Agroforestry treatment are 6%
(p<0.05) less likely to fallow. In the Agroforestry + Land Tenure group, we find that youth are less likely 
to leave their fields fallow (6% (p<0.10)) and elder-headed households are 8% (p<0.05) less likely to use 
zero tillage than their control counterparts. 
Given that negative investment results were unexpected, we checked the subgroup balance on 
investment indicators at baseline and found several cases of imbalance that might contribute to the 
explanation of these results. 
In terms of the within-treatment analysis, the findings do not reveal systematic evidence of a difference 
between the average treatment effects and subgroup results for female-headed, elderly, poor or land 
constrained households. The exception to this is the results for youth. Compared to the average 
treatment effect for ‘non-youth,’ youth are less likely to have fallowed in the past 3 years, or to have 
invested in CSA in the Agroforestry + Land Tenure condition. 
68
9.0 OUTCOME FAMILY V,
LONG TERM OUTCOMES: 
AGRICULTURAL 
PRODUCTIVITY & 
LIVELIHOODS
Hypothesis 9: Households in villages receiving the TGCC intervention have higher agricultural productivity and 
crop yields.
Hypothesis 10:Households in villages receiving the TGCC intervention have different livelihood and welfare 
outcomes.
SUMMARY OF KEY FINDINGS 
• There is no evidence of treatment impacts on outcomes that may take several years to materialize, 
including changes in agricultural productivity, land transactions or livelihoods. 
TABLE 9-1: AGRICULTURAL PRODUCTIVITY AND LIVELIHOODS*
(HOUSEHOLDS OVERALL)
Ag LT Ag+LT
HH reported experiencing improved crop growth as a 
result of agroforestry (P)
HH reported experiencing higher overall crop yield as a 
result of agroforestry (S)
Asset-based wealth index (P)
HHs obtained formal loan from bank or microcredit 
institution (S)
HH engages in land rental activity as landlord (S)
RESULTS 
As Table 9-1 shows above, there is no evidence of a treatment impact for our long term primary 
indicators. This is not surprising given the short time period between program completion and follow-up
data collection. The evaluation team recommends another round of data collection in 2–3 years to 
investigate the impact of TGCC on these development priorities. 
69
ANNEX 1. GEOSPATIAL 
CONTEXT
OVERVIEW
CDLA provided geospatial data on parcel boundaries across the four treatment chiefdoms as well as 
claimant details pertaining to each parcel. An examination of claimant details and spatial relationships in 
land treatment areas provides contextual information to enable better interpretation of the evaluation 
findings and help identify how program implementation and context variation should be accounted for. It 
is important to note that this data is only available for villages in the Land and Ag+Land treatment 
groups.
Another important distinction between CDLA’s geospatial parcel data and ERC’s non-spatial field data is 
how they are defined. ERC’s field data is defined by how fields are used for cultivation—for example a 
field used for maize cultivation versus a field that has been left fallow. In contrast, CDLA’s parcel data is 
defined by ownership. CDLA allowed households to delineate each of their fields as a single parcel or by 
multiple parcels depending on who claimed ownership. CDLA’s geospatial data shows that most 
claimants own contiguous parcels that are all located within the same village. For individuals that own 
parcels in different villages, this is generally since a village boundary is separating two adjunct parcels.
CERTIFICATE REGISTRATION 
According to CDLA’s certificate registration details, there are 5,911 land parcels that are registered 
across the four treatment chiefdoms and that match the treatment villages in this analysis, shown in 
Figure A1-1 below. The majority of these parcels (87%, N=4,799) are held by a household with the 
remaining (13%, N=732) are held by extended family. The majority of landholders’ claims to the land are 
based on inheritance (73%, N=4,354) followed by land given by a family member (15%, N=863) and land 
given by the chief or headman (11%, N=697). 
73%
15%
11%
1%
0%
10%
20%
30%
40%
50%
60%
70%
80%
Land Acquired from 
Inheritance
Land Given by Family 
Member
Land Given by Chief 
or Headman
Other
FIGURE A1-1: PARCEL CLAIMANT BASIS
70
The total number of landholders per parcel ranges from one to 11. The most common number of 
landholders per parcel is two (67%, N=3,949), followed by one (20%, N=1,186) and then three 
landholders (8%, N=487). Overall, most land parcels were registered jointly (80%, N=4,725). Of the 
individually-owned parcels, men (65%, N=771) are almost twice as likely to own the land parcel than 
women (35%, N=415), shown in Figure A1-2. 
Since inheritance is the most common basis for claims to land, it’s important to assess the gender 
division in parcel ownership to understand any potential differentiation in inheritance rights across 
women and men. Across jointly owned parcels, 73% (N=3,439) are male-female owned, 21% (N=1,002) 
are male-male owned and 6% (N=284) are female-female owned. It is expected that male-female is the 
most common type of joint ownership, specifically between spouses, which the certificate registration 
details confirm. Across jointly-owned male-female parcels, 48% (N=1,641) are indeed between spouses, 
followed by 32% (N=1,081) between parent and child and 8% (N=282) between siblings. Parent-child 
joint ownership is the most common type of joint ownership for both male-male parcels (53%, N=528) 
and female-female parcels (55%, N=156).
PARCEL CHARACTERISTICS
PARCEL TOTAL AREA
The average parcel size across all treatment groups is .028 sq. km with a minimum parcel size of 0.007 
sq. km and a maximum of 0.16 sq. km. Figure A1-3 shows individually owned parcels have a slightly 
smaller mean size (0.025 sq. km) than the overall average and jointly owned parcels are in line with this 
mean size (0.028 sq. km). For Individually owned parcels, male-owned parcels have a greater average 
size than females (0.025 sq. km vs 0.023 sq. km). For jointly owned parcels, male-male landholders 
overall have the greatest average parcel size (0.03 square kilometers) compared to male-female (0.028 
sq. km) and female-female (0.027 sq. km). 
13%
7%
17%
58%
5%
0%
10%
20%
30%
40%
50%
60%
70%
Male Female Male - Male Male - Female Female - Female
Individual-owned parcels Jointly-owned parcels
FIGURE A1-2: PARCEL CLAIMANT TYPE BY GENDER
71
Across the male-male and male-female owned parcels, siblings have the greatest average parcel sizes 
(0.036 sq. km and 0.036 sq. km, respectively). Female-female sibling landholders also have the largest 
parcel sizes (0.03 sq. km) out of all the female-female groups. For male-female spouse owned parcels, 
the average parcel size is 0.029 sq. km. Parcel size across parent-child owned parcels is largest among 
the male-male group (0.03 sq. km) followed by the male-female group (0.028 sq. km) and lastly the 
female-female group (0.025 sq. km). These larger than average parcel sizes for siblings may be an 
indication that larger, more desirable land parcels are inherited and shared within a family while smaller 
parcels may be used for other purposes such as sharing with spouses or other distant relatives. 
