PHOTO CREDIT: THE CLOUDBURST GROIP 
RESILIENCYCAMBODIA IE FINDINGS 
REPORT
Findings From a Civil Society RCT in Cambodia: 
Local Organizations—Movement Towards Self￾Reliance (LO-MTSR) Activity
JULY 2022 
DISCLAIMER: The authors' views expressed in this publication do not necessarily reflect the views of the United States Agency for 
International Development or the United States Government. 
Written and prepared by Jeremy Springman (DevLab@Penn), Ryan Hatano (Cloudburst), Aleta 
Starosta (Cloudburst), and Erik Wibbel (DevLab@Penn). 
Prepared for the United States Agency for International Development (USAID) Cambodia, USAID 
Contract Number 72044219C00004, Local Organizations—Movement Towards Self-Reliance 
(LO-MTSR) Activity. 
Implemented by: 
The Cloudburst Group
8400 Corporate Drive, Suite 550
Landover, MD 20785-2238
Tel: 301-918-4400
USAID.GOV LO-MTSR RESILENCYCAMBODIA IE FINDINGS REPORT | iii
TABLE OF CONTENTS
_Toc112426267TABLE OF FIGURES IV
TABLE OF TABLES VI
ACRONYMS VII
EXECUTIVE SUMMARY 1
EVALUATION PURPOSE AND EVALUATION QUESTIONS 1
METHODS 2
FINDINGS OUTCOME 1: MANAGERIAL AND ADMINISTRATIVE CAPACITY OF TREATED CSOS 2
FINDINGS OUTCOME 2: NUMBER OF CONNECTIONS BETWEEN TREATED CSOS AND CITIZENS, THE PRIVATE FOR-PROFIT 
SECTOR, AND OTHER CSOS 2
FINDINGS OUTCOME 3: FINANCIAL RESILIENCY OF TREATED CSOS 3
FINDINGS OUTCOME 4: RESILIENCY TO CHANGING CIVIC SPACE OF TREATED CSOS 3
RECOMMENDATIONS AND LESSONS LEARNED 3
1.0. EVALUATION PURPOSE AND QUESTIONS 5
1.1 EVALUATION PURPOSE 5
1.2 EVALUATION QUESTIONS AND THEORY OF CHANGE 6
1.3 PROJECT BACKGROUND 6
Current challenges facing civil society in Cambodia 6
2.0 RESILIENCYCAMBODIA OVERVIEW 8
RESILIENCY ORIENTATION & ROC ASSESSMENT 9
RESILIENCY ROADMAP 9
COACHING AND MENTORING 10
GROUP TRAININGS 10
RESILIENCY RESOURCES AND TOOLKITS 10
SOCIAL LAB 11
RESOURCES FOR INSTITUTIONAL DEVELOPMENT 11
3.0 EVALUATION METHODS 11
3.1 METHODS 11
3.2 OUTCOME FAMILIES, HYPOTHESES, AND INDICATORS 12
3.3 SAMPLING AND RANDOMIZATION 13
3.4 DATA SOURCES 14
3.5 SAMPLE CHARACTERISTICS 19
4.0 ANALYSIS 20
MAIN OUTCOMES ANALYSIS 20
5.0 FINDINGS—MANAGERIAL AND ADMINISTRATIVE CAPACITY OF TREATED CSOS 21
PRIMARY OUTCOME 1.1: MANAGERIAL CAPACITY INDEX: INTERNAL CHALLENGES AND TIME ALLOCATION 22
PRIMARY OUTCOME 1.2: COVID-19 ADAPTATION INDEX 26
PRIMARY OUTCOME 1.3: ADMINISTRATIVE CAPACITY 28
USAID.GOV LO-MTSR RESILENCYCAMBODIA IE FINDINGS REPORT | iv
6.0 FINDINGS—CONNECTIONS BETWEEN TREATED CSOS AND CITIZENS, THE PRIVATE 
FOR-PROFIT SECTOR, AND OTHER CSOS 31
PRIMARY OUTCOME 2.1: COUNT OF PARTNERSHIPS INDEX 32
PRIMARY OUTCOME 2.2: SOCIAL MEDIA INTERACTIONS INDEX 36
SECONDARY OUTCOME 2.3: SELF-REPORTED NETWORK STRENGTH INDEX 38
SECONDARY OUTCOME 2.4: PARTNERSHIP-SEEKING BEHAVIOR INDEX 39
SECONDARY OUTCOME 2.5: NETWORK DENSITY AMONG TREATMENT AND CONTROL SAMPLE 40
7.0 FINDINGS—FINANCIAL RESILIENCY OF TREATED CSOS 42
EXPLORATORY ANALYSIS: REVENUE SOURCES AND VALUE 42
PRIMARY OUTCOME 3.1: REVENUE GENERATION INDEX 45
PRIMARY OUTCOME 3.2: REVENUE DIVERSIFICATION INDEX: HERFINDAHL–HIRSCHMAN INDEX 47
SECONDARY OUTCOME 3.3: FINANCIAL HEALTH 50
SECONDARY OUTCOME 3.4: DIVERSIFICATION AWAY FROM AID: LOCAL REVENUE INDEX 52
SECONDARY OUTCOME 3.5: SHARE OF REVENUE FROM FOREIGN SOURCES 54
8.0 FINDINGS—RESILIENCY TO CHANGING CIVIC SPACE OF TREATED CSOS 56
PRIMARY OUTCOME 4.1: EXTERNAL CHALLENGES INDEX 56
PRIMARY OUTCOME 4.2: CSO NETWORK DIVERSIFICATION: HHI 59
PRIMARY OUTCOME 4.3: SHARE OF TIME ENGAGING IN ADVOCACY 60
PRIMARY OUTCOME 4.4: SHARE OF TIME ENGAGING IN ADVOCACY 63
9.0 RECOMMENDATIONS AND LESSONS LEARNED 64
RECOMMENDATIONS 65
LESSONS LEARNED 67
APPENDIX A. DESIGN REPORT 69
APPENDIX B. BASELINE REPORT 70
APPENDIX C. PRE-ANALYSIS PLAN 71
APPENDIX D. DEPARTURES FROM PRE-ANALYSIS PLAN 72
APPENDIX E. BUDGET CATEGORIES 74
APPENDIX F. REFERENCES 76
USAID.GOV LO-MTSR RESILENCYCAMBODIA IE FINDINGS REPORT | iv
TABLE OF FIGURES
FIGURE 1. CAUSAL MODEL 8
FIGURE 2. TIMELINE OF RESILIENCYCAMBODIA INTERVENTION ACTIVITIES 9
FIGURE 3. CSO SAMPLING METHODOLOGY 14
FIGURE 4. TYPE OF RECORDS PROVIDED AND LEVEL OF DETAIL ACROSS GROUPS 16
FIGURE 5. TYPE OF RECORDS PROVIDED AND LEVEL OF DETAIL ACROSS GROUPS 17
FIGURE 6. ATTRITION BY TREATMENT AND CONTROL GROUPS 18
FIGURE 7. MAP OF CSO LOCATIONS 19
FIGURE 8. SUMMARY OF RESULTS FOR OUTCOME FAMILY 1 22
FIGURE 9. SHARE OF CSOS REPORTING THAT EACH OF THESE INTERNAL CHALLENGES INTERFERES 
WITH THEIR ABILITY TO FULFILL THEIR MISSION 24
FIGURE 10. SHARE OF TIME SPENT ENGAGING IN POLITICAL ADVOCACY, COMMUNITY OUTREACH, 
AND SERVICE DELIVERY 24
FIGURE 11. DISTRIBUTION OF AVERAGED Z-SCORE INDEX VARIABLES FOR THE FULL INDEX 
COMBINING ALL COMPONENT VARIABLES (ROW 1), AN AVERAGED Z-SCORE COMBINING 
COMPONENTS OF THE INTERNAL CHALLENGES IN FIGURE 9 (ROW 2), AND AN AVERAGED Z-SCORE 
COMBINING COMPONENTS OF THE TIME UTILIZATION IN FIGURE 10 (ROW 3) 25
FIGURE 12. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON MANAGERIAL 
CAPACITY OUTCOMES 26
FIGURE 13. SHARE OF CSOS REPORTING THAT THEY ENGAGED IN EACH OF THE COVID-19 
ADAPTATION ACTIVITIES 27
FIGURE 14. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON COVID-19 
ADAPTATIONS OUTCOMES 27
FIGURE 15. COUNT OF OBSERVED OCCURRENCES OF 10–99 AS THE LEADING TWO DIGITS FOR THE 
FULL SAMPLE OF CSO FINANCIAL ENTRIES 29
FIGURE 16. COMPARISON OF OBSERVED VALUES OF THE KS SCORE FOR EACH CSO BETWEEN THE 
BASELINE AND ENDLINE PERIODS 30
FIGURE 17. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON MANAGERIAL 
CAPACITY OUTCOMES 30
FIGURE 18. SUMMARY OF RESULTS FOR OUTCOME FAMILY 2 31
FIGURE 19. AVERAGE NUMBER OF NON-PROFIT PARTNERS (TOTAL AND BY SECTOR) REPORTED BY 
CSOS 33
FIGURE 20. AVERAGE NUMBER OF OTHER PRIVATE PARTNERS (TOTAL AND BY SECTOR) REPORTED 
BY CSOS 34
FIGURE 21. DISTRIBUTION OF COMBINED INDEX MEASURING THE TOTAL NUMBER OF NON-PROFIT 
AND OTHER PRIVATE PARTNERS REPORTED BY CSOS 35
FIGURE 22. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON PARTNERSHIP 
OUTCOMES 35
FIGURE 23. AVERAGE NUMBER OF SOCIAL MEDIA INTERACTIONS PER MONTH 36
FIGURE 24. AVERAGED Z-SCORES INDEX COMBINING AVERAGE SOCIAL MEDIA INTERACTIONS PER 
MONTH 37
FIGURE 25. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON SOCIAL MEDIA 
OUTCOMES 37
USAID.GOV LO-MTSR RESILENCYCAMBODIA IE FINDINGS REPORT | v
FIGURE 26. SHARE OF CSOS REPORTING THAT NETWORK-RELATED INTERNAL CHALLENGES 
INTERFERE WITH THEIR ABILITY TO FULFILL THEIR MISSION 38
FIGURE 27. AVERAGE Z-SCORE INDEX COMBINING NETWORK-RELATED INTERNAL CHALLENGES 39
FIGURE 28. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON SELF-REPORTED 
NETWORK STRENGTH OUTCOMES 39
FIGURE 29. THE Z-SCORES FOR PARTNERSHIP-SEEKING BEHAVIOR COMPONENT VARIABLES AND 
THE AVERAGED Z-SCORE INDEX 40
FIGURE 30. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON PARTNERSHIP￾SEEKING BEHAVIOR OUTCOMES 40
FIGURE 31. FULL NETWORK OF DIRECT CONNECTIONS BETWEEN TREATMENT AND CONTROL CSOS 
IN THE SAMPLE AT BASELINE AND ENDLINE 41
FIGURE 32. SUMMARY OF RESULTS FOR OUTCOME FAMILY 3 42
FIGURE 33. COMPARISON OF THE NUMBER OF REVENUE SOURCES FOR EACH CSO BETWEEN THE 
BASELINE AND ENDLINE PERIODS 43
FIGURE 34. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON THE NUMBER 
OF SOURCES OF REVENUE OUTCOMES 44
FIGURE 35. COMPARISON OF THE TOTAL VALUE OF REVENUE FOR EACH CSO BETWEEN THE 
BASELINE AND ENDLINE PERIODS 44
FIGURE 36. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON THE VALUE OF 
REVENUE OUTCOMES 45
FIGURE 37. DISTRIBUTION OF AVERAGED Z-SCORE INDEX VARIABLES FOR AN INDEX COMBINING 
AWARD-SEEKING COMPONENTS, AN INDEX COMBINING INCOME-SEEKING COMPONENTS, AND AN 
INDEX COMBINING ALL OF THE REVENUE-SEEKING COMPONENTS 46
FIGURE 38. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON REVENUE 
GENERATION OUTCOMES 47
FIGURE 39. COMPARISON OF THE CONCENTRATION OF THE NUMBER OF REVENUE SOURCES 
ACROSS CATEGORIES AND THE VALUE OF REVENUE ACROSS CATEGORIES FOR EACH CSO 
BETWEEN THE BASELINE AND ENDLINE PERIODS 49
FIGURE 40. COMPARISON OF THE CONCENTRATION OF REVENUE ACROSS ALL REVENUE LINE ITEMS 
FOR EACH CSO BETWEEN THE BASELINE AND ENDLINE PERIODS 49
FIGURE 41. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON REVENUE 
DIVERSIFICATION OUTCOMES 50
FIGURE 42. COMPARISON OF THE INVESTMENTS, REVENUE, OPERATING MARGIN, AND OVERALL 
FINANCIAL HEALTH FOR EACH CSO BETWEEN THE BASELINE AND ENDLINE PERIODS 51
FIGURE 43. COMPARISON OF THE OVERALL OPERATING MARGIN AND ITS TWO COMPONENTS 
(EXPENSES AND REVENUE) FOR EACH CSO BETWEEN THE BASELINE AND ENDLINE PERIODS51
FIGURE 44. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON FINANCIAL 
HEALTH OUTCOMES 52
FIGURE 45. COMPARISON OF THE TOTAL REVENUE FROM LOCAL SOURCES AND THE TOTAL 
REVENUE FROM LOCAL SOURCES AS A SHARE OF TOTAL REVENUE FOR EACH CSO BETWEEN THE 
BASELINE AND ENDLINE PERIODS 53
FIGURE 46. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON LOCAL REVENUE 
OUTCOMES 54
FIGURE 47. COMPARISON OF THE TOTAL REVENUE FROM FOREIGN SOURCES AND THE TOTAL 
REVENUE FROM FOREIGN SOURCES AS A SHARE OF TOTAL REVENUE FOR EACH CSO BETWEEN THE 
BASELINE AND ENDLINE PERIODS 55
USAID.GOV LO-MTSR RESILENCYCAMBODIA IE FINDINGS REPORT | vi
FIGURE 48. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON FOREIGN 
REVENUE OUTCOMES 55
FIGURE 49. SUMMARY OF RESULTS FOR OUTCOME FAMILY 4 56
FIGURE 50. SHARE OF CSOS REPORTING THAT EACH OF THESE EXTERNAL CHALLENGES 
INTERFERES WITH THEIR ABILITY TO FULFILL THEIR MISSION 57
FIGURE 51. VALUES OF AN AVERAGED Z-SCORE COMBINING THESE COMPONENTS INTO A SINGLE 
SUMMARY VARIABLE 58
FIGURE 52. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON EXTERNAL 
CHALLENGES OUTCOMES 58
FIGURE 53. COMPARISON OF THE CONCENTRATION OF PARTNERS ACROSS SECTORS FOR EACH 
CSO BETWEEN THE BASELINE AND ENDLINE PERIODS. 60
FIGURE 54. PLOT COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON NETWORK 
DIVERSIFICATION OUTCOMES 60
FIGURE 55. DISTRIBUTION OF TIME SPENT ON POLITICAL ADVOCACY 62
FIGURE 56. COEFFICIENTS ESTIMATING THE IMPACT OF THE TREATMENT ON CHANGES IN THE 
SHARE OF TIME SPENT ON POLITICAL ADVOCACY BY TREATMENT ORGANIZATIONS 62
FIGURE 57. COEFFICIENT ESTIMATES FOR THE EFFECT OF COMMUNITY, LG, AND GRANT 
CHARACTERISTICS ON CSOS’ PREFERENCE FOR WORKING IN A GIVEN COMMUNITY 64
TABLE OF TABLES 
TABLE 1. SAMPLE TRAINING TOPICS 10
TABLE 2. RESILIENCYCAMBODIA MANUALS 10
TABLE 3. SAMPLE SIZE BY DATA COLLECTION ROUND 14
TABLE 4. CHARACTERISTICS OF ORGANIZATIONS 20
TABLE 5. SAMPLE SIZE BY USAID SECTOR 20
USAID.GOV LO-MTSR RESILENCYCAMBODIA IE FINDINGS REPORT | vii
ACRONYMS 
CDCS Country Development Cooperation Strategy
CSO Civil society organization 
DRG Democracy, human rights, and governance
HHI Herfindahl–Hirschman Index
HI High-intensity 
IE Impact Evaluation
