PREG
ENDLINE REPORT
JUNE 2022
PARTNERSHIP FOR 
RESILIENCE AND 
ECONOMIC GROWTH IN 
NORTHERN KENYA
ii
ABOUT THE REAL AWARD
The Resilience Evaluation, Analysis and Learning (REAL) Associate Award is a consortium-led 
effort funded by the USAID Center for Resilience. It was established to respond to growing 
demand among USAID Missions, host governments, implementing organizations, and other key 
stakeholders for rigorous, yet practical, monitoring, evaluation, strategic analysis, and capacity 
building support. Led by Save the Children, REAL draws on the expertise of its partners: Food for 
the Hungry, Mercy Corps, and TANGO International.
Prepared for
United States Agency for International Development’s Center for Resilience.
Recommended citation 
TANGO International. 2022. Partnership for Resilience and Economic Growth (PREG) Initiative in 
Northern Kenya Endline Report. Washington, DC: Resilience, Evaluation, Analysis and Learning 
(REAL) Associate Award. 
Disclaimer 
This report is made possible by the generous support and contribution of the American people 
through the United States Agency for International Development (USAID). The contents of the 
materials produced through the REAL Award do not necessarily reflect the views of USAID or the 
United States Government.
Acknowledgements
This publication was prepared by Maryada Vallet, Karyn Fox, and Tim Frankenberger for the 
REAL Associate Award. 
Cover photo credit
Maryada Vallet for TANGO International 
Contact Information
The REAL Associate Award
c/o Save the Children
899 North Capitol Street NE
Washington, DC 20002
Email: real@savechildren.org
TANGO International
Tim Frankenberger
406 South Fourth Avenue
Tucson, AZ 85701
Email: tim@tangointernational.com 
PREG Initiative in Northern Kenya Endline Report | i
CONTENTS
CONTENTS ............................................................................................................................................................... I
LIST OF TABLES ..................................................................................................................................................... III
LIST OF FIGURES ................................................................................................................................................... IV
LIST OF ACRONYMS............................................................................................................................................. V
EXECUTIVE SUMMARY ....................................................................................................................................... VI
Study Design / Methods........................................................................................................................................ vi
Key Findings........................................................................................................................................................... viii
1 INTRODUCTION ......................................................................................................................................... 1
1.1 Project Area Profiles ............................................................................................................................. 2
1.1.1 REGAL Theory of Change .................................................................................................................. 3
1.1.2 REGAL-IR................................................................................................................................................. 4
1.1.3 REGAL-AG .............................................................................................................................................. 4
1.2 Description of the PREG Project Context ....................................................................................... 5
1.3 Objective of the Study ........................................................................................................................... 8
2 METHODOLOGY.......................................................................................................................................... 9
2.1 Quantitative Survey ............................................................................................................................... 9
2.1.1 Sample Design ........................................................................................................................................ 9
2.1.2 Resilience Information Collected..................................................................................................... 11
2.1.3 Field Procedures................................................................................................................................... 15
2.1.4 Data Analysis....................................................................................................................................... 15
2.1.5 Limitations of the Quantitative Surveys........................................................................................ 16
2.2 Qualitative Study .................................................................................................................................. 18
2.2.1 Survey Sample Design ........................................................................................................................ 18
2.2.2 Fieldwork ............................................................................................................................................... 19
2.2.3 Data Management and Security...................................................................................................... 20
2.2.4 Data Analysis....................................................................................................................................... 20
2.2.5 Limitations of the Qualitative Component................................................................................... 20
3 ANALYSIS OF FINDINGS .......................................................................................................................... 22
3.1 Participation in PREG Interventions ................................................................................................ 22
3.2 Livelihood Activities ............................................................................................................................. 23
3.2.1 Livelihood Diversification and Education ...................................................................................... 26
3.3 Coping Strategies ................................................................................................................................. 27
3.4 Resilience Capacities ........................................................................................................................... 29
3.4.1 Assets and Savings............................................................................................................................... 29
3.4.2 Livelihood Diversity............................................................................................................................. 29
3.4.3 Reliance on Relief ................................................................................................................................ 29
ii | REAL | Resilience, Evaluation, Analysis & Learning
3.4.4 Social Capital........................................................................................................................................ 29
3.4.5 Agency.................................................................................................................................................... 34
3.5 Household Well-Being Outcomes ................................................................................................... 34
3.5.1 Relationship between Household Characteristics and Subjective 
Resilience................................................................................................................................................ 35
3.6 Child Nutrition Status ......................................................................................................................... 37
4 OBSERVATIONS AND CONCLUSIONS ............................................................................................. 39
4.1 Key Findings Related to Resilience Analysis ................................................................................. 39
4.1.1 Livelihood Activities ............................................................................................................................ 39
4.1.2 Coping Strategies................................................................................................................................ 39
4.1.3 Resilience Capacities.......................................................................................................................... 40
4.1.4 Well-Being Outcomes........................................................................................................................ 41
4.2 Program Implications .......................................................................................................................... 41
5 RECOMMENDATIONS .............................................................................................................................. 43
REFERENCES .......................................................................................................................................................... 44
ANNEX 1: FOCUS GROUP TOPICAL OUTLINE ....................................................................................... 47
ANNEX 2: KEY INFORMANT INTERVIEW TOPICAL OUTLINE .......................................................... 51
ANNEX 3: FINDINGS FROM NORTHERN KENYA BASELINE AND ENDLINE SURVEYS: 
OVERLAPPING HIGH INTENSITY AND LOW INTENSITY SUBLOCATIONS ................................. 52
ANNEX 4: SAMPLE DESIGN FOR THE 2013, 2015 AND 2018 SURVEYS............................................ 58
ANNEX 5: PARTICIPATION IN PREG INTERVENTIONS BY COUNTY ............................................. 69
PREG Initiative in Northern Kenya Endline Report | iii
LIST OF TABLES
Table 1: Kenya livestock production over past ten years plus past reference year (1,000 metric 
tons) ............................................................................................................................................................................ 6
Table 2: Gross marketed production value of livestock products at current prices (KSh million), by 
year ............................................................................................................................................................................. 7
Table 3: Sample size (disaggregated by county), by survey round .......................................................... 11
Table 4: Sites visited by qualitative team ........................................................................................................ 19
Table 5: PREG-supported trainings and groups ............................................................................................ 22
Table 6: Percent of households reporting participation in project activities, HI and LI counties ..... 23
Table 7: Household engagement in various livelihood activities, by survey round ............................. 23
Table 8: Household coping strategies and asset recovery, by survey round ........................................ 27
Table 9: Use and availability of social capital, by survey round ............................................................... 31
Table 10: Household well-being outcomes, by survey round .................................................................... 34
Table 11: Resilience capacity components for three categories of subjective resilience, by survey 
round ......................................................................................................................................................................... 36
Table 12: Household engagement in livelihoods activities by subjective resilience category, by 
survey round (% households) .............................................................................................................................. 37
Table 13: Prevalence of stunting, underweight and wasting across survey rounds for children 
under 5 years of age ............................................................................................................................................. 38
Table 14. Household engagement in select livelihood activities ............................................................... 53
Table 15. Household use of coping strategies ............................................................................................... 53
Table 16. Assets, livelihood diversification, relief .......................................................................................... 54
Table 17. Use and availability of social capital ............................................................................................. 55
Table 18. Well-being indicators ......................................................................................................................... 55
Table 19. Children's anthropometric indicators ............................................................................................ 56
Table 20: 2013 survey sample ............................................................................................................................ 61
Table 21: Initial and realized 2013 survey sample ........................................................................................ 62
Table 22: 2015 survey sample ............................................................................................................................ 64
Table 23: 2018 survey sample ............................................................................................................................ 66
Table 24: Initial and realized PREG II baseline sample ................................................................................ 68
Table 25: Participation in PREG-supported trainings ................................................................................... 69
Table 26: Participation in PREG-supported groups ...................................................................................... 70
iv | REAL | Resilience, Evaluation, Analysis & Learning
LIST OF FIGURES
Figure 1: PREG I program areas in northern Kenya ...................................................................................... 1
Figure 2: REGAL Theory of Change ................................................................................................................... 3
Figure 3: 2013 Seasonal calendar ..................................................................................................................... 59
PREG Initiative in Northern Kenya Endline Report | v
LIST OF ACRONYMS
ADESO African Development Solutions
ASAL Arid and semi-arid lands
EA Enumeration Area
DID Difference in Differences
FGD Focus Group Discussion
FTF Feed the Future
GDP Gross Domestic Product
GoK Government of Kenya
HA Humanitarian Assistance
HHS Household Hunger Scale
IE Impact Evaluation
KII Key Informant Interview
KSh Kenyan Shilling
NDMA National Drought Management Authority
NGO Non-Governmental Organization
PBS Population-Based Survey
PPP Purchasing Power Parity
PREG Partnership for Resilience and Economic Growth
REAL Resilience Evaluation, Analysis and Learning
REGAL Resilience and Economic Growth in the Arid Lands
TANGO Technical Assistance to Non-Governmental Organizations
USAID United States Agency for International Development
USD United States Dollar
VSLA Village Savings and Loan Association
ZOI Zone of Influence
vi | REAL | Resilience, Evaluation, Analysis & Learning
EXECUTIVE SUMMARY
In Kenya, 23 of 47 counties are considered arid and semi-arid lands (ASALs). Many constraints 
have and continue to negatively impact the resilience, food security situation and socio-economic 
development of the ASALs of Kenya. Counties in the ASAL region have been adversely impacted 
by a legacy of marginalization by the government, drought emergencies, human conflict, and 
poverty. USAID’s Partnership for Resilience and Economic Growth (PREG) brings together 
multiple humanitarian and development partners that work with the Kenya National Drought 
Management Authority (NDMA) and county governments to coordinate resilience and economic 
growth activities to strengthen resilience among the vulnerable pastoralist communities in nine 
ASAL counties. 
As a mechanism for coordinating all resilience programming in the ASALs in northern Kenya, the 
PREG model of collaboration is grounded in sequencing, layering, and integrating interventions, 
which has enabled partners to minimize overlap, promote synergies, and achieve multi-partner 
collaboration. By 2014, USAID-supported resilience programs implemented in the ASALs in 
northern Kenya were subsumed under the PREG Partnership umbrella. This included the 
Resilience and Economic Growth in the Arid Lands (REGAL) projects: REGAL-Improving 
Resilience (REGAL-IR) and REGAL-Accelerated Growth (REGAL-AG). As a result, the PREG I 
endline survey draws heavily from the REGAL Impact Evaluation design, but with several 
significant adaptations.
The main body of this report presents findings from a pre-post analysis of PREG resilience 
programming activities for its first five-year phase (i.e., PREG I), which ostensibly represents a 
transition from REGAL to PREG. 
Study Design / Methods
As mentioned above, the PREG I study was originally designed as an impact evaluation of the 
REGAL project (i.e., REGAL-IR and REGAL-AG) in northern Kenya, which was designed to 
understand the differences in outcomes between households that received high-intensity 
(treatment group) and low-intensity (comparison group) resilience programming. Numerous 
issues emerged over the course of the transition from REGAL to PREG that made clear an impact 
evaluation—as originally designed—would not be possible; a key limitation was that some 
households within the treatment group had not, in fact, directly participated in program activities 
(i.e., had not received any treatment). Thus, PREG I endline study findings are based on a pre-post 
analysis, analyzing change over time to detect differences in key outcomes between baseline and 
endline surveys. Note that this pre-post study does not allow for attribution of benefits to PREG 
households from PREG programming. 
Data for the PREG I baseline (originally the REGAL impact evaluation baseline) were captured by 
integrating resilience modules into the population-based survey for the Feed the Future Zone of 
Influence. The PREG I endline utilizes a mixed-methods design that includes both quantitative and 
qualitative components; the baseline involved quantitative data collection only. Baseline data were 
collected in January/February 2013 and endline data in August/September 2018 in six of the nine 
PREG Initiative in Northern Kenya Endline Report | vii
counties in which PREG I operated, including Isiolo, Marsabit, Turkana, Mandera, Samburu, and 
Tana River.1 The timing of data collection for the baseline occurred post-harvest. In contrast,
data collection for the endline occurred during the lean season for pastoral areas. The timing of 
data collection should ideally be conducted at the same time of year for each survey round in 
order to more accurately assess the effects of programming (versus effects due to seasonality, 
for example) on any changes in food security that might occur over time. 
Both the baseline and endline survey samples were drawn independently and use cross-sectional 
samples of the population within the PREG area of intervention. However, the definition of the 
PREG area of intervention changed between baseline and endline. The baseline sample was 
divided into two geographic strata defined at the county level and was based on planned PREG 
programming with low-intensity (LI) counties receiving only humanitarian assistance through the 
USAID-supported PREG partnership: Baringo, Samburu, and Tana River. High-intensity (HI) 
counties were defined as areas receiving (or that would receive) PREG resilience-enhancing 
investments in addition to humanitarian assistance: Isiolo, Marsabit, and Turkana. Based on 
guidance for Feed the Future Zone of Influence surveys, the samples drawn from the counties in 
these two strata were population-based, that is, all households within the counties constituted the 
sampling frame from which surveyed households were drawn. However, midline survey data 
showed that a large number of sampled households from the HI counties (treatment group) did 
not directly participate in any resilience programming supported by USAID, so the sampling 
strategy for the endline was modified.2
The sampling strategy for the endline survey was modified from the one used at baseline in order 
to address this sampling limitation. Rather than identifying the HI sampling frame on the basis of 
county only (i.e., as in the baseline), information from project records was used to identify much 
smaller geographic areas (sublocations) in which PREG projects had been carried out or were 
currently engaged in specific interventions at the time of the PREG I endline survey. The LI endline 
sampling frame was also changed from county level to sublocations within the LI counties where 
WFP-supported humanitarian assistance (and other non-resilience-building) activities had been 
carried out. As a result, the PREG I endline is a subset of the baseline sample; the subset consists 
of overlapping sublocations. 
Because households sampled in the HI areas for the PREG I endline (i.e., at sublocations) were 
more likely to have been exposed to resilience interventions than HI sampled households for the 
PREG I baseline (i.e., at county level), the two populations are not comparable. Endline 
sublocations (at least in HI areas) reflect project selection criteria and therefore it is possible—
and likely—that there would be systematic differences between the selected sublocations and the 
county as a whole. To mitigate this bias, the findings for the PREG I endline are based on a pre￾post comparison of data from the purposively selected endline sublocations where data were also 
collected at baseline. In addition, a limited difference-in-differences impact analysis was conducted 
comparing baseline-to-endline changes in the overlapping HI sublocations with changes in the 
1 Due to security issues at baseline, data were not collected in Garissa, Mandera and Wajir counties. A mid-term survey was conducted in 
2015 to collect point estimates of the Feed the Future Zone of Influence indicators, that sample was not powered to statistically measure 
changes from the baseline survey. Therefore, the mid-term survey information is not included in this endline report. 2 Based on baseline and mid-term survey data about program participation. See Section 2.1.1 for more details. 
viii | REAL | Resilience, Evaluation, Analysis & Learning
overlapping LI sublocations. Qualitative insights help explain and provide more depth to the 
quantitative analyses.
Key Findings3
Livelihoods. Livelihood diversity was generally low and decreased slightly over time, from 
approximately two sources of income or food, to one. While livestock production remained the 
primary livelihood, qualitative focus group participants across all counties reported increased 
interest in diversifying into small-scale farming and vegetable gardening. In multiple communities, 
however, drought, lack of sufficient water resources, and flooding in some cases (e.g., Garissa), 
have often undermined these efforts. The quantitative data show household participation in 
farming, wage labor and self-employment decreased significantly over time.
Coping Strategies. At endline, roughly 8 percent of households relied on migration to cope with 
shocks/stressors, a significant increase from baseline. Reliance on selling off both large and small 
assets decreased significantly since the baseline; selling of large assets remained a coping strategy 
at endline for 30 percent of households. Neither the quantitative nor the qualitative data provide 
sufficient insights into whether households shifted to more negative coping strategies over time, 
though this often occurs in cases of recurrent shocks requiring more dire action. 
Resilience Capacities
Social capital. By endline, more households had, or thought they could, seek help from others 
than at baseline. Although relatives are still the primary sources of support in times of need, there 
was a significant increase in reaching out to non-relatives in other tribes and ethnic groups 
(p<0.05). The qualitative findings suggest this may be partly due to an erosion of bonding social 
capital (i.e., links between families, friends and neighbors within a household’s community) over 
time, driving greater reliance on geographically distant sources of social support, such as bridging 
social capital (i.e., links between family, friends and neighbors outside a households’ community) 
and linking social capital (i.e., vertical linkages to people with access to resources, such as leaders, 
political figures, business owners). The qualitative data also indicate that even wealthier 
households were greatly affected by recent shocks and incurred debt that inhibited their 
traditional role of providing loans, food and other support within their communities during times 
of need. Shop owners were less willing or able to allow people to purchase food on credit than in 
the past. This is consistent with other findings that the rate of borrowing dropped at endline.
Humanitarian assistance. Household reliance on humanitarian assistance was very low in both 
survey rounds: 15 percent at baseline and 6 percent at endline; the change over the two rounds is 
not statistically significant. The qualitative data suggest that most communities still received 
(although diminished) some form of assistance during times of crises, including food aid, cash, 
fodder, seedlings, vaccines, malaria nets, and livestock restocking. While some communities noted 
that this type of assistance was insufficient to meet the needs of households within an affected 
area (e.g., Wajir), others indicated that emergency humanitarian assistance layered with 
3 Findings are based on comparisons across time and not between HI and LO groups.
PREG Initiative in Northern Kenya Endline Report | ix
development activities was critical for dealing with shocks and stresses. A female FGD in Garissa, 
for example, described food assistance as critical to reducing “starvation” during drought.
Asset ownership. Households owned fewer assets at endline compared to baseline, but the 
decrease was not statistically significant. 
Well-being Outcomes 
Economic indicators. Household expenditures significantly decreased and poverty increased, 
although the latter change was not statistically significant.
Food security. Moderate or severe household hunger increased significantly over time. Some of 
this increase may be because the baseline data were collected in the post-harvest season (when 
we would expect hunger levels to be lower) and the endline data were collected in the lean 
season (when we would expect hunger to increase).
Child nutritional status. Generally, the trend was toward improvements in all three measures of 
child nutrition by endline, though no changes were statistically significant. 
Ability to cope (subjective resilience). The perceived ability to cope with future shocks is also 
referred to as subjective resilience. The percentage of surveyed households who report that they 
are able to cope with future shocks or times of stress decreased significantly. Reliance on relief 
assistance is lower for households with greater reported ability to cope with future shocks. 
Households that report inability to cope with future shocks relied most heavily on natural￾resource-based livelihoods (livestock and harvesting wild products) and wage labor, although at 
endline the difference with salaried labor was minimal. Those that reported ability to cope with 
future shocks with some adjustments had the most widely diversified portfolio of livelihoods, while 
the households that reported ability to cope without adjustments were more likely to focus on 
livestock livelihoods and salary work.
Summary. Most results at endline do not differ significantly from those collected at baseline. The 
lack of statistical significance overall may be related to the relatively small sample size. Insights 
from the qualitative data offer a somewhat more nuanced perspective; people are struggling but 
have some capacity—although diminished—to cope with shocks and stressors, particularly if 
afforded some external assistance. Some respondents reported improvements in livelihood 
diversification; improved nutrition among children; increased collective action and community 
cohesion, particularly as regards natural resource management; increased access to services 
(e.g., markets, veterinary and human health services) and technologies (e.g., farming practices); 
and improvements in and access to infrastructure (e.g., markets, water points). 
Such seemingly contradictory results may, in part, reflect differences in the nature of the data 
reported; the quantitative results (primarily neutral or negative) reflect the average across all 
sampled households while the qualitative data (somewhat more positive) reflect the opinions and 
perspectives of a considerably smaller sample size of diverse but purposively selected key 
informants and focus group participants. Additionally, participants in focus groups or key 
informant interviews are asked to provide their perceptions of conditions in their communities 
x | REAL | Resilience, Evaluation, Analysis & Learning
overall, whereas the information collected in quantitative surveys reflects the specific conditions of 
the respondents’ households. The important point is that both the quantitative and qualitative 
results are valid even when seemingly contradictory; one must simply keep the context for each in 
mind when interpreting results. In particular, the qualitative results represent the opinions of the 
individuals who provided them and may not be broadly applicable beyond their community, in 
contrast to quantitative results.
PREG Initiative in Northern Kenya Endline Report | 1
1 INTRODUCTION
The USAID Partnership for Resilience and Economic Growth (PREG) program was formed as a 
collaborative and coordinated effort between the USAID Kenya and East Africa missions, and 
brings together humanitarian and development partners to build the resilience of vulnerable 
pastoralist communities in 
northern Kenya. The Partnership 
targets communities in nine of the 
arid and semi-arid land (ASAL) 
counties by layering and 
sequencing multiple USAID 
activities and implementing 
partners, and working with the 
Kenya National Drought 
Management Authority (NDMA) 
and county governments to 
coordinate resilience and 
economic growth activities. 