PARCEL DISTANCE TO VILLAGE CENTER
The mean distance from land parcels to the center of the village is 1.1 km with a minimum distance of 
0.15 km and a max distance of 4.9 km. Village centers were calculated using the mean center of 
household GPS coordinates. Individually owned parcels are slightly farther from the village center (1.2 
km) than the overall average and jointly owned parcels are just below the average distance (1 km). For 
individually owned parcels, male-owned parcels are the same mean distance to the village center as 
female-owned parcels (1.2 km). For jointly owned parcels, all gender subgroups also have the same 
mean distance of 1 km from the village center. These parcel distance calculations indicate that there is 
no substantial differences across gender in parcel distance from the center of the village.
PARCEL DISTANCE TO WELL OR BOREHOLE
The average distance to wells or boreholes is 1.27 km with a minimum distance of 0.02 km and a 
maximum distance of 6.5 km. Parcels in the Ag+Land treatment group (N=3,296) have an average 
distance of 1.38 km to the nearest well or borehole, which is farther than villages in the Land treatment 
group (1.14 km, N=2,547).
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.035
Male Female Male - Male Male - Female Female - Female
Individual-owned parcels Jointly-owned parcels
Average Parcel Area (Sq. KM)
FIGURE A1-3: AVERAGE PARCEL AREA BY GENDER
72
The average distances from a parcel to a well or borehole across parcels that are jointly owned (1.26 
km) and individually owned (1.28 km) are almost equal. The mean distance of individually owned male 
parcels to a well or borehole (1.29 km) is slightly greater compared to individually owed female parcels 
(1.21 km). Of the jointly owned parcels, male-male owned parcels have the farthest average distance 
from wells (1.32 km) compared to male-female owned parcels (1.27 km) and female-female owned 
parcels (1.26 km).
Across male-female owned parcels, spouses have parcels that are the farthest away from wells or 
boreholes (1.35 km) compared to siblings (1.19 km) and parent and child owned parcels (1.13 km). 
Female-female sibling owned parcels have an average closer parcel distance to wells (1.19 km) than 
female-female parent and child owned parcels (1.24 km).
PARCEL DISTANCE TO STREAMS AND DAMBOS
The average distance from a parcel to a stream is 0.61 km with a minimum distance of 0 km and a 
maximum distance of 3.7 km. Parcels in the Ag+Land treatment group are farther on average (0.71 km) 
from streams than parcels in the Land treatment group (0.47 km). Parcels that are individually owned 
are farther away on average (0.74 km) from streams than jointly owned parcels (0.57 km). Similar to the 
distances to wells or boreholes, female-female jointly owned parcels are closer to streams (0.49 km) 
than male-male (0.58 km) or male-female (0.57) jointly owned parcels. Of the female-female owned 
parcels, siblings have parcels that are farther away (0.55 km) compared to parent and child landholders 
(0.45 km).
The average distance from a parcel to a dambo is 1.1 km with a minimum distance of 0.04 km and a 
maximum distance of 15 km. This average distance is consistent across parcels in both the Ag+Land and 
Land treatment groups though there are large standard deviations within chiefdoms across the Ag+Land 
treatment group, particularly in Maguya (1.9 km) and Mnukwa (1.5 km).
The averages across all resources including wells or boreholes, streams and dambos is shown in Figure 
A1-4. Overall, parcels in the Ag+Land treatment group shows slightly farther distances away from local 
resources.
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60
Land
AgLand
Dambos Streams Well or Boreholes
FIGURE A1-4: AVERAGE DISTANCE (KM) FROM PARCEL TO RESOURCES BY 
TREATMENT GROUP
73
VILLAGE CHARACTERISTICS
VILLAGE AREA AND NUMBER OF PARCELS
The average village size across all villages is 2.5 sq. km. Villages in the Ag+Land treatment group are 
overall slightly larger (2.7 sq. km, N=62) than villages in the land treatment group (2.4 sq. km, N=53). 
The variation in total village land area across chiefdoms varies substantially by treatment groups (see 
Table A1-1 and Figure A1-5). In Mshawa, villages in the Ag+Land treatment group (3.2 sq. km, N=23) 
are about double the total average size of villages in the Land treatment group (1.5 sq. km, N=19). 
Conversely, villages in the Ag+Land treatment group in Mkanda are half the size (0.8 sq. km, N=14) of 
those in the Land treatment group (1.3 sq. km, N=12). In Mnukwa, the average size of villages in the 
Land treatment group have a very high standard deviation (7 sq. km), as shown in Table A-1. The mean 
number of parcels by village is 51 with the Ag+Land group averaging a slightly greater number of parcels 
(53) compared to the Land group (48).
TABLE A1-1: MEAN VILLAGE SIZE BY TREATMENT GROUP
FIGURE A1-5: VILLAGE BOUNDARIES BY TREATMENT 
GROUP ACROSS CHIEFDOMS
74
Ag+Land Land
Area Std. N Area Std. N
Maguya 2.8 2.0 13 2.9 2.5 11
Mkanda 0.8 1.2 14 1.3 1.2 12
Mnukwa 3.9 3.9 12 4.5 7.0 11
Mshawa 3.2 2.7 23 1.5 1.1 19
DAMBOS
The average distance from village centers to dambos is 0.36 km with a minimum distance of 0 km and a 
maximum distance of 2.8 km. Both villages in the Land and Ag+Land treatment group are also equal to 
the overall average distance so there is little variation across these two treatment groups (see Figure 
A1-6). 
FIGURE A1-6: DISTANCE FROM VILLAGE TO NEAREST DAMBO
75
ANNEX 2. WITH CERTIFICATE 
VERSUS WITHOUT 
CERTIFICATE
AGGREGATE 
All tenure security indicators that were significant for the aggregate population remain significant for 
households who report having a customary land certificate. Several of these indicators increase from a 
5% level of significance to a 1% level of significance, particularly for the Ag+Land group. This is not 
surprising, as we would expect households who hold a physical customary land certificate (CLC) to 
perceive their tenure security to be higher than households with certificates, even if those households 
had undergone the certification process. The significance of the field investment indicators related to 
long-term investments also increases in significance from 10% to 1% for the Ag+Land group. 
When the sample is restricted to only households who have received CLCs, a number of governance 
indicators gain significance. For the tenure security group, these include an index of overall governance, 
an index of leader satisfaction and indicators about household’s perception of their leaders, including 
that they are trustworthy, protect natural resources and make fair decisions. For both the tenure 
security group and the Ag+Land group, having a certificate increases the significance of indicators about 
leader’s decisions being transparent and rules being clear. Interestingly, having a certificate also increases 
the significance of minority tribes being disadvantaged for Ag+Land households. 