IWE Interaction-weighted estimator
KS Kolmogorov–Smirnov
LANGO Law on Associations and Non-Governmental Organizations
LG Local government
LI Low-intensity 
LO-MTSR Local Organizations—Movement Towards Self-Reliance
MAD Mean absolute deviation
NGO Non-governmental organization
PAP Pre-analysis plan
R+ Resiliency+
RCT Randomized Control Trial
RGC Royal Government of Cambodia
ROC Resilient Organizations in Cambodia
SBAR Small Business Applied Research 
USAID United States Agency for International Development
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 1
EXECUTIVE SUMMARY
In Cambodia, nearly 5,000 local and international civil society organizations (CSOs) play a critical 
role as service providers in the health, education, democracy, governance, and agriculture 
sectors, especially in remote communities. However, shrinking civic space and volatile donor 
funding threaten the ability of civil society to do this important work. The United States Agency for 
International Development (USAID)/Cambodia relies heavily on local CSOs to achieve its 
development objectives. This findings report describes an innovative impact evaluation (IE) 
testing an intervention designed to build CSOs’ resiliency to closing civic space and funding 
uncertainties through capacity-building, network-strengthening, and financial diversification.
This IE is an innovative, first-of-its-kind experimental evaluation to measure the key development 
impacts of the ResiliencyCambodia program under the Local Organizations—Movement 
Towards Self-Reliance (LO-MTSR) activity. ResiliencyCambodia draws on the PartnersGlobal
Resiliency+ (R+) model, expanded to focus on revenue diversification and “scaled up” to reach 
a larger, more diverse group of CSOs. The IE represents the first attempt by USAID to design 
and evaluate CSO programming using the highest scientific standards.
EVALUATION PURPOSE AND EVALUATION QUESTIONS
USAID/Cambodia identified improved organizational capacity and 
expanded networks as avenues to help organizations access new 
sources of funding. Furthermore, the combination of improved 
capacity, expanded networks, and more diverse revenue streams 
has the potential to increase the ability of CSOs to adapt to 
increasing legal and informal restrictions on their resources and 
activities. 
Drawing on this theory, the overarching policy question underlying 
the evaluation is: 
Does building organizational capacity, networks, and 
financial diversification increase resilience to closing civic 
space? 
Specifically, the evaluation will investigate the extent to which the 
activities under the LO-MTSR activity generate the following 
outcomes: 
Components of the 
ResiliencyCambodia 
Intervention
● Resiliency Orientation
● Resilient Organizations 
in Cambodia (ROC) 
Assessment
● Resiliency Roadmaps
● Coaching and Mentoring
● Capacity Development 
Funds
● Trainings
● Resiliency Resources and 
Toolkits 
● Social Lab
1. Increase the managerial and administrative capacity of CSOs. 
2. Increase the size and strength of CSO networks. 
3. Increase the financial diversification of CSOs. 
4. Increase the resiliency of CSOs to closing civic space. 
These four evaluation objectives form the basis for a series of testable development hypotheses 
and indicators on the impact of the LO-MTSR activity. Under each of these outcomes, the 
Cloudburst Group (Cloudburst) and the DevLab@Penn (DevLab—formerly the DevLab@Duke), 
in consultation with PartnersGlobal, identified a variety of measures to assess the success of the 
program. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 2
METHODS 
The ResilencyCambodia IE is a randomized control trial (RCT). RCTs estimate the impact of an 
intervention (referred to as the “treatment”) by comparing outcomes for participants that receive 
the treatment against outcomes for a “control” group of similar participants that were randomly 
selected to not receive the treatment. This technique gives implementers and donors a 
straightforward way to understand program effects and how outcomes would be different if the 
intervention had not taken place. The LO-MTSR team recruited 105 Cambodian CSOs to 
participate in the ResiliencyCambodia program. The team then assigned each CSO to a 
treatment arm using matched-quadruplet randomization, resulting in a sample of 49 treatment
and 53 control organizations. 
To investigate each evaluation objective, the evaluation utilizes organization-level data measuring 
changes in organizational administrative and managerial practices, challenges that affect their 
operations, networks and networking efforts, and revenues and revenue-seeking activities. Data 
sources include a baseline and endline survey, comprehensive data on Facebook posts and 
interactions during the project period, and detailed data on all revenues and expenditures 
collected by digitizing and classifying each CSO’s financial records for the 2019 and 2021 fiscal 
years. These data allow the team to identify the size of any differential improvements on these 
outcomes over the course of the projective that are attributable to the ResiliencyCambodia 
intervention. The team collected baseline data from April–June 2020 and endline data from April–
June 2022. 
FINDINGS OUTCOME 1: MANAGERIAL AND ADMINISTRATIVE CAPACITY OF TREATED 
CSOS
Overall, there is little evidence of an impact of the treatment on the managerial and administrative 
capacity of CSOs that participated in ResiliencyCambodia. While CSOs in the treatment group 
do report spending a greater share of their time on core programmatic activities (including political 
advocacy and community outreach), this difference is substantively small. Furthermore, there is
no evidence for a differential decrease in their reporting of internal challenges that the intervention 
was designed to address, a differential increase in their administrative capacity (measured by the 
accuracy of their financial records), or a differential increase in the success of their adaptations 
to COVID-19.
FINDINGS OUTCOME 2: NUMBER OF CONNECTIONS BETWEEN TREATED CSOS AND 
CITIZENS, THE PRIVATE FOR-PROFIT SECTOR, AND OTHER CSOS 
Overall, there is no evidence of an increase in the size or strength of CSO networks among 
members of the treatment group. Looking at both self-reported and objective measures of network 
size and strength, as well as measures of effort to expand or strengthen networks, there are no
significant differences between the treatment and control groups. Although treatment CSOs do 
report a greater increase in the self-reported strength of their networks, this difference is 
substantively small and is not statistically significant. As discussed in greater detail in the final 
section, this may be due in part to the program’s heavy reliance on remote activities during the 
first year of COVID-19 lockdowns. Many activities were initially designed to provide in-person 
opportunities for networking and partnership-building, and the switch to remove workshops and 
trainings precluded these activities. These findings do suggest that the social media trainings 
offered to treatment CSOs were not effective at improving their ability to foster public engagement 
with their Facebook content. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 3
FINDINGS OUTCOME 3: FINANCIAL RESILIENCY OF TREATED CSOS 
Overall, there is little evidence that the treatment increased the financial resiliency of CSOs. 
Looking at both objective measures drawn from current budget data as well as self-reported 
measures of recent and future planned behavior, there are no significant differences between the 
treatment and control groups. There is some evidence for a small increase in the total value of 
revenue from donations and earned income and in the total value of revenue from local sources. 
However, these results are imprecisely estimated due to the small sample size and are only 
apparent when removing outlier observations from the sample. 
FINDINGS OUTCOME 4: RESILIENCY TO CHANGING CIVIC SPACE OF TREATED CSOS 
There is mixed evidence for the impact of the treatment on CSO resiliency to changing civic space. 
Counter to expectations, reporting of external challenges increased more for members of the 
treatment group. Furthermore, the effect size is moderate at 0.3 standard deviations and 
statistically significant. These unexpected findings may suggest that the treatment increased 
CSOs’ awareness of civic space issues. There is also evidence for a greater increase in the share 
of time treatment CSOs spend on political advocacy compared to control CSOs. Increased time 
spent on advocacy may be driving increases in the number of external challenges CSOs face in 
a heavily restrictive environment like Cambodia. Alternatively, the intervention’s emphasis on civic 
space issues may have increased the salience of these issues for treatment CSOs. 
RECOMMENDATIONS AND LESSONS LEARNED
Unfortunately, the bulk of the evidence suggests that ResiliencyCambodia did not cause 
meaningful improvements in CSO capacity, networks, finances, or resiliency to closing civic 
space. The team’s analyses estimate small differences in how outcomes for treatment and control 
CSOs changed over, suggesting that these results are not attributable to small sample sizes or 
attrition from the evaluation sample. Furthermore, the prevalence of these null results across both 
objective and self-reported measures, indicating that treatment and control CSOs changed in very 
similar ways over the course of the program on the key outcomes, strengthens the team’s 
conclusion that the program largely failed to achieve its objectives. 
However, the ResiliencyCambodia intervention still provides many lessons and 
recommendations for future USAID capacity-building programs. Recommendations from the IE
include: 
• USAID should invest in programs to help organizations across all sectors combat 
closing civic space. IE data show that the diverse sample of CSOs participating in 
ResiliencyCambodia reported high levels of external challenges related to closing 
civic space but were also willing and able to increase the share of time spent on 
political advocacy. The LO-MTSR activity represented an important effort to design 
programming that directs civic space programming across technical areas to bolster 
organizations working not just on democracy, human rights, and governance (DRG) 
issues like political advocacy on human rights but also ostensibly apolitical sectors like 
health, education, and agriculture. The team recommends that USAID/Cambodia 
continue their support for such work and encourage other missions to build on these 
efforts.