This report presents findings from 
a study of the first five-year phase 
of PREG (PREG I) in Kenya’s 
northern ASALs. The primary aim 
of the study is to document 
changes over the five-year period 
(2013–2018) in selected household 
resilience capacities, coping 
strategies, and well-being 
outcomes, including food security 
and “subjective resilience” (i.e., 
households’ perceived ability to 
cope with future shocks) among households surveyed in Baringo, Isiolo, Marsabit, Turkana, 
Samburu, and Tana River (Figure 1).4 Well-being is measured by four indicators: 1) daily per 
capita expenditures measured in 2010 U.S. dollars (USD) after adjusting for 2005 purchasing 
power parity (PPP); 2) prevalence of poverty (the percentage of households living on less than 
$1.25 per day); 3) depth of poverty; 4) moderate to severe hunger (Household Hunger Scale); and 
5) subjective resilience.
4 Due to security issues at baseline, data were not collected in Garissa, Mandera and Wajir counties. A mid-term survey was conducted in 
2015 to collect point estimates of the Feed the Future Zone of Influence indicators, and was not powered to statistically measure changes 
from the baseline survey. Therefore, the mid-term survey information is not included in this endline report. 
Figure 1: PREG I program areas in northern Kenya
Map Source: ©Nairobi123 (June 2013) / Kenya_Updated_location_map.jpg 
/ Wikimedia Commons / CC-BY-SA-3.0 / Original map has been modified.
FFA/HA = Food for Assets/Humanitarian Assistance
REGAL-IR and REGAL-AG are resilience projects brought under the PREG 
umbrella and on which the PREG I baseline is based.
 2 | REAL | Resilience, Evaluation, Analysis & Learning
connected pastoralists and agro-pastoralists to 
markets in two of the REGAL-IR counties (Isiolo 
and Marsabit).7 Both programs were to be layered 
on top of humanitarian assistance programming 
implemented by USAID partners within the nine￾county ZOI.
The REGAL program, referred to as the 9-5-2 Club, initially involved USAID, World Food 
Programme (WFP), and REGAL (AG and IR). Its purpose was to “coordinate and harmonize 
USAID direct-funded resilience-building activities amongst key humanitarian and development 
actors in the arid and semi-arid lands in Kenya, in collaboration with the National Drought 
Management Authority and the devolved county government structures.”8 By 2014, the original 
9-5-2 Club had expanded its membership and transitioned to the Partnership for Resilience and
Economic Growth (PREG), with a focus on bringing together humanitarian and development
partners to “build resilience among pastoral communities in northern Kenya.” That is, REGAL
was subsumed under the PREG umbrella and as such, the PREG I endline includes a final
evaluation of the REGAL projects.
5 USAID. 2016. USAID Partnership for Resilience and Economic Growth Learning Event. Accessed at: https://www.africaleadftf.org/wp￾content/uploads/2016/09/PREG-Learning-Event-Slide-Deck-Aug-2016.pdf. 6 See: TOS_Snapshot_#xx_Kenya (usaid.gov). 7 See: TOS_Snapshot_#xx_Kenya (usaid.gov). 8 USAID. 2016. USAID Partnership for Resilience and Economic Growth Learning Event. Accessed at: https://www.africaleadftf.org/wp￾content/uploads/2016/09/PREG-Learning-Event-Slide-Deck-Aug-2016.pdf. 
1.1 PROJECT AREA PROFILES
Prolonged and recurring drought (2008–2011, 2017, 2019), combined with localized vulnerabilities, 
represent key challenges affecting development progress in the northern ASALs.5
In 2012, the 
USAID Feed the Future Resilience and Economic Growth in the Arid Lands (REGAL) program was 
initiated in five of the nine counties within USAID’s Zone of Influence (ZOI). 
REGAL aimed to reduce poverty and hunger among vulnerable Kenyans in the northern ASALs 
through two five-year projects, REGAL-IR (Improved Resilience) and REGAL-AG (Accelerated 
Growth), implemented respectively by African Development Solutions (Adeso) and ACDI/VOCA. 
The REGAL-IR program sought to engage in broad investment to improve resilience of vulnerable 
households in Garissa, Isiolo, Marsabit, Turkana, and Wajir counties. Working with communities 
to build their capacity to cope with shocks, the interventions supported local structures to 
improve social, economic, and environmental
resilience.6 The REGAL-AG program provided 
additional investment in livestock value chains and 
USAID DEFINITION OF RESILIENCE
“THE ABILITY OF PEOPLE, HOUSEHOLDS, 
COMMUNITIES, COUNTRIES, AND SYSTEMS 
TO MITIGATE, ADAPT TO, AND RECOVER 
FROM SHOCKS AND STRESSES IN A 
MANNER THAT REDUCES CHRONIC 
VULNERABILITY AND FACILITATES 
INCLUSIVE GROWTH.”
PREG Initiative in Northern Kenya Endline Report | 3
1.1.1 REGAL Theory of Change
To generate the economic growth needed to reduce poverty and hunger, and to achieve the 
Government of Kenya’s (GoK’s) vision of a commercial and modern agricultural sector, Feed the 
Future in Kenya has invested in transforming livestock production through improved 
competitiveness of high-potential value chains (e.g., milk, meat) and the promotion of 
diversification into higher-return activities. As documented by the International Food Policy 
Research Institute for maize in Ethiopia, the development of selected value chains can have 
multiplier effects that, in turn, can create employment opportunities.9 
Value chain investment in livestock markets with lower risk and lower entry barriers has been one 
way of encouraging the participation of poorer rural households in expanding economic activities. 
Feed the Future in Kenya believed that while these investments in economic growth are necessary 
to reduce poverty and hunger, by themselves they would be insufficient. Beyond growth, poverty 
reduction requires targeted interventions that address the needs of agro-pastoralists (the rural 
poor) as well as more vulnerable populations, women and youth. REGAL interventions used both 
“push” and “pull” strategies to engage vulnerable households in market-oriented activities and 
promote economic growth (Figure 2). The value chain programs of REGAL-AG aimed to “pull” 
rural households into income-raising activities by improving links to markets and input access, 
providing affordable business development skills and financial services, and promoting greater 
diversification—specifically tailored to the needs of agro-pastoralists, women, and youth. 
Activities implemented under REGAL-IR were designed to “push” individuals toward increased 
engagement in markets through reduced risk, better nutrition, improved access to information 
and resources for small business entrepreneurship, and better access to and management of the 
natural resources on which many small businesses depend. 
Figure 2: REGAL Theory of Change
9 International Food Policy Research Institute (IFPRI). 2010. Maize Value Chain Potential in Ethiopia Constraints and Opportunities for 
Enhancing the System. 
 4 | REAL | Resilience, Evaluation, Analysis & Learning
1.1.2 REGAL-IR
As the “push” project, the REGAL-IR aimed to reduce hunger and poverty by strengthening social, 
economic, and environmental resilience in pastoral and transitioning communities in Isiolo, 
Garissa, Wajir, Marsabit, and Turkana. Project goals were to:
• Improve capacity of individuals and community-based enterprises to become more
competitive in business and non-pastoral activities;
• Support community structures to better manage natural resources and relieve pressure on
the environment;
• Support community and entrepreneur access to market information and produce markets;
• Strengthen capacity to manage conflict; and
• Improve household consumption of nutritious foods.
REGAL-IR implemented activities and sought to strengthen social, economic and environmental 
resilience for 558,000 people (93,000 households), out of a total population of approximately 2.6 
million in these counties in 2009,10 through community engagement and strengthening of local 
institutions. Reported key accomplishments include working with 293 self-help groups, with a total 
of 6,701 members, including 4,816 women (72 percent).11 Communities were able to leverage 
USD 12.7 million from the GoK and other development actors toward implementation of 
community development action plan (CDAP) activities; 77 CDAP oversight committees were 
formed, with a total of 1002 members, 35 percent of which are women. The activity also supports 
Sidai Africa, Ltd. Super Service Centers (SSCs), which provide animal health products and services 
in northern Kenya,12 including establishing at least four SSCs and 18 franchises in Marsabit, Isiolo, 
Wajir and Turkana.
1.1.3 REGAL-AG
The REGAL-AG project goal was to increase economic growth and social stability in Marsabit 
and Isiolo counties by expanding and strengthening competitive livestock value chains and 
markets. Project goals were to:
• Improve the enabling environment for livestock value chain development;
• Improve market linkages and livestock productivity;
• Expand existing and develop new livestock service and input markets; and
• Expand livestock-related economic opportunities that engage and benefit men and women.
10 Kenya Open Data: https://www.opendata.go.ke/datasets/4b8b2da624ba4cdf9cd38b5f359eecf3/explore. 11 Accessed at: https://www.usaid.gov/sites/default/files/documents/1860/REGAL_IR.pdf. 12 Accessed at: https://www.usaid.gov/sites/default/files/documents/1860/REGAL%20IR%20%20fact%20sheet%20August%202014.pdf. 
PREG Initiative in Northern Kenya Endline Report | 5
As the “pull” component, the REGAL-AG project made upgrades to 12 livestock markets, benefiting 
24,597 rural households over the five-year project lifecycle, and constructed eight additional 
markets, which expanded market opportunities and helped to eliminate value chain constraints.13
The project supported the establishment of 36 enterprises through business development grants, 
and 31 microenterprises and 17 small enterprises received business support services. Grants helped 
to identify and build the capacity of change agents within pastoral communities who could drive 
further investments, upgrades, and increased economic competitiveness. Finally, over 2,600 
individuals were trained in agricultural sector productivity or food security.
1.2 DESCRIPTION OF THE PREG PROJECT CONTEXT
Kenya’s political, structural and economic reforms over the past decades have been successful in 
driving economic growth and progress towards sustainability goals.14 The most recent and 
significant political reform involved a new constitution in 2010, which included a major provision 
for devolution to improve governance and services at local levels. The country experienced 
political uncertainty related to the 2017 presidential election and re-election along with a severe 
drought the same year, and growth in the gross domestic product (GDP) slowed to 4.9 percent.15
GDP growth rebounded to 5.7 percent in 2018 and there has been significant recovery, though 
not evenly across sectors, in early 2021.16 
Agriculture is a backbone of the country’s economy, key to national food security and to poverty 
reduction. The sector continues to be supported by the GoK, USAID, and development partners 
through a collaborative effort to improve agricultural and pastoralist markets and value chains. 
While urban and peri-urban areas of Kenya are growing, 73 percent of the population still lives in 
rural areas.17
Development challenges remain, as “Kenya’s reliance on smallholder, rain-fed agriculture and its 
high poverty rates render the country particularly vulnerable to climate risks,”18 particularly in 
the ASALs of northern Kenya. Recurring and prolonged drought (e.g., 2008–2011, 2017, and 
2019) continues to be one of several factors contributing to increasing vulnerability in northern 
Kenya. The drought in 2017 resulted in 3.6 million acutely food insecure people in the ASALs. The 
drought continued into 2019, with northern Kenya classified as “stressed” per the IPC Food 
Insecurity Phase Classification (IPC), with 700,000 people severely food insecure.19 Other factors 
impacting rural development include population growth, natural resource degradation, land 
fragmentation, human and animal disease, and tribal conflict. 
13 See: https://www.acdivoca.org/projects/resilience-and-economic-growth-in-the-arid-lands-accelerated-growth-PREG I-ag/. 14 World Bank. (2018). The World Bank in Kenya. April. Accessed at: http://www.worldbank.org/en/country/kenya/overview. 15 World Bank. (2018). Press Release: Kenya’s Economy Poised to Rebound in 2018 and Remain Robust through 2020, 10 October. 16 Ibid. 17 World Bank. (2019). Data: Rural population (% of total population). Estimates based on the United Nations Population Division's World 
Urbanization Prospects: 2018 Revision. 18 World Bank Group. (2015). Kenya Agricultural Risk Assessment. Agriculture Global Practice Note 17, October. 19 OCHA. (2019). Greater Horn of Africa Region Humanitarian Snapshot (November–December 2018), 31 January.
 6 | REAL | Resilience, Evaluation, Analysis & Learning
The main livelihood in PREG intervention areas is raising livestock, which is also important for the 
overall economy of Kenya. Livestock production contributes 13.4 percent (USD 3.1 billion) to 
agricultural value added in Kenya, with cattle being the most significant.20 The beef industry is the 
number-one contributor to agricultural GDP in Kenya, at around 35 percent, followed by dairy 
cattle production,21 with crop production being Kenya’s largest economic sector.22 The PREG 
program area and surrounding counties provide the vast majority (approximately 80–90 percent) 
of Kenya’s beef,23 and much of Kenya’s meat for export as well. Besides meat, Kenya exports live 
animals, milk, and animal hides and skins. In terms of long-term trends, Kenya’s total cattle 
population in semi-arid regions has decreased by more than 26 percent in the past 30 years, 
according to a recent study funded by the Canada-based International Development Research 
Centre and the UK Department for International Development, and Turkana County is the most 
harshly affected.24 The study attributes this sharp decrease in cattle population to rising 
temperatures and reduced or unpredictable rainfall.
Table 1 and Table 2 show that cattle meat and cow milk are most important in terms of 
production and gross marketed production value from 2008–2017. However, camel meat and milk 
production and goat milk production have dramatically increased over the past 15 years, using 
2002 as a reference point (Table 1). For instance, 18,000 tons of camel meat were produced in 
2002, increasing to 73,000 tons in 2017.
Table 1: Kenya livestock production over past ten years plus past reference year (1,000 metric tons) 
2002
(ref.) 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Meat, 
cattle 319 458 483 462 458 411 425 443 487 529 589
Meat, goat 39 45 46 47 56 56 58 68 68 50 64
Meat, 
sheep 39 39 41 42 41 40 42 30 31 38 33
Meat, 
camel 18 75 62 65 65 65 66 60 66 70 73
Milk, cow 2,891 3,209 3,567 3,639 3,711 3,733 3,686 3,425 3,444 4,115 3,561
Milk, goat 104 136 258 260 263 268 269 280 260 274 256
Milk, sheep 58 57 57 58 60 61 62 63 62 70 67
Milk, camel 225 854 877 892 890 911 871 749 812 850 876
Source: FAOSTAT. 2019 (data from January 31, 2019). Based on FAO estimates and imputation methodology.
20 KNBS. (2017). Kenya Economic Survey 2016. 21 FAO. (2018). Africa Sustainable Livestock 2050: Kenya Cattle and Poultry Sectors. Livestock production systems spotlight. 22 IGAD. (2013). 23 Farmer, E. and J. Mbwika. (2012). 24 Kenya Markets Trust. (2018). “Cattle Population in Decline in Arid Areas, but there are Opportunities to Explore—Study” 17 February. 
Accessed at: http://www.kenyamarkets.org/press/cattle-population-in-decline-in-asals/. 
PREG Initiative in Northern Kenya Endline Report | 7
Table 2: Gross marketed production value of livestock products at current prices (KSh million), by year
Livestock 
and related 
products
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017*
Meat, cattle 22,834 24,057 31,893 48,943 54,141 58,237 59,273 66,217 84,701 93,630
Meat, goat, 
sheep, lamb 4,147 3,097 5,160 9,632 10,025 7,532 9,208 4,855 5,364 6,782
Meat, pigs 626 804 754 828 1,079 1,307 1,398 1,317 1,838 2,216
Poultry and 
eggs 2,789 4,345 5,012 5,553 6,482 7,086 7,441 6,006 8,788 10,675
Wool 4.9 1.8 3.1 3.3 2.9 2.9 0.8 0.2 0.2 0.3
Hides and 
skin 1,201 1,249 1,092 1,392 1,525 1,888 1,884 1,006 1,287 1,418
Dairy 
products 8,369 11,497 11,346 14,548 15,416 16,777 18,785 21,205 23,020 20,878
* Provisional data
KNBS. Statistical Abstract. Reports from 2018, 2017, 2016. 2015.
As described in Section 1.1, PREG represents an expansion of the REGAL project, both in terms of 
geographic coverage in the ASALs and implementing partners. Building on—and subsuming—
REGAL, the PREG initiative “brings together humanitarian and development partners to build 
resilience among vulnerable pastoralist communities in northern Kenya”.25 PREG includes USAID 
programs and implementing partners operating in nine ASAL counties, with WFP in all nine and 
additional partners in Turkana, Isiolo, Marsabit, Wajir, and Garissa. 
The PREG initiative is intended to:
• Sequence, layer, and integrate programming in order to avoid duplication and
maximize impact;
• Leverage and partner strategically;
• Mutually reinforce activities that are different but still coordinated;
• Measure and evaluate impact by collecting data and measuring results consistently
across all partners;
• Build sustainability through all stages of the project cycle and promote ownership by
engaging with governments; and
• Share on the global stage a picture of USAID Feed the Future Synergy and
Complementarity through partnerships.
25 USAID. 2016. USAID Partnership for Resilience and Economic Growth Learning Event. Accessed at: https://www.africaleadftf.org/wp￾content/uploads/2016/09/PREG-Learning-Event-Slide-Deck-Aug-2016.pdf. 
 8 | REAL | Resilience, Evaluation, Analysis & Learning
1.3 OBJECTIVE OF THE STUDY
The initial research plan designed for the REGAL—and subsequently the PREG I—project 
evaluation was to conduct an impact evaluation that would allow for estimating the effect of 
differing levels of resilience programming investments (high- and low-intensity) on household 
resilience capacity. However, several issues emerged during the transition from REGAL to PREG 
that preempted an impact evaluation approach. As a result, data were analyzed using a pre-post 
analysis that traces changes in outcomes over time (i.e., between baseline and endline). Details of 
the methodological challenges and solutions are discussed in Section 2.
Because PREG subsumed REGAL, the PREG endline utilizes research questions designed for the 
REGAL IE:
1. What impact do investments to enhance resilience (REGAL) have on livelihood
outcome indicators (HHS, as well as stunting, wasting, underweight, poverty
prevalence and income)? What impact do they have on adaptive capacity?
2. What is the additive/multiplicative value of layering resilience (REGAL-IR) and
economic growth (REGAL-AG) investments in relation to the indicators noted above?
3. What is the combined impact of REGAL-IR and REGAL-AG on depth of poverty
(derived from expenditure data used to determine poverty prevalence) and other well￾being outcomes?
4. What are the relationships between household and community resilience (derived
from the qualitative data)?
5. Have interventions strengthened risk-reduction strategies pursued by men and women
to cope with shocks (agro-climatic, health, economic, and socio-political)?
Overall, several of the original research questions developed for the REGAL impact evaluation 
that required comparison groups (Q1–4, and Q6) are only addressed to a limited degree in the 
current analysis (see Annex 3). Findings that address Q5 are integrally tied to the qualitative 
findings throughout the report and as such, Q5 is not addressed as a separate research question. 
In addition to providing a pre-post analysis of change over time in PREG programming areas, the 
analysis in the main body of this report answers the following research question from the 
baseline: 
“How does resilience capacity affect household perceived ability to cope with future shocks?”
PREG Initiative in Northern Kenya Endline Report | 9
2 METHODOLOGY
2.1 QUANTITATIVE SURVEY
The 2013 Feed the Future baseline survey served two purposes: 
• As a population-based survey (PBS) designed to measure Feed the Future performance
indicators for the nine ASAL counties of the ZOI in northern Kenya, and,
• As an impact evaluation baseline survey of PREG I to measure the changes in
household resilience capacities and outcomes in areas with different levels of PREG I
resilience programming intensity.
Similarly, the 2018 PREG I endline survey served two purposes:
• As the impact evaluation endline survey for PREG I, and
• As the impact evaluation baseline survey of the second five-year phase of PREG (PREG II).
Note that a mid-term survey was also conducted in 2015. Although this survey collected resilience 
information, it was primarily designed to collect point estimates of Feed the Future ZOI indicators 
and was not statistically powered to measure changes from the baseline survey. The results from 
the mid-term survey are not included in this endline study. For a full description of the midline 
survey sampling, please see Annex 4.
2.1.1 Sample Design
As previously noted, the sample design changed from the baseline to the endline survey rounds. 
The baseline sample was divided into two geographic strata defined at the county level, based on 
existing and planned REGAL programming intensity. The LI stratum was identified as areas where 
only humanitarian assistance was provided through USAID resources supporting REGAL (and 
subsequently PREG), and covered Baringo, Mandera, Samburu, and Tana River counties. The HI 
stratum was defined as areas that received other resilience-enhancing investments (initially areas 
covered by REGAL, but later extended to all PREG programming) in Garissa, Isiolo, Marsabit, 
Wajir, and Turkana counties. The baseline samples drawn from the counties in these two strata 
were population-based, that is, all households within the counties constituted the sampling frame 
from which surveyed households were drawn. Note that for reasons of insecurity, one LI county 
(Mandera) and two HI counties (Garissa and Wajir) were excluded from the baseline survey 
sample. 