FHH
We also see an increase in significance for tenure security indicators for female-headed households who 
report having land certificates in the Land Tenure and Ag+Land treatments. Female-headed households 
are more likely to say their land is safe from encroachment in both the long and short term. An 
indicator for the threat of encroachment by neighboring villages also becomes significant for female￾headed households. 
Like the results for aggregate households, when the sample of female-headed households is restricted to 
those who have received CLCs, a number of governance indicators gain significance for the Land Tenure 
group, but not the Ag+Land group. Female-headed households with certificates are more likely to 
participate in community meetings, more likely to say that village rules are clear, decisions are 
transparent and leaders are trustworthy. Three index variables that capture overall governance, leader 
satisfaction and the risk of allocation also become significant.
Indicators for long-term field investments and an asset index that measures livelihoods also gained 
significance amongst female-headed households in the land tenure group. This suggests that once 
certificates are distributed for the entire sample, we may expect to see some positive benefits from the 
land tenure program across a few of these long-term indicators, at least for female-headed households. 
76
YOUTH 
Possession of certificates appears to have a large effect on the perceived tenure security for youths in 
the Land Tenure group compared to youths without a certificate. Every single tenure security indicator 
is significant, and many at the 1% level. In the Ag+Land group, youth are more likely to believe their 
fields are safe from encroachment from other households in the village, from family members, from 
elites and from the chief. Youth in the Land Tenure group are also more likely to invest in climate-smart 
agricultural practices.
Youth with certificates do not show the same increase in significance in indicators associated with good 
governance that we see with other groups, with a few exceptions. Youth in Land Tenure households 
with certificates are more likely to say their leaders are trustworthy than those without certificates, and 
in the Ag+Land group, youth with certificates are more likely to say decisions are transparent, though 
both are only significant at the 10% level. 
Five indicators related to the treatment of vulnerable populations, including women, youth and the 
elderly, all gain significance for the Land Tenure group. In the Ag+Land group, an index related to the 
treatment of marginalized groups gains significance; thoseother indicators remain the same as for the 
aggregate group. 
ELDERLY 
The significance of land tenure indicators for elderly households with land certificates remains the same 
as they are for the aggregate population, as do most of the governance and agroforestry indicators. 
However, as with female-headed households and households overall, governance indicators gain 
significance for households in the Land Tenure group. These include indicators for overall governance, 
trustworthiness of leaders, clarity of rules about land management and participation in land meetings. 
Elderly households with land certificates in both the land tenure and Ag+Land group are more likely to 
invest in long-term climate-smart agriculture technologies, and in the Ag+Land elderly households are 
more likely to participate in climate-smart agriculture overall. 
POOR
Poor households with land certificates are similar to the overall sample across all land tenure indicators, 
as well as many governance indicators. Poor households with certificates in the Land Tenure group are 
more likely than the sample as a whole to say that land management decisions are transparent. 
There are some differences in indicators related to vulnerable groups in the community. Land Tenure 
elderly households with certificates are more likely to say that households of a different tribe than the 
headman were disadvantaged, and more likely to say that vulnerable groups in general as 
disadvantaged—but less likely to say that elderly are disadvantaged. In the Ag+Land group, households 
with certificates are more likely to believe that vulnerable groups are disadvantaged overall, as 
determined by a PCA index. 
LAND CONSTRAINED 
Land constrained households in the Land Tenure group experience the same increase in significance 
across all the tenure security indicators that the aggregate sample, female-headed households youth and 
elderly households. Land constrained households in the Land Tenure group are also more likely 
participate in land management meetings. There are also differences on perceptions of treatment of 
77
vulnerable groups. In the Land Tenure group, elderly households are more likely to say women are 
disadvantaged, and in the Ag+Land group believe that vulnerable groups in general are more 
disadvantaged.
78
ANNEX 3. AGROFORESTRY -
Follow-up sample 
Note that this Annex provides statistics for the full follow-up sample versus the household panel sample. 
Within treatment areas, the seedling and tree survival rates are low. Figure A3-1 shows seedling survival 
rates across all tree species from 2014–2016 for treatment and control households. Figure A3-2 displays 
the survival rates of each village in the sample. These low survival rates do not appear to correlate with 
the geographical location of the village, or with the location of wells or water points. It is important to 
highlight that the entire Chipata district experienced droughts during the 2014–2015 and 2015–2016 
growing seasons.
Across all years of the intervention, the majority of households who engaged in agroforestry report that 
less than 50% of their Musangu seedlings survived, as shown in Figure A3-3. However, treatment 
households have slightly higher survival rates than control households. Since the overall sample size of 
control households who planted seedlings is low, the percentages may be misleading. In 2014, the first 
year of the COMACO program, just under a quarter of households in treatment communities (23%, 
FIGURE A3-1: SEEDLING SURVIVAL 
RATES BY TREATMENT GROUP
FIGURE A3-2: SEEDLING SURVIVORSHIP BY 
VILLAGE
79
N=59) report that between 76–100% of seedlings are alive today. This is slightly higher than households 
in control communities (18%, N=24). Nine percent (N=23) of households in treatment communities 
have a zero percent survival rate, compared to 16% (N=21) of control communities. 
Survival rates for Musangu trees planted in 2015, the second year of the program, are slightly lower than 
they were for the first year of the program. In the second year of the program, COMACO discontinued 
extension services and only provided seeds, which may explain the lower survival rate. Only 19%
(N=37) of treatment households have a survival rate above 75%. Still, this is higher than control 
households, where only three households have a survival rate that high. It is also important to note that 
the 2014–2015 and 2015–2016 growing seasons were years with historically bad rainfalls and low crop 
yields. 
In 2016, after COMACO ended agroforestry support in treatment villages, survival rates for Musangu 
trees fell below the survival rates of trees in control villages. Only 8% of treatment households report a 
survival rate higher than 75%, (N=6), the same number of households as in the control group (30%, 
N=6). Table A3-1 shows the average household survival rate for Musangu trees in 2014–2016 by 
treatment group. 