• IEs are a smart investment to ensure taxpayer dollars are spent on maximizing
program impacts. In this IE, as in any applied research endeavor, even null results 
are helpful for providing concrete feedback on program successes and failures, 
potential unexpected consequences, and the need to adapt or redesign activities. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 4
Without the rigorous design of an RCT, it would have been easy to rely only on the 
qualitative feedback from organizations that reported overall satisfaction with the 
program and decide that the ResiliencyCambodia model should be scaled. 
• There is a need to build the evidence base around each component of the 
program’s theory of change. The original theory that motivated this project states 
the expectation that organizations would increase their organizational capacity through 
increasing financial diversification, networks and partnerships, and communications 
strategies. The null results for this initial, necessary outcome are one potential 
explanation for the null results across other outcomes. Each of these outcomes 
requires its own researched theory of change to show that these outcomes are backed 
by evidence to promote organizational capacity.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 5
1.0. EVALUATION PURPOSE AND QUESTIONS
1.1 EVALUATION PURPOSE
This report presents findings from an IE of USAID/Cambodia’s LO-MTSR activity. LO-MTSR is a 
three-year, $2,230,510 USD project funded by USAID/Cambodia as part of the Small Business 
Applied Research (SBAR) pilot mechanism. The LO-MTSR activity includes two primary 
components. The first component is the implementation of a civil society capacity-building 
program, known as ResiliencyCambodia. The second component is an RCT evaluating the 
impact of ResiliencyCambodia on key development outcomes. Cloudburst is conducting both 
ResiliencyCambodia and the IE in partnership with PartnersGlobal and DevLab. 
ResiliencyCambodia draws on the PartnersGlobal R+ model, which has been implemented in 
countries around the world. However, ResiliencyCambodia was modified in several important 
ways to meet the needs of USAID/Cambodia and Cambodian civil society. First, whereas R+
focuses more narrowly on civic space issues, the content of ResiliencyCambodia is expanded to 
address other dimensions of organizational capacity, including revenue diversification. Second, 
while R+ focuses on engaging a small number of CSOs working on civic space issues with intense 
coaching and mentoring, ResiliencyCambodia is “scaled up” to reach a larger and more diverse 
group of CSOs. 
The IE is an innovative, first-of-its-kind experimental evaluation to measure the key development 
impacts of the ResiliencyCambodia program under the LO-MTSR activity. It represents USAID’s
first attempt to design and evaluate CSO programming using the highest scientific standards. The 
project also demonstrates several novel approaches to research design and data collection that 
should be used in future evaluations of CSO programming. By testing the impact of 
ResiliencyCambodia on the outcomes it was designed to improve, this evaluation provides an 
evidence base for improved policymaking and programming. 
Importantly, this evaluation estimates the impact of receiving access to the entire 
ResiliencyCambodia program. The program is a “bundled treatment” containing numerous 
components. This means that the team is not able to identify the impact of individual components, 
such as the coaching CSOs received or the specific trainings they attended. 
The bulk of the evidence in this report suggests that ResiliencyCambodia did not cause 
improvements in the resiliency of participating CSOs. However, it is important to note that the 
program was implemented during the first two years of COVID-19, which posed unprecedented 
challenges to the implementation team. Specifically, much of the programming that was designed 
to take place in person had to be moved online, which potentially reduced engagement and limited 
opportunities for networking.
Although these challenges do not limit the team’s ability to evaluate the impact of this 
implementation, they do limit the team’s ability to make inferences about the expected 
effectiveness of similar programming implemented during more typical conditions. Importantly, 
the results suggest several lessons for the design of future CSO capacity-building interventions, 
including the limitations of remote implementation. The final section discusses these lessons in 
greater detail and offers recommendations for the design of future CSO and resiliency programs.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 6
1.2 EVALUATION QUESTIONS AND THEORY OF CHANGE
The LO-MTSR IE was intended to design, implement, and rigorously evaluate the impact of 
programming designed to decrease CSOs’ reliance on foreign donors and make CSOs more 
resilient to an increasingly difficult external environment. USAID/Cambodia identified improved 
organizational capacity and expanded networks as avenues to help organizations access new 
sources of funding. Furthermore, the combination of improved capacity, expanded networks, and 
more diverse revenue streams has the potential to increase the ability of CSOs to adapt to 
increasing legal and informal restrictions on their resources and activities. 
Drawing on this theory, the overarching policy question underlying the evaluation is: 
Does building organizational capacity, networks, and financial diversification increase 
resilience to closing civic space? 
Specifically, the evaluation investigates the extent to which the activities under the LO-MTSR 
activity generate the following outcomes: 
1. Increase the managerial and administrative capacity of CSOs. 
2. Increase the size and strength of CSO networks. 
3. Increase the financial diversification of CSOs. 
4. Increase the resiliency of CSOs to closing civic space. 
These four evaluation objectives form the basis for a series of testable development hypotheses 
and indicators on the impact of the LO-MTSR activity. Under each of these outcomes, Cloudburst 
and DevLab, in consultation with PartnersGlobal, identified a variety of measures to assess the 
success of the program. 
1.3 PROJECT BACKGROUND 
In Cambodia, nearly 5,000 local and international CSOs play a critical role as service providers in 
the health, education, democracy, governance, and agriculture sectors, especially in remote 
areas and communities. For USAID/Cambodia to achieve its development objectives, CSOs must
be resilient and self-reliant. As donor funding becomes more limited or tied to specific 
governmental objectives, growing competition for these resources threatens to divide local 
organizations working toward development goals. While Cambodian organizations have 
“graduated” to become direct and successful USAID grantees, they have failed to thrive as 
independent, Mission-oriented organizations with diverse funding sources to insulate them from 
external shocks. Furthermore, while governments around the world strive to shrink civic space 
and restrict the operations of organizations promoting accountability and good governance, 
reduced access to international funding will decrease the independence of CSOs and erode their 
ability to provide checks on governments. Encouraging CSOs to shift to more sustainable 
financing models is important for them to continue fulfilling this purpose. 
CURRENT CHALLENGES FACING CIVIL SOCIETY IN CAMBODIA 
In recent years, the Royal Government of Cambodia (RGC) has intensified its interference in CSO 
activities. While this increased scrutiny has resulted in burdensome registration requirements and 
invasive monitoring practices that affect the entire sector, the brunt of this interference has been 
targeted toward CSOs receiving foreign support for rights-based advocacy and democracy 
promotion, as well as local organizations focused on land rights and environmental protection 
(Springman et al., 2022). Much of this increased interference has found a legal basis in the Law 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 7
on Associations and Non-Governmental Organizations (LANGO) enacted in August 2015 (Curley, 
2018), which was met with widespread criticism from civil society and the international community. 
Chief concerns about the LANGO are mandatory registration for all domestic and international 
associations, unfettered discretion by the Ministry of Interior over registration, and the requirement 
that all associations and organizations be “politically neutral.” 
Legal restrictions on civil society are part of a broader push by the RGC to stifle dissent and
undermine political competition. LANGO was accompanied by similar legal restrictions on the 
media and the freedoms of expression and assembly, including the use of social media. In recent 
years, the RGC has used legal channels to limit the main opposition party, expel foreign non￾governmental organizations (NGOs), and eliminate independent media critical of the ruling party. 
These moves come despite increased pressure from the European Union and other international 
actors to discourage the Cambodian People’s Party’s turn toward closing space. While these 
actions represent a purposeful closure, Cambodian CSOs’ reliance on international funding, lack 
of social embeddedness, divisions within civil society, and insufficient CSO capacity make CSOs 
in Cambodia especially vulnerable to this type of restrictive legislation. 
Cambodian CSOs also struggled to operate in a fractionalized civil society sector. Using baseline 
data from this project, Springman and Wibbels (2021) document a relatively low level of 
connectivity and a strong desire to improve connections and engagement with both the public and 
other CSOs. The concept of civil society implies a dense network of individuals and organizations 
capable of engaging in collective action in pursuit of shared goals (Viterna, Clough, and Clarke, 
2015; Petrova and Tarrow, 2007). Networks also convey material benefits to individuals and 
organizations, including facilitating the flow of resources, whether these resources are material, 
legal, political, or technological (Dalaibuyan, 2013; Marshall and Suárez, 2014; Suárez and 
Marshall, 2014; Beaman et al., 2018; Cruz, Labonne, and Querubin, 2020). This lack of 
connectivity among Cambodian CSOs likely limits their ability to organize and operate. For more 
details about the current circumstances of CSOs in Cambodia, see Appendix A. Design Report. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 8
2.0 RESILIENCYCAMBODIA OVERVIEW
This section describes the ResiliencyCambodia 
intervention under evaluation. The goal of the 
ResiliencyCambodia program is to build the 
organizational resiliency of targeted CSOs so they are 
better able to strategically plan for and expand network 
connections outside of primary donors. In doing so, 
organizations will be less dependent on government or 
large international donor funding that may be at odds 
with USAID’s Country Development Cooperation 
Strategy (CDCS) objectives. 
The primary development hypothesis driving the 
LO-MTSR activity is that by increasing the 
organizational resiliency of targeted CSOs through 
the ResiliencyCambodia suite of activities—
including the ROC workshop and assessment, 
one-on-one coaching, resources for leadership 
development, and training on entrepreneurship, 
communications, and social media—organizations 
will be better positioned to seek out and gain 
access to funding streams outside traditional 
international donors (such as USAID) or other 
funding that may be at odds with USAID’s CDCS 
objectives. This increased financial diversity will 
empower local organizations to be increasingly 
independent of donor funds. Figure 1, to the right, 
outlines the causal model approach to 
implementing performance metrics of the LO￾MTSR activity.
The ResiliencyCambodia program consists of a 
bundled intervention of capacity-building activities 
targeted at improving organizations’ organizational 
capacity, reliance on USAID and other donors, and 
networks. Interventions took place over 24 months 
from August 2020 to August 2022, but the IE only 
captures changes through April 2022. Figure 2 below shows the timeline of the various interventions, and 
each intervention is described in greater detail below. 
FIGURE 1. CAUSAL MODEL
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 9
Figure 2. Timeline of ResiliencyCambodia intervention activities 
RESILIENCY ORIENTATION & ROC ASSESSMENT
The team invited all organizations to send two staff representing various levels of seniority to 
participate in a Resiliency Orientation. The one-day small group workshop introduced 
organizations to the ResiliencyCambodia Framework, including discussions of the most 
prevalent civic space threats and common vulnerabilities that make organizations susceptible to 
civic space shifts. During the workshops, CSOs implemented the ROC Self-Assessment Tool.
RESILIENCY ROADMAP
Following the Resiliency Orientation, each of the 52 CSOs created a Resiliency Roadmap with 
assistance from PartnersGlobal and each organization’s Resiliency Coach. The roadmap was 
comprised of strategies, tools, tactics, and approaches that correlate to the challenges identified 
through ROC. Organizations worked with their coach to develop personalized targets and action 
plans and continued to update their roadmap throughout the project. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 10
COACHING AND MENTORING 
Treatment organizations benefited from one day per month of on-demand coaching services by 
trained Resiliency Coaches from October 2020–September 2021.1 Organizations called on their 
coaches for support when writing funding proposals, developing social enterprise plans, re￾designing their organization’s websites, and making connections to new partners, among many 
other things. 
GROUP TRAININGS 
A core component of the intervention was a series of capacity-building trainings conducted by 
PartnersGlobal to help each organization achieve its roadmap goals. Originally planned to be 
conducted in person, the COVID-19 pandemic forced training to be conducted virtually via zoom. 
Trainings included Revenue Generation Training, Wellbeing in the Workplace, Social Media 
Strategy, Enterprise Strategies 2.0, and Strategic Communication. For more information on the 
objectives and content of each training, please refer to Appendix C. Pre-Analysis Plan. 
TABLE 1. SAMPLE TRAINING TOPICS
TRAINING MODULE DATES # OF ORGANIZATIONS 
IN ATTENDANCE
Revenue Generation Training January 18–March 5, 2021 49
Wellbeing in the Workplace June 1–June 9, 2021 24
Social Media Engagement and Digital Security August 9–August 27, 2021 34
Enterprise Strategies 2.0 September 7–September 21, 2021 41
RESILIENCY RESOURCES AND TOOLKITS
In addition to group trainings, ResiliencyCambodia also developed three stand-alone manuals 
on topics of interest identified in each organization’s ROC assessment and Resiliency Roadmap 
priorities. Each toolkit was customized for the Cambodian context, and available in Khmer and 
English. Topics included revenue diversification, partnership and networking, and contingency 
planning. 