At the time of the analysis of the midline survey, it became clear that the definition of the LI and 
HI strata by counties did not align with the reality of programming on the ground: because the 
sample of households was drawn from the total populations of the counties in accordance with 
the PBS design and the initial understanding of program intentions, many of the households that 
were selected from the HI counties did not directly participate in (i.e., benefit from) the resilience 
 10 | REAL | Resilience, Evaluation, Analysis & Learning
programming supported by USAID. While the REGAL projects did include systems-level 
interventions to support value chains, the geographic reach of these indirect interventions was not 
known. Thus, it is impossible to determine whether respondents in HI (or LI) strata were indirect 
beneficiaries of these systems-level interventions. 
The sampling strategy for the endline sample was modified in order to address these sampling 
and programming limitations. Rather than identifying the HI sampling frame on the basis of 
county only (as in the baseline), information from project records was used to identify much 
smaller geographic areas (sublocations) in which PREG projects had been carried out or were 
currently engaged in specific interventions at the time of the PREG I endline survey. HI 
sublocations were identified as those with reported interventions under REGAL (or a subsequent 
Feed the Future Livestock Market Systems project) or at least two other PREG activities (from the 
PREG intervention data set maintained by USAID). 
The LI sampling frame was also changed from county level to sublocations within the LI counties 
where WFP-supported humanitarian assistance (and other non-resilience-building) activities had 
been carried out.26 Thus, while the HI households selected for the endline survey were much more 
likely to have been directly exposed to resilience interventions, at least some households in the 
baseline sample from the HI treatment group were much less likely to have been direct 
beneficiaries, since the baseline sampling frame included sublocations where no PREG direct 
interventions occurred. 
Because of this change in sampling strategy, the baseline and endline sample populations are no 
longer comparable: the sampling frame for the baseline covered the entirety of each county while 
the sampling frame at endline only covered part of each county (i.e., sublocations). Although the 
initial understanding of the intended high-intensity and low-intensity programming areas, defined
at the level of counties, was applied for the baseline sample design, the actual area coverage of 
REGAL/PREG interventions did not conform to the original sampling design. The sampling frame 
for the endline survey was therefore modified to include only those sublocations where 
REGAL/PREG interventions were reported to have occurred. 
In order to be able to make the most appropriate comparisons across time, the sample was 
restricted to sublocations that were covered in both the baseline and endline samples. This 
resulted in an 81 percent decrease in the size of the baseline sample and a 60 percent decrease in 
the size of the endline sample available for analysis (Table 3). 
In some counties, the change in sample design resulted in a larger (e.g., Samburu, Tana River) or 
equal (e.g., Turkana) number of households because of a change in the sizes of the sampling 
frames relative to the baseline. Specifically, enumeration areas (EAs) were selected at both 
baseline and endline using Probability Proportional to Size (PPS). However, at baseline, the 
sampling frame comprised all EAs in a county, whereas the endline sampling frame only included 
sublocations (and their EAs) where PREG activities were being implemented. As the relative 
proportions of households in each county are different in the sampling frames of the two rounds, 
26 The USAID PREG team indicated that PREG II interventions are expected to continue in the same sublocations covered under the first 
phase of PREG.
PREG Initiative in Northern Kenya Endline Report | 11
the number of sampled households is different in each round. In particular, the proportion of 
households in the total sampling frame was higher at endline than baseline for both Samburu and 
Tana River. 
Table 3: Sample size (disaggregated by county), by survey round
Baseline (Jan/Feb 2013) Endline (Aug/Sept 2018)
County Total sample 
(# HHs)
# HHs in 
sublocations that 
overlap with 
endline
Total sample
(# HHs)
# HHs in 
sublocations that 
overlap with 
baseline
Baringo 234 10 58 29
Samburu 158 61 354 116
Tana River 180 34 382 84
Turkana 364 96 364 152
Isiolo 191 71 188 160
Marsabit 588 43 216 89
Total 1,715 315 1,562 630
Because of the limitations imposed by the change in sampling strategies for the LI and HI strata 
between baseline and endline, rigorous impact analysis techniques are not possible. Rather, a 
more basic pre-post analysis of the quantitative data from baseline to endline is conducted. More 
insights about the impacts of resilience programming on households will be drawn from the 
qualitative information. A comparable descriptive analysis of key indicators across the two rounds 
is provided in Annex 3, where the analyzed samples include only sublocations that were included 
in both the baseline and endline samples. 
2.1.2 Resilience Information Collected
The 2013 PBS baseline survey27 questionnaire included information to measure resilience 
outcomes (food security measures, expenditures and poverty, recovery from previous drought,28
ability to cope with future shocks/stresses), as well as a brief module to measure selected 
dimensions of resilience capacities. It is important to note that at the time of the 2013 baseline 
survey, USAID had not yet developed a set of tools to fully measure all dimensions of resilience 
capacities. Since that time, USAID has developed a more comprehensive framework for resilience 
measurement and has identified a wider range of sources to measure resilience capacities and 
outcomes.29 Additionally, the baseline did not collect information about the shocks/stresses 
households’ experience, which is important for resilience analysis. Although USAID has since 
developed a shock/stresses module and the Shock Exposure Index (SEI), the lack of data on 
27 https://pdf.usaid.gov/pdf_docs/PA00KPHV.pdf 28 Although there was no drought in the year prior to the baseline, respondents were asked about recovery from the last drought, which 
occurred in 2011.
29 USAID/TOPS. 2017. Resilience Methodological Guide: Full Model. Prepared by TANGO. 25 October.
 12 | REAL | Resilience, Evaluation, Analysis & Learning
shocks/stresses at baseline means we do not have a clear idea of the types of shocks households 
faced over the period of the evaluation, which is a limitation for the current analysis. 
Building resilience requires an integrated approach and a long-term commitment to improving 
three critical capacities: absorptive capacity, adaptive capacity, and transformative capacity.
30 Each 
capacity is an index constructed from relevant indicators, some of which are themselves indices. 
Below is a list of the three resilience capacity indices and their respective components. The 2018 
endline survey collected information on some—but not all—components of each index, since 
comparable information was not collected at baseline (i.e., this resilience measurement 
framework had not been fully developed at the time of the baseline). Thus, only a subset of 
resilience capacity dimensions from the resilience measurement framework is included in the 
analysis presented in this report (included components are indicated in bold). Further, it should be 
noted that some of the indices are used to calculate multiple resilience capacity indices because 
they are relevant to both. For example, bridging and linking social capital are both used in the 
calculations of adaptive and transformative capacities.31 
Absorptive capacity index. Absorptive capacity is the ability to minimize exposure to shocks 
and stresses through preventative measures and appropriate coping strategies to avoid 
permanent, negative impacts. The eight indicators are:
• Availability of informal safety nets
• Bonding social capital (social networks within a respondent’s community)
• Access to cash savings32
• Access to remittances
• Asset ownership
• Shock preparedness and mitigation
• Availability of/access to insurance
• Availability of/access to humanitarian assistance
Adaptive capacity index. Adaptive capacity is the ability to make proactive and informed 
choices about alternative livelihood strategies based on an understanding of changing 
conditions. The ten indicators are: 
• Bridging social capital (social networks outside a respondent’s community)
• Linking social capital
30 Béné et al. 2016. 31 The use of resilience capacity dimensions in more than one index does not affect calculations of the individual resilience capacity indices, as 
they are independent. 
32 Savings was not calculated for the endline analysis. Although the data were collected, we determined that the results would be potentially 
biased because the wording of the question was different in the baseline (retrospective) and endline (prospective) surveys. 
PREG Initiative in Northern Kenya Endline Report | 13
• Social network index
• Education/training
• Livelihood diversification
• Exposure to information
• Adoption of improved practices
• Asset ownership
• Availability of financial services
• Aspirations/confidence to adapt index33
Transformative capacity index. Transformative capacity involves the governance 
mechanisms, policies/regulations, infrastructure, community networks, and formal and informal 
social protection mechanisms that constitute the enabling environment for systemic change. The 
fifteen indicators are:
• Availability of/access to formal safety nets
• Availability of markets
• Availability of/access to communal natural resources
• Availability of/access to basic services
• Availability of/access to infrastructure
• Availability of/access to agricultural services
• Availability of/access to livestock services
• Bridging social capital
• Linking social capital
• Collective action
• Social cohesion
• Gender equitable decision-making index
• Participation in local decision-making
• Local government responsiveness
• Gender index
Recovery from shock and coping strategies. Subjective reports of households’ ability to recover 
from shocks they experience are a key source of information on the strength of their resilience. 
33 Agency is one of several components of the aspirations/confidence to adapt index. Information on household agency is not included in this 
report due to measurement error (see Section 3.3.5).
 14 | REAL | Resilience, Evaluation, Analysis & Learning
Coping strategies are household responses to shocks and may include selling off livestock and 
other assets, reducing food consumption, taking loans from family members, or sending one or 
more household members elsewhere to work.
Baseline resilience-related information. As mentioned above, the 2013 PBS baseline survey had 
a much more abbreviated set of questions that captured only a portion of all the resilience 
components described above. They include the following questions that measure aspects of 
resilience capacity, coping strategies, and subjective resilience (measured as the perceived ability 
to cope with future shock/stress): 
• Resilience capacity components
o Livelihood diversification
o Adult education
o Savings
o Asset ownership (a count of the number of types of assets a household currently
owns; range: 0 to 34)
o Social capital (reliance on financial or in-kind food support from relatives and non￾relatives within (bonding) and outside (bridging) the respondents’ communities
during past and future shocks)
o Agency (perception of whether one is responsible for his/her success or failure or if
this is a matter of destiny)34
• Coping strategies
o Change income or food sources
o Add income or food sources
o Sell large or small assets
o Migration of one or more household members
• Perceived ability to cope with future shocks35
For more detail on how the resilience capacity indices and each dimension of resilience capacity 
are measured, refer to USAID Resilience and Resilience Capacities Measurement Options 
guidance.36
34 Information on “household agency” is not reported here, due to measurement error (see Section 3.3.5). 35 The baseline survey included a question about recovery from the latest drought but it did not collect information about the degree to 
which households were affected by the latest drought, which is optimal in order to accurately estimate recovery. However, perceived ability 
to recover from future shocks is a valid alternative, subjective measure of resilience.
36 https://www.fsnnetwork.org/resource/resilience-and-resilience-capacities-measurement-options. 
PREG Initiative in Northern Kenya Endline Report | 15
2.1.3 Field Procedures
Training of supervisors and master trainers for the baseline survey started on December 28, 2012, 
and was followed immediately on January 6, 2013, by a twelve-day training of enumerators. 
Supervisors and enumerators traveled to Nairobi for training. Training methods were in 
alignment with those developed for other Feed the Future FEEDBACK countries where a PBS data 
collection activity had taken place.37 Training covered careful review of the quantitative 
questionnaire (including translation into and/or discussion in local languages), use of electronic 
tablets, use of weighing and measuring equipment, detailed instruction on interview methods, and 
human subjects’ protection. Based on a thorough understanding of the questionnaires during 
training, enumerators translated questions in the field and recorded responses in the English 
version of the questionnaire. Supervisors received additional training on checking data quality, 
uploading data to the Westat server, and making and tracking enumerator assignments. All 
enumerators and supervisors reviewed and signed confidentiality forms. 
For the endline survey, the local data collection subcontractor, Kimetrica, led a 15-day training 
(including supervisors and enumerators) in August 2018 in Nakuru, Kenya, which again included a 
careful review of the quantitative questionnaire, use of electronic tablets, use of weighing and 
measuring equipment, detailed instruction on interview methods, and human subjects protection. 
Teams consisted of five enumerators and a quality control interviewer. Staff from TANGO 
attended the training to provide technical assistance as needed. Each enumerator was provided 
with an Android-based tablet running CSPro programmed with the survey questionnaire including 
the necessary skip patterns and validation rules. In addition to English, tablets were programmed 
with questionnaires in Kiswahili, Somali, or Turkana and enumerators selected the relevant 
language for the household. Team supervisors received additional training on data quality 
procedures (e.g., daily review of each questionnaire in the field for completeness and accuracy), 
uploading data to the remote server, and enumerator assignments. During data collection and 
upon successful review, data were transmitted via cellular network to a remote secure server 
where it was aggregated and reviewed daily by Kimetrica staff. Kimetrica and TANGO staff 
developed field check tables to provide feedback to supervisors and enumerators during data 
collection so that inconsistencies could be corrected quickly. 
2.1.4 Data Analysis
The quantitative data analysis was conducted with the statistical software STATA version 15 
using descriptive analysis techniques. 
Descriptive Analysis of Trends
In this report, the baseline and endline survey data are used to conduct descriptive analysis of 
change over time in indicators describing household coping strategies, poverty and food security 
status, and resilience capacities. Indicator values are reported as percentages and means.
37 These methods are documented in manuals, reports and PowerPoint presentations. Copies of the training manuals can be obtained from 
the Development Experience Clearinghouse or by contacting the Feed the Future FEEDBACK project.
 16 | REAL | Resilience, Evaluation, Analysis & Learning
• Percentages. For values provided in nominal scales (e.g., yes/no responses),
percentages are computed using the weighted number of cases that provided a given
response as the numerator, and the total weighted number of cases with valid data for
the indicator as the denominator.
• Means. For continuous variables (e.g., number of livelihood activities), means are
computed using the weighted sum of values as the numerator and the total weighted
number of cases with valid data as the denominator.
Indicators are reported by time period (baseline and endline). Tests for statistically significant 
differences in the indicators between baseline and endline are reported for p<0.05 or less.
Data Management and Security
Prior to release of data from the baseline and endline surveys, any direct or indirect identifiers 
that could be used to identify individuals, such as geographic information at lower administrative 
levels, was removed and/or modified to prevent disclosure of individual identities.
2.1.5 Limitations of the Quantitative Surveys
Multiple purposes of the survey. The PREG I baseline survey was initially designed as part of the 
population-based Feed the Future ZOI baseline survey for northern Kenya. The ZOI survey is 
designed as a performance monitoring survey to track changes in Feed the Future indicators at 
the level of the population within the ZOI, and is not designed to be able to make any assessment 
of the causes of these changes. The ZOI survey is population-based, drawn from the entire 
population within the ZOI. The impact evaluation, on the other hand, is intended to assess the 
extent to which resilience programming interventions contributed to enhancing the resilience of 
households, and therefore requires the inclusion of a counterfactual in the survey design. Because 
the survey was to serve as the Feed the Future ZOI baseline, the sample had to be representative 
at the level of the population as a whole. This requirement, along with the fact that the PREG I 
projects had not identified their precise geographic areas of intervention at the time of the 
baseline survey in 2013, placed limitations on the ability to delineate treatment and control strata 
within the total population-based sample of the ZOI. 
The sampling strategy of defining control areas (areas that received only humanitarian 
assistance) and treatment areas (areas that received resilience programming in addition to 
humanitarian assistance) geographically by county proved to be an inaccurate way to identify the 
treatment and control populations. First, the PREG I project reached only parts of the populations 
in the counties designated as treatment, and since the sample was selected randomly from the 
entire populations of the counties, many sampled households in the designated treatment stratum 
did not, in fact, receive the ‘treatment’ of being exposed to resilience programming. Secondly, the 
survey design did not control for possible underlying differences in household characteristics 
between the control and treatment areas (e.g., level of education, wealth). Underlying contextual 
differences between implementation areas weakens the ability to attribute the observed changes 
between the treatment and control groups over time to the ‘treatment’ of resilience interventions 
PREG Initiative in Northern Kenya Endline Report | 17
and as such, does not allow for a robust impact evaluation.38 Instead, this report presents findings 
from a more general pre-post analysis that tracks changes over time. A limited difference-in￾differences impact analysis is provided in Annex 3.
Inability to attribute changes to PREG interventions. The study does not allow for clear 
attribution of resilience programming to PREG I households as this was not a targeted beneficiary 
survey and many agencies were providing similar programming during the same time period as 
PREG I.
Sample size. Another limitation to the current study is related to the significantly reduced sample 
size available for the pre-post comparative analysis (see Annex 3 for the pre-post comparative 
analysis using a Difference in Differences (DID) approach on the significantly smaller overlapping 
sample that resulted from the change in sample design). 
Timing. Fieldwork for the baseline and endline surveys took place at different times of year, 
which poses a potential bias in the data. In order to detect change over time in food insecurity 
and nutrition outcomes, ideally, data are collected at the same time each year and typically 
during the most vulnerable season (e.g., the lean season). Fieldwork for the baseline occurred in 
January/February 2013, when harvesting had begun (see Figure 3 in Annex 4). Fieldwork for the 
endline survey, which was contingent on timing for the PREG II baseline survey, ran from August 
to September of 2018, during the lean season. The inconsistent timing of the baseline and endline 
surveys is important to consider when interpreting the results, particularly as they relate to food 
security indicators, which are highly sensitive to seasonal variation. 
Resilience measurement development. Since 2012, TANGO has been developing and refining its 
conceptual model and measures of resilience. The Feed the Future PBS baseline survey offered the 
opportunity to include questions to try to capture preliminary thoughts on resilience measurement 
and as such, reflects the nascent stage of the model’s development. In the baseline survey, data on 
households’ perceived level of severity of drought (i.e., the drought two years prior) were not 
collected. Without this information, we are unable to measure recovery from the drought two 
years prior and subsequently, were not able to include recovery as an outcome indicator. Rather, 
households’ perceived ability to cope with future shocks/stresses is considered the primary 
subjective resilience outcome. Lack of information on shocks/stresses (or their severity) also makes 
it difficult to interpret data on coping strategies or ability to recover in the future. 
Participation in project interventions other than PREG. A possible limitation of the PREG I 
analysis is the presence of interventions implemented by other projects. Ideally, the baseline 
survey would have included questions to collect information on household participation in any 
resilience programming activities generally, not just those implemented by PREG. Without this
crucial information, it is not possible to estimate potential bias associated with other 
programming. 
Security-related restrictions on access to survey areas. At the time of the baseline survey, 
Garissa, Mandera and Wajir counties were considered “no-go areas” by the Kenya National 
38 In PREG II it should be noted that the matched sample will overcome such potential differences.
 18 | REAL | Resilience, Evaluation, Analysis & Learning
Bureau of Statistics (KNBS) due to the presence of the known terrorist group al-Shabab, and 
were removed from the baseline sampling frame. Thus, the sample available for comparison 
across survey rounds is not fully representative of all PREG programming areas.
2.2 QUALITATIVE STUDY
Qualitative data collection for this report took place simultaneously with the endline quantitative 
field work (September 13–22, 2018). No qualitative data collection was conducted for the 2013 
baseline. Thus, all discussion presented in this section refers to the PREG I endline.
The primary focus of the qualitative component was to collect perceptions of change in livelihoods 
and well-being at the community level. Data provided by community groups and key actors form 
the basis for identifying signals of PREG I contributions to this change process. In these interviews 
and focus group discussions (FGDs), explicit attention was placed on changes affecting women 
and youth. Insights garnered from the qualitative data provide context for observed quantitative 
patterns and, in some cases, illustrate the motivations and constraints that underlie different 
pathways of resilience strengthening. The analysis of qualitative data also suggests areas for 
further quantitative inquiry. Qualitative data is explanatory and is not meant to be generalizable 
to a broader population, thus it is not quantified. However, qualitative responses are triangulated 
during analysis across focus groups, key informants, and secondary data, and atypical responses 
are carefully reviewed to see if they add value to the understanding of participant responses.
Qualitative data-gathering took place in the six PREG I counties to meet the needs of the PREG II 
impact evaluation baseline survey (Marsabit, Isiolo, Garissa, Turkana, Wajir, Tana River). A team 
comprised of Kimetrica and TANGO staff trained local field researchers assigned to each county 
targeted in the study. 
2.2.1 Survey Sample Design
Qualitative teams for the endline consisted of four-person teams, each consisting of two gender￾balanced interview teams including one not taker and one facilitator. The teams were 
multidisciplinary and included both international consultants and local Kenyan consultants with 
knowledge of the PREG I implementation areas. Qualitative teams visited 16 villages that were 
selected to represent PREG I program distribution across counties, program intensity levels, and 
the variety of programming present in each sublocation. Security considerations also impacted 
the selection of community sites, with final site selection based on Kimetrica’s security ratings, 
which were informed by quantitative fieldwork experience in the study areas. The list of 
communities visited in each county is presented in Table 4. 