FIGURE A3-3: MUSANGU SURVIVORSHIP BY YEAR
80
TABLE A3-1: MUSANGU TREE SURVIVAL RATES
Musangu Trees Ag Ag+LT Control
Percentage of trees planted in 2014 that are alive now
0% 9% (8) 10% (15) 26% (6)
1–25% 37% (35) 24% (38) 22% (5)
26–50% 18% (17) 23% (36) 30% (7)
51–75% 11% (10) 14% (22) 0% (0)
76–100% 19% (18) 26% (41) 17% (4)
Percentage of trees planted in 2015 that are alive now
0% 8% (7) 10% (11) 23% (6)
1–25% 35% (29) 25% (28) 31% (8)
26–50% 14% (12) 26% (29) 15% (4)
51–75% 18% (15) 14% (15) 15% (4)
76–100% 20% (17) 18% (20) 12% (3)
Percentage of trees planted in 2016 that are alive now
0% 3% (1) 22% (9) 10% (2)
1–25% 31% (11) 22% (9) 25% (11)
26–50% 20% (7) 15% (6) 0% (0)
51–75% 31% (11) 24% (10) 30% (6)
76–100% 11% (4) 5% (2) 30% (6)
Survival rates for Gliricidia trees are slightly more optimistic, though the majority of households still 
report survival rates under 50%, as shown in Figure A3-4. Of the seedlings planted in 2014, 27% (N=65) 
of treatment households say that more than 75% of them are alive today, compared to only two of 
control households. The figures are nearly identical for seedlings that treatment households planted in 
2015 (26%, N=50) and 2016 (26%, N=31). The survival rate of control households increased to 28%
(N=12) in 2015, and to 40% (N=17) in 2016, surpassing the survival rate of the treatment households. 
Table A3-2 shows the average household survival rate for Gliricidia trees in 2014–2016 by treatment 
group. 
FIGURE A3-4: GLIRICIDIA SURVIVORSHIP BY YEAR
81
TABLE A3-2: GLIRICIDIA TREE SURVIVAL RATES
Gliricidia Trees Ag Ag+LT Control
Percentage of trees planted in 2014 that are alive now
0% 12% (14) 12% (15) 36% (5)
1–25% 23% (27) 20% (26) 14% (2)
26–50% 19% (22) 21% (27) 21% (3)
51–75% 17% (20) 15% (19) 14% (2)
76–100% 26% (30) 28% (35) 14% (2)
Percentage of trees planted in 2015 that are alive now
0% 9% (9) 11% (10) 19% (8)
1–25% 26% (26) 23% (21) 9% (4)
26–50% 17% (17) 18% (17) 21% (9)
51–75% 17% (17) 16% (14) 21% (9)
76–100% 28% (28) 24% (22) 28% (12)
Percentage of trees planted in 2016 that are alive now
0% 4% (2) 13% (9) 12% (5)
1–25% 20% (10) 20% (14) 7% (3)
26–50% 16% (8) 10% (7) 19% (8)
51–75% 22% (11) 23% (16) 17% (7)
76–100% 31% (15) 23% (16) 40% (17)
Overall, seedling survival for Musangu and Gliricidia was the lowest in 2016. This suggests that the 
support provided by COMACO during the agroforestry program was critical to the survival of seedlings, 
and withdrawing this support has had negative impacts on seedling survival rates.
DISCUSSION 
This section explores the factors driving program participation, expected benefits and the main 
challenges to agroforestry, in order to better understand uptake and seedling survival. 
PARTICIPATION
In communities where the agroforestry extension was offered, roughly a third of households had at least 
one household member participate in the program. Male-headed households are 7% more likely than 
female-headed households to have had a member participate. The benefits of agroforestry are well￾known to households in Chipata district. Focus group participants easily articulated the benefits, even if 
they did not participate in the agroforestry extension intervention. 
An Agroforestry participant in Mnukwa chiefdom described the benefits this way: “Growing agroforestry 
in our fields will help us save tremendously on the purchases of fertilizers. At least we will have money 
to spend on other necessities, like our children’s education… It will also help us maintain the fertility in 
our fields, rather than depending on these fertilizers that even end up destroying the fertility of the soils 
in our fields.”
A female FGD participant in Maguya chiefdom described how she expects to benefit from agroforestry. 
“Growing little food has troubled us here in this village. If you don’t have fertilizer and at home you have 
a big family including grandchildren and the field is not productive so to hear that there are trees that 
can make soil fertile… so when fertility returns to the soil, people can grow enough food for their 
82
families. That is why we want agroforestry trees. Most of us can’t afford to buy fertilizer. So we want 
fertility to return to the soil so that we can reduce hunger in our homes.”
Table A3-3 illustrates the many reasons households chose to participate in the Agroforestry 
intervention. The primary reason households participate is to reap the agricultural benefits from 
agroforestry (76%, N=1,117), such as improving soil fertility and reducing the need for fertilizer. Other 
common reasons included wanting to learn new farming techniques (37%, N=547) and wanting advice 
about agroforestry (36%, N=529). Other draws to the program were receiving free inputs (9%, N=132), 
and being encouraged by the headman to participate (3%, N=9). A small number of households also 
participated in the program because they wanted fuelwood (2%, N=27). 
TABLE A3-3: WHY PARTICIPATE IN COMACO PROGRAM
Reasons for participating Ag Ag+LT
Want to higher soil fertility and other benefits from 
agroforestry 
77% (288) 76% (349)
Want free inputs 7% (25) 10% (44)
Want advice about agroforestry 33% (125) 40% (181)
Want to learn about new farming techniques 37% (136) 36% (165)
My friend/neighbor/relative urged me to participate <1% (2) 2% (7)
The headman urged me to participate 5% (2) 2% (9)
Want a guaranteed buyer/better price for my crops 1% (4) 1% (4)
Want fuelwood <1% (2) 3% (12)
Want fodder for livestock 0% (0) 0% (0)
Want to benefit from fuel-efficient cookstoves 1% (3) <1% (2)
Table A3-4 illustrates the reasons that households did not participate in the agroforestry extension 
program. The biggest reason for nonparticipation was inability to attend the initial meeting (34%, 
N=428). This suggests that modifications to recruitment protocols, such as greater outreach before 
signups began, or allowing for multiple rounds of program signups as news of the program permeates 
the villages, might engender higher participation rates. 
According to FGDs, there was some confusion about the recruitment process. As one respondent 
described, "When COMACO first came, it appears as though they came to a group (of participants) that 
had already been formed. Yes. And the required number had already been met. It is not as though we 
did not want to participate; it is just that the required number for the group had already been met. 
Otherwise, we also wanted to participate in agroforestry."
A FGD participant from Maguya chiefdom who did not participate in the agroforestry extension 
program echoed this observation. “I thought that maybe this organization only comes only for those 
who are in the COMACO program—that is the reason why we didn’t take part but we have a heart to 
join.” 
Other reasons households did not participate include an overall lack of interest (5%, N=62), a desire to 
see if other households were successful before trying agroforestry (4%, N=56), and because 
agroforestry requires too much labor for the household (3%, N=43). Not being allowed to plant trees is 
almost never a reason households do not participate (1%, N=15), suggesting that for most households, 
there are no rules prohibiting them from planting trees.
83
In the FGDs, a lack of incentive to counteract the high labor and time costs is a consistent theme for 
households who chose not to participate. They would prefer a program that provided inputs that 
materialized in the short term, such as seeds, to the long-term benefits associated with agroforestry. 