TABLE 2. RESILIENCYCAMBODIA MANUALS
RESILIENCY RESOURCES DATE DISTRIBUTED # OF ORGANIZATIONS THAT
RECEIVED THE RESOURCE 
Contingency Planning Guide December 2021 49
Revenue Diversification Toolkit April 2021 49
Partnership and Networking Toolkit June 2021 49
1 Originally, coaching support for CSOs assigned to the low-impact (LI) treatment arm was to be capped at one day per 
quarter, but the availability of additional resources allowed for the tripling of the amount of coaching offered to this 
group.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 11
SOCIAL LAB 
In Year 3, the LO-MTSR team added a new intervention to the bundle of interventions specifically 
to strengthen organizations’ networks and partnership-building. The “Social Lab” intervention 
included a Building Systems Resilience workshop focused on collecting data for systems mapping 
and conducting relationship- and trust-building activities to break down organizational and 
sectoral barriers. The second component was a Social Lab for groups of 6–10+ organizations to 
vote on the most important ideas for cross-collaborations, form teams on those topics, and begin 
to design lean experiments (small activities for learning or testing assumptions about those ideas). 
Thirty organizations ultimately chose to participate in the Social Lab experiment. The 
ResiliencyCambodia activity provided each of the three teams with $3,000 in funding to help 
facilitate learning meetings and activities. After the initial workshop in March, the organizations 
participating in Social Lab teams continued to meet and participate in monthly reflection meetings. 
This intervention was not part of the initial IE design, and none of the evaluation activities were 
designed to measure its impact. 
RESOURCES FOR INSTITUTIONAL DEVELOPMENT 
CSOs randomly assigned to a high-intensity (HI) treatment arm were eligible to apply for a $3,000 
grant for institutional development, paid in two installments over 18 months. The funds provide 
leadership development and institutional support needed to implement their Resiliency 
Roadmaps. Resiliency Coaches helped organizations identify appropriate leadership and 
organizational coaching opportunities, focusing on building organizational capacity or helping 
implement identified improvements. The funds could be used to hire consultants for grant or 
proposal writing, hire consultants to develop monitoring, evaluation, and learning protocols, 
purchase computers or other office equipment, or attend other trainings or networking 
opportunities. Of the 30 HI organizations, 23 ultimately applied for and received these funds. 
Organizations experienced significant challenges complying with USAID regulations on the 
purpose of the funds, as well as the Cloudburst regulations on how to apply for and receive each 
payment. This delayed the receipt of funding for many organizations until the last quarters of the 
program, which limits the IE’s ability to detect any change due to receiving the funds. 
3.0 EVALUATION METHODS 
3.1 METHODS 
The ResilencyCambodia IE, under the LO-MTSR activity, is an RCT. RCTs estimate the impact 
of an intervention by comparing outcomes for treated units against outcomes for a “counterfactual” 
group that was randomly selected to not receive the treatment. This technique gives implementers 
and donors a straightforward way to understand program effects and how outcomes would be 
different if the intervention had not taken place. The random assignment of treatment and control 
units is the most scientifically rigorous way to establish a causal relationship between an 
intervention and outcome; it is considered the “gold standard” in policy evaluation. Random 
assignment is also a normatively fair method for assigning CSOs to programming, given that a 
limited budget necessarily implies that the programming can only be provided to a modest number 
of CSOs. There have been very few rigorous evaluations of programming on CSOs, so this 
approach provides an enormous opportunity for USAID to learn crucial, rigorous lessons that 
might help civil society programming writ large.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 12
The IE was originally designed as a tiered intervention with two treatments “arms”: one receiving 
an LI treatment and the other receiving an HI treatment. A third group serves as a control. The 
design called for 30 CSOs to be assigned to the LI cohort, 30 CSOs to be assigned to the HI
cohort, and 60 to be assigned to the control group. The tiered approach was designed primarily 
due to resource constraints, but it also provided an opportunity to learn whether, and to what 
extent, CSO resilience can be improved with a light touch, or if it requires larger, more sustained 
programming and coaching. The tiered design would also allow for a cost-benefit analysis of LI
vs. HI programming. However, when the ResiliencyCambodia intervention adapted its 
intervention to COVID-19, the LI cohort was able to receive equally intensive coaching, mentoring, 
and training. After these adaptations, the only difference between the two cohorts was the 
availability of capacity-building funds, which remained exclusive to the HI group. Because of this, 
the endline analysis combines the HI and LI groups into a single treatment group. 
3.2 OUTCOME FAMILIES, HYPOTHESES, AND INDICATORS 
This IE report presents results testing the impact of the ResiliencyCambodia intervention on the 
four families of outcomes targeted by the intervention. The findings section describes each 
outcome family and the specific intervention activities designed to target that outcome and
presents formal hypotheses specifying the expected impact on each outcome. Primary outcomes 
are those that will be interpreted as the strongest evidence for each objective and secondary 
outcomes represent measures that are either less likely to be affected by the treatment or are 
less directly related to the objective under consideration. 
Outcome families and their associated primary and secondary outcomes are listed below. 
● Outcome Family 1: Increased managerial and administrative capacity of treated CSOs
− Primary Outcome 1.1: Managerial Capacity Index
− Primary Outcome 1.2: COVID-19 Adaptation Index
− Primary Outcome 1.3: Administrative Capacity
● Outcome Family 2: Increased number of connections between treated CSOs and citizens, 
the private for-profit sector, and other CSOs
− Primary Outcome 2.1: Count of Partnerships Index
− Primary Outcome 2.2: Social Media Interactions Index
− Secondary Outcome 2.3: Self-Reported Network Strength Index
− Secondary Outcome 2.4: Partnership-Seeking Behavior Index
− Secondary Outcome 2.5: Network Density Among Treatment and Control Sample
● Outcome Family 3: Increased financial resiliency of treated CSOs
− Primary Outcome 3.1: Revenue Generation Index
− Primary Outcome 3.2: Revenue Diversification Index
− Secondary Outcome 3.3: Financial Health
− Secondary Outcome 3.4: Diversification Away from Aid: Local Revenue Index
− Secondary Outcome 3.5: Share of Revenue from Foreign Sources
● Outcome Family 4: Increased resiliency of treated CSOs to changing civic space
− Primary Outcome 4.1: External Challenges Index
− Primary Outcome 4.2: CSO Network Diversification
− Secondary Outcome 4.3: Share of Time Engaging in Advocacy
For more details on each outcome family and the data that is used to construct each, please refer 
to Appendix C. Pre-Analysis Plan. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 13
3.3 SAMPLING AND RANDOMIZATION
To secure a sufficiently large sample of local organizations, Cloudburst attempted to recruit at 
least 120 organizations through a call for applications distributed by the partner organizations in 
Cambodia. The organizations were filtered to include only those CSOs with at least three full-time 
staff members, that existed for at least three years, that are registered with the RGC as NGOs, 
and that have been approved by USAID/Cambodia. 
Challenges during recruitment ultimately led to 105 organizations applying to and being selected 
for the program. While Cloudburst ultimately received 127 completed applications, seven 
organizations were rejected for not meeting the program requirements and another 15 
organizations did not complete the necessary surveys by the deadline. The team then asked each 
of the 105 organizations to complete one survey about their organization’s management, 
characteristics, and behavior and one survey asking for detailed financial data. 
Using data collected from these surveys, the team divided the sample into 25 groups of very 
similar CSOs using a non-bipartite matching algorithm. To identify similar organizations, the team 
focused on characteristics related to either the outcomes of interest or how CSOs might respond 
to the intervention. Within each block, the team randomly selected half of the units to receive the 
treatment and the other half to serve as controls. The final randomization resulted in 32 HI 
organizations, 20 LI organizations, and 53 control organizations. 
Matching organizations on relevant characteristics and randomizing within these matched groups 
allows the team to compare outcomes across very similar organizations. This technique can 
dramatically reduce variance and increase statistical power. In short, this matching exercise 
ensures that the team’s R+ cohorts and control group are as similar as possible, thereby 
improving the team’s capacity to detect differences at endline between organizations that did and 
did not receive programming. This permits a more precise measurement of measure program 
effects. This also helps to ensure that important organizational characteristics that shape how 
organizations respond to the program are distributed equally between the treatment and control 
groups. 
Of the 105 CSOs that applied to participate in ResiliencyCambodia, six CSOs dropped out prior 
to the start of intervention activities. These CSOs either withdrew over concerns about the amount 
of staff time required to participate or ceased responding to the team’s attempts to contact them. 
These six CSOs were distributed equally between the treatment and control arms, eliminating 
concerns that this attrition may bias the results. For this reason, the team dropped these CSOs 
from the sample prior to estimation.
A detailed description of the measures, algorithms, randomization procedures, and software used 
to perform the matching and randomization is available in the Baseline Report referenced in 
Appendix B. Results presented in the pre-analysis plan (PAP) referenced in Appendix C
demonstrate that the sample is well-balanced and the design is powered to detect medium-sized 
effects between 0.34 and 0.37 standard deviations.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 14
Figure 3. CSO sampling methodology
3.4 DATA SOURCES
The evaluation utilizes three primary sources of organizational-level data to investigate outcomes 
of interest, including organizational characteristics, networks, revenue-generating activities, 
employee perspectives of political and financial independence, funding sources, and revenue. 
The three sources of data were:
1. CSO survey: The survey firm collected organization-level data from 105 organizations at 
baseline and 82 organizations at endline. A CSO employee completed an approximately one￾hour-long, self-administered Qualtrics online survey. For most organizations, the executive 
director completed the survey. The survey focused on organization characteristics, budget and 
fundraising information, networking, and challenges faced. 
2. Budgets: The survey firm collected detailed budget data for 2019 and 2021 from 94 
organizations at baseline and 78 organizations at endline. Teams of two enumerators visited the 
offices of 85 organizations, spending approximately four hours conducting an interview with a 
member of the NGO’s leadership team and a member of the NGO’s finance team. The finance 
member provided complete budgets from 2019 and 2021, with information on income, expenses, 
and assets. Enumerators classified budget line items into a series of categories created by the 
research team, asking clarification questions to organizations’ finance teams whenever 
necessary. Budget categories included grant sources (USAID, the European Union, China, etc.) 
while expense categories included labor and various types of project material or expenses. The 
financial manager remained available to answer any questions while the enumerator classified 
each line item in the budget. Income was classified by source, while expenditures were classified 
by type. The full list of budget categories and subcategories is detailed in Appendix E. Budget 
Categories. To the best of the authors’ knowledge, this represents the most detailed financial data 
ever collected from a sample of non-profit organizations in the global south. 
3. Social media data scraping: Cloudburst and DevLab collected organization-level data on 
each post made to a CSO’s Facebook page from 87 organizations at baseline and 89 
organizations at endline. Data collected included the date of the post, the content (text and 
images), and the number of likes, comments, and shares that the post received. This information 
provided multiple measures for each organization’s Facebook activity and the extent of its online 
following. This information was collected through an automated Python script. In addition to the 
automated data scraping, Cloudburst hand-collected information about the organizations’ Mission 
statements, number of page likes, number of page followers, number of events (in 2021 and in 
the future), number of top fans, and if the page was running ads. 
TABLE 3. SAMPLE SIZE BY DATA COLLECTION ROUND
DATA SOURCE BASELINE ENDLINE
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 15
CSO survey 105 87
Budget (2019 and 2021) 94 78
Social media data scraping 87 89
BASELINE DATA COLLECTION
The team collected data through an online survey hosted by the online survey platform Qualtrics. 
After applying and being selected to receive the baseline survey, organizations received an initial 
Qualtrics contact form requesting the primary point of contact at the organization to nominate two 
staff members to complete the CSO survey and a third staff member to complete the budget 
survey. The contact form requested the names, positions, and email addresses of the employees 
who would complete the survey. This information became the sampling list for each survey 
instrument. 
Baseline data collection took place between April 20 and July 10, 2020. Respondents entered 
data directly into the online form, which the team then exported into comma-separated value files 
and cleaned and analyzed in R. Once contact information was collected, Qualtrics generated an 
email to link each survey respondent to its survey. This allowed Cloudburst to track who had 
opened each email, who had started a survey, and who had finished a survey, and then follow up 
appropriately. Each week, respondents who had not yet completed their survey received a 
personalized email reminder, and as the survey end date approached, respondents received 
phone call reminders from a ResiliencyCambodia coach. Respondents were offered a $10 
payment, sent through the mobile money app Wing, in exchange for completing the survey. 
ENDLINE DATA COLLECTION
Cloudburst and DevLab designed endline data collection protocols and instruments that rely on 
in-person enumeration. After soliciting applications from research firms working in Cambodia to 
lead the implementation of in-person enumeration, Cloudburst contracted the local survey firm, 
IndoChina Research Limited (IRL), to manage all aspects of endline data collection through a 
competitive request for proposal process. IRL has extensive experience working in the social 
research sector over the past decade, including work on cross-cutting issues such as political 
issues, health and maternal rights, feasibility, and access to WATSAN products and services. The 
IRL team consisted of eight enumerators, each with extensive knowledge of the Cambodian NGO 
space. 