The study team gathered data from separate FGDs with men and women who are recognized as 
engaged and active members of their community. FGDs consisted of 6–8 participants. Recruitment 
criteria aimed to include a breadth of perspectives by engaging representatives from different 
clans and ethnic groups, and across various socio-economic groups within each community. 
Particular emphasis during the FGDs was given to individual and household participation in 
livestock markets; livelihoods; main shocks and stressors in the community; changes in household 
PREG Initiative in Northern Kenya Endline Report | 19
and community responses to shocks and stresses; impacts of shocks on the community; and 
aspirations for the future. The topical outlines used to guide the discussions are presented in 
Annex 1.
The research team also conducted 54 interviews with a range of key informants with special 
knowledge of at least one aspect of the project’s topics of interest. For PREG I, the key informants 
included community and clan leaders, school headmasters, private livestock traders, local 
merchants, implementing partner staff, representatives of local financial institutions, livestock 
insurance representatives, livestock association managers, NDMA staff, agriculture and livestock 
extension workers, health and nutrition staff, youth leaders, and slaughterhouse officials. The KIIs 
generated in-depth information on impact and the dynamics of local change, and functioned as a 
form of triangulation to improve the interpretation of the FGD results and quantitative trends. 
The major themes explored in the KIIs were the roles and responsibilities of the informant’s 
position, major pathways of change in the region (related to the informant’s area of expertise), 
the factors that drive such change, and the perception of future trends. The key informants were 
also asked about their understanding of and participation in PREG I interventions. The KII tools 
are provided in Annex 2. 
Table 4: Sites visited by qualitative team
County Sublocation Setting Type
Tana River Maroni Rural
Mikinduni Urban
Garissa
Waberi Urban
Bour-Algi Urban
Iftin Urban
Nanighi Rural
Turkana
Lokichar Rural
Lopur Rural
Nadapal Kakuma Rural
Lodwar Township Urban
Wajir El Nur Rural
Wagberi Urban
Marsabit Karare Rural
North Horr Rural
Isiolo Sericho Rural
Bula Pesa Urban
2.2.2 Fieldwork
Key to the effectiveness of the qualitative component was an intensive training of the field 
researchers. TANGO and Kimetrica staff conducted a five-day training for qualitative researchers
in Nakuru, Kenya. Participants were oriented to the overall purpose of the research, the central 
 20 | REAL | Resilience, Evaluation, Analysis & Learning
research questions, a discussion of resilience concepts and the REGAL theory of change. Training 
also involved in-depth discussion of the topical outlines for the FGDs and KIIs, including the 
translation of these tools into local languages. A tool for recording and documenting the results 
of FGDs and KIIs was presented and analyzed at length. A practical, flexible guideline for 
successful fieldwork was prepared collectively, and a pilot field test was conducted outside of 
Nakuru. Qualitative data collection occurred over a three-day period in each site; thus, field 
teams were able to observe the local context and gather complementary observational 
information. Day One consisted of visiting the county seat and conducting KIIs, followed by 
separate FGDs with men and women and local-level KIIs on Day Two. Day Three allowed teams 
to discuss their findings and enter data into data collection matrices before moving to the next 
site, helping to ensure that qualitative information was reliable and representative of the project 
area from which it was collected. Qualitative researchers were equipped with voice recording 
devices and had the option to record interviews after first gaining permission from participants. 
Supervisors in each team coordinated logistics, FGD recruitment, and identified key informants. 
2.2.3 Data Management and Security
Each team entered data into analytical data matrices. Field supervision was heavily oriented 
toward recording rich, complete data in a consistent format. During the first round of site visits, 
the qualitative research team leaders provided feedback to interviewers on their qualitative data 
notes and worked directly with teams to refine and improve data collection and data entry. 
Research team leads reviewed each data matrix over a minimum of two rounds to ensure 
completeness before the dataset was finalized. 
Data files were stored using Drobox cloud storage and backed up on flash drives, which were 
submitted to Kimetrica once data collection ended. Prior to release of data, any direct or indirect 
identifiers that could be used to identify individuals, such as geographic information at lower 
administrative levels, was removed and/or modified to prevent disclosure of individual identities.
2.2.4 Data Analysis
The study team entered the qualitative dataset, consisting of FGDs and KIIs, into a spreadsheet 
matrix organized by county, level of intensity of programming, FGD type (i.e., male or female), 
and designation as a rural or urban/peri-urban site. The analysts then carried out an initial 
content analysis to identify patterns of response by program intensity level and gender within and 
across counties. These patterns reflect local experiences with respect to changes in livelihood, 
livestock management, household and community adaptations to shocks and stresses, and 
reliance on external support. This report incorporates insights from the qualitative analysis in 
relevant sections. 
2.2.5 Limitations of the Qualitative Component
The team of researchers was highly competent, with good knowledge of the communities visited 
and many years of experience working in the survey areas. In the final assessment, the team was 
able to capture the processes and drivers of change necessary to inform the analysis of the 
PREG Initiative in Northern Kenya Endline Report | 21
impacts of the first five years of the PREG initiative and to add richness to the interpretation of 
quantitative measures of resilience. 
The major limitations of this component involve the restrictions of movement due to security 
risks. The teams only operated in locations that had been assessed and deemed a low security 
risk, potentially creating bias in the results. In other words, the qualitative findings are limited to 
reflecting the views and experiences of a subset of the project’s population, at least some of 
whom reside in more secure communities. That said, the field teams did encounter incidences of 
insecurity in the study areas. TANGO staff, for reasons of security, could not visit all the 
communities selected in the sampling process, so the importance of the training, including team 
supervisor training, was critical. Where possible, regular contact through daily phone check-ins, 
text and email communication was used to support the progress of each team for quality 
assurance and to ensure adequate coverage and representativeness across KIIs. A second 
limitation is that the qualitative sample was drawn at the county and sublocation levels in order 
to meet the needs of the PREG II impact evaluation baseline survey and was not based on 
participation in PREG I activities at the site (i.e., community) level.
 22 | REAL | Resilience, Evaluation, Analysis & Learning
3 ANALYSIS OF FINDINGS
This section provides information from the PREG I baseline and endline surveys about participation 
in PREG interventions (from the endline survey only), livelihood activities, coping strategies, 
household resilience capacities, and well-being outcomes. The analysis provides insight into how 
livelihoods, coping strategies, resilience capacity, and well-being outcomes, including poverty, food 
security, and perceived ability to cope with future shocks/stresses, have changed over time in six of 
the nine PREG I counties (Isiolo, Marsabit, Turkana, Mandera, Samburu, and Tana River).39 
3.1 PARTICIPATION IN PREG INTERVENTIONS
The endline survey included information about the participation of respondents or their household 
members in specific resilience interventions supported by PREG programs. Table 5 describes the 
categories of trainings and participation in project-supported groups that were captured in the 
household questionnaire.
Table 5: PREG-supported trainings and groups
Trainings Groups
Alternative livelihoods/income-generating activities Savings and loan groups (VSLA, REAP groups, BOMA 
groups, etc.)
Livestock production practices/health/management Livestock management associations (LMA, etc.)
Business/financial/accounting practices Women groups (GIRL model, etc.)
Crop production practices Mothers’ groups
Crop or livestock marketing Livestock producer groups
Rangeland management Self-help groups (ADESO groups, etc.)
Nutrition training Village committees (water management, NRM, 
conflict management health, school, etc.)
Savings/microfinance Ward level committees (WARD adaptation planning 
committee, etc.)
WASH practices
Conflict management
Life skills (hygiene/family planning, etc.)
Youth skills training (apprenticeship/internship, etc.)
Table 6 provides information about households’ reported participation in PREG development 
activities and trainings, and participation in project-supported groups in the endline survey. 
Overall participation in any activity (either trainings or groups) is quite low, and very similar 
across the HI and LI counties; overall participation is 15 percent in HI counties and 13 percent in 
LI counties. Participation in trainings is actually somewhat higher in the LI counties than the HI 
counties. More detailed information about participation by county is provided in Annex 5.
39 Due to security issues at baseline, data were not collected in Garissa, Mandera and Wajir counties, hence the comparison is for six 
counties only.
PREG Initiative in Northern Kenya Endline Report | 23
Table 6: Percent of households reporting participation in project activities, HI and LI counties
High Intensity Low Intensity Total Sample
Any participation (group and/or training) 15.1 13.1 14.1
Training participation 5.5 8.9* 7.2
Group participation 11.4 8.2 9.8
N 1,396 1,410 2,806
* Statistically significant difference between LI and HI groups at p<0.05
3.2 LIVELIHOOD ACTIVITIES 
Table 7 presents findings on the types of activities in which households engaged to generate 
sources of food and/or income over the course of the PREG I project. At endline, households 
remain largely dependent on livestock production and sales, with participation in farming, wage 
labor and self-employment decreasing significantly between baseline and endline.
Table 7: Household engagement in various livelihood activities, by survey round
Baseline
(Jan/Feb 2013)
Endline
(Aug/Sept 2018)
Livelihood activities (% HH)
Farming/crop production and sales 19.2 6.4*
Livestock production and sales 32.4 41.0
Wage labor 24.2 8.4*
Salaried work 27.3 19.7
Sale of wild/bush products (including charcoal) 13.6 26.6
Other self-employment/own business 23.2 7.4*
N 315 627
Traditionally, the primary livelihood system for most of northern Kenya has been based on 
livestock and rangeland management. Livestock herds moved between rainy-season and dry￾season pastures, with entire families establishing temporary residence near the herds. In these 
male-dominated pastoral societies, the size of the herd carried great status, and animals played a 
fundamental role in marriage and in the maintenance of pastoral households. Competition for 
pasturelands among ethnic groups, and raiding of animals from other rangelands, was a common 
strategy for increasing herd size. Individual herder families were reluctant to sell their animals on 
a regular basis and did so only when family necessity dictated it or in times of crisis. 
However, years of drought and banditry have decimated herds in the ASALs, resulting in major 
livelihood transformations. Many households are being “pushed out” of traditional livelihood 
systems, and entering into a semi-sedentary existence. Little (2016)40 describes the varying 
processes that livelihoods are experiencing as “stepping up” where households are intensifying 
40 Little’s analysis draws on: Catley, A. and Y. Aklilu (2012). Moving up or moving out? Commercialization, growth and destitution in 
pastoralist areas. In Pastoralism and development in Africa: Dynamic change at the margins, ed. A. Catley, J. Lind, and I. Scoones, 85–97. 
Abingdon, UK: Routledge.
 24 | REAL | Resilience, Evaluation, Analysis & Learning
their livestock activities, “stepping out” where households are complementing their livestock 
activities with other forms of income generation (agriculture, small business, etc.), and “moving 
out” where households are abandoning livestock production altogether. 
FGDs reveal that a combination of frequent drought since 2010, ever-more violent raiding, 
tension between land use for pasture and farming, modernization, and government policy has 
reduced the viability of traditional livestock activities and has ushered in an altered livelihood 
scenario. According to FGDs, the predominant pathways of change are characterized by 
decreasing engagement in livestock production, smaller herd sizes, herd compositions favoring 
more resistant breeds (e.g., goats, camels), reduced mobility and greater sedentarization, and 
shifting gender roles. Restocking programs initiated through the government and NGOs have 
helped some pastoral communities rebuild herds and re-engage in milk and meat processing 
activities (e.g., making dried camel steaks), particularly for women. 
Findings from the quantitative survey show an upward trend of households selling wild/bush 
products. Data from FGDs in both urban and rural communities suggest that charcoal
production, which is illegal, has become a primary source of income, particularly during the dry 
season and for the poorest households and communities. Despite the knowledge and risk of 
engaging in an illegal activity, multiple FGs expressed the perception that they “had no other 
option.” It is relevant to add that gathering firewood and making charcoal are onerous and taxing 
activities that require travel to remote areas with significant risks of criminal charges. It is neither 
a preferred livelihood strategy nor a specialized profession, but rather a type of “fallback” 
strategy that expands during difficult times. Further research into the specific types of wild 
products gathered for sale might help shed light on whether this can be considered a positive or 
negative shift in livelihood options.
The livelihood transformation over the last five years was evidenced across all communities and 
groups in the qualitative component of the survey, although it has not occurred in a uniform 
manner, and has different resilience outcomes for different communities. In large part, the specific 
context has influenced the nature of the transformation. For example, proximity to the massive 
Kakuma refugee camp in Turkana provides women with the opportunity to exchange washing
services for food, while others sell miraa (khat), firewood, and charcoal in camp markets. 
Elsewhere in Turkana, oil was discovered some years ago, creating a “boom” economy with the 
inflow of outsiders searching for employment. The Tullow Oil company created opportunities for 
local men and women to service the growing population, but the company abandoned the region 
in 2017, resulting in a localized depression. 
Similarly, urban and peri-urban communities across each of the six qualitative sample counties are 
experiencing significant livelihood changes toward modernization and away from more traditional 
pastoral livelihoods. In communities situated on the edge of urban areas or in close proximity to 
large markets and trading centers, FGD participants characterized most people now as “business 
people, [who] used to only depend on livestock.” At the same time, the quantitative findings indicate a 
reduction in self-employment. This apparent “discrepancy” between the quantitative and 
qualitative results may reflect—at least in part—the difference in perspective between 
respondents in the household survey and the considerably smaller qualitative sample size, as well 
PREG Initiative in Northern Kenya Endline Report | 25
as individual perspectives of what constitutes a “business” person. Differences in understanding—
or translating—respondent replies on the part of field teams can also contribute to perceived 
differences between qualitative and quantitative data.
While households still keep livestock, if possible, these are primarily one to two small livestock 
near the homestead and small herds (of 10–40 livestock) outside of town, where extended kinship 
networks, or hired herders in wealthier households manage livestock on nearby communal 
grazing lands. During periods of drought, households migrate further away to “greener pastures” 
or “hired rangeland.” 
Some households keep poultry and “kitchen gardens” if water is available, and may have two to 
three hectares of rainfed farmland for cash-crop production of vegetable, fruit and maize 
production (particularly when closer to rivers/water sources) (e.g., Tana River). Vegetables, milk, 
textiles, handicrafts, and household goods are sold in small shops and traded at nearby markets, 
primarily by women. In some communities of Tana River, for example, riverbank ecology provides
water resources for river-fed agricultural production, including cash-crop production of fruits and 
vegetables, as well as maize. Riverine resources also produce raw plant material for handicrafts. 
As part of the changes perceived by FG participants, it was repeatedly stated that gender roles 
had shifted in ways that would suggest an improvement in women’s status and an expansion of 
their economic roles in the household economy. Women not only engage more fully in livestock 
activities, including owning and marketing livestock, they are also the key leaders in diversifying 
the household economy. In Tana River, for example, where agriculture has become increasingly 
important, women make decisions on crop choice, provide labor, and market their products in 
urban areas. In times of crisis, women gather fodder for animals and sell their milk to urban 
clients. In Turkana, women harvest palm fronds to make and sell handicraft items, such as hats. 
Women in refugee camps also work as laborers, and sell and a variety of products (including 
miraa and fuelwood). 
In some rural communities that retain high dependence on livestock and pastoral livelihoods (e.g., 
Marsabit), women have become less involved with livestock, shifting into care for children at the 
homestead, in “new appreciation for education and a shift away from pastoralism” characterized 
by the seasonal migration of family groups and herds. In most communities, both male and female 
FG participants described women as adapting to the “modern lifestyle”. In an eroded livestock￾based livelihood, women have accepted more of the household income-generating burden, thus 
increasing their control, their mobility, and their status. In Isiolo, one women’s FG reflected on 
these shifting gender relations, noting that women have more of a voice, with greater 
engagement in the market and more opportunities for education. 
Qualitative insights from FGDs and KIIs suggest a process of adaptation to drought and other 
shocks/stresses, with livelihood diversification considered a key factor. Most households and 
communities are seeking alternatives to the livestock economy. For some households, this process 
appears to be driven by exposure to information or education, or linkages with external actors. 
For most, however, the qualitative data indicate a low level of resilience capacity, despite a 
general desire to diversify livelihoods and seek greater access to education, training, and financial 
 26 | REAL | Resilience, Evaluation, Analysis & Learning
services. The poorest households may be moving out, abandoning herding altogether and 
switching to low-paying labor markets, such as working in refugee camps or selling charcoal. 
According to FGDs and KIIs, some relatively better-off households—for example, those with more 
animals and access to markets and infrastructure— are complementing their livestock activity 
with small-scale farming oriented toward the market, even as farming has decreased generally 
between baseline and endline across the PREG I project areas. While communities often described 
experience with or intentions to diversify into farming, in some areas this effort at both the 
household and community levels has been compromised by insufficient access to water (e.g., 
Marsabit, Isiolo), and in other areas to pests, flooding or conflict over crop land (e.g., Turkana, 
Garissa). Others are diversifying into small business enterprise, according to some FGD 
participants, including some handicraft production. Some herders, considered wealthier, are 
choosing to sell most of their herd to invest in an urban-based dwelling and open a shop. 
Although insights from the qualitative data appear to contradict quantitative findings presented in 
Table 7, it should be noted that the qualitative insights are from a small sample size of people 
selected to participate in FGDs or KIIs from a small sample of sublocations across the PREG I 
implementation area and are not necessarily generally representative of the average PREG I 
beneficiary population, which is represented by the quantitative results. However, even as the 
quantitative data show certain trends or directional changes over time, insights from FGDs and 
KIIs can still provide relevant—though more localized—insights regarding household responses 
and behaviors, at least for some households.
3.2.1 Livelihood Diversification and Education 
FGD participants tended to closely link the ability of a household to diversify their livelihood 
strategies with education, training and access to financial capital. FGD participants generally 
perceived a great deal of household interest in education, increased enrollment, and a widespread 
desire for greater educational and training opportunities for youth in general and girls in 
particular. In one Garissa community, for example, women FGD participants emphasized the 
importance of education, adding that they “don’t want [our] girls to have the same fate [we] had.” 
Education is viewed as the pathway to livelihood diversification and adaptation to “modern life,” 
moving away from pastoralism and toward education. Diversification into farming, small business, 
and the establishment of savings groups is also often attributed to skills training. However, 
migration or dislocation of the household as a response to shock disrupts educational enrollment, 
as well as the more persistent lack of access to schools and teachers in the most rural, isolated, 
or insecure areas.
According to qualitative findings, young people and their families rely more on education as a 
pathway to a non-herding career than in the past. Even the most traditionally pastoral 
communities appear to favor a “modern” pathway of education and employment. Across FGDs, 
participants expressed a vision for the future that focused on access to primary and high school, 
even higher education for some. 
The aspiration to educate children for sustained employment is, however, tempered by the local 
economic reality. Especially in urban sublocations, FGD participants lamented a rise in social 
PREG Initiative in Northern Kenya Endline Report | 27
dysfunction. Many high school graduates do not find suitable employment and people complain of 
high rates of petty crime, particularly theft and prostitution. In one urban sublocation in Isiolo, 
female focus group participants talked of the social problem of “single mothers” (understood as 
sex workers who become pregnant) and increased rates of HIV infection. Thus, qualitative 
discussions suggest that some of those who benefitted from the pathway of education may have 
lacked opportunities to benefit from some of the advantages it affords.
3.3 COPING STRATEGIES
Coping strategies are actions households adopt to reduce the negative impact of shocks and 
stressors. The number of households that relied on at least one coping strategy in the face of 
shocks or stressors decreased between baseline and endline, but was not statistically significant 
(Table 8). At endline, fewer households than at baseline changed their source of income or food, 
or sold assets, both large and small. In contrast, more households at endline than baseline 
reported that one or more household member migrated as a way of coping with a shock or 
stressor. It should be noted that this is a partial list of coping strategies (thus percentages may 
not sum to 100 percent), multiple responses are possible (percentages may sum to more than 100 
percent), and some households “did nothing.”
Table 8: Household coping strategies and asset recovery, by survey round
Baseline
(Jan/Feb 2013)
Endline
(Aug/Sept 2018)
Households who adopted at least one coping strategy (%) 48.4 42.0
N 314 627
Types of adaptations (% HH)*
Changed income or food sources 36.1 1.4*
Added income or food sources 12.0 6.8
Migration of one or more household members 5.4 19.3*
Sold large asset 48.5 29.9*
Able to recoup some or all large asset(s) sold 0.3 0.3
Sold small asset 55.7 0.3*
Able to recoup some or all small asset(s) sold 0.2 0.0
HHs who adopted at least one coping strategy 152 269
ŧ Partial list of all possible coping strategies; full list includes “did nothing.” The sum of strategies may exceed 100% because 
this was a multiple-response question and households could identify more than one strategy.
* Statistically significant difference at p<0.05
As noted above, sales of large assets dropped significantly at endline, and there was no change 
between baseline and endline in households’ ability to recoup those assets once sold: at both 
baseline and endline, very few households reported they were able to recoup their assets—large 
or small—after selling them. The ability of a household to recoup an asset (e.g., livestock) after 
selling it as a way to deal with a shock or stressor implies some degree of “recovery” but is not a 
sufficiently robust measure of recovery for resilience analysis and is not used as such here. 