This is especially true in areas that currently have virgin land or relatively fertile soil. The two quotes 
below illustrate this sentiment:
“The reason why I did not take part is that doesn’t this project go with other additional crops such as 
beans and cowpeas? So you find that you as a farmer can’t start with planting trees without first planting 
crops because as a farmer you first need a seed that will lift you out of your problems” (Mshawa, 
Agroforestry non-respondent).
“How can a program be only about trees! It takes a lot of years before you can cut a ‘beam’ out of 
it?...We would be glad if this program will go with seeds that we plant in the fields. These same seeds 
will also be planted in the same field but now if we have only planted trees, what benefit will we achieve 
if there is no seed to plant in it? Then poverty will continue but trees and seeds ought to go together so 
that we can also sell” (Mshawa, male Agroforestry non-participant).
TABLE A3-4: WHY NOT PARTICIPATE
Reasons for not participating Ag Ag+LT
Wanted to participate but failed to attend the 
meetings 40% (80) 33% (79)
Not interested in trying a new farming method 7% (14) 5% (11)
Wanted to see if other households had success 
before trying 6% (12) 6% (13)
Required too much labor 6% (12) 5% (10)
No benefits to agroforestry 3% (6) 3% (7)
Not allowed to plant trees on my land <1% (1) 1% (4)
AGROFORESTRY UPTAKE 
Households in Agroforestry treatment villages (54%, N=782) are more likely to have engaged in 
agroforestry than control households (24%, N=156). Agroforestry + Land treatment households have 
the highest participation rate (56%, N=425). Agroforestry households have a participation rate of 52%
(N=357).
The most common tree species planted by households in the treatment group is the Musangu tree, as 
shown Table A3-5. It can be grown amidst any crop, but COMACO encouraged farmers to plant their 
fields where Musangu seedlings were being grown with low-growing crops such as groundnuts, to 
ensure that the seedlings would get enough sunlight. Households were provided with 25 Musangu 
seedlings. A third of households (33%, N=227) receiving the Agroforestry treatment and (42%, N=315) 
of households receiving the Agroforestry + Land Tenure treatment planted Musangu trees. Uptake for 
the control group is substantially lower (14%, N=93). 
The second most common species of agroforestry tree is Gliricidia. Gliricidia is also grown from a seedling 
and intercropped. Households struggled to keep the seedlings alive in the nursery, as Gliricidia is 
particularly sensitive to water shortages. This may explain the slightly lower rates of Gliricidia adoption, 
though households who participated in the extension program received both species. Households in the 
84
treatment group were provided with 100 Gliricidia seedlings and have a rate of adoption slightly under 
40% (288+270). The rate for control villages is significantly lower (14%, N=93). Female headed 
households in treatment communities are slightly less likely to have planted Gliricidia trees than male 
headed households (FHH: 35%, MHH: 40%). Based on Monitoring and Evaluation (M&E) conversations 
with the COMACO team, female-headed households were more likely to struggle to transport their 
seedlings from the nursery to their fields, and the large number of Gliricidia seedlings (100 Gliricidia vs 25 
Musangu) may have been too much of an obstacle for female-headed households. 
The third species of tree provided by the program was pigeon peas, though the provision of these seeds 
does not appear to have been uniform. The original design called for 500g of pigeon pea seeds to be 
distributed to households, which would be planted directly in their fields. Pigeon peas grow into a bush 
that produces an edible fruit (the pigeon pea), and are highly desirable for consumption. Despite their 
desirability, less than 5% of treatment households planted pigeon peas.33
TABLE A3-5: HH TREE SPECIES PLANTED
Species Ag (N=680) Ag+LT (N=759) Control (N=582)
Musangu 33% (227) 42% (315) 14% (93)
Gliricidia 40% (270) 38% (288) 14% (93)
Pigeon Peas 5% (34) 3% (19) 1% (6)
EXPECTED BENEFITS FROM AGROFORESTRY UPTAKE
Households have a clear understanding of the expected benefits for agroforestry trees. The most 
common expected benefit cited by treatment groups both now and in the future is improved soil 
fertility (36%, N=339 for Musangu; 38%, N=364 for Gliricidia). The number of households who expect to 
see that benefit in the future is double the number of households who currently see the benefit (67%, 
N=636 for Musangu; 66%, N=608 for Gliricidia), which seems to indicate households understand that the 
benefits of agroforestry accrue in the future. 
Similarly, improved crop growth around trees is another benefit cited for both species, particularly in 
the future. Between 17% and 21% of households say they expect to see improved crop growth in the 
future for Musangu trees, as do 15–19% of households for Gliricidia. This is a lower percentage than we 
would expect, since higher yields and/or improved crop growth should be the ultimate benefit to 
households. Also worth noting, no households believed that either agroforestry species would reduce 
their need for chemical fertilizer, although reducing chemical fertilizer use is a major objective of 
climate-smart agriculture and conservation farming. 
A benefit emphasized in focus group discussions is the increased availability of fuel wood. It appears that 
at least some households view their agroforestry trees as an opportunity for a woodlot as opposed to 
caring for their trees to improve their field’s agricultural productivity. This is further supported by the 
number of households who do not believe that there are any yield-related benefits to the trees. The 
household statistics are less striking than the qualitative analysis, but are worth noting—8% (N=71) of 
households believe their Musangu trees will increase the availability of fuelwood in the future. This is 
 33 COMACO did not distribute pigeon peas in all villages due to a lack of seeds.
85
highest in the Agroforestry group (9%, N=21). The percent of households who believe that Gliricidia
trees will increase the availability of fuelwood in the future is slightly higher (10%, N=88).
As Table A3-6 below illustrates, agroforestry adoption is not associated with greater perceived tenure 
security. Almost no households believe that planting agroforestry trees reduce the fear of their land 
being taken, or raise the value of their land for collateral, either now or in the future. This coincides 
with the regression results presented in Section 5 that show no significant difference between the Land 
Tenure and Agroforestry + Land Tenure findings for perceived tenure security.