After a week of enumerator training and discussion of the objectives of data collection, the team 
finalized the data collection protocols and instruments based on feedback from the enumerator 
team and three pilot organizations. Enumerators became familiarized with the survey and budget 
data collection protocols and instruments through a detailed overview of the ResiliencyCambodia 
project, the CSO Qualtrics Survey, budget data categories, and the budget data collection 
spreadsheet. 
Endline data collection began on April 4, 2022 and continued until the end of June, 2022. Teams 
of two enumerators visited the offices of 85 organizations, spending approximately four hours 
conducting an interview with a member of the NGOs leadership team and a member of the NGOs 
finance team. Organizations completed an online Qualtrics survey on the organization’s activities 
and fundraising. The finance member provided complete budgets from 2019 and 2021, with 
information on income, expenses, and assets. Enumerators classified budget line items into a 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 16
series of categories created by the research team, asking clarification questions to organizations’ 
finance teams whenever necessary. Budget categories included grant sources (USAID, the 
European Union, China, etc.) while expense categories included labor and various types of project 
material or expenses. Throughout the data collection process, enumerators addressed any 
questions or concerns organizations had. Organizations were paid $75 for their time, and 
individuals who completed each survey were paid $15 each. 
DATA QUALITY
Throughout the data collection, Cloudburst and the DevLab monitored for data quality through 
individual reviews of each CSO Qualtrics survey and the budget spreadsheets. These checks 
searched for systematic errors, including short survey times, missing responses, outliers, and 
illogical responses in the CSO survey as well as missing responses, outliers, incorrect 
classifications, and improperly aggregated responses in the budget spreadsheets. To reconcile 
differences, the team sent follow-up emails to organizations requesting clarification. The research 
team was in constant contact with the IRL enumerators, providing support and addressing any 
questions they had. The team also compared results to the CSO baseline survey and CSO 
budgets, if available. Organizations could also clarify any points of confusion by contacting 
Cloudburst or Duke personnel directly through email, WhatsApp, and Telegram. 
Organizations varied widely on how their budgets were formatted, stored, and shared. While some 
organizations kept their transactions coded by project in accessible Excel spreadsheets, other 
organizations required substantial effort to decipher paper records. Some organizations kept a 
detailed record for every expense in a project’s budget, while others only had highly aggregated 
sums for broad categories. The assessment team was mindful of possible biases that may have 
arisen from these different formats. Figure 4 below shows that roughly 10 percent of organizations 
kept only nonelectronic records, with little difference between treatment and control organizations. 
Organizations generally kept the same method of record-keeping from 2019 to 2021, so there 
was little variation from year to year. 
Figure 4. Type of records provided and level of detail across groups
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 17
Roughly half of the organizations provided detailed transaction-level detail. The assessment team 
impressed on the enumerators the importance of receiving budgets as detailed as possible and 
provided guidance on how to assuage any privacy concerns organizations had. The assessment 
team also followed up directly with organizations on these concerns, providing reassurances of 
confidentiality and clarifying the project’s goals. 
Figure 5 below shows the level of aggregation provided across five of the budget categories, 
broken down by treatment group and by year. Control organizations were less likely to provide 
detailed, transaction-level budgets than the treatment organizations. This difference is especially 
notable for earned income and expenses. This may be due to lower levels of familiarity and trust 
with the ResiliencyCambodia project. It is important to note that this could bias the team’s primary 
measure of financial misreporting (Outcome 1.3) or line-item measures of revenue concentration 
(one measure discussed under Outcome 3.2) to make treatment CSOs look like they have lower 
levels of misreporting and concentration. However, because this difference in reporting between 
the treatment and control groups is present at both baseline and endline, this should not bias the
statistical analysis, which identifies differential changes in these outcomes between baseline and 
endline for both treatment assignment groups.
Figure 5. Type of records provided and level of detail across groups
PANEL ATTRITION
Attrition was a significant problem in this project. In addition to reducing sample size and statistical 
power, attrition may bias estimation when there is differential attrition across treatment 
assignment. Of the 53 control and 52 treatment CSOs, 16 control and 14 treatment CSOs did not 
fully participate in endline data collection. Because attrition was relatively balanced between 
treatment and control arms and most estimated treatment effects are close to zero, the 
substantive impact of these limitations on the findings in this report is minimal.
Of the 105 CSOs that applied to participate in ResiliencyCambodia, six CSOs dropped out prior 
to the start of intervention activities. These CSOs either withdrew over concerns about the amount 
of staff time required to participate or ceased responding to the team’s attempts to contact them. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 18
These six CSOs were distributed equally between the treatment and control arms, eliminating 
concerns that this attrition may bias the results. One additional treatment CSO dropped out after 
the start of intervention activities.
The team expected further attrition to be a significant problem for endline data collection due to 
the time-intensive nature of data collection, the potential for NGOs to close or experience changes 
in leadership, and factors like confusion about why control groups are expected to participate in 
the evaluation. For this reason, the team designed the endline data collection activities to 
minimize further attrition of NGOs from the sample. First, rather than distributing fully self￾administered online data collection instruments by email (as at baseline), the team hired a survey 
firm to make contact with each NGO prior to data collection, schedule a meeting with the 
necessary employees, and travel to their offices to administer data collection in person. This 
dramatically reduced the amount of effort required by respondents and allowed the research team 
to build trust with respondents and communicate with those that had reservations about allocating 
staff time or sharing potentially sensitive information. Second, the team substantially increased 
financial incentives. 
Despite these efforts, attrition was high. In total, two treatment and three control CSOs closed 
between 2019 and 2021, the survey firm was unable to make contact with two control CSOs that 
also likely closed, and six treatment and seven control CSOs refused to participate in at least one 
component of endline data collection. Four treatment CSOs completed the survey but did not 
complete budget data collection. To maintain balance in attrition, the results presented in this 
report exclude those CSOs from all analyses. Due to the relative balance in attrition, the already￾small sample size, and the prevalence of null results, the team does not estimate results with and 
without blocks from which attrition occurred. The team also does not estimate Lee bounds for the 
treatment effect. These changes are also noted in the section on departures from the PAP 
referenced in Appendix C. 
Figure 6. Attrition by treatment and control groups 
One additional treatment and one additional control CSO did not finalize endline data collection. 
Both provided only one year of budget data to enumerators. This results in a final sample of 39 
treatment and 37 control NGOs for which there is complete data. This reduces statistical power 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 19
and limits the ability to conduct several tests and test certain hypotheses (noted in the section on 
departures from the PAP). However, the substantive impact of these limitations on the findings in 
this report is minimal. In general, the models estimate very small differences between the 
treatment and control groups. For the majority of the outcomes, these estimates would indicate a 
small impact of the intervention, even if differences were statistically significant. The issues of
each outcome are discussed in the relevant section of the text. Due to the reduced sample size, 
the team also refrains from trying to estimate differential treatment effects for the LI and HI 
treatment arms which decision the team also made in light of the low levels and late timing of 
updates, as discussed earlier.
3.5 SAMPLE CHARACTERISTICS
Participants in the ResiliencyCambodia program originated from 16 of Cambodia’s 25 provinces. 
This represented every major region of the country. Just under two-thirds of CSOs (60 percent, 
N=63) in the ResilencyCambodia program were headquartered in Phnom Penh, reflecting the 
large clustering of CSOs nationally in the capital. A distant second was Battambang, a major 
commercial hub and agriculture center located in the northwest of the country. Siem Reap, a 
tourist and commercial hub also located in the northwest of the country was the third largest. 
These provinces contained nine and eight organizations that participated in the program, 
respectively. Other provinces represented included (but were not limited to) Kampong Cham, 
Kampong Thom, Ratana Kiri, and Kratie. Figure 7 shows each province and the number of 
participating organizations from each. Colors represent the treatment status of the organizations, 
which control in grey, HI in dark blue, LI in light blue, and the three dropouts in red. 
Figure 7. Map of CSO locations
Organizations in the sample were diverse, ranging from small, relatively recently founded 
organizations to large, well-established organizations. Table 4 below shows the total number of 
employees—broken into full-time employees, part-time employees, and volunteers—at endline. 
Full time is defined as being paid for over 30 hours a week. There was a wide range in organization 
staff size, with the largest organization (Cambodian Children's Fund) having 464 employees and 
the smallest (Rehabilitation and Development for Cambodians Organization) having three. 
Organizations had an average of 44.5 total employees, which represents an increase from the 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 20
baseline mean of 35 employees. Except for a few much larger organizations, organizations 
typically do not have many part-time employees (an average of 2.5). The average number of
volunteers at an organization is 6.4, although the larger organizations can employ over 100.
TABLE 4. CHARACTERISTICS OF ORGANIZATIONS
MEAN (STANDARD DEVIATION) MIN–MAX
TOTAL EMPLOYEES 44.5 (60.4) 3–464
FULL-TIME EMPLOYEES 33.0 (51.6) 1–399
PART-TIME EMPLOYEES 2 .5 (5.8) 0–30
VOLUNTEERS 6.4 (19.9) 0–150
The sample for baseline data collection included 105 CSOs working in sectors across all three of USAID’s 
technical areas: agriculture/food security, health/education, and DRG. Table 5 illustrates the disaggregation 
of the sample across these sectors. Please note that because organizations frequently work across multiple 
sectors, some organizations are counted more than once.
TABLE 5. SAMPLE SIZE BY USAID SECTOR
AGRICULTURE/FOOD SECURITY HEALTH/EDUCATION DRG
Baseline Endline Baseline Endline Baseline Endline
Treatment 29 28 36 33 23 20
Control 21 15 27 36 20 14
Total 51 43 73 59 43 34
4.0 ANALYSIS
MAIN OUTCOMES ANALYSIS
To estimate the effect of ResiliencyCambodia on the outcomes of interest, the team estimates 
Equation 1 using ordinary least squares. 
Equation 1: Yi = B1*Zi + B2*lag(Yi) + B3*α + ei
Yi is the endline value of an outcome variable, Zi is an indicator for each CSO’s treatment status, 
lag(Yi) is the baseline value of the outcome variable, α indicates fixed effects for each matched 
quadruplet “block,” and ei is an error term. To account for heteroskedasticity, the team calculates 
robust standard errors. Specifically, the team implements more conservative HC3 standard errors 
for their superior performance with small sample sizes (Long and Erwin, 2000). Due to the 
potential for HC3 standard errors to over-correct in some cases, the team highlights and 
discusses estimates that are substantively meaningful even if they are not statistically significant.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 21
The team opted for ANCOVA rather than a difference-in-differences estimator due to the greater 
statistical power when autocorrelation of outcome variables is relatively low without reduced 
power when autocorrelation is high (McKenzie, 2012). Although respondents are assigned to 
treatment arms with the same probability across blocks, attrition is not perfectly balanced between 
the treatment and control arms, introducing the possibility of some bias in estimation. 
Furthermore, the potential for heterogeneous effects across treatment blocks also creates the 
potential for bias in the estimates. To address these concerns, the team reports findings with and 
without the interaction-weighted estimator (IWE) suggested by Gibbons et al. (2018). Further 
discussion of these decisions is available in the PAP referenced in Appendix C.
As is indicated in Figure 6, there are lower levels of attrition for the CSO survey data compared 
to the budget data. As a result, attrition from the survey is much less balanced between the 
treatment and control groups. For all outcomes measured using survey data, the team reports 
results excluding CSOs that completed the survey but did not complete budget data collection. 
However, results are unaffected when including these CSOs in the analyses.
When looking at outcomes measured using data from CSO budgets, some results are sensitive 
to the inclusion of data from a small number of CSOs that reported extremely large changes in 
outcome values between baseline and endline. These outliers may represent errors in reporting 
or highly anomalous financial circumstances. With a small sample, outlier values can drive 
estimates. In these cases, the team reports results obtained when excluding the most extreme 
outliers. The team identifies these cases when reporting the results and notes that this adds 
additional uncertainty to the results. However, there are no cases where findings that include 
these outliers suggest a positive impact of the treatment, reinforcing the team’s conclusions about 
the failure of the treatment to accomplish its objectives. 
When possible, the team tests each hypothesis using a single index variable that summarizes 
variation across individual measures. This strategy reduces concerns about multiple hypothesis 
testing. For each index, the team first converts all measures included in the index (component 
variables) into a z-score and then combines these into a single average z-score. Z-scores are 
constructed by subtracting the mean of the control group from each observation and dividing by 
the standard deviation of the control group. The averaged z-score index is constructed by 
averaging the z-scores across component variables. Prior to construction, component variables 
are re-scaled to ensure that positive/negative values have the same direction. 