 28 | REAL | Resilience, Evaluation, Analysis & Learning
More dramatically, the practice of selling small household items plunged to nearly zero at endline. 
This may be a result of households no longer having assets to sell, having divested of assets in 
response to shocks over time or having lost assets due to shocks. In some qualitative FGDs, 
participants described severe loss of livestock, in particular, as a result of drought, as well as 
shelter and productive assets (e.g., water pumps, farm implements) due to flooding (Marsabit, 
Wajir, Tana River, Garissa).
The relationship between asset sales and the ability to recoup those assets is quite complex in 
light of the ongoing transformations occurring in northern Kenya. The qualitative research found 
that some communities (e.g., Wajir, Garissa, Isiolo) indicated they have taken up destocking as a 
strategy to acquire cash for savings or purchase food and fodder and “save remaining livestock” 
during a drought. While not widespread, FGD participants described this strategy as distinct from 
and more effective than government destocking programs (e.g., Isiolo) or distress sales. More 
commonly, however, the divestiture of a major asset, such as part of the livestock herd, 
represents a coping response to a shock, such as drought. In several communities across all 
counties, FGDs reported selling livestock to meet “urgent” needs—specifically school fees and 
medical expenses, which in turn has led to decreased herd sizes. Some indicated that selling 
livestock to meet basic needs is a more recent practice, most common among those with smaller 
herd sizes (i.e., the less well-off). 
The qualitative findings also suggest that, in some cases, livestock asset sales may represent the 
process of transforming to another livelihood configuration: livestock sales could have other—or 
multiple—drivers beyond shocks. In contrast to the findings presented in Table 7 on livelihood 
practices at baseline and endline, several FGDs note that some households with livestock sell 
animals, sometimes the entire herd, to invest in a less risky activity such as a small business. 
Again, differences between quantitative results and insights from qualitative FGDs and KIIs may 
simply reflect dramatically different samples—one a very small subset of the other; qualitative 
findings are considerably more local than quantitative averages across the entire project sample.
Those households in a more precarious economic situation may similarly divest themselves of
productive assets not only as a response to a shock, but also as part of a wider strategy of 
diversification. The FGDs also revealed a strategy of more advantaged households in which part 
or all of the herd is sold at time of crisis, and the resources are set aside to purchase of a new 
herd once the stressors have diminished. 
Multiple FGDs described the widespread and devastating loss of herds in recent droughts (i.e., 
2017, 2019), as well as reductions in herd size due to conflict, banditry and disease. Among those 
with the means to retain herd sizes, low prices (e.g., Garissa) are a disincentive to sell livestock.
PREG Initiative in Northern Kenya Endline Report | 29
3.4 RESILIENCE CAPACITIES
This section provides information on a subset of household resilience capacities. Households with 
strong resilience capacity often diversify their livelihoods and adapt their farming and pastoral 
systems to future uncertainty or variable conditions, such as climate change.41 
3.4.1 Assets and Savings
The average number of assets owned decreased from 5.9 to 4.5 over the survey rounds, though 
the change is not statistically significant. From this finding, we can infer that the decrease in sales 
of large and small assets as a coping strategy (Sec. 3.3, Table 8) was not necessarily due to lack 
of assets to sell. 
3.4.2 Livelihood Diversity
Livelihood diversification provides flexibility, resources, and skills, all of which contribute to 
household capacity to deal with shocks/stressors. Between the baseline and endline, households 
reported engaging in slightly fewer different livelihood activities, dropping from 1.5 to 1.2 
(p<0.05). Engaging in fewer livelihood strategies may limit households’ ability to adapt to 
changing conditions—even if they were highly motivated or confident to do so. 
3.4.3 Reliance on Relief
Households were asked if they relied on relief or assistance (e.g., cash, food aid, cash/food for 
work, asset transfer, etc.) as a source of income or food. Reliance on relief decreased from 14.8 
percent to 6.5 percent, though the change is not statistically significant. Qualitative findings 
suggest that most communities received some form of relief (e.g., food aid, cash, fodder, seedlings, 
vaccines, security, malaria nets) during times of crisis over the previous five years, if not during 
the last 12 months. In some cases, FGDs credited school feeding with minimizing “starvation,” and 
government and NGO disaster response was described as a “great help” in the face of drought. 
In Marsabit, for example, FGDs credited relief in response to drought and floods, layered with 
development activities (e.g., investments in water infrastructure and education, provision of health 
care and veterinary services, savings groups, and livelihood diversification such as farming—as 
critical to their overall capacity to cope with shocks and stresses. Other communities in the PREG 
I area, however, reported that while they perceive a need for relief, there has been little to no aid 
of any kind available and where provided, that it was often too slow, too limited in duration, and 
insufficient relative to need. Food aid is often described as helpful but insufficient, particularly for 
the poorest households who depend almost entirely on outside assistance.
3.4.4 Social Capital
Social capital refers to the bonds between community members and across communities. It 
involves principles and norms such as trust, reciprocity, and cooperation, and is often drawn on in 
41 Béné et al. 2012.
 30 | REAL | Resilience, Evaluation, Analysis & Learning
the disaster context, when survivors work closely to help each other to cope and recover.42 In the 
baseline survey, respondents were asked if their household relied on others for food support 
(financial or in-kind) during the last drought (which occurred in the year prior to the baseline), 
and if so, who were the sources of support. We refer to this as “use of social capital”. A second 
indicator of social capital includes a measure of whether respondents indicate they think they 
could rely on others in the future during times of stress; it differs from the first indicator in that it 
measures the respondent’s perceived potential for using social capital. For both questions, a 
positive response prompts them to indicate the sources of support, including:
• Relatives in my village/community;
• Relatives outside my village/community;
• Non-relatives in my village/community;
• Non-relatives outside my village/community; and
• Non-relatives outside my tribe/ethnic group.
Table 9 (next page) shows the percentage of households that report relying on others in the past 
(use of social capital) and the percentage that believe they can rely on others in the future 
(perceived availability of social capital), and of those households responding positively to each 
question, the types of social networks they rely on. The table also reports the average number of 
sources of support. 
About thirty percent of households at baseline and forty percent of households at endline report 
they had been able to rely on someone else for financial or in-kind support when needed. While 
relying on relatives within a household’s own village or community was the most commonly used 
form of support at both the baseline and endline, there was a significant shift in households relying 
on those outside one’s tribe or ethnic group. Correspondingly, the number of sources of social 
support that households relied on increased by 0.5. 
The statistically significant, and substantial, increase in the percentage of households reporting 
they sought support from non-relatives outside of their tribe or ethnic group, from 6.7 percent at 
baseline to 31.2 percent at endline, may reflect people needing to “cast their nets” wider than 
their immediate family. This explanation is consistent with other studies showing that households’ 
ability to share food, livestock, etc. during times of stress can diminish over time if the shock/stress 
(e.g., drought) is prolonged;
43 available resources simply run low, or run out completely. 
Numerous FGDs (e.g., Isiolo, Marsabit, and Turkana) reported that households were, in general, 
not less reliant on relief efforts and external actors—rather, for many, it was not enough and 
households sought additional support through their social networks to cover gaps. It is also 
possible that households have more sources of support from which to draw at the endline than 
42 Frankenberger, T., Mueller, M., Spangler, T., and Alexander S. (2013). Community Resilience: Conceptual Framework and Measurement 
Feed the Future Learning Agenda. Rockville, MD: Westat. https://www.fsnnetwork.org/community-resilience-conceptual-framework-and￾measurement-feed-future-learning-agenda. 43 Frankenberger T and L Smith. 2015. Ethiopia Pastoralist Areas Resilience Improvement and Market Expansion (PRIME) Project Impact 
Evaluation: Report of the Interim Monitoring Survey 2014–2015. November 2015.
PREG Initiative in Northern Kenya Endline Report | 31
they did at the baseline, for example through engagement in markets, livelihoods activities, or 
collective action, all of which can expand households’ social networks and build social cohesion. 
There was a statistically significant increase between baseline and endline in people’s perceptions 
of being able to rely on support from others if they needed it in the future (perceived availability 
of social capital), in almost all types of sources of social support. Of those who reported they 
could rely on others to help them deal with a future shock or stress, “relatives in my 
village/community” was again the most commonly reported source of perceived future support, at 
both baseline and endline. Moreover, the number of different types of social support households 
perceived they could rely on in the future increased from 1.7 to 2.8, a statistically significant 
increase. This suggests that people are fairly “optimistic” about where they think they can seek 
support when they need it; it is possible that this is due to the expanded social networks resulting 
from participation in PREG activities.
Table 9: Use and availability of social capital, by survey round
Baseline
(Jan/Feb 2013)
Endline
(Aug/Sept 2018)
% of households that relied on others 
(use of social capital)
31.5 40.3
N 315 627
Types of social networks (%)
Relatives in my village/community 69.6 67.1
Relatives outside my village/community 39.5 51.1
Non-relative in my village/community 30.2 42.1
Non-relatives outside my village/community 19.7 21.1
Non-relatives outside of my tribe or ethnic group 6.7 34.2*
Average # of sources on which household relied (max range: 
0–5) 1.7 2.2*
HHs that relied on others 130 280
% of households that can rely on others in future shocks 
(availability of social capital)
57.0 88.5*
N 315 625
Types of social networks (%)
Relatives in my village/community 77.1 82.3
Relatives outside my village/community 45.0 70.9*
Non-relative in my village/community 29.2 46.6*
Non-relatives outside my village/community 12.9 46.0*
Non-relatives outside of my tribe or ethnic group 6.3 35.6*
Average # sources on which household can rely in future (max 
range: 0–5) 1.7 2.8*
HHs that can rely on others in future shocks 197 558
* Statistically significant difference at p<0.05
PREG Initiative in Northern Kenya Endline Report | 35
The qualitative data indicate that the strength and nature of social capital and collective action 
vary across communities over time and in relation to other sources of external support. Multiple 
communities, both urban and rural, reported numerous examples of collective action and coping 
and adaptation strategies that they rely on and that are, in many cases, embedded in intra￾community traditions of resource sharing within family and community groups. “The community 
depends on each other as families, close clan members and good neighbors” (Marsabit). For example, 
communities support each other by providing food, clothing, fodder, water sharing, loans of 
livestock for milk, and housing materials. In some areas, community leaders mobilize food sharing 
and fundraising efforts, and organize wage labor activities for the poorest families. Savings 
groups, namely VSLAs, were cited as a common way for “savings sharing.” 
In some cases, however, social networks have deteriorated over time. The qualitative findings 
suggest that for some communities, critical bonding social capital (e.g., relatives within the 
respondents’ community) that represents a first line of defense in times of crisis may have become 
overtaxed due to recurrent climatic shocks, conflict over increasingly scarce natural resources, 
and the loss of productive household assets. This may underlie—at least to some degree—the 
quantitative findings presented in Table 7, which indicate that: i) more households sought help 
from non-relatives outside of their tribe or ethnic group, and ii) more households think they will 
need to use support sources beyond their immediate relatives, both within and outside of their 
own community, to help them deal with shocks/stresses in the future. As stated in one Turkana 
FGD: 
Inter-ethnic conflict and insecurity, theft, and assault also contribute to the breakdown of social 
bonds. Participants in some communities reported increasing levels of “mistrust,” “crime,” 
“prostitution” and “single mothers” as warning signs of social dysfunction. Multiple FGDs, both 
rural and urban, reported that distrust within the community is heightened when natural 
resources are stressed. 
In some areas, collective action centers on community initiatives to manage natural resources 
such as water catchment points and enclosures, manage communal pasture to buffer against 
drought, as protect forests. However, only three of 16 communities (Turkana, Garissa, and 
Wajir) stated that they conduct annual disaster risk reduction meetings and develop planning and 
response plans. According to FGDs, a number of areas engage in “collective problem solving” and 
maintain clearly delineated mechanisms to address intra-community and intertribal conflict. 
Community-based engagement with government officials and NGOs to advocate for services or a 
role in local development planning was reported in some FGDs (e.g., Tana River, Isiolo). In some 
areas (e.g., urban areas), local peace committees and security teams, comprised primarily of local 
youth and community leaders, address issues of insecurity and assault through patrols and dispute 
mediation. 
“When we had livestock, we used to help each other; but now since we are in the same 
economic level, this is no longer possible.”
 36 | REAL | Resilience, Evaluation, Analysis & Learning
FGDs also indicate that while people continue to borrow from family members and others within 
their communities—namely from better-off households and shop-owners on credit—these sources 
of support have been drawn down, as one FGD in Tana River summarized:
In Turkana, for example, FGDs noted that in some cases better-off households affected by recent 
shocks were less able to fulfill their traditional role of providing loans, food and other support for 
community members during times of need. In some sites, FGDs reported growing mistrust within 
the community around borrowing and debt, with shopkeepers less willing or able to allow people 
to purchase food and basic goods on credit during times of stress, described as common practice 
in the past. In a number of study sites, opportunities to borrow within the community during 
periods of droughts or floods have been constrained by stress on lenders (both cash and in-kind 
commodities and foodstuffs).
The extent to which these proximal sources of social capital have waned over time may have 
triggered the need to access sources of social support outside of one’s family, community, or 
ethnic group, as indicated in the quantitative findings. However, there is also some indication that 
investments in market systems may have contributed to broader social networks, at least for 
some households. Although there is no clear evidence of such an expansion of social networks in 
the community-level FGD data, some KIIs and market visits (Isiolo, Marsabit) suggest that new 
markets and market linkages provide an opportunity for increased interaction and networking 
among previously more disparate groups. Though plausible, current data (quantitative and 
qualitative) do not capture whether this translates into people feeling comfortable enough, or 
willing, to rely on what may be newly established networks for helping them deal with a future 
shock. 
3.4.5 Agency
Information on household agency was measured at baseline and endline. Agency is measured by 
views related to the perception that one has little control over one’s future. It was organically 
measured because perceiving that one has little control over one’s future has been highly 
correlated with negative outcomes in the face of recurrent droughts in other studies in the Horn 
of Africa.44 Results are not included in this report due to measurement error potentially related to 
differences in translation of the survey questions in the baseline and endline.
44 Frankenberger et al. 2013.
“There are limits on sharing resources because every household feels the stress, they do receive 
food aid during the crisis .... In the past, it was easy to solicit help from neighbors, but not so 
today, and loans only make the situation worse. People do not share as they once did, except in 
time of death.” 
PREG Initiative in Northern Kenya Endline Report | 37
3.5 HOUSEHOLD WELL-BEING OUTCOMES 
Table 10 presents the results for four measures of well-being: expenditures, poverty level, 
moderate/severe hunger, and perceived ability to cope with future shocks. At the end of the first 
five-year PREG initiative, households had less money to spend, their hunger status (HHS) had 
significantly increased, and fewer households indicated being able to cope without difficulty in 
facing future shocks. More households were in poverty at endline, but this was not a statistically 
significant change. 
Table 10: Household well-being outcomes, by survey round
Baseline
(Jan/Feb 2013)
Endline
(Aug/Sept 2018)
Expenditures (USD mean) 2.07 1.28*
N 315 627
Poverty (% HH below 1.25USD) 59.9 74.9
N 315 627
Moderate or severe hunger (HHS, % HH)45 43.2 70.3*
N 277 615
Perceived ability to cope with future shocks (% HH)
Unable to cope 30.4 46.5
Able to cope, with adjustments 36.9 43.1
Able to cope, without adjustments 32.7 10.5*
N 315 614
* Statistically significant difference at p<0.05
The qualitative data show that the frequent, persistent and compounding nature of the shock 
context has, in many cases, undermined efforts of households and communities to effectively 
recover: shocks and stresses are a constant feature of their lives, rather than a single-event 
episode. While there appears to be a great deal of variability in the experience of shocks across 
households, a number of communities reported experiencing intense flooding, drought and 
livestock disease during 2017, among other shocks, compounded by the erosion of assets and 
resources (e.g., homes, livestock, productive technologies, grazing land) that have collectively 
reduced their capacity to recover. Across the qualitative sample, for example, livestock disease 
was widely reported and indicated as a driver, along with climatic shocks, of reduced herd size or 
a shift out of pastoralism. Most communities indicated lack of veterinary services as a major 
challenge, with services described as “inconsistent” and typically provided only in response to 
disease outbreaks (e.g., rift valley fever, foot and mouth disease)—a response often described as 
inadequate or too late to prevent livestock death or support timely recovery.
45 The Household Hunger Scale (HHS) is a perception-based food deprivation scale that measures the prevalence of households with 
moderate or severe hunger. Answers to three of nine frequency-of-occurrence questions pertaining to the most severe forms of food 
insecurity are used to construct a score on a scale of 0 to 6. The prevalence of hunger is then calculated as the percentage of households 
whose scale value is greater than or equal to two, which represents “moderate to severe hunger.”
 38 | REAL | Resilience, Evaluation, Analysis & Learning
3.5.1 Relationship between Household Characteristics and Subjective 
Resilience 
Table 11 shows the values of various resilience capacity components for households with 
different perceived abilities for coping with future shocks (i.e., subjective resilience) at baseline 
and endline. For each component, levels of perceived ability to cope with future shocks are 
categorized according to participants’ responses as “unable,” “able, with adjustment” or “able, 
without adjustment.” Tests of statistical significance were conducted relative to the “unable” 
category for each survey round, i.e., comparing i) “unable” and “able, with adjustment” values at 
baseline and ii) “unable” and “able, with adjustment,” values at baseline; the same tests were 
performed for endline values. The asterisks represent significant results for those comparisons. In 
addition, the baseline and endline values for each level of ability were tested for statistical 
significance, e.g., “unable” values at baseline are compared to “unable” values at endline. 
Significant differences for the latter comparison are indicated with a superscript “a.” 
The trends in these data align with our hypothesis that higher levels of resilience capacity 
correspond to greater subjective resilience. At baseline, households with greater subjective 
resilience (able to cope without making adjustments) had more types of assets and higher human 
capital (as indicated by the percentage of households with at least one adult with at least a 
primary education) and tended not to rely on emergency relief (e.g., food aid, cash, fodder, 
seedlings, vaccines, malaria nets, etc., but were less likely to report they could ask others for 
support in times of need compared to households who reported being unable to cope with future 
shocks. It is notable that considerably fewer households at endline reported being able to cope 
with shocks/stressors without making any adjustments, compared to baseline. The results 
suggest that by endline, households generally perceived themselves as being less able to deal 
with future shocks/stressors but also felt more strongly that they could rely on others for help. 
As previously noted, the qualitative data indicate that even those households indicating they are 
able to cope with future shocks without adjustments were affected by recent shocks and incurred 
debt that interfered with their traditional role of providing loans, food and other support within 
their communities during times of need. 
Table 11: Resilience capacity components for three categories of subjective resilience, by survey round 
Resilience capacity 
components
Ability to cope with future shocks
Baseline
Ability to cope with future shocks
Endline
Unable
Able,
with 
adjustment
Able, 
without 
adjustment
Unable
Able,
with 
adjustment
Able, 
without 
adjustment
Perceived ability to 
cope with future shocks 
(% HH)
30.4 36.9 32.7 46.5 43.1 10.5a
Able to rely on others 
in future (% HH) 65.6 67.1 37.3* 83.5 92.6** 96.4***
Livelihood diversity 
(avg. # livelihoods) 1.4 1.8** 1.3 1.1 1.2 1.2
PREG Initiative in Northern Kenya Endline Report | 39
Resilience capacity 
components
Ability to cope with future shocks
Baseline
Ability to cope with future shocks
Endline
Unable
Able,
with 
adjustment
Able, 
without 
adjustment
Unable
Able,
with 
adjustment
Able, 
without 
adjustment
Asset index (avg. # 
asset types) 3.4 5.8** 8.4*** 3.2 5.4*** 6.3***
Adult education (% HH 
with at least one adult 
with at least primary 
education)
37.1 63.7* 73.1*** 41.0 67.6* 77.7**
Relied on relief as a 
source of income or 
food in the past 12 
months (% HH)
21.3 16.6 6.7*** 9.1 3.7 6.4
N 123 118 74 234 307 71
a Statistically significant difference (p<0.05) between baseline and endline. 
*, **, and *** represent statistically significant differences at p<0.05, p<0.01, and p<0.001, respectively, compared to the 
“unable” reference category within each survey round. 