TABLE A3-6: MUSANGU AND GLIRICIDIA TREE BENEFITS 
Benefits Present Future Present Future
Musangu trees Ag (N=227) Ag+LT (N=315)
No benefits 38% (87) 14% (32) 44% (141) 15% (47)
Improved soil fertility 37% (83) 65% (148) 32% (102) 67% (212)
Improved crop growth around trees 11% (25) 18% (40) 8% (25) 21% (67)
Higher overall crop yield 7% (15) 22% (50) 6% (18) 23% (72)
Increased fuel wood availability 3% (7) 9% (21) 2% (6) 7% (21)
Reduced labor time on weeding 1% (2) 2% (5) 1% (2) 1% (4)
Reduced weeds 0% (0) 1% (2) 1% (3) 4% (11)
Reduced fear of land being taken 0% (0) <1% (1) 0% (0) <1% (1)
Reduced need for fertilizer 0% (0) 0% (0) 0% (0) 0% (0)
Raised value of the land for collateral 0% (0) 0% (0) 0% (0) 0% (0)
Gliricidia trees Ag (N=270) Ag+LT (N=288)
Improved soil fertility 42% (112) 65% (175) 38% (108) 63% (181)
No benefits 37% (100) 15% (40) 43% (124) 18% (51)
Improved crop growth around trees 12% (32) 17% (45) 13% (38) 19% (54)
Higher overall crop yield 9% (23) 24% (65) 9% (27) 25% (71)
Increased fuel wood availability 3% (7) 9% (25) 4% (11) 9% (27)
Reduced labor time on weeding <1% (1) 2% (4) 1% (4) 5% (13)
Reduced weeds <1% (1) 1% (4) 1% (2) 5% (5)
Reduced fear of land being taken <1% (1) 1% (2) <1% (1) 1% (2)
Raised value of the land for collateral <1% (1) 0% (0) 0% (0) 0% (0)
Reduced need for fertilizer 0% (0) 0% (0) 0% (0) 0% (0)
Taken together, these findings suggest that the program helped households understand the major 
ecological benefits of agroforestry trees, as well as the fact that these benefits will likely manifest in the 
future as opposed to the present.
Despite understanding the benefits, households seem skeptical that the benefits will actually materialize 
for them, either now or in the next 3–5 years. At present, a quarter of treatment households see no 
benefits to their agroforestry trees (24%, N=228 for Musangu; 24%, N=224 for Gliricidia), though as 
expected, this number drops substantially when households are asked about benefits they expect in the 
next 3–5 years (8%, N=79 for Musangu; 9%, N=91 for Gliricidia). Still, the number of people who do not 
expect any benefits from their agroforestry trees seems high, considering the amount of time and labor 
households put into their trees, and raises the question of why these households are engaging in 
86
agroforestry at all. There is nothing definitive in the data to answer this question, but focus group 
responses do allude to households joining the program from a sense of obligation.
CHALLENGES TO SURVIVAL 
Table A3-7 shows the types of challenges to agroforestry tree survival that households encountered. 
The most common challenges to agroforestry seedling survival, for both Gliricidia and Musangu, include 
a lack of water for seedlings, fires burning trees, pests killing the trees and animals grazing in the field.34
Given the drought conditions during implementation, lack of water for seedlings was the number one 
challenge for both Musangu and Gliricidia seedlings, identified by 37% (N=201) of treatment households 
for Musangu trees, and 23% (N=209) of treatment households for Gliricidia trees. 14% of treatment 
households (N=114) say that their nursery never had enough water, and 12% (N=97) say their nursery 
only had enough water part of the year.35 COMACO attempted to address this challenge by installing 
wells in villages with significant water scarcity, however, it does not appear that the wells improved 
seedling survival in subsequent years. 
FGDs tell a similar story. A participant in the agroforestry extension from Mnukwa chiefdom 
commented: “The challenge we are facing is lack of water. Even if they brought more seed for us to 
plant, once the rains stop, we work in vain. So we don’t know how you are going to help us, once you 
help us with water, aah we will have trees all over here." 
Traditional land management practices also present a challenge to the survival of agroforestry trees. In 
particular, animals grazing trees, which occurs when animals are allowed to graze anywhere in the fields 
after crops have been harvested, was cited as a challenge to growing Musangu by 7% of treatment 
households (N=68). Another traditional practice, setting fires to fields to clear crops after harvest or to 
harvest mice to eat, resulted in 13% of treatment households (N=72) reporting that their Musangu 
seedlings were burned by fires. 
Pests killing and damaging trees was also a challenge for households growing both Musangu and Gliricidia , 
especially in the control areas. The fact that treatment communities were slightly less likely to have their 
trees killed by pests may speak to the success of COMACO’s training about how to protect their trees. 
Finally, it is worth noting that lack of knowledge was not a prevalent challenge for households. 
Additional agroforestry trainings in the future would likely have limited impact on uptake or 
survivorship. 
 34 The greatest challenge facing households planting pigeon peas is not lack of water, but animals grazing in the field, followed by pests killing or 
damaging the plants. Lack of water is the third most common challenge. 
35 Thirty-one percent (N=456) of household respondents did not know if their nurseries had water during the year.
87
TABLE A3-7: AGROFORESTRY CHALLENGES
Musangu Gliricidia
Ag Ag+LT Control Ag Ag+LT Control
Lack of water 39% (89) 36% (112) 28% (26) 39% (104) 36% (105) 31% (30)
Pests killed/damaged 
trees 22% (51) 27% (85) 38% (35) 23% (62) 23% (65) 26% (25)
Animals grazed trees 15% (34) 11% (34) 8% (7) 15% (43) 10% (31) 12% (12)
Fires burned the 
trees 11% (24) 15% (48) 15% (14) 14% (38) 11% (33) 16% (16)
Trees were chopped 
down by mistake 6% (13) 8% (26) 12% (11) 6% (16) 5% (15) 6% (6)
Lack of knowledge 5% (11) 4% (11) 9% (8) 6% (15) 5% (15) 11% (11)
Difficulty transporting 
seedlings to field 1% (2) 1% (4) 0% (0) 3% (8) 3% (6) 2% (2)
Did not see benefits 
so stopped caring for 
trees 
1% (2) <1% (1) 2% (2) 0% (0) 1% (3) 1% (1)
Not enough labor to 
care for trees 1% (3) 2% (7) 2% (2) 1% (3) 2% (7) 0% (0)
Could not access 
seeds <1% (2) <1% (2) 1% (1) 2% (5) 1% (4) 1% (1)
VILLAGE LAND MANAGEMENT PRACTICES 
Land management rules in a village play an important role in the success or failure of agroforestry. Based 
on the key informant YGL responses, the agroforestry program had a limited effect on the existence or 
adoption of village rules related to land management. The descriptive results do not indicate a 
substantive difference in rule adoption or the types of rules between treatment and control 
communities. 
Rules about setting fires and grazing livestock are particularly important, since seedlings are susceptible 
to being grazed or trampled by livestock or burned by fires, as discussed above. Half of villages have a 
rule that forbids lighting fires in fields at any point in the year (53%, N=146). The other half of villages 
either have no rule about lighting fires, or allow fires after the crops are harvested. Figure A3-5 shows 
seedling survival in villages that have a rule that forbids fires on fields at any time in the year. The results 
show a potential association between the presence of rules and higher survival rates for agroforestry 
trees.