One advantage of z-scores is that they can be interpreted in terms of standard deviations, allowing 
for direct comparisons of magnitudes across different outcomes. Generally, effect sizes around 
0.2 are considered small, around 0.5 are considered medium, and around 0.8 are considered 
large (Cohen, 1992). For results that are reported as a z-score, the team discusses their 
magnitude in terms of standard deviations to facilitate comparison. As shown in the PAP, pre￾registered power calculations indicate that the team can detect treatment effects of between 0.34
and 0.37 standard deviations. This means that the team will be able to detect medium-sized 
effects as statistically significant but are unlikely to detect small effects. Additional details about 
multiple-hypothesis testing and index construction are available in the PAP referenced in 
Appendix C. 
5.0 FINDINGS—MANAGERIAL AND ADMINISTRATIVE CAPACITY OF 
TREATED CSOS
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 22
Figure 8. Summary of results for Outcome Family 1
Overall, there is little evidence of an impact of the treatment on the managerial and administrative 
capacity of CSOs that participated in ResiliencyCambodia. The ResiliencyCambodia 
intervention includes organizational planning, management coaching, and targeted skills training 
to improve the management, adaptation, and administrative capacity of treatment CSOs. While 
CSOs in the treatment group do report spending a greater share of their time on core 
programmatic activities (including political advocacy and community outreach), this difference is 
substantively small. Furthermore, there is no evidence of a differential decrease in their reporting 
of internal challenges that the intervention was designed to address, a differential increase in their 
administrative capacity (measured by the accuracy of their financial records), or a differential 
increase in their adaptations to challenges related to COVID-19.
PRIMARY OUTCOME 1.1: MANAGERIAL CAPACITY INDEX: INTERNAL CHALLENGES AND 
TIME ALLOCATION
ResiliencyCambodia trainings in Organizational Planning, Management Coaching, and 
Wellbeing in the Workplace were designed to help treatment CSOs identify and address internal 
challenges that inhibit the organization’s resiliency and capacity. All components of the 
Management Capacity Index are drawn from the CSO baseline and endline surveys. The first set 
of components is drawn from a question that asks respondents to identify internal challenges that 
inhibit the ability of the organization to achieve its goals or fulfill its Mission. Specifically, the team
selects six internal challenges that the Wellbeing in the Workplace webinars are designed to 
address. 
The second set of components is drawn from a question that asks about the share of management 
and staff time spent on a series of activities during a typical month. Specifically, the team selects 
three activities that constitute the core of most CSOs’ Missions. Although the intervention is likely 
to increase the share of time spent on some non-core activities, such as capacity-building and 
fundraising, the team expects that more effective management will allow CSOs to increase the 
amount of time spent on core activities.
The team hypothesizes that the Managerial Capacity Index will increase more for CSOs in 
the treatment group than for CSOs in the control group. The team combines the following 
variables into a single index variable as described in Section 4, with higher values of the index 
indicating a more desirable outcome. The team then estimates Equation 1 taking this index as 
the outcome variable.
Internal challenges:
● Unable to adapt the organization in response to environmental changes. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 23
● Lack of inter-generational leadership. 
● Unclear Mission and goals. 
● Lack of staff. 
● Overworked staff. 
● Lack of staff training. 
Time utilization: share of time spent on each activity: 
● Directly providing services.
● Advocating or raising awareness.
● Conducting community outreach and communication.
Figures 9 and 10 below plot the observed values for each component of the Managerial Capacity 
Index. Figure 9 plots the values for the internal challenges components, showing that although 
the share of CSOs reporting that each of these challenges are an obstacle decreases between 
baseline and endline, it decreases more for the control group. Figure 10 plots values for the time 
utilization components, where the share of time spent on advocacy and outreach increases for 
the treatment group and decreases for the control group while the share of time spent on service 
delivery decreases for both groups, but the decrease is greater for the treatment group. In Figure 
11, the team plots values for index variables created by combining the internal challenges 
components only, the time utilization components only, and an index that combines all of these 
variables. In the final combined index, internal challenges components are inverted to ensure that 
a positive value indicates an increase in organizational capacity.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 24
Figure 9. Share of CSOs reporting that each of these internal challenges interferes with their ability to fulfill their Mission
Figure 10. Share of time spent engaging in political advocacy, community outreach, and service delivery
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 25
Figure 11. Distribution of averaged z-score index variables for the full index combining all component variables (Row 1), an 
averaged z-score combining components of the internal challenges in Figure 9 (Row 2), and an averaged z-score combining 
components of the time utilization in Figure 10 (Row 3)
Figure 12 plots the estimated coefficients measuring the impact of the treatment on each outcome 
and 90 percent and 95 percent confidence intervals. Overall, there is no evidence of an increase 
in overall managerial capacity. In fact, treatment CSOs are slightly more likely to report that these 
internal challenges are interfering with their ability to fulfill their Missions. However, there is some 
evidence for a small increase in the share of time spent on core programming activities, including 
political advocacy, community outreach, and service delivery. At about 0.2 standard deviations, 
this difference is small but statistically significant. This increase corresponds roughly with a 3 
percent increase in time spent on these core programmatic activities. Looking at Figure 10, the 
share of time spent on service delivery decreases more for the treatment than for the control 
group, suggesting that this overall increase across the three time utilization categories is driven 
by increases in time spent on advocacy and community outreach. Importantly, this is only one 
component of the pre-registered index, so one must be cautious when interpreting this finding.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 26
Figure 12. Plot coefficients estimating the impact of the treatment on managerial capacity outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals
PRIMARY OUTCOME 1.2: COVID-19 ADAPTATION INDEX
For many CSOs, the COVID-19 situation in Cambodia shifted organizational priorities away from 
those established at the outset of the intervention and toward financial survival. Many 
organizations expressed concerns that donors were almost exclusively focused on either COVID￾19 or major NGOs within the country and not small, locally based ones. To assess whether 
ResiliencyCambodia trainings and coaching put treatment CSOs in a better position to adapt, 
the team added a short module to the endline survey asking about organizational experiences 
with and responses to COVID-19. 
This outcome is drawn from a question that asks respondents to select from a list which changes 
their organization made as a direct result of financial pressures created by COVID-19. The team 
hypothesizes that the COVID-19 Adaptation Index will increase more for CSOs in the 
treatment group than for CSOs in the control group over the intervention period. The team
combines the following survey questions (listed below) into a single index variable using the 
method described in Section 4. The team then estimates Equation 1 taking this index as the 
outcome variable (excluding the control for baseline values). When combined into the index, the 
values for negative components that may harm organizational resiliency and indicate immediate 
distress are inverted to ensure that positive values of the index are associated with improved 
adaptation to COVID-19. 
CSOs made the following positive adaptations to financial pressures created by COVID-19: 
● Changed program activities to increased or purely remote work. 
● Formed new partnerships. 
● Pursued new revenue-seeking activities. 
● Planned or hosted events online (for example, on Zoom or Facebook Live) because they could not 
be held in person.
● Pivoted projects to focus on COVID-19-related activities. 
CSOs made the following negative adaptations to financial pressures created by COVID-19:
● Permanently reduced the number of staff or staff working hours.
● Reduced expenditures on other non-staff recurrent expenses, such as office space.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 27
● Permanently suspended or reduced any programming.
● Sold assets to cover expenses.
Figure 13 below plots the share of CSOs reporting that they engaged in each of the COVID-19
adaptation activities. Importantly, these variables are measured at endline only, meaning that the 
team cannot control for pre-randomization baseline differences between treatment and control 
organizations. 
Figure 13. Share of CSOs reporting that they engaged in each of the COVID-19 adaptation activities 
Figure 14 below plots the estimated coefficients measuring the impact of the treatment on the 
index and 90 percent and 95 percent confidence intervals. There is no evidence that the treatment 
increased the quality of CSOs’ COVID-19 adaptations. Although the confidence intervals are 
wide, the extremely small size of the coefficient (~0.025 standard deviations) indicates that there 
is no meaningful difference between treatment and control CSOs.
Figure 14. Plot coefficients estimating the impact of the treatment on COVID-19 adaptations outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 28
PRIMARY OUTCOME 1.3: ADMINISTRATIVE CAPACITY 
To assess administrative capacity, the team uses a technique from forensic accounting to 
measure the ability of CSOs to maintain accurate financial reports. Maintaining an accurate 
budget is essential to gaining access to new sources of revenue and navigating more stringent 
regulations in the NGO sector. Benford's Law holds that the observed distribution of digits in many 
numeric distributions will follow a logarithmic distribution. This empirical phenomenon has been 
shown to apply to sources of data ranging from the surface area of rivers and the molecular 
weights of chemical compounds to house prices and the luminance values of pixels in 
photographs.
A Benford analysis of financial records compares the distribution of leading digits (typically, the 
first one or two digits) of all numeric entries in an organization’s financial data with the Benford 
distribution. The greater the deviation of observed leading digits from the expected Benford 
distribution, the less likely it is that the data are being reported accurately. Like other financial 
data, CSOs' financial data are expected to conform to the Benford distribution (Nigrini, 1999; Qu, 
Steinberg, and Burger, 2020; Dang, Burger, and Owens, 2020).
Following Dang and Owens (2020), the team adheres to the Benford distribution as a measure of 
administrative capacity for CSOs. Hill (1995) states that data must meet three conditions in order 
to be expected to adhere to the Benford distribution: Data should not have a natural maximum or 
minimum, the values should be naturally occurring rather than externally assigned, and the 
distribution should be positively skewed with a median that is lower than the mean. The team’s 
data meet all of these criteria. The team hypothesizes that the size of the deviation from the 
Benford distribution will decrease more for the treatment group than for the control group.
The team begins by combining the responses from all CSOs into a single list of numeric values. 
This yields 71,477 distinct financial entries. To assess the adherence of these numbers to the 
Benford distribution, the team plots the count of observed first two digits against the count of first 
digits expected by the Benford distribution, disaggregated by year and treatment assignment. To 
provide a statistical test of conformity with the Benford distribution, the team also calculates the 
mean absolute deviation (MAD), which quantifies the absolute value of the difference among 
observed and expected frequencies of leading digits. MAD scores between 0.004 and 0.008 
indicate “acceptable” conformity to the Benford distribution (Drake and Nigriri, 2000). Scores 
below this range indicate extremely high conformity while higher scores above this range indicate 
potential misreporting. Figure 15 plots the count of observed occurrences of 10–99 as the leading 
two digits for each financial entry in the data. Grey bars indicate the observed distribution of 
leading digits of each financial entry. The red line indicates the distribution of leading digits that is 
expected by Benford's Law. Looking at the MAD scores printed in each facet of the plot, there is 
little evidence of non-conformity in these aggregated data. There is a small increase in conformity 
in the treatment group along with a small decrease in conformity among the control group. 
However, these differences are small and suggest high levels of conformity for both groups at 
baseline and endline.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 29
Figure 15. Count of observed occurrences of 10–99 as the leading two digits for the full sample of CSO financial entries
To investigate how widespread misreporting is within the sample, the team now tests the 
conformity of financial entries to the Benford distribution for each CSO individually. Due to the 
small number of financial entries for individual CSOs, the team limits its analysis to the first digit 
only and drops the MAD test statistic in favor of the Kolmogorov–Smirnov (KS) test, which is better 
suited for applications with fewer than 100 entries (Dang and Owens, 2020). Similar to MAD, the 
KS test assigns a score that measures the extent to which the observed distribution of digits 
departs from the distribution expected by Benford’s Law. All but 10 CSOs have at least 20 distinct 
financial entries, suggesting a large enough sample to apply the KS test to each CSOs’ financial 
records.
Figure 16 plots a comparison of observed values of the KS score for each CSO between the 
baseline and endline periods. The red line indicates zero changes between baseline and endline. 
The black text indicates the change in the mean value of the KS-score for each treatment group 
between the baseline and endline periods.2 A higher score indicates a larger departure from the 
Benford distribution. Values above the red line indicate an increase in the KS score between 
baseline and endline. The KS score increases between baseline and endline for both the 
treatment and control groups. This change is slightly larger for the control group. However, the 
average increase in the KS score for both the treatment and control groups is relatively small 
(~0.2 standard deviations). In the sample, KS scores range from 0.19 to 8 with a standard 
deviation of 1.33. Furthermore, the increase observed in the treatment group is only 0.1 greater 
than the increase in the control group (0.08 standard deviations).
2 At the individual level, the KS test flags 63 out of 160 budgets across 44 out of 80 CSOs as exhibiting a statistically 
significant deviation from the Benford distribution.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 30
Figure 16. Comparison of observed values of the KS score for each CSO between the baseline and endline periods
Looking at Figure 17, there is no evidence of a statistically significant change in the KS score between 
treatment and control. The two models produce conflicting evidence about the direction of the impact 
and both coefficients are extremely small at around 0.01 standard deviations.