Given that promotion of livelihood diversity is a prominent element of PREG, in Table 12 we 
present a more in-depth look into how subjective resilience is associated with the types of 
livelihoods in which households are engaged. As in Table 11, levels of perceived ability to cope 
with future shocks are categorized according to participants’ responses as “unable,” “able, with 
adjustment” or “able, without adjustment.” Tests of statistical significance were conducted relative 
to the “unable” category for each survey round, i.e., comparing i) “unable” and “able, with 
adjustment” values at baseline and ii) “unable” and “able, with adjustment,” values at baseline; the 
same tests were performed for endline values. The asterisks represent significant results for those 
comparisons. In addition, the baseline and endline values for each level of ability were tested for 
statistical significance, e.g., “unable” values at baseline are compared to “unable” values at 
endline. Significant differences for the latter comparison are indicated with a superscript “a.” 
At both baseline and endline, households that perceive themselves as unable to cope with future 
shocks were predominantly engaged in livestock production/sales and sales of bush products. This 
perception was also relatively common among wage laborers at baseline. 
Comparing households that are unable to cope with those that are able to cope with or without 
adjustments, the former tend to rely on selling bush products, a high-risk labor-intensive livelihood 
activity. This is true at both baseline and endline. Conversely, a much smaller proportion of 
households that are unable to cope are engaged in salaried work, self-employment, or crop 
production and sales.46
46 At endline, few households are engaged in crop production and sales, irrespective of their perceived ability to cope with future shocks. 
 40 | REAL | Resilience, Evaluation, Analysis & Learning
Livelihood activity
Ability to cope with future shocks
Baseline
Ability to cope with future shocks
Endline
Unable
Able,
with 
adjustment
Able, 
without 
adjustment
Unable
Able,
with 
adjustment
Able, 
without 
adjustment
Livestock production & 
sales 39.5 36.5 21.7* 33.7 48.4 36.7
Crop production & 
sales 10.8 33.3* 11.6 4.0 8.0 9.6
Wage labor 25.6 34.8 11.3* 9.3 8.8 0.4*
Self-employment 13.4 33.0* 21.4 2.6 12.0* 8.1
Salary work 7.7 20.9* 51.8* 8.8 23.2** 56.1*
Sales of bush products 26.8 11.4* 4.0* 48.1 9.8* 4.0*
N 123 117 74 234 309 71
Note: Significance tests were performed on baseline and endline values but there were no significant differences.
*, **, and *** represent statistically significant differences at p<0.05, p<0.01, and p<0.001, respectively, compared to the 
“unable” reference category within each survey round. 
3.6 CHILD NUTRITION STATUS
This section reviews three anthropometric measures of undernutrition in children under age five in 
the study area: stunting, wasting), and underweight. Child undernourishment can have serious 
short-term and long-term consequences, including higher mortality risk. 
Stunting is an indicator of linear growth retardation, most often due to a prolonged inadequate 
diet and poor health. Reducing the prevalence of stunting among children, particularly age 0–23 
months, is important because linear growth deficits accrued early in life are associated with 
cognitive impairments, poor educational performance, and decreased work productivity as adults. 
Stunting is a height-for-age measurement that reflects chronic undernutrition. Wasting is an 
indicator of acute malnutrition; children who are wasted (i.e., low weight-for-height 
measurement) are too thin for their height and have a much greater risk of dying than children 
who are not wasted. Underweight is a weight-for-age measurement and reflects acute and/or 
chronic undernutrition. 
Table 13 shows no statistically significant differences between baseline and endline for any of the 
anthropometric indicators. 
Table 12: Household engagement in livelihoods activities by subjective resilience category, by survey round (% households) 
PREG Initiative in Northern Kenya Endline Report | 41
Baseline
(Jan/Feb 2013)
Endline
(Aug/Sept 2018)
Stunting
Moderate-to-severe stunting, <-2SD (%) 28.8 22.9
Severe stunting, <-3SD (%) 12.0 6.5
Underweight
Moderate-to-severe underweight, <-2SD (%) 20.5 18.7
Severe underweight, <-3SD (%) 4.5 2.0
Wasting
Moderate-to-severe wasting, <-2SD (%) 12.3 11.3
Severe wasting, <-3SD (%) 4.3 1.3
HHs with children under 5 212 262
* Statistically significant difference at p<0.05
Table 13: Prevalence of stunting, underweight and wasting across survey rounds for children under 5 years of age 
PREG Initiative in Northern Kenya Endline Report | 39
4 OBSERVATIONS AND CONCLUSIONS
This report presents findings from the PREG I northern Kenya baseline and endline surveys. The 
pre-post analysis compares endline values to those estimated in the sample of households in the 
six county sublocations surveyed at baseline. The report provides estimates of household 
resilience measures and uses qualitative data for context and to help explain the quantitative 
findings. It also provides results from analysis using a DID approach to compare differences 
between survey rounds for LI households to differences between survey rounds for HI households
(see Annex 3 for more details).
4.1 KEY FINDINGS RELATED TO RESILIENCE ANALYSIS
4.1.1 Livelihood Activities
Livestock production and sales was the most common livelihood activity of all households in both 
surveys. At endline, fewer households cited farming, wage labor and self-employment as a 
livelihood activity. Qualitative focus groups indicated widespread desire to diversify into small￾scale farming; for many, however, recent investments in small-scale agriculture have been 
undermined by droughts and floods. Of note was a slight but significant increase in reliance on 
wild harvested products at endline and qualitative findings also indicate some reliance on wild 
foods. Harvesting wood for charcoal production has become a primary activity across study sites, 
despite its illegality and advocacy against charcoal production. 
The quantitative data suggest a shift in livelihood strategies among households in the 
programming areas. In particular, households are “stepping out” and “moving out” of livestock 
activities; either complementing them with additional sources of income or abandoning livestock 
as a source of income altogether. This finding is supported by qualitative data, which suggest that 
frequent drought, violent raiding, tension between land use for pasture and farming, 
modernization, and government policy have combined to reduce the viability of the traditional 
pastoral livelihood and to usher in an altered livelihood scenario. 
Findings from the DID analysis showed that the decrease from baseline to endline in households 
engaging in wage labor was significantly greater for HI households than for LI households (Table 
14; Annex 3). Otherwise, differences between survey rounds for HI households were not 
significantly different than those for LI households for any other livelihood activity.
4.1.2 Coping Strategies
A common household strategy for dealing with shocks/stresses involves selling assets. Sales of 
large assets are considered “distress sales” and exclude routine livestock sales. Households relied 
less on selling large assets at endline, although almost 30 percent of households continued to 
engage in this practice. Given the large-scale loss of livestock over the life of the activity, 
however, this is not necessarily surprising, as households may simply have had fewer livestock to 
sell at endline. According to some focus groups, their communities still had not recovered from 
the 2010 drought in which entire communities lost nearly all their livestock. Households were 
 40 | REAL | Resilience, Evaluation, Analysis & Learning
significantly less likely to sell small assets at endline compared to baseline, and significantly more 
likely to rely on migration (although the prevalence of migration was lower than that of selling off 
large assets). The DID analysis indicated that the difference between survey rounds for HI 
households was not significantly different than that for LI households (Table 15; Annex 3). 
4.1.3 Resilience Capacities
Livelihood diversification. Overall, livelihood diversification was very low in the six PREG 
counties. At baseline, households relied on approximately 1.5 sources of income and/or food; this 
decreased to 1.2 at endline. There was no significant difference in how livelihood diversification 
between survey rounds changed for HI and LI households (Table 16, Annex 3). 
Assets. Asset ownership decreased over time, but the change is not statistically significant. DID 
analysis showed no significant difference in the change in asset ownership for HI and LI 
households over time (Table 16, Annex 3). 
Social capital. When asked whether households could rely on others for financial or in-kind 
support during the drought prior to each survey round, about 30 percent during the baseline and 
40 percent during the endline indicated they could. Households reported that they could rely on 
approximately two different sources of support at endline, a significant increase from the first 
survey round. The most commonly cited source of support was relatives within a household’s 
village/community. Though relatives outside of a household’s village/community was were a fairly 
common source, there was a noticeable increase in reaching out to non-relatives and to people 
outside of a household’s immediate circle, including individuals in other tribes or ethnic groups. 
Qualitative findings suggest this may be due in part to an erosion of bonding social capital, driving 
greater reliance on more-distant sources of social support, including new markets and market 
linkages. The qualitative data also indicate that households that were able to cope with future 
shocks without adjustments were greatly affected by recent shocks and incurred debt that 
inhibited their traditional role of providing loans, food and other support within their communities 
during times of need. Shop owners were less willing or able than in the past to allow people to 
purchase food on credit. This is consistent with other findings that the rate of borrowing dropped 
drastically at endline. 
Between baseline and endline, the increase in reliance on others was significantly more for LI 
households compared to HI households, specifically among non-relatives both within and outside 
of one’s community (Table 17, Annex 3). The number of sources on which a household relied 
during the last drought increased more for LI households compared to HI households. Although 
there was no significant difference between the change from baseline to endline in whether a 
household would be able to rely on others in the future among HI and LI households, LI 
households were more likely than HI households to report they could rely on relatives outside of 
their village or community and on non-relatives within their village or community.
Humanitarian assistance. Reliance on humanitarian assistance was very low in both survey 
rounds (15 percent at baseline and 6 percent at endline), although the changes between survey 
rounds for HI and for LI households were not statistically different (Table 16, Annex 3). 
PREG Initiative in Northern Kenya Endline Report | 41
Qualitative data suggest that—although diminished—most communities still received some form 
of assistance during times of crises, including food aid, cash, fodder, seedlings, vaccines, malaria 
nets, etc. The qualitative data suggest that although often too slow, too limited in duration, and 
insufficient overall, emergency humanitarian assistance layered with development activities was 
critical for dealing with shocks and stresses. 
4.1.4 Well-Being Outcomes
Economic indicators. The number of households living at or below the poverty line of USD 1.25 
per day was higher at endline, though the difference from the baseline was not significant. Per 
capita daily expenditures (i.e., a proxy for income) declined significantly. 
Food security. Moderate or severe household hunger increased significantly from baseline to 
endline. 
Child nutritional status. Generally, the trend was toward improvements in all three measures of 
child nutrition (stunting, wasting and underweight) by endline, though none of the changes were 
statistically significant. 
Household ability to cope with future shocks or times of stress. When asked about the 
household’s ability to cope with future droughts and other stresses, households felt more 
confident at endline than at baseline in their ability to cope with shocks to their sources of income 
or food. At endline, households are more likely to believe in their ability to cope if they are able 
to rely on others, and own more assets. 
Table 18 and Table 19 (Annex 3) show that the differences in well-being outcomes between 
survey rounds for HI and LI households were not statistically different for any of the four 
measures of well-being (expenditures, poverty level, moderate/severe hunger, and perceived 
ability to cope with future shocks). 
4.2 PROGRAM IMPLICATIONS
Across the PREG I program area, there were large increases in reliance on people outside of 
one’s own village, community, and/or ethnic group for financial and in-kind support during the 
latest drought. This pattern could reflect the impact of PREG I initiatives linking communities to 
markets, services, and support institutions. It could also signal that bonding social capital has 
weakened in that everyone’s capacity for sharing (e.g., low food stocks, fodder, etc.) has been 
exhausted and people need to reach further out than their immediate—and more traditional—
circles for support.
The insights from the qualitative indicate that the process of development envisioned in the theory 
of change has been experienced in different ways by different communities, but not necessarily 
influenced by PREG I activities. The qualitative sample included some communities that seemed 
more engaged in transformative changes, as captured by the level of local governance and 
collective action, the intensity of presence of local county agencies, and the range of activities 
 42 | REAL | Resilience, Evaluation, Analysis & Learning
offered by NGOs and other actors. In these seemingly more-resilient communities, study 
participants described improvements in livelihood diversification leading to improved child nutrition; 
increased collective action and community cohesion; increased access to services (e.g., veterinary 
and human health care, insurance), technologies (e.g., mobile phones, refrigeration, farming 
practices), and education; and improvements in infrastructure (e.g., markets, roads, water points). 
In more resilient communities, as described in the previous paragraph, FGDs reported a strong role 
of county government and a range of external actors providing support over the previous five to ten 
years, layered with emergency humanitarian assistance (e.g., Wajir, Marsabit). In one site in Wajir, 
for example, study participants described the role of county government to improve community 
infrastructure and technical assistance, including training, along with consistent support from NGOs, 
coupled with support from government and NGOs during droughts over the previous five years as 
resulting in a positive change in their response capacity. This external support leverages community￾based pasture and natural resource management, annual disaster risk reduction planning meetings 
(at least in some communities), intra-community support systems (e.g., contributions for home 
construction), and informal safety nets such as local-level cash-for-work initiatives to support the 
neediest members of the community. These communities show a commitment to collective action 
and an ability to form effective partnerships with external resource providers, and project a vision of 
the future. It is not possible to determine at this point why one community holds this perspective, 
while others do not. The FGDs, however, do suggest that location close to a more urban area favors 
access to resources and project interventions. Although such insights may suggest that proximity to 
an urban area means a community might be more resilient, additional analysis with quantitative 
data would be needed to verify or refute such a possibility.
The goal of PREG I to enable the development of a flourishing livestock industry consistent with 
the vision of the Kenyan government is a difficult one to achieve. If PREG I meant to target the 
most vulnerable groups in northern Kenya, its intervention strategies, particularly the livestock 
value chains, were only realistically available to a limited number of households. The challenge 
facing the development community is to also accommodate those who are shifting into more 
diverse portfolios of livelihood activities, such as farming and small business. Many communities 
expressed a desire to educate their children, to expand training opportunities that could support 
livelihood diversification, and to support investments in technologies and infrastructure to support 
these aspirations (e.g., roads, electrification, market access, drought-tolerant seeds and livestock, 
health care facilities, access to water / irrigation). While many community FGDs articulated the 
capacity and desire to address problems and constraints through collective action, the data show 
a general need and desire for external support to facilitate a process of transforming livelihoods 
and managing shocks and stresses.
PREG Initiative in Northern Kenya Endline Report | 43
5 RECOMMENDATIONS
The section highlights key observations and associated recommendations relevant to future 
resilience programming in the ASALs of northern Kenya. 
• In the ASALs of northern Kenya, the frequency of droughts, inter-ethnic conflict over a
diminishing resource base, and devolution of government responsibilities have all
accelerated a major transformation of the traditional pastoral livelihood. Thus,
potential program activities that focus on income generation and diversification should
be prioritized.
• Devolution has increased the presence of government across the ASALs, including
increased investment in infrastructure such as livestock markets, services, and safety
nets. The county development planning system and other NGO and development
actors in the area should be leveraged for more effective PREG coordination and
implementation, particularly around strengthening community-level disaster planning
and response mechanisms (e.g., early warning systems, DRR strategies).
• Where PREG employs “system-level” interventions to strengthen transformative
capacity, such as improving market infrastructure and training public service providers,
strategies of inclusion should be prominent so that a broader range of households and
communities are able to participate and perceive these interventions as
transformative. This is particularly important with educational opportunities provided
through the program, as there is widespread interest in education—particularly of
children—and in training in business practices and farming as a means of acquiring
sustained employment and income.
 44 | REAL | Resilience, Evaluation, Analysis & Learning
REFERENCES
ACDI/VOCA. (2015). Kenya—Resilience and Economic Growth in the Arid Lands–Accelerated Growth. 
http://acdivoca.org/our-programs/project-profiles/kenya-resilience-and-economic-growth￾arid-lands-accelerated-growth 
ACDI/VOCA. (n.d.) Resilience and Economic Growth in the Arid Lands–Accelerated Growth (REGAL-AG). 
https://www.acdivoca.org/projects/resilience-and-economic-growth-in-the-arid-lands￾accelerated-growth-regal-ag/. 
Ballard, T., Coates, J., Swindale, A., and Deitchler, M. (2011). Household hunger scale: Indicator 
definitions and measurement guide. Washington, DC: Food and Nutrition Technical 
Assistance II Project, FHI 360.
https://www.fsnnetwork.org/sites/default/files/hhs_indicator_guide_03212012.pdf 
Béné, C., Godfrey Wood, R., Newsham, A., and Davies, M. (2012). Resilience: New Utopia or New 
Tyranny? Reflection about the Potentials and Limits of the Concept of Resilience in Relation to 
Vulnerability Reduction Programmes. [IDS Working Paper]. 405. 
https://doi.org/10.1111/j.2040-0209.2012.00405.x 
Catley, A. and Aklilu, Y. (2012). Moving up or moving out? Commercialization, growth and 
destitution in pastoralist areas. In A. Catley, J. Lind, and I. Scoones (Eds.), Pastoralism and 
development in Africa: Dynamic change at the margins. (pp. 85–97). Abingdon, UK: Routledge.
https://doi.org/10.4324/9780203105979 
D’Alessandro, S. P., Caballero, J., Lichte, J., Simpkin, S. (2015). Kenya: Agricultural Sector Risk 
Assessment. Agriculture global practice technical assistance paper;. World Bank, 
Washington, DC. © World Bank. 
https://openknowledge.worldbank.org/handle/10986/23350 License: CC BY 3.0 IGO.
FAO. (2018). Africa Sustainable Livestock 2050: Kenya Cattle and Poultry Sectors. Livestock production 
systems spotlight. Rome, Italy. https://www.fao.org/documents/card/en/c/I8978EN/ 
Farmer, E. and Mbwika, J. (2012). End market analysis of Kenyan livestock and meat: Adesk study. 
Micro report 184. USAID: Washington, DC. https://www.marketlinks.org/resources/end￾market-analysis-kenyan-livestock-and-meat-desk-study 
Feed the Future. (2017). GFSS Technical Guidance: Diversifying livelihoods, resilience, and pathways out 
of poverty. USAID. Washington, DC. https://www.feedthefuture.gov/wp￾content/uploads/2018/03/GFSS_TechnicalGuidance_Diversifying%20Livelihoods-
%20Resilience-%20and%20Pathways%20Out%20of%20Poverty_18Sep2017.pdf 
Frankenberger, T. and Smith, L. (2015). Ethiopia Pastoralist Areas Resilience Improvement and Market 
Expansion (PRIME) Project Impact Evaluation: Report of the Interim Monitoring Survey 2014–
2015. November 2015. https://pdf.usaid.gov/pdf_docs/PA00MGHS.pdf 
PREG Initiative in Northern Kenya Endline Report | 45
Frankenberger, T.R., Langworthy, M., Spangler, T., and Nelson, S. (2012). Enhancing resilience to 
food security shocks in Africa. [Discussion paper]. 
http://www.fsnnetwork.org/sites/default/files/discussion_paper_usaid_dfid_wb_nov._8_2012
.pdf 
International Food Policy Research Institute (IFPRI). 2010. Maize Value Chain Potential in Ethiopia 
Constraints and Opportunities for Enhancing the System. 
https://www.ifpri.org/publication/maize-value-chain-potential-ethiopia
Kenya Markets Trust. (2018). Cattle Population in Decline in Arid Areas, but there are Opportunities to 
Explore—Study. 17 February. http://www.kenyamarkets.org/press/cattle-population-in￾decline-in-asals/ 
Kenya National Bureau of Statistics. (2010a). The 2009 Kenya Population and Housing Census: 
Volume 1A, Population Distribution by Administrative Units. Nairobi, Kenya: KNBS.
Kenya National Bureau of Statistics. (2010b). The 2009 Kenya Population and Housing Census: 
Volume 1B, Population Distribution by Political Units. Nairobi, Kenya: KNBS.
Kenya National Bureau of Statistics. (2010c). The 2009 Kenya Population and Housing Census: 
Volume 2, Population and housing distribution by socio-economic conditions. Nairobi, Kenya: 
KNBS. 
Kenya National Bureau of Statistics. (2017). Kenya Economic Survey 2016. 
Little, P. D. (2016). Overview: Recent Trends in Diversified and Alternative Livelihoods among 
Pastoralists in Eastern Africa. In P. Little (Ed.), Resilience and Risk in Pastoralist Areas: Recent 
Trends in Diversified and Alternative Livelihoods. USAID/East Africa Resilience Learning 
Project. Washington, D.C.: USAID Feed the Future. 
https://pdf.usaid.gov/pdf_docs/pa00m1pz.pdf
Nelson, S., Frankenberger, T., Langworthy, M., Finan, T., and Bower, T. (2016). The effect of 
livelihood diversity on recovery and shock impact in Ethiopia, Kenya, and Uganda. Feed 
the Future, Washington, DC. http://www.fsnnetwork.org/effect-livelihood-diversity￾recovery-and-shock-impact-ethiopia-kenya-and-uganda-0 
OCHA. (2019). Greater Horn of Africa Region Humanitarian Snapshot (November–December 2018). 31 
January. https://reliefweb.int/report/ethiopia/greater-horn-africa-region-humanitarian￾snapshot-november-december-2018 
Schreiner, M. (2011). Simple Poverty Scorecard® Poverty-Assessment Tool Kenya. 10 March. 
https://www.simplepovertyscorecard.com/KEN_2005_ENG.pdf
 46 | REAL | Resilience, Evaluation, Analysis & Learning
SubSahara African water resources and food security. Bulletin of the American Meteorological 
Society. 95, (pp. 861–882). https://doi.org/10.1175/BAMS-D-12-00124.1 
USAID. (2015). Resilience and Economic Growth in Arid Lands—Improving Resilience (REGAL IR) Snap 
Shot. USAID/Kenya. 
https://www.usaid.gov/sites/default/files/documents/1860/REGAL_IR.pdf. 