Grazing rules are another possible means for increasing seedling survival. Traditionally, animals are 
allowed to graze in fields after crops are harvested, which leaves seedlings vulnerable to being grazed or 
trampled by cattle, goats and other livestock. We examined the seedling survival rates for villages that 
forbid livestock grazing on fields, or allow it only if the livestock are accompanied (68%, N=187), and 
compared to villages that either have no rule about grazing or allow livestock to graze unaccompanied 
on fields (32%, N=86). Figure A3-6 shows the results of the comparison. There does not appear to be as 
strong of an association between grazing rules and seedling survival, compared to the rules about 
starting fires. One reason that grazing rules may not have the same effect on survival as fires is that even 
if supervised, it is not always possible to control livestock, and seedlings may still be grazed or trampled 
even if the rule is being obeyed. Another possible explanation is a lower rate of adherence to rules 
about grazing than rules about fire, as reported by households. Sixty-two percent of households 
88
(N=2,123) say that households do not follow rules about grazing, compared to 57% of households 
(N=1,952) who say households do not follow rules about fires.
Focus group discussions further highlight the importance of good grazing practices. A participant from 
Mnukwa described the challenge this way: “We will first need to have proper prevention plan from 
livestock. Sometimes we are not around, we may be attending a funeral at a neighboring village, so if we 
do not secure these trees from livestock then there is nothing that will happen. The chief need to teach 
the headmen how to look after livestock and the headmen should in turn teach their subjects, because if 
these livestock are left like this destroying plants… we need strict rules on livestock, if not it will 
continue to be a serious problem."
FIGURE A3-5: SURVIVORSHIP RATE & FIRE 
RULES BY VILLAGE
FIGURE A3-6: SURVIVORSHIP RATE & 
GRAZING RULES BY VILLAGE
89
ANNEX 4. MULTIPLE TEST 
CORRECTION
To examine the robustness of the impact estimates, a ‘false discovery rate’ adjustment was used, to 
correct p-values from each test for the fact that multiple tests were run within each outcome family and 
across subgroups (Benjamini and Hockberg, 2000). Given the number of tests that were run, some 
portion of the significant results obtained would be expected to be simply due to chance. Put differently, 
the more tests that are run, the higher the likelihood that some of them will come back significant, but 
some of these are likely to be false positives. Results that maintained their significance even after the p￾values were adjusted via the ‘false discovery rate’ correction are considered highly robust.
Using multiple test correction on a subset of key indicators reveals that the three tenure security 
indexes for long term, short term and overall perceptions of tenure security are significant at the 5%
level for households in the land tenure group. 
90
ANNEX 5. SUPPLEMENTAL 
SUMMARY STATISTICS AND 
REGRESSION TABLES
For the supplemental summary statistics and regressions tables, please see the Annex 5 folder included 
in the zipped follow-up report package. The tables included in the Annex 5 folder are:
Household Panel Summary Statistics
Household Panel and Subgroup Regressions
91
ANNEX 6. SUPPLEMENTAL 
EVALUATION 
DOCUMENTATION
The following documentation and datasets for the TGCC Impact Evaluation (IE) in Zambia can found on 
LandLinks “https://land-links.org/evaluation/tenure-global-climate-change-tgcc-zambia/” 
1. TGCC IE Design Report
2. TGCC IE Baseline Survey Instruments
3. TGCC IE Baseline Household Dataset
4. TGCC IE Baseline Headperson Dataset
5. TGCC IE Baseline Leader Geospatial Dataset
6. TGCC IE Baseline Household Codebook
7. TGCC IE Baseline Headperson Codebook
8. TGCC IE Baseline Analysis Report
9. TGCC IE Endline Pre-analysis Plan 
The following documentation and datasets will be forthcoming through the same URL
1. TGCC IE Endline Household Dataset
2. TGCC IE Endline Headperson Dataset
3. TGCC IE Endline Yield Group Leader (YGL) Dataset
4. TGCC IE Endline Village Land Committee (VLC) Dataset
5. TGCC IE Endline Leader Geospatial Dataset
6. TGCC IE Endline Household Codebook
7. TGCC IE Endline Headperson Codebook
8. TGCC IE Endline Yield Group Leader (YGL) Codebook
9. TGCC IE Endline Village Land Committee (VLC) Codebook
92
ANNEX 7. ATTRITION TABLE
TABLE A7-1: MUSANGU TREE SURVIVAL RATES
Variable
Control 
Mean
Agro -
Control
Land 
Tenure -
Control
Agro X 
Land 
Tenure -
Control
Saili -
Control P
tensec_index_hh_pca 0.22 0.02 -0.03 -0.04 -0.04 0.913
tensec_index_hh_long_pca 0.20 0.02 -0.01 -0.03 -0.02 0.977
tensec_index_hh_short_pca 0.23 0.01 -0.05 -0.04 -0.05 0.790
tensec_internal_pca 0.23 -0.03 -0.04 -0.08 -0.06 0.804
tensec_external_pca 0.18 0.06 -0.01 0.02 -0.01 0.828
dispute 0.24 0.01 -0.01 -0.00 0.00 0.937
landgov_overall_pca 0.14 0.05 -0.04 0.09 0.07 0.243
leadsatindex_pca 0.20 0.14 0.04 0.13 0.13 0.090
agroinv_csa 0.47 -0.01 -0.01 -0.02 -0.01 0.665
agroinv_aginvest_short 0.37 -0.01 -0.00 -0.00 0.02 0.244
agroinv_aginvest_long 0.45 -0.03 -0.03 -0.02 0.03 0.000
fallow_seasons_log_min 2.22 0.03 0.08 0.17 -0.20 0.310
agforben_4 0.13 -0.02 0.02 0.10 0.09 0.264
score_assetindex1 0.12 0.10 0.04 0.00 -0.30 0.000
attrit 0.13 0.05 0.03 0.03 0.05 0.200
93
ANNEX 8. DISCLOSURE OF 
ANY CONFLICTS OF 
INTEREST
There are no conflicts of interest in relation to the TGCC IE Zambia.
94
REFERENCES
Ajayi, O. C. (2007). User acceptability of sustainable soil fertility technologies: Lessons from farmers’ 
knowledge, attitude and practice in Southern Africa. Journal of Sustainable Agriculture, 30(3), 21–
40. http://doi.org/10.1300/J064v30n03_04
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful 
approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological), 289–
300. 
Benjamini, Y., & Hochberg, Y. (2000). On the adaptive control of the false discovery rate in multiple 
testing with independent statistics. Journal of Educational and Behavioral Statistics, 25, 60–83. 
http://doi.org/10.3102/10769986025001060.
Besley, T. (1995). Property rights and investment incentives: Theory and evidence from Ghana. The 
Journal of Political Economy, 103(5), 903–937.