Figure 17. Plot coefficients estimating the impact of the treatment on managerial capacity outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 31
6.0 FINDINGS—CONNECTIONS BETWEEN TREATED CSOS AND 
CITIZENS, THE PRIVATE FOR-PROFIT SECTOR, AND OTHER CSOS 
Figure 18. Summary of results for Outcome Family 2
The concept of civil society implies a dense network of individuals and organizations capable of 
engaging in collective action in pursuit of shared goals (Viterna, Clough, and Clarke, 2015; 
Petrova and Tarrow, 2007). Networks also convey material benefits to individuals and 
organizations, including facilitating the flow of resources, whether these resources are material, 
legal, political, or technological (Dalaibuyan 2013; Marshall and Suárez 2014; Suárez and 
Marshall 2014; Beaman et al. 2018; Cruz, Labonne, and Querubin 2020). Among 
ResiliencyCambodia CSOs, Springman and Wibbels (2021) document a relatively low level of 
connectivity and a strong desire to improve connections and engagement with both the public and 
other CSOs.
The ResiliencyCambodia intervention includes organizational planning, intensive management 
coaching, and targeted skills trainings to help NGOs expand and better use their networks. This 
includes a detailed guide and coaching on building networks and developing partnerships, a 
three-module training course on using social media for strategic communications, and nine 
remote interactive discussion sessions and one in-person workshop where organizations could 
meet with other organizations within and outside their traditional sectors to learn from one another. 
To estimate the impact of the ResiliencyCambodia intervention on CSO networks, the team
utilizes two primary and four secondary measures. The primary measures focus on objective 
measures of the size of NGO networks, including the number of partnerships a CSO has formed 
and the amount of social media interaction a CSO has. The team tests the impact of 
ResiliencyCambodia on these two outcome measures separately because they capture different 
types of networks and different intervention activities are designed to target them. This provides 
evidence regarding the effectiveness of specific clusters of intervention activities on their 
associated outcome. The secondary measures rely on subjective reports on the quality of these 
networks or the CSO’s ability to communicate or utilize network connections.
Overall, there is no evidence for an increase in the size or strength of CSO networks among 
members of the treatment group. Looking at both self-reported and objective measures of network 
size and strength, as well as measures of effort to expand or strengthen networks, there are no
significant differences between the treatment and control groups. Although treatment CSOs do 
report a greater increase in the self-reported strength of their networks, this difference is 
substantively small and is not statistically significant. As discussed in greater detail in the final 
section, this may be due in part to the program’s heavy reliance on remote activities during the 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 32
first year of COVID-19 lockdowns. Many activities were initially designed to provide in-person
opportunities for networking and partnership-building, and the switch to remove workshops and 
trainings precluded these activities. These findings do suggest that the social media trainings 
offered to treatment CSOs were not effective at improving their ability to foster public engagement 
with their Facebook content. 
PRIMARY OUTCOME 2.1: COUNT OF PARTNERSHIPS INDEX
The first primary outcome under CSO networks is the change in the number of partnerships 
reported between January and December 2019 (12-month pre-treatment period before the call 
for applications) and April 2021 and March 2022 (12-month post-treatment period before endline).
The team measures organizational networks by asking respondents to report the names and 
sectors of up to 15 non-profit or other private organizations with which they have partnered over 
the past year.3 The team hypothesizes that the number of connections reported will 
increase more for the treatment group than for the control group. The team combines the 
following variables into a single count of partnerships to use as an outcome variable in Equation 
1:
● The number of non-profit organization partners reported (Cambodian NGOs, foreign NGOs).
● The number of non-profit organization partners named (Cambodian NGOs, foreign NGOs).
● The number of other private organization partners reported (domestic corporations, foreign 
corporations, local businesses, professional associations, unions).
● The number of other private organization partners named (domestic corporations, foreign 
corporations, local businesses, professional associations, unions).
Figures 19 and 20 plot the average number of non-profit and other private partners (total and by
sector) reported by CSOs in the treatment and control groups in both the baseline and endline 
surveys. Both figures show substantial variation in the change in the number of partnerships 
across sectors and treatment assignments. Figure 21 plots the coefficients estimating whether 
the treatment group experienced a significantly larger increase in partnerships between baseline 
and endline. There is no evidence of a significant difference in the change between baseline and 
endline for the treatment group. 
3 The team provides respondents with the following definition of partnerships: A partnership could mean that your 
organizations have signed a Memorandum of Understanding, submitted a funding proposal together, served as a 
primary or sub-contractor on a contract or grant together, shared resources or expertise, or worked together on a 
specific project.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 33
Figure 19. Average number of non-profit partners (total and by sector) reported by CSOs
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 34
Figure 20. Average number of other private partners (total and by sector) reported by CSOs
The team combines the count of non-profit and other private partners named by each CSO with 
two similar measures asking CSOs to report the total number of non-profit and other private 
partners. The former measure forces respondents to name and classify each partner, reducing 
the chances of misreporting. However, these measures limit the number of potential partners to 
15. The second set of measures has no upper limit and therefore may capture larger increases 
for CSOs with more than 15 partners.
Figure 21 plots the distribution of an average z-score across these four measures of the number 
of partners for each CSO. There is a small decrease in the average number of partners for the 
control group and a slight increase for the treatment group. However, these changes between 
baseline and endline are very small in magnitude at 0.01 to 0.04 standard deviations. Figure 21 
plots coefficients estimating whether the treatment group experienced a significantly larger 
increase in partnerships between baseline and endline for the combined index in Figure 22 and 
for individual index variables for non-profit and other private partnerships, respectively. There is 
no evidence of a significant impact of the treatment on the increase in partnerships. Although the 
coefficient on the combined index is positive, it is very small in magnitude at around 0.02 to 0.1 
standard deviations.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 35
Figure 21. Distribution of combined index measuring the total number of non-profit and other private partners reported by 
CSOs
Figure 22. Plot coefficients estimating the impact of the treatment on partnership outcomes—points indicate coefficient point
estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 36
PRIMARY OUTCOME 2.2: SOCIAL MEDIA INTERACTIONS INDEX
The team’s second primary outcome under CSO networks is the change in the volume of social 
media followers and interactions. ResiliencyCambodia included a designated social media 
training designed to increase CSOs’ social media presence. The team hypothesizes that the 
number of followers and interactions will increase more for the treatment group than for 
the control group. The team combines the following variables into a single index variable to use 
as an outcome variable in Equation 1:
● Total Facebook page followers.
● Total Facebook page likes.
● Total Facebook post likes per month.
● Total Facebook post shares per month.
● Total Facebook post comments per month.
Figure 23 plots the average number of social media interactions per month across the five 
measures. Figure 24 plots the averaged z-scores index combining these five social media 
interactions. Figure 25 plots coefficients estimating whether the treatment caused a larger 
increase in social media interactions for CSOs in the treatment group.4 There is no evidence of a 
positive impact of the treatment on social media interactions.5 There is some evidence for a 
differential increase in the average number of monthly post likes received by treatment CSOs. 
However, this effect is very small, indicating a less than 0.5 increase in likes per month.
Figure 23. Average number of social media interactions per month
4 The results presented here exclude four control CSOs with extremely large changes in outcome values between 
midline and endline. However, results are similar when these four outlier observations are included. 5 Figure 24 plots results from a model comparing midline and endline values only. This approach maximizes the sample 
size and statistical power because baseline data was unavailable for several CSOs. However, the team re-estimates 
these models controlling for both midline and baseline values and controlling for unit-specific trends between baseline 
and endline. Results are unaffected.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 37
Figure 24. Averaged z-scores index combining average social media interactions per month
Figure 25. Plot coefficients estimating the impact of the treatment on social media outcomes—points indicate coefficient point
estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 38
SECONDARY OUTCOME 2.3: SELF-REPORTED NETWORK STRENGTH INDEX
The team’s first secondary outcome combines four components from a question that asks 
respondents to identify internal challenges that inhibit the ability of the organization to achieve its 
goals or fulfill its Mission. Specifically, the team includes internal challenges that indicate weak 
connections or engagement with the public or other CSOs. The team hypothesizes that the 
number of challenges inhibiting the CSO will decrease more for the treatment group than 
for the control group. The team combines the following variables into a single count variable to 
use as an outcome variable in Equation 1. 
Internal challenges:
● Weak leverage of networks.
● Not sufficiently connected to constituents/beneficiaries.
● Inability to communicate effectively with constituents.
● Inability to communicate effectively with other NGOs.
Figure 26 plots the share of CSOs reporting that network-related internal challenges interfere with 
their ability to fulfill their Mission. Figure 26 plots an averaged z-score index combining network￾related internal challenges. Higher values indicate more reporting of network-related challenges. 
Figure 27 plots coefficients estimating whether the treatment caused a larger decrease in 
network-related internal challenges for CSOs in the treatment group. There is no evidence that 
the treatment decreased reporting of network-related internal challenges. In fact, the treatment 
group reports a greater increase in levels of network-related challenges, although this difference 
is substantively small (between 0.15 and 0.2 standard deviations) and not statistically significant. 
Figure 26. Share of CSOs reporting that network-related internal challenges interfere with their ability to fulfill their Mission—
higher values indicate worse outcomes (i.e., insufficient beneficiary connections)
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 39
Figure 27. Average z-score index combining network-related internal challenges—higher values indicate more reporting of 
network-related challenges
Figure 28. Plot coefficients estimating the impact of the treatment on self-reported network strength outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals
SECONDARY OUTCOME 2.4: PARTNERSHIP-SEEKING BEHAVIOR INDEX
The team’s fourth secondary outcome is the number of new partnerships being pursued by CSOs 
in the treatment and control groups. Although they were not included in the baseline, the following 
questions will be included in the endline CSO survey and will be added as indicators to the PAP. 
The team hypothesizes that the partnership-seeking efforts of CSOs will be higher for the 
treatment group than for the control group. The team combines the following variables into a 
single index variable to use as an outcome variable in Equation 1:
● The number of new partners listed as seeking awards or donations within the last 12 months.
● The number of new partners listed as seeking awards or donations within the next 12 months.
● The number of new donors listed as seeking awards or donations within the last 12 months.
● The number of new donors listed as seeking awards or donations within the next 12 months.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 40
Figure 29 plots the z-scores for the two components of the Partnership-Seeking Behavior Index 
and the Averaged Z-Score Index. Figure 30 plots coefficients estimating the impact of the 
treatment on levels of partnership-seeking behavior for the treatment group. There is no evidence 
of higher levels of partnership-seeking behavior in the treatment group. In fact, there is a 0.2 
standard deviation decrease in the index measuring the number of new partners, corresponding 
to a decrease of about 0.5 new partners listed as seeking an award or donation.
Figure 29. The z-scores for partnership-seeking behavior component variables and the Averaged Z-Score Index
Figure 30. Plot coefficients estimating the impact of the treatment on partnership-seeking behavior outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals
SECONDARY OUTCOME 2.5: NETWORK DENSITY AMONG TREATMENT AND CONTROL 
SAMPLE
The team’s fifth secondary measure is the change in network density among CSOs in the treatment and 
control groups. A tie, or connection, occurs when one CSO in the sample reports that they have 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 41
partnered with another CSO in the sample in the last year.6 Network density is the number of ties that 
are present in a network relative to the total number of possible ties in that network. If every node is 
directly connected to every other node, then the network density is 100 percent. The team 
hypothesizes that the network density will increase more for the treatment group than for 
the control group. 
To test this hypothesis, the team calculates the density of direct connections between CSOs 
assigned to the treatment and CSOs assigned to the control at both baseline and endline. If the 
treatment is effective at encouraging CSOs in the treatment group to form partnerships with one 
another through opportunities to network, the density of ties between treatment CSOs should
increase more for that of control CSOs. There is little evidence that the treatment group became 
more connected between 2020 and 2022 relative to the control group. In fact, both groups became 
less dense between baseline and endline. Although the treatment group experienced a slightly 
smaller decrease (57 percent) in density than the control group (59 percent), this difference is 
extremely small. This sample includes the 39 control CSOs that completed the survey and the 47 
treatment CSOs that did so. Both the treatment and the control group contained 13 CSOs 
connected to another CSO at baseline and only 10 CSOs connected to another at endline. 
Figure 31. Full network of direct connections between treatment and control CSOs in the sample at baseline and endline—
node colors report the treatment status, an arrow pointing from Node A to Node B indicates that Node A listed Node B as a 
partner, bidirectional arrows indicate that both Node A and Node B reported each other as a partner, unconnected nodes did 
not report another CSO in the sample as a partner and were not reported as a partner by any other CSO.