USAID. (2015). Resilience and Economic Growth in Arid Lands—Accelerated Growth (REGAL-AG) Snap 
Shot. USAID/Kenya. 
https://www.usaid.gov/sites/default/files/documents/1860/REGAL_AG.pdf. 
World Bank. (2019). Data: Rural population (% of total population). Estimates based on the United 
Nations Population Division's World Urbanization Prospects. 2018 Revision.
World Bank. (2018, October 10). Kenya’s Economy Poised to Rebound in 2018 and Remain Robust 
through 2020. [Press release]. https://www.worldbank.org/en/news/press￾release/2018/10/10/kenyas-economy-poised-to-rebound-in-2018-and-remain-robust￾through-2020 
Sheffield, J., Wood, E.F., Chaney, N., Guan K., Sadri, S., Yuan, X., Olang, L., Amani, A., Ali, A., 
Demuth, S., and Ogallo, L. (2014). A drought monitoring and forecasting system for 
PREG Initiative in Northern Kenya Endline Report | 47
ANNEX 1: FOCUS GROUP TOPICAL OUTLINE
Livelihoods (How do People Make a Living)
1. Characteristics of current livelihoods.
• Diversity of activities (primary and secondary activities)
• Resources used to do these activities (e.g., land, water, rangeland, assets)
• How people use these resources
2. Access to and the use of financial resources, such as micro-credit and savings.
3. Changes in these livelihoods over the last 5–10 years.
4. Extent of linkages to external actors and external investment changes over the last 5–10
years. Focus on impacts. Probes can include:
• GoK actions in the region (i.e. available government services)
• NGO/UN/project activities in the region
• Investments schools and health care
• Investment in communication technology, rural infrastructure
Participation in Livestock Markets
5. Describe common livestock practices in this region. Probes include:
• Herd and pasture management
• Variability in herd size across household
• Veterinary practices
• Fodder and purchased inputs
• Principal livestock products
• Normal patterns of selling animals/animal products
6. Changes in livestock and livestock product markets over the last 5–10 years? Possible
probes include:
• More or less market activity
• More or less infrastructure investment (slaughterhouses, holding pens, dip tanks,
milk collection centers, dairy processing plants)
• Investment by government/private sector/NGOs
• Technical assistance and training (e.g., milk quality preservation/sanitation
practices, meat slaughtering/processing, poultry production)
 48 | REAL | Resilience, Evaluation, Analysis & Learning
7. Availability and use of livestock and/or crop insurance. Who are the principal users?
8. Constraints to participation of households in these livestock markets. Possible probes
include:
• Prices, traders, market information
• Poor infrastructure/transport
• Animal disease, lack of community animal health workers, professional veterinary
products and services
• Lack of necessary feed inputs
• Lack of training in business skills and management
• Government of Kenya regulations and controls
9. Changes in the role of women in livestock production and processing.
Shocks and Stressors
10. Describe the main shocks this community has experienced over the last 5–10 years.
11. With reference to the last major shock, please describe the main effects on households and
the community? Probes might include:
• Crop and livestock losses
• Loss of household assets
• Increase in prices
• Health problems
12. Different impacts of shock among community groups: among men and women, youths,
elderly, wealthier and poorer, etc.
Household Responses to Shocks, Stresses, Challenges
13. Types of responses by households. Probes might include:
• Reduced food consumption: Less/ate lower quality food
• Sought casual wage labor
• Sold/slaughtered livestock
• Borrowed money/used savings
• Sought help from family/friends: local and abroad, remittances
• Migration—temporary for work; temporary for pasture; permanent
• Received outside assistance from GoK/NGO/UN agency
PREG Initiative in Northern Kenya Endline Report | 49
14. Differences in responses for different households in the community: which households and
why?
15. Changes in household responses over the last 5–10 years.
Community Responses to Shocks / Impacts of Shocks on Community
16. With reference to the most recent shock, kinds of community responses:
• Sharing of resources among families, collective labor to support household losses
• Collective action to build or improve community infrastructure (e.g., roads,
markets, schools, water, health care facilities, etc.)
• Collective management of natural resources (e.g., rangelands protection, water
and soil conservation)
• Cooperative actions with other communities to reduce/respond to shocks that
affect multiple communities
17. Roles of community leaders in organizing collective action. Differences between men and
women, youth and elderly, etc. in terms of who participates in which types of collective
action.
18. Changes in community responses over the last 5–10 years.
19. Ways shocks have affected community solidarity. (Do these shocks have an impact on 
community solidarity and social capital? On the ability of the community to collaborate and work 
together and help each other?) Possible areas to probe:
• Changes in trust within community among households and with community
leadership
• Changes in community cooperation and mutual help (formal and informal)
• Changes in conflict/security inside and outside the community
20. Outside responses to shocks (e.g. GoK, NGO) and changes over the last 5–10 years?
Possible probes include:
• Type of response (e.g., food, cash, fodder, seeds)
• Effectiveness of response to help the community
21. Community suggestions to improve response capacity. (How can the community become 
better prepared?)
• More collective action
• More outside support
 50 | REAL | Resilience, Evaluation, Analysis & Learning
Aspirations / Planning for the Future
22. Anticipated changes over the next 10 years. (What do community members think will change 
over these 10 years?) 
23. Community plans or projects over the next 10 years (Does the community have any projects 
planned for the future? What is the community discussing or doing now to prepare for future 
changes?) Possible prompts include:
• Community disaster preparedness and response plans
• Investments in infrastructure, irrigation, health, education
• Natural resource management
• Collective action projects
24. Changes through time in the roles of youth and women/girls (Does the community think that 
its youth and women/girls will have a different future?) Probe might include:
• The role of education for youths/girls
• The role of youths in pastoralist culture and society
25. Perception of the community’s ability to achieve the vision for the future (Does the 
community think that it has the ability to make changes and better prepare for the future? What is 
needed to solve problems into the future?) Probe might include:
• Actions that the community can take to move forward
• The role of outsiders (GoK, NGOs, etc.)
Final Comments: Do you have any questions for the team, or anything else you would like us to 
document? 
Asante Sana!
PREG Initiative in Northern Kenya Endline Report | 51
ANNEX 2: KEY INFORMANT INTERVIEW TOPICAL
OUTLINE
Recruitment Criteria: KIIs (Village, Sublocation, Contry)
1. Community leader, clan chiefs, imam/religious leader (VILLAGE)
2. Headmaster/mistress/ principal, head teacher (VILLAGE)
3. Youth leaders, Age grade leader (VILLAGE)
4. Health worker (VILLAGE, SUBLOCATION)
5. Local merchants (inputs, livestock products, etc.) (SUBLOCATION)
6. Livestock trader (SUBLOCATION)
7. Local implementing partner staff (NGOs, WFP Staff Member) (SUBLOCATION/COUNTY)
8. Local financial institutions (SUBLOCATION, COUNTY)
9. Insurance representatives (SUBLOCATION, COUNTY)
10. NDMA staff (COUNTY)
11. Agriculture/livestock extension agents/staff (SUBLOCATION, COUNTY)
12. Slaughterhouse manager (SUBLOCATION COUNTY)
Generic Key Informant Interview Topical Guide—Tailored to Each KI
Instructions: The team supervisor will inquire in the region and decide on which KIs are available and 
feasible to meet. Some are available in the village (local leader), some at the sublocation (merchants), and 
some at the county level (NDMA staff). The supervisor will then decide the distribution of KIIs among the 
team members, then adapt the knowledge categories below to fit the particular key informant. For each 
key informant, write out topics to be covered in the interview, and attach to the interview report. 
1. What do you do? What are your roles/responsibilities?
2. Who do you work with? Please describe the population(s) you work with.
3. How have your activities and responsibilities changed in the last 5–10 years?
4. How has the region changed in the last 5–10 years (how and why)?
5. What was the cause or reason for that change?
6. Describe what changes you see you in the future
7. Are you familiar with REGAL/PREG projects? (If appropriate. OR What projects do you
work with?)
 52 | REAL | Resilience, Evaluation, Analysis & Learning
ANNEX 3: FINDINGS FROM NORTHERN KENYA 
BASELINE AND ENDLINE SURVEYS: 
OVERLAPPING HIGH INTENSITY AND LOW 
INTENSITY SUBLOCATIONS
This annex presents results of analyses using DID methods to compare differences between 
survey rounds for low-intensity (LI) households to differences between survey rounds for high￾intensity (HI) households. HI counties (REGAL-AG and REGAL-IR) at baseline are Marsabit, Isiolo, 
and Turkana. Two counties (Garissa and Wajir) were omitted from the baseline due to security 
issues in those areas. The LI counties (FFA/HA) at baseline are Tana River, Samburu, Mandera 
and Baringo. As described above in Section 2.1.1 of the main report, the sample design was 
modified for the endline to better capture program participation. In particular, the sampling frame 
used for the endline was defined into HI and LI categories based on information about project 
interventions at the level of sublocations within the PREG counties. In order to maintain a 
consistent sampling frame across the two survey rounds, the DID analysis presented in this annex 
is conducted only on the sample of sublocations that were identified by PREG as being in HI or LI 
categories that were included in both baseline and endline surveys.
Tables in this annex show percentages or means for LI and HI in baseline and endline survey 
rounds. The DID values in the tables are the estimated effect of high-intensity program 
participation on HI households compared to low-intensity program participation on LI 
households. Tests of statistical significance show whether the effect is significantly different from 
zero (no effect). Mathematically, DID = (HI_endline–HI_baseline) – (LI_endline–LI_baseline). 
In order to incorporate survey weights, DID estimation used multivariate regression equations of 
the following form: 
𝑌𝑌 = 𝛼𝛼 + 𝛽𝛽1𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 + 𝛽𝛽2𝑅𝑅𝑅𝑅𝑅𝑅 𝑅𝑅 + 𝛽𝛽3( ∗ 𝑟𝑟 ) + 𝜀𝜀
where 
Y is the outcome of interest 
Intensity is programming intensity (LI or HI), 
Round is survey round (baseline or endline), and
Intensity * round is the interaction of the two terms.
The Logit regression estimator was used to estimate nominal (0–1) outcomes, Ordinary Least 
Squares (OLS) was used to estimate interval outcomes, and Generalized Linear Model (GLM) was 
used to estimate the logarithm of expenditures. 
PREG Initiative in Northern Kenya Endline Report | 53
Limitations
The application of the DID analysis on the restricted sample of households that fall in the HI and 
LI sublocations as defined by the reported locations of PREG interventions is intended to improve 
the comparability of the sampled households across the two rounds, as the samples from the two 
rounds are drawn from the same sampling frame. However, this restriction of the sample does 
not completely remove the possibility of selection bias of sampled households across the two 
groups. It is still possible that there may be non-random differences in household characteristics 
between households in the LI and HI sublocations, perhaps due to targeting strategies of PREG 
interventions. 
One implication of applying DID analysis to this restricted sample is that the sample sizes included 
in the analysis for both rounds is much smaller than the full sample. This reduction of the sample 
sizes reduces the accuracy of the sample estimates, and correspondingly reduces the ability to 
statistically detect difference in differences across the two survey rounds.
Findings
Table 14 shows the percentage of LI and HI households engaged in various livelihood activities at 
baseline and endline. The table includes activities reported by at least 5 percent of households in 
either round. The decrease in wage labor was significantly greater for HI households than LI 
households i.e., HI households reported a larger drop in wage labor than did LI households. 
Table 14. Household engagement in select livelihood activities
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
Livelihood activities (%)1
Farming/crop production and 
sales 
36.3 30.8 10.4 0.7 -4.2 ns
Livestock production and sales 35.0 52.0 31.1 38.4 -9.7 ns
Wage labor 24.3 22.4 24.1 5.1 -17.2 *
Salaried work 35.9 20.6 22.8 19.5 12.0 ns
Sale of wild/bush products 3.5 5.0 18.9 31.6 11.2 ns
Other self-employment/own 
business
23.5 16.6 23.1 5.2 -11.0 ns
nŧ 105 226 210 401
* Statistically significant difference at p<0.05; ns denotes no statistically significant difference.
1
Includes activities reported by at least 5% of households in either round.
ŧ Sublocations common to both survey rounds.
Table 15 shows the percentage of households engaging in various coping strategies in the face of 
shocks and/or stresses, comparing LI at baseline and endline to HI at baseline and endline. There 
were no statistically significant differences. Changes in the use of coping strategies were not 
significantly different for HI compared to LI households. 
 54 | REAL | Resilience, Evaluation, Analysis & Learning
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
Households reporting at least 
one coping strategy (%) 37.1 40.6 40.3 42.3 -1.5 ns
nŧ 105 226 209 401
Types of adaptations (%)
Changed income or food 
sources 37.1 0.3 35.7 1.6 2.7 ns
Added income or food sources 19.5 21.7 8.4 3.5 -7.1 ns
Migration of one or more HH 
members 3.9 10.2 6.1 21.4 8.9 ns
Sold large assets 48.7 25.3 48.3 30.9 6.0 ns
Sold small assets 57.1 0.0 55.1 0.4 2.4 ns
HHs using at least one coping 
strategy 52 95 100 174
ns denotes no statistically significant difference.
ŧ Sublocations common to both survey rounds.
Table 16 shows the results of DID analysis comparing baseline-to-endline changes in assets, 
livelihood diversification and use of humanitarian assistance for LI households to HI households. 
The table presents the mean number of assets and livelihood activities and the percentage of 
households reporting the use of relief as a source for food. There were no statistically significant 
differences. Changes between rounds were not significantly different for HI compared to LI 
households. 
Table 16. Assets, livelihood diversification, relief
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
Assets index (mean, 0–34) 6.5 6.5 5.6 4.0 -1.5 ns
Number of livelihood 
activities (mean, 0–12) 1.6 1.5 1.4 1.1 -0.2 ns
Relief as a source of cash or 
food (%) 9.0 3.7 17.8 7.2 -5.3 ns
nŧ 105 226 210 401
ns denotes no statistically significant difference.
ŧ Sublocations common to both survey rounds.
Table 17 shows the results of DID analysis of social capital. The first row of the table shows the 
percentage of households reporting that they relied on others. The increase for HI households 
was lower than for LI households, a difference of 25.5 percent. Of the types of social networks, 
the increase in reliance on non-relatives in the community was 36.3 percent lower for HI 
households than LI households. The increase in reliance on non-relatives outside the community 
was 19.8 percent lower for HI than LI households. HI households also reported less of an increase 
in the number of sources on which they relied during the last drought. When asked about 
Table 15. Household use of coping strategies 
PREG Initiative in Northern Kenya Endline Report | 55
potential sources of support in the future, the increase in HI households reporting that they could 
rely on relatives outside of their community was 28.6 percent lower than for LI households. The 
increase in the percentage of HI households reporting that they could rely on non-relatives within 
the community was 26.0 percent less than for LI households. 
Table 17. Use and availability of social capital
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
Households that relied on others 22.6 50.0 36.1 38.0 -25.5 +
nŧ 105 226 210 401
Types of social networks (% HHs)
Relatives in my community 69.9 76.3 69.6 64.4 -11.6 ns
Relatives outside community 31.4 55.5 42.2 49.8 -16.5 ns
Non-relatives in my community 10.4 49.9 36.6 39.7 -36.3 **
Non-relatives outside community 3.2 19.5 25.1 21.6 -19.8 *
Non-relatives outside of 
tribe/ethnic group 0.0 16.9 8.9 39.5 13.7 ns
Count of sources on which 
household relied during last 
drought 1.1 2.2 1.8 2.1 -0.7 *
HHs that relied on others 31 118 99 162
Households reporting that they will 
be able to rely on others in the 
future 46.6 93.5 62.3 87.3 -21.9 ns
nŧ 105 225 210 400
Relatives in my village/community 35.3 81.2 48.3 70.9 -7.6 ns
Relatives outside my 
village/community 15.6 75.8 30.8 59.8 -28.6 *
Non-relatives in my 
village/community 7.2 48.7 21.5 39.5 -26.0 *
Non-relatives outside 
village/community 5.2 37.0 8.4 41.6 5.5 ns
Non-relatives outside of 
tribe/ethnic group 1.4 27.2 4.7 32.5 3.6 ns
HHs that can rely on others in the 
future 57 208 140 350 ŧ Sublocations common to both survey rounds.
Statistically significant difference at *p<0.05 and **p<0.01; ns denotes no statistically significant difference.
Table 18 shows changes in well-being indicators for LI and HI households between survey rounds. 
There were no statistically significant differences. 
 56 | REAL | Resilience, Evaluation, Analysis & Learning
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
Households reporting moderate or 
severe hunger (%) 33.4 42.2 48.4 76.9 19.7 ns
nŧ 93 225 184 390
Expenditures and poverty
PCD consumption expenditure in 
USD in 2010 prices at 2005 PPP 3.86 2.61 3.34 2.05 -0.1 ns
PCD consumption expenditure in 
USD in 2010 prices at 2011 PPP 2.27 1.54 1.97 1.21 -0.1 ns
% HH living below $1.25 poverty 
line at 2005 PPP 51.7 64.8 64.1 77.5 0.3 ns
% HH living below $1.90 poverty 
line at 2011 PPP 37.9 50.3 53.9 68.5 2.1 ns
Households’ ability to cope with 
future shocks (%)
Unable to cope 20.2 27.8 35.7 50.9 7.5 ns
Able to cope at significant cost to 
well-being 39.4 60.5 35.5 39.0 -17.7 ns
Able to cope without significant 
cost 40.4 11.7 28.7 10.2 10.1 ns
nŧ 105 223 210 391
ŧ Sublocations common to both survey rounds.
ns denotes no statistically significant difference.
Table 19 shows changes in children’s anthropometric indicators between survey rounds, 
comparing LI to HI households. There were no statistically significant differences. 
Table 19. Children's anthropometric indicators
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
Stunting
Moderate to severe stunting 
(Z<-2SD) 31.1 25.7 31.1 20.6 -5.0 ns
Severe stunting (Z<-3SD) 15.9 12.4 13.1 5.4 -4.1 ns
Underweight
Moderate to severe underweight 
(Z-2SD) 14.6 17.7 21.5 18.2 -6.3 ns
Severe underweight (Z-3SD) 10.4 1.9 4.1 1.7 6.2 ns
Wasting
Moderate to severe wasting 
(Z<-2SD) 14.6 8.5 15.0 12.3 3.3 ns
Severe wasting (Z<=-3SD) 5.9 0.9 4.3 1.7 2.4 ns
Table 18. Well-being indicators 
PREG Initiative in Northern Kenya Endline Report | 57
Low intensity High intensity
Baseline Endline Baseline Endline DID sig
nŧ 61 145 151 262
ŧ Sublocations common to both survey rounds.
ns denotes no statistically significant difference.
Summary
The DID analysis compares changes in outcomes between baseline and endline for HI (treatment 
group) and LI (control group) households. Overall, there were few statistically significant results; 
changes in most outcomes between baseline and endline for HI households were not significantly 
different than those for LI households. A notable exception involves social capital, where LI 
households show a much higher degree of reliance on an expanded circle of relatives and non￾relatives for help in dealing with a shock or stressor than HI households. The lack of significant 
difference between changes experienced by HI and by LI households over time may, in part, be 
influenced by the small sample sizes available for this analysis, which restricted the baseline 
sample to those sublocations that overlapped with the endline sample. 
With few exceptions, the direction of change between baseline and endline was the same for HI 
and LI households. In many cases, outcomes in both groups got worse over time. The only 
statistically significant exception was LI households relying more heavily on non-relatives outside 
of their community at endline than baseline; in contrast, HI households reported a small decrease 
in reliance on this source of social capital across survey rounds. 
 58 | REAL | Resilience, Evaluation, Analysis & Learning
ANNEX 4: SAMPLE DESIGN FOR THE 2013, 2015 
AND 2018 SURVEYS
4.1 PURPOSE OF THE SURVEYS, AND ZOI AND 
PROGRAMMING INTENSITY AREA DEFINITIONS 
4.1.1. 2013 Survey
The 2013 survey served two purposes: 
1. As a population-based survey (PBS) designed to measure baseline values for Feed the
Future performance indicators for the area covered by nine ASAL counties of the Feed the
Future ZOI in northern Kenya; and
2. As a baseline for a resilience impact evaluation designed to measure the changes in
household resilience capacities and outcomes in areas with different levels of planned
resilience programming intensity.47
The nine counties identified in the northern Kenya ZOI in 2013 were Baringo, Garissa, Isiolo, 
Mandera, Marsabit, Samburu, Tana River, Turkana, and Wajir. 