Deininger, K., & Chamorro, J. S. (2004). Investment and equity effects of land regularisation: The case of 
Nicaragua. Agricultural Economics, (30), 101–116.
Deininger, K., & Jin, S. (2006). Tenure security and land-related investment: Evidence from Ethiopia. 
European Economic Review, 50, 1245–1277. http://doi.org/10.1016/j.euroecorev.2005.02.001
Deininger, K., Ali, D. A., & Alemu, T. (2011). impacts of land certification on tenure security, investment, 
and land market participation: Evidence from Ethiopia. Land Economics, 87(2), 312–334. 
Retrieved from http://le.uwpress.org/content/87/2/312.abstract
Deininger, K., Byerlee, D., Lindsay, J., Norton, A., Selod, H., & Stickler, M. (2011). Rising global interest in 
farmland: Can it yield sustainable and equitable benefits? (p. 264). The World Bank. 
http://doi.org/10.1596/978-0-8213-8591-3
Feder, G., Onchan, T., Chalamwong, Y., & Hongladarom, C. (1988). Land policies and farm productivity in 
Thailand. Baltimore, MD: Johns Hopkins University Press.
Franzel, S., Coe, R., Cooper, P., Place, F., & Scherr, S. J. (2001). Assessing the adoption potential of 
agroforestry practices in sub-Saharan Africa. Agricultural Systems, 69, 37–62. 
http://doi.org/10.1016/S0308-521X(01)00017-8
Gelman, A., Hill, J. & Yajima, M. (2012). Why we (usually) don’t have to worry about multiple 
comparisons. Journal of Research on Educational Effectiveness, 5, 189–211. 
http://doi.org/10.1080/19345747.2011.618213
Goldstein, M., et al. (2018). Land tenure security for men and women: Evidence from a randomized 
controlled trial in Uganda. AEA RCT Registry. April 22.
Holden, S. T., Deininger, K., & Ghebru, H. (2009). Impacts of low-cost land certification on investment 
and productivity. American Journal of Agricultural Economics, 91(2), 359–373.
95
Huntington, H., Stickler, M., & Stevens, C. (2018). Customary land registration: A pathway to agricultural 
transformation and inclusive economic growth for rural Zambians? (Working paper).
Jack, B. K., Oliva, P., Bell, S., Severen, C., & Walker, E. (2016). Technology adoption under uncertainty: Take 
up and subsequent investment in Zambia (Working paper).
Jacoby, H. G., Li, G., & Rozelle, S. (2002). Hazards of expropriation: Tenure insecurity and investment in 
rural China . American Economic Review, 92(5), 1420–1447.
Jayne, T. S., Zulu, B., Kajoba, G., & Weber, M. T. (2009). Access to land and poverty reduction in rural 
Zambia: Connecting the policy issues. Retrieved from http://purl.umn.edu/55054
Joireman, S. F. (2008). The Mystery of capital formation in Sub-Saharan Africa: Women, property rights 
and customary law. World Development, 36(7), 1233–1246.
Kiptot, E., Franzel, S., & Degrande, A. (2014). Gender, agroforestry and food security in Africa. Science 
Direct: Current Opinion in Environmental Sustainability, 6, 104–109. 
http://doi.org/10.1016/j.cosust.2013.10.019
Lawry, S., Samii, C., Hall, R., Leopold, A., Hornby, D., & Mtero, F. (2014). The impact of land property 
rights interventions on investment and agricultural productivity in developing countries. Campbell 
Systematic Reviews. https://doi.org/10.4073/csr.2014.1
Mbow, C., Van Noordwijk, M., Luedeling, E., Neufeldt, H., Minang, P. a, & Kowero, G. (2014). 
Agroforestry solutions to address food security and climate change challenges in Africa. Science 
Direct: Current Opinion in Environmental Sustainability, 6, 61–67. 
http://doi.org/10.1016/j.cosust.2013.10.014
Mercer, D. E. (2004). Adoption of agroforestry innovations in the tropics: A review. Agroforestry Systems. 
http://doi.org/10.1023/B:AGFO.0000029007.85754.70
Pattanayak, S. K., Mercer, D. E., Sills, E., & Yang, J. (2003). Taking stock of agroforestry adoption studies.
Agroforestry Systems, 57, 173–186. http://doi.org/10.1200/JCO.2003.11.022
Persha, L., & Huntington, H. (2016). To what extent does tenure security shape agroforestry investment 
and impacts on agriculture and livelihoods in strong customary land systems? Evidence from 
early agroforestry adopters in Zambia’s Eastern Province. In 2016 World Bank Conference on 
Land and Poverty. Washington, D.C.
Phiri, D., Franzel, S., Mafongoya, P., Jere, I., Katanga, R., & Phiri, S. (2004). Who is using the new 
technology? The association of wealth status and gender with the planting of improved tree 
fallows in Eastern Province, Zambia. Agricultural Systems, 79, 131–144. 
http://doi.org/10.1016/S0308-521X(03)00055-6
Place, F. (2009). Land Tenure and agricultural productivity in Africa: A comparative analysis of the 
economics literature and recent policy strategies and reforms. World Development, 37, 1326–
1336. http://doi.org/10.1016/j.worlddev.2008.08.020
Place, F., & Otsuka, K. (2001). Tenure, agricultural investment, and productivity in the customary tenure 
sector of Malawi. Economic Development and Cultural Change. http://doi.org/10.1086/321918
96
Rozelle, S., & Swinnen, J. F. M. (2004). Success and failure of reform: Insights from the transition of 
agriculture. Journal of Economic Literature, 42(2), 404–456.
Sirrine, D., Shennan, C., & Sirrine, J. R. (2010). Comparing agroforestry systems’ ex ante adoption 
potential and ex post adoption: on-farm participatory research from southern Malawi. 
Agroforestry Systems. http://doi.org/10.1007/s10457-010-9304-0
Sitko, N. J., Chapoto, A., Kabwe, S., Tembo, S., Hichaambwa, M., Lubinda, R., … Nthani, D. (2011). 
Technical compendium: Descriptive agricultural statistics and analysis for Zambia in support of the 
USAID Mission’s Feed the Future strategic review (Food Security Collaborative Working Paper No. 
104016). Michigan State University, Department of Agricultural, Food, and Resource Economics.
Smith, R. E. (2004). Land tenure, fixed investment, and farm productivity: Evidence from Zambia’s 
Southern Province. World Development, 32, 1641–1661. 
http://doi.org/10.1016/j.worlddev.2004.05.006
Tetra Tech. (2014). Zambia Tenure and Climate Change Scoping Report: Opportunities for land tenure and 
agroforestry intervention.
97
U.S. Agency for International Development
1300 Pennsylvania Avenue, NW
Washington, D.C. 20523
Tel: (202) 712-0000
Fax: (202) 216-3524
www.usaid.gov