6 In many instances, CSO A will report a connection to CSO B, but CSO B will not report a connection with CSO A. In this case, there 
are two possible ties, but only one actual tie recorded.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 42
7.0 FINDINGS—FINANCIAL RESILIENCY OF TREATED CSOS
Figure 32. Summary of results for Outcome Family 3 
As with for-profit organizations, attracting revenue is at the core of organizational survival for 
CSOs. The ResiliencyCambodia intervention includes organizational planning and revenue 
generation training to improve financial resiliency. Specifically, the team tests the impact of 
ResiliencyCambodia on two outcomes that shape financial resiliency: their ability to increase 
revenue and the diversification of that revenue. Because strengthening revenue generation, 
diversification, and financial health was the objective of different intervention activities, the team 
tests the impact of ResiliencyCambodia on these outcomes as separate hypotheses. The team 
also specifies four secondary outcomes that are alternative ways of measuring financial 
diversification. 
Overall, there is little evidence that the treatment increased the financial resiliency of CSOs. 
Looking at both objective measures as well as self-reported measures of future planned behavior, 
there are no significant differences between the treatment and control groups. There is some 
evidence for a small increase in the total value of revenue from donations and earned income and 
in the total value of revenue from local sources. However, these results are imprecisely estimated 
due to the small sample size and are only apparent when removing outlier observations from the 
sample. Notably, there is no evidence for increases in the pre-registered index variables that 
serve as the main text of impact on financial resiliency.
EXPLORATORY ANALYSIS: REVENUE SOURCES AND VALUE
Before turning to the results of the pre-registered hypotheses, the team looks at whether NGOs 
in the treatment group saw greater increases in either the number of sources from which they 
received revenue or in the amount of revenue they received. The team presents both results after 
removing large outliers from the sample. For outcomes measuring the number of revenue 
sources, the team removes three control CSOs and one treatment CSO with extremely large 
changes in the number of sources of revenue. For outcomes measuring the value of revenue, the 
team removes one treatment CSO with extremely large changes in the value of awards, 
donations, and income between baseline and endline.
Figures 32 and 34 compare the number of revenue sources and the total value of revenue for 
each CSO between the baseline and endline periods. The red line indicates zero changes 
between baseline and endline. The black text indicates the change in the mean value for each 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 43
treatment group between the baseline and endline periods. Figures 33 and 35 report coefficients 
measuring the impact of the intervention on the outcome. 
There is no evidence of an increase in the number of sources of revenue. Across the measures, 
the treatment group gained an average of between 0.2 and 0.3 new sources of total revenue, 
awards, and donations (less than 0.05 standard deviations). Treatment CSOs experienced a 
similarly sized decrease in sources of earned income. 
Similarly, there is little evidence for an increase in the value of total revenue or awards for 
treatment CSOs. Although the average increase in total revenues is around $25,000, this is less 
than 0.1 standard deviations and less than 6 percent of the average total revenue for CSOs in the 
sample.
There is some evidence for an increase in the value of revenue received from donations and from 
earned income. Specifically, there is an average increase of about $27,000 in donations for the 
treatment group, which indicates an almost 0.3 standard deviation increase. This estimate falls 
slightly short of statistical significance. For earned income, there is an average increase for the 
treatment group of about $13,000, which indicates a more than 0.3 standard deviation increase. 
However, this result is very imprecisely estimated and does not approach significant at 
conventional levels. 
Importantly, both results are substantially smaller when including outlier observations from the 
sample. Although these findings should only be taken as suggestive, they provide some evidence 
that the treatment may have caused a small increase in donations and earned income for 
treatment CSOs.
Figure 33. Comparison of the number of revenue sources for each CSO between the baseline and endline periods
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 44
Figure 34. Plot coefficients estimating the impact of the treatment on the number of sources of revenue outcomes—points 
indicate coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent
confidence intervals
Figure 35. Comparison of the total value of revenue for each CSO between the baseline and endline periods 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 45
Figure 36. Plot coefficients estimating the impact of the treatment on the value of revenue outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals
PRIMARY OUTCOME 3.1: REVENUE GENERATION INDEX
ResiliencyCambodia provides individualized coaching to help CSOs identify new funding 
opportunities and skills training to help them capitalize on those opportunities. The effect of 
organizational planning and skills training may require more than one year to impact the 
composition of CSO budgets. In this case, relying exclusively on endline budget data from 2021 
may fail to capture the true impact of ResiliencyCambodia. 
To capture changes in CSO behavior that may be a leading indicator of changes in budget 
composition, the team asks respondents to identify the number and value of specific new awards 
that they have applied to over the past 12 months, that they received over the past 12 months, 
and that their organization plans to apply for over the following 12 months. Similarly, the team
asks CSOs to list the number of new donors they have sought or received awards or donations 
from or plan to seek awards or donations from and whether CSOs have started using earned 
income strategies over the past 12 months or whether they have plans to start over the next 12 
months. By requiring respondents to list specific awards, donors, or earned income strategies, 
the team hopes to guard against overreporting.
Although they were not included in the baseline, the following questions were included in the 
endline budget survey. The team hypothesizes that the revenue generation efforts of CSOs 
will be higher for the treatment group than for the control group. The team combines the 
following variables into a single index variable to use as an outcome variable in Equation 1 
(substituting controls for baseline levels of revenue diversification and financial health for the 
nonexistent baseline values).
Award-seeking (endline only):
● The number of new awards listed as applied for in the last 12 months.
● The number of new awards listed as received in the last 12 months.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 46
● The value of new awards received in the last 12 months.
● The number of new awards listed to be applied for in the next 12 months.
● The number of new donors listed as having sought awards or donations in the last 12 months.
● The number of new donors listed as having received awards or donations in the last 12 months.
● The number of new donors listed as seeking awards or donations in the next 12 months.
Earned income-seeking behavior (endline only):
● Began charging fees for any services rendered to CSOs in the past 12 months.
● Plans to begin charging fees for any services rendered to CSOs in the next 12 months.
● Began charging fees for any services rendered to the government in the past 12 months.
● Plans to begin charging fees for any services rendered to the government in the next 12 months.
● Started a new social enterprise in the past 12 months.
● Plans to start a new social enterprise in the next 12 months.
● Started charging membership fees in the past 12 months.
● Plans to start charging membership fees in the next 12 months.
Figure 36 plots values for index variables created from the award-seeking components only, the 
earned income-seeking behavior components only, and an index created by combining all of 
these variables. Among treatment CSOs, there is a small decrease in reported award-seeking 
behavior and a larger increase in income-seeking behavior. Figure 37 reports coefficients 
estimating the impact of the treatment on changes in revenue-seeking behavior among treatment 
CSOs between baseline and endline. There is no evidence of a significant change in revenue￾seeking behavior. There is a slight increase in income-seeking behavior, although this increase 
is very small at less than 0.1 standard deviations and not statistically significant.
Figure 37. Distribution of averaged z-score index variables for an index combining award-seeking components, an index 
combining income-seeking components, and an index combining all of the revenue-seeking components
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 47
Figure 38. Plot coefficients estimating the impact of the treatment on revenue generation outcomes—points indicate coefficient 
point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals. 
PRIMARY OUTCOME 3.2: REVENUE DIVERSIFICATION INDEX: HERFINDAHL–HIRSCHMAN 
INDEX
The literature on non-profit organizations has found links between financial diversification and a 
variety of positive outcomes, including overall financial health (Hung and Hager, 2019). In addition 
to focusing on new funding opportunities, ResiliencyCambodia provides individualized coaching 
to help CSOs identify new revenue sources and skills training to help them tap into those sources. 
These activities are designed to promote financial diversification that can help CSOs become 
more embedded in their communities and more resilient to external financial shocks. Because 
measuring financial diversification requires comprehensive budget data, it is not possible to 
measure diversification prospectively. Instead, the team uses budget data from the most recently 
completed budget cycle. 
Following standard practice in the portfolio theory literature, the team measures revenue 
diversification at baseline and endline using a Herfindahl–Hirschman Index (HHI). Specifically, 
the team measures the concentration of both the number and value of revenues across revenue 
sources (sources correspond to each potential granting organization, donor, or income source). 
HHI scores are measured on a scale of 0–10,000, with 10,000 indicating complete concentration. 
In this application, complete concentration would mean that all of a CSOs’ revenues come from 
a single funding source (for example, one specific grant-making organization or donor). The team 
hypothesizes that the financial diversification of CSOs will increase more for the treatment 
group than for the control group. The team combines the following variables into a single HHI 
to use as an outcome variable in Equation 1:
● Grants and awards from:
− USAID. 
− Other foreign government aid agencies or multilateral organizations. 
− International NGOs and foundations. 
− Cambodian NGOs and foundations. 
− RGC. 
● Donations from:
− Foreign businesses.
− Cambodian businesses.
− Foreign individuals.
− Cambodian individuals.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 48
● Earned income from:
− Membership fees and dues.
− Fees paid by recipients of services rendered by the organization.
− The sale of handicrafts.
− The sale of food services.
− The sale of publications or print materials.
− Performance arts and cultural programs.
− Income from services rendered to the government.
− Income from services rendered to another NGO/community-based organization.
− Income from fundraisers or other special events.
− Property income/endowment income.
Figure 38 compares the HHI score for both the value of revenue from these sources as well as 
the HHI of the distribution of the number of sources across these categories. This approach uses 
the sum of revenue in each of the categories listed above, often combining multiple line items 
from each category and sub-category. 
To complement this analysis, the team also looks at HHI across every line item of revenue. In this 
application, complete concentration would mean that all revenues came from a single line item 
(for instance, one specific grant or donation). Figure 39 compares the concentration of revenue 
across all revenue line items for each CSO between the baseline and endline periods. Figure 40 
plots the coefficients estimating the impact of the treatment on changes in these HHI scores 
between baseline and endline. 
There is evidence of modest decreases in revenue concentration (increases in diversification) 
among the treatment group. However, these changes are substantively small and not statistically 
significant at conventional levels. The largest coefficient, capturing the impact of the intervention 
on decreases in the concentration of revenue sources, estimates a decrease in concentration of 
about 0.2 standard deviations.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 49
Figure 39. Comparison of the concentration of the number of revenue sources across categories and the value of revenue 
across categories for each CSO between the baseline and endline periods—the red line indicates zero changes between 
baseline and endline, the black text indicates the change in the mean value for each treatment group between the baseline 
and endline periods
Figure 40. Comparison of the concentration of revenue across all revenue line items for each CSO between the baseline and 
endline periods—the red line indicates zero changes between baseline and endline, the black text indicates the change in the 
mean value for each treatment group between the baseline and endline periods
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 50
Figure 41. Plot coefficients estimating the impact of the treatment on revenue diversification outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals.
SECONDARY OUTCOME 3.3: FINANCIAL HEALTH
The team’s first secondary measure of financial resiliency combines several common measures 
of overall financial health (Hung and Hager, 2018). ResiliencyCambodia provides skills training 
to help CSOs track and understand important financial metrics. Furthermore, it is also important 
to document that increases in diversification are not the result of decreases in total revenues or 
the liquidation of assets. The team hypothesizes that the financial health of CSOs will 
increase more for the treatment group than for the control group. The team combines the 
following variables into a single index variable to be used as an outcome variable in Equation 1:
● Total investments. 
● Total revenue. 
● Operating margin (revenue/expenditures). 
Figure 41 compares the value of investments and revenue, the operating margin, and an 
averaged z-score of these values capturing the overall financial health for each CSO between the 
baseline and endline periods. The red line indicates zero changes between baseline and endline. 
The black text indicates the change in the mean value for each treatment group between the 
baseline and endline periods. Figure 42 compares the values for operating margin and its two 
component variables, the value of total expenses and total revenues. Figure 43 reports 
coefficients estimating the impact of the treatment on the changes in these outcomes for the 
control group between baseline and endline. There is no evidence for an overall increase in 
financial health or any of its component variables. Results generally suggest an increase of 
between 0.06 and 0.12 standard deviations for investments, revenue, and financial health, and a 
decrease of 0.17 standard deviations for operating margin.7
7 For expenditures, the team removed two control CSOs and for investments, the team removed two treatment CSOs 
due to extremely large changes between baseline and endline values. However, results are similar when including 
these outliers.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 51
Figure 42. Comparison of the investments, revenue, operating margin, and overall financial health for each CSO between the 
baseline and endline periods
Figure 43. Comparison of the overall operating margin and its two components (expenses and revenue) for each CSO between 
the baseline and endline periods—the red line indicates zero changes between baseline and endline, the black text indicates 
the change in the mean value for each treatment group between the baseline and endline periods
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 52
Figure 44. Plot coefficients estimating the impact of the treatment on financial health outcomes—