The levels of planned resilience programming intensity were as follows: The Low-Intensity (LI) 
area was where only USAID-funded World Food Programme humanitarian assistance would 
be/was provided and included Baringo, Mandera, Samburu and Tana River counties. The Medium￾Intensity (MI) stratum area would receive humanitarian assistance and USAID-funded Resilience 
and Economic Growth in the Arid Lands Improving Resilience (REGAL-IR) support and included 
Isiolo, Turkana and Wajir counties. The High Intensity (HI) area would receive HA, REGAL-IR, 
and REGAL Accelerating Growth (REGAL-AG) investments and included Garissa and Marsabit 
counties. Table 20 shows the distribution of counties into program intensity categories as defined 
prior to 2013 data collection and the population in each county provided in the 2013 baseline 
report. 
The 2013 survey data were collected then January–February 2013, before the long rains in the 
pastoral areas when livestock are migrating to dry-season grazing areas (Figure 3). 
47 Resilience programming initially referred to that of the Resilience and Economic Growth in the Arid Lands (REGAL) project, then was 
expanded to include all resilience programming under the USAID-Kenya Partnership for Resilience and Economic Growth (PREG) I project.
PREG Initiative in Northern Kenya Endline Report | 59
Figure 3: 2013 Seasonal calendar
Source: Famine Early Warning System (FEWS Net). Food Security Brief. December 2013, in Feed the Future Northern 
Kenya 2015 Zone of Influence Interim Assessment Report, page 15
4.1.2. 2015 Survey
The 2015 survey served two purposes: 
1. As a midline population-based survey (PBS) designed to provide point estimates for Feed
the Future performance indicators for the five ASAL counties now defined as comprising
the ZOI in northern Kenya; and
2. As a midline for the PREG I impact evaluation to compare indicator values between HI
and MI areas, HI and LI intensity areas, and MI and LI areas, covering the original nine￾county area.
The Feed the Future ZOI was redefined in the 2015 survey to encompass only five counties: 
Garissa, Isiolo, Marsabit, Turkana, and Wajir. The five counties were all in the MI or HI 
programming areas. Garissa, Turkana, and Wajir in MI areas; and Isiolo and Marsabit in HI 
areas. The remaining four counties, Baringo, Mandera, Samburu, and Tana River, were in the LI 
program area, which was no longer considered part of the Feed the Future ZOI. The LI counties 
were included in the survey because they are considered the control group for the PREG I IE.
The 2015 survey data were collected in May–June 2015, during the long rains in pastoral areas.
 60 | REAL | Resilience, Evaluation, Analysis & Learning
4.1.3. 2018 Survey
The 2018 survey served two purposes:
1. As the endline survey for the initial (PREG I) impact evaluation; and
2. As the baseline survey for an impact evaluation of the new round of PREG resilience
programming (PREG II).
Given that the purposes of the 2018 survey were exclusively related to impact evaluation needs, it 
did not collect data at the Feed the Future ZOI-level. 
The programming intensity strata were reduced to two in the 2018 survey: The LI (humanitarian 
assistance only) stratum included sublocations in Baringo, Mandera, Samburu, and Tana River 
counties. MI and HI were collapsed into a single HI stratum, defined as sublocations that received 
humanitarian assistance plus other resilience-enhancing interventions within Garissa, Isiolo, Marsabit, 
Turkana and Wajir counties. Sublocations in these five counties that only received humanitarian 
assistance were not included in the LI stratum and thus were excluded entirely from the 2018 survey. 
The 2018 survey data were collected in September 2018, in the lean season.
4.2 SAMPLE DESIGN
4.2.1. 2013 Survey
Due to security issues near the Somali border at the time of the 2013 survey, Garissa, Mandera 
and Wajir counties were excluded from the sampling frame. Thus, the PBS and the PREG I impact 
evaluation baseline sample only covered six of the nine ASAL counties of the ZOI in northern 
Kenya: Baringo, Isiolo, Marsabit, Samburu, Tana River and Turkana. The six HI counties in the 
sample were divided into three programming intensity strata defined at the county level, based on 
the programming intensity area described in Section 4.1.1. The LI stratum included Baringo, 
Samburu and Tana River counties. The MI stratum included Isiolo and Turkana counties. The HI 
stratum included Marsabit County. 
Each stratum was stratified by county and within county, by urban and rural. The overall sample 
of 2,136 households in 140 clusters was allocated among the three strata (LI–724 households, 46 
clusters; MI–705 households, 47 clusters; HI–705 households, 47 clusters), then EAs within each 
stratum were selected using probability-proportional-to-size methodology (see Table 20). The 
samples drawn in each of the three strata were population-based, that is, all households within the 
counties constituted the sampling frame from which surveyed households were drawn. In order to 
maximize the number of clusters while at the same time conforming to time constraints and 
logistics requirements, KNBS used existing listings and protocols to select between 14 and 18 
households per cluster, resulting in a total projected sample of 2,136 households over 140 clusters.
Table 20 presents the distribution of the 2013 survey sample.
PREG Initiative in Northern Kenya Endline Report | 61
Table 20: 2013 survey sample
Strata/program County Population Sample
Clusters Households
People Households Distribution of
households % 
within stratum 
or among 
strata
Rural Urban Total Rural Urban Total Distribution of 
total sample 
households % 
within stratum 
or among strata
LI: HA only Baringo 555,561 110,649 54% 12 5 17 216 90 306 42%
Samburu 223,947 47,354 23% 11 3 14 165 45 210 29%
Tana River 240,075 47,414 23% 13 2 15 182 28 210 29%
LI sub-total 1,019,583 205,417 49% 36 10 46 563 163 726 34%
MI: HA+REGAL IR Isiolo 143,294 31,326 20% 9 7 16 135 105 240 34%
Turkana 855,399 123,191 80% 22 9 31 330 135 465 66%
MI sub-total 998,693 154,517 37% 31 16 47 465 240 705 33%
HI: HA+REGAL￾IR+REGAL-EG
Marsabit 291,166 56,941 100% 37 10 47 555 150 705 100%
HI sub-total 291,166 56,941 14% 37 10 47 555 150 705 33%
TOTAL 2,309,442 416,875 104 36 140 1,583 553 2,136
 62 | REAL | Resilience, Evaluation, Analysis & Learning
Field teams were unable to access five clusters (78 households). Three of these clusters were 
excluded because they were inaccessible due to flooding, and two were excluded due to long 
travel times (five-day journey to one cluster). Two of the clusters were in Turkana, and one each 
were in Baringo, Isiolo and Marsabit. 
Of the remaining 2,058 households, field teams were unable to locate 158 households. 
Enumerators contacted 1,900 households, of which 1,760 consented to be interviewed, for an 
overall non-response rate of 17.6 percent. No information is provided in the baseline report on 
the distribution of non-response by stratum. 
The 2018 survey draft report includes a table that provides the 2013 realized sample size, which is 
less than the 1,760 documented in the 2013 report (Table 21). However, based on the data 
provided, it does not appear that there is any large bias among the counties in terms of non￾response. 
Table 21: Initial and realized 2013 survey sample
County Initial Sample Realized Sample Realized sample / 
Initial sample
Baringo 306 234 76%
Samburu 210 158 75%
Tana River 210 180 86%
Turkana 465 364 78%
Isiolo 240 191 80%
Marsabit 705 588 83%
Total 2,136 1,715 80%
4.2.2. 2015 Survey
The 2015 survey collected data over the full nine-county area: Baringo, Garissa, Isiolo, Mandera, 
Marsabit, Samburu, Tana River, Turkana, and Wajir. The sample was designed to be large 
enough to measure differences among intensity levels of REGAL and humanitarian assistance
programming for the impact evaluation, and provide point estimates for Feed the Future 
indicators at the new five-county ZOI level. 
The three counties excluded at baseline due to security concerns in 2013, Garissa, Mandera and 
Wajir, were included in the 2015 sampling frame and sample. There had also been shifts in where 
REGAL-AG was programming after 2013 data were collected, which led to Garissa being 
redefined as MI (although no data were collected from Garissa in 2013), and Isiolo moved from 
MI to HI. 
The nine counties in the sample were thus divided into three slightly modified programming 
intensity strata defined at the county level. The LI HA-only stratum included Baringo, Mandera, 
Samburu, and Tana River. The MI HA+REGAL-IR stratum included Garissa, Turkana, and Wajir. 
The HI HA+REGAL-IR+REGAL-EG program area included Isiolo and Marsabit. 
PREG Initiative in Northern Kenya Endline Report | 63
Each stratum was stratified by county and within county, by urban and rural. The overall target 
sample of 2,100 households in 132 clusters was allocated equally among the three strata with 44 
EAs per stratum. Then EAs within stratum were selected using PPS. In the LI and MI areas, KNBS 
selected EAs from the 2009 Population and Housing Census of Kenya. In the HI area, Feed the 
Future FEEDBACK drew a subsample of the Kenya DHS EAs. Two different frames were the 
source for EAs in the three areas. Initially, the sample design distributed the 2,100-household 
sample over the three intensity areas with the number of EAs in each intensity level proportional 
to population. Accordingly, KNBS selected EAs in March 2015. Listing took place in April, several 
weeks before survey fieldwork, so that Westat analysts could select the sample and include 
household identification information in ODK prior to fieldwork in May. However, PPS sampling 
across the intensity areas meant that the LI areas had the most EAs and the HI area had the least 
EAs (17). In late April, it was determined that the sample in the HI area was insufficient for 
analysis. KNBS was concerned about re-sampling to supplement the 17 EAs. In May, it was 
decided to not use the 17 EAs and to select 44 EAs from the recently completed Demographic 
and Health Survey. Kimetrica conducted fieldwork for that survey and had access to the sample 
information. 
In the second stage, 16 households within each selected EA were selected randomly from a list of 
eligible households in the MI and LI areas, while 18 households were selected in the HI area due 
to potentially higher non-response. 
Table 22 presents the distribution of the 2015 survey sample.
 64 | REAL | Resilience, Evaluation, Analysis & Learning
Table 22: 2015 survey sample
Strata/program County Population Sample
Clusters Households
People Households Distribution 
of HHs % 
within 
stratum or 
among 
strata
Rural Urban Total Rural Urban Total Distribution of 
total sample 
households % 
within 
stratum or 
among strata
LI - HA only Baringo 555,561 110,649 33% 11 2 13 176 32 208 30%
Mandera 1,025,756 125,497 38% 12 3 15 192 48 240 34%
Samburu 223,947 47,354 14% 6 2 8 96 32 128 18%
Tana River 240,075 47,414 14% 6 2 8 96 32 128 18%
LI sub-total 2,045,339 330,914 50% 35 9 44 560 144 704 33%
MI - HA+REGAL IR Garissa 143,294 31,326 13% 9 4 13 144 64 208 30%
Turkana 855,399 123,191 51% 13 4 17 208 64 272 38%
Wajir 661,941 88,574 36% 11 3 14 176 48 224 32%
MI sub-total 1,660,634 243,091 37% 33 11 44 528 176 704 33%
HI - HA+REGAL￾IR+REGAL-EG
Isiolo 143,294 31,326 35% 11 10 21 198 180 378 51%
Marsabit 291,166 56,941 65% 16 4 20 287 72 359 49%
HI sub-total 434,460 88,267 13% 27 14 41 415 252 737 34%
TOTAL 4,140,433 662,272 95 34 129 1503 572 2154
PREG Initiative in Northern Kenya Endline Report | 65
4.2.3. 2018 Survey
The LI stratum sampling frame included all households in the sublocations within the four counties 
(Baringo, Mandera, Samburu, and Tana River). The new HI stratum sampling frame only included 
households in purposively selected sublocations within the five counties (Garissa, Isiolo, Marsabit, 
Turkana and Wajir). The HI stratum was further categorized into Low, Medium, and High strata 
to indicate intensity of resilience programming based on information provided by USAID/Kenya. 
The sublocations selected were those in which the PREG project was or is carrying out resilience 
programming. Of all the activities that had taken place over the five years of PREG I, some were 
completed by the time of the 2018 survey, and some were still on-going and anticipated to 
continue under PREG II. The sublocations are all at the KNBS-defined Admin 5 level (Admin 1 = 
County, Admin 2 = District, Admin 3 = Division, Admin 4 = Location, Admin 5 = Sublocation). 
The overall sample of 3,116 households in 107 clusters was allocated to the LI strata (1,520 
households) and HI strata (1,596 households), and the three resilience programming intensity 
strata: Low–541 households, 19 clusters; Medium–505 households, 19 clusters; High–550 
households, 19 clusters). EAs within each HI and LI stratum were selected using PPS. 
Table 23 presents the distribution of the 2018 survey sample.
 66 | REAL | Resilience, Evaluation, Analysis & Learning
Table 23: 2018 survey sample
Strata/ 
program
County Population Sample
Clusters Households
People Households Distribution 
of HHs % 
within 
stratum or 
among 
strata
Rural Urban Total Rural Urban Total Distribution of 
total sample 
households % 
within 
stratum or 
among strata
LI: HA only Baringo PREG 
sublocations
6,635 1,301 1% 2 0 2 60 0 60 4%
Mandera PREG 
sublocations
304,751 59,755 47% 19 3 21 547 90 637 42%
Samburu PREG 
sublocations
163,690 32,096 25% 9 4 13 270 119 389 26%
Tana River PREG
sublocations
175,644 34,440 27% 12 3 15 344 90 434 29%
LI sub-total 650,719 127,592 42% 42 10 51 1221 299 1520 49%
HI: HA + other 
resilience 
interventions
Garissa PREG 
sublocations
267,541 52,459 30% 5 9 14 147 242 389 24%
Isiolo PREG 
sublocations
106,238 20,831 12% 4 3 7 120 90 210 13%
Marsabit PREG 
sublocations
83,992 16,469 9% 8 1 9 235 27 262 16%
PREG Initiative in Northern Kenya Endline Report | 67
Strata/ 
program
County Population Sample
Clusters Households
People Households Distribution 
of HHs % 
within 
stratum or 
among 
strata
Rural Urban Total Rural Urban Total Distribution of 
total sample 
households % 
within 
stratum or 
among strata
Turkana PREG 
sublocations
215,801 42,314 24% 12 3 15 319 90 409 26%
Wajir PREG 
sublocations
230,127 45,123 26% 10 1 11 296 30 326 20%
HI sub-total 903,699 177,196 58% 39 17 56 1117 479 1596 51%
TOTAL 1,554,41 304,788 81 27 108 2338 778 3116
 68 | REAL | Resilience, Evaluation, Analysis & Learning
Table 24: Initial and realized PREG II baseline sample
County Initial Sample Realized Sample Realized sample / 
Initial sample
Baringo 60 58 97%
Mandera 637 608 95%
Samburu 389 354 91%
Tana River 434 382 88%
Garissa 389 341 88%
Isiolo 210 188 90%
Marsabit 262 216 82%
Turkana 409 364 89%
Wajir 326 309 95%
Total 3,116 2,820 91%
PREG Initiative in Northern Kenya Endline Report | 69
ANNEX 5: PARTICIPATION IN PREG INTERVENTIONS BY COUNTY
Table 25: Participation in PREG-supported trainings
Low Intensity Counties High Intensity Counties
Total 
Sample Project Participation 
Category All Baringo Mandera Samburu Tana 
River All Turkana Wajir Garissa Isiolo Marsabit
Percent of households
Trainings
Alternative 
livelihoods/income￾generating activities 5.2 25.0 7.1 2.3 6.3 13.5 9.5 3.7 23.8 17.6 15.0 10.3
Livestock production 
practices/health/
management 10.4 0.0 14.3 14.0 0.0 9.5 14.3 11.1 4.8 0.0 12.5 9.9
Crop production practices 9.1 75.0 0.0 9.3 0.0 2.4 0.0 3.7 9.5 0.0 0.0 4.9
Crop or livestock marketing 2.6 0.0 7.1 2.3 0.0 3.2 0.0 0.0 4.8 0.0 7.5 3.0
Business/financial/accounting 
practices 18.2 50.0 14.3 23.3 0.0 7.9 4.8 3.7 19.0 5.9 7.5 11.8
Rangeland management 2.6 0.0 0.0 4.7 0.0 4.8 4.8 0.0 4.8 5.9 7.5 3.9
Nutrition training 16.9 0.0 28.6 20.9 0.0 29.4 14.3 37.0 38.1 5.9 37.5 24.6
Savings/microfinance 3.9 25.0 0.0 2.3 6.3 5.6 0.0 7.4 0.0 5.9 10.0 4.9
WASH practices 33.8 25.0 7.1 37.2 50.0 30.2 33.3 25.9 14.3 29.4 40.0 31.5
Conflict management 6.5 0.0 7.1 4.7 12.5 10.3 0.0 7.4 19.0 17.6 10.0 8.9
Life skills (hygiene/family 
planning, etc.) 32.5 25.0 28.6 34.9 31.3 27.8 28.6 29.6 23.8 11.8 35.0 29.6
Youth skills training 
(apprenticeship/internship, 
etc.) 2.6 0.0 0.0 4.7 0.0 6.3 4.8 3.7 19.0 5.9 2.5 4.9
Other (specify) 11.7 0.0 21.4 4.7 25.0 20.6 28.6 29.6 9.5 35.3 10.0 17.2
 70 | REAL | Resilience, Evaluation, Analysis & Learning
Low Intensity Counties High Intensity Counties
Total 
Sample Project Participation 
Category All Baringo Mandera Samburu Tana 
River All Turkana Wajir Garissa Isiolo Marsabit
Percent of households
N (number of cases 
reporting any training) 77 4 14 43 16 126 21 27 21 17 40 203
Table 26: Participation in PREG-supported groups
Low Intensity Counties High Intensity Counties
Total 
Sample Project Participation 
Category All Baringo Mandera Samburu Tana 
River All Turkana Wajir Garissa Isiolo Marsabit
Percent of households
Groups
Savings and loan groups 
(VSLA, REAP groups, 
BOMA groups, etc.) 45.6 58.6 0.0 40.6 52.6 24.8 38.5 3.8 38.9 25.0 34.8 36.8
Livestock management 
associations (LMA, etc.) 1.3 0.0 10.0 1.6 0.0 4.3 0.0 7.7 25.0 0.0 0.0 2.5
Livestock producer groups 2.5 0.0 10.0 3.1 1.8 2.6 0.0 3.8 8.3 0.0 2.2 2.5
Self-help groups (ADESO 
groups, etc.) 23.1 34.5 10.0 25.0 17.5 23.1 38.5 7.7 41.7 25.0 21.7 23.1
Women groups (GIRL 
model, etc.) 18.1 17.2 30.0 25.0 8.8 33.3 46.2 34.6 41.7 20.0 32.6 24.5
Mothers’ groups 9.4 0.0 20.0 10.9 10.5 14.5 30.8 19.2 16.7 15.0 6.5 11.6
Village committees (water 
management, NRM, conflict 
management health, 
school, e 5.6 10.3 0.0 4.7 5.3 5.1 7.7 7.7 8.3 5.0 2.2 5.4
PREG Initiative in Northern Kenya Endline Report | 71
Low Intensity Counties High Intensity Counties
Total 
Sample Project Participation 
Category All Baringo Mandera Samburu Tana 
River All Turkana Wajir Garissa Isiolo Marsabit
Percent of households
Ward level committees 
(WARD adaptation 
planning committee, etc.) 1.3 0.0 10.0 1.6 0.0 1.7 7.7 0.0 0.0 5.0 0.0 1.4
Other (specify) 13.1 6.9 20.0 7.8 21.1 21.4 23.1 26.9 8.3 15.0 23.9 16.6
N (number of cases reporting 
group membership) 160 29 10 64 57 117 13 26 12 20 46 277
Any participation (group 
and/or training) 15.1 51.7 3.5 26.1 17.6 13.1 7.4 13.7 7.7 15.5 27.6 14.1
Any training participation 5.5 6.9 2.3 11.9 4.2 8.9 5.8 8.8 6.2 9.1 17.8 7.2
Any group participation 11.4 50.0 1.7 18.2 15.0 8.2 3.6 8.2 3.5 10.7 21.5 9.8
n 1,396 58 605 352 381 1,410 364 306 339 187 214 2,806
ABOUT THE REAL AWARD
The Resilience Evaluation, Analysis and Learning (REAL) Associate Award is a consortium-led effort funded 
by the USAID Center for Resilience. It was established to respond to growing demand among USAID 
Missions, host governments, implementing organizations, and other key stakeholders for rigorous, yet 
practical, monitoring, evaluation, strategic analysis, and capacity building support. Led by Save the Children, 
REAL draws on the expertise of its partners: Mercy Corps, and TANGO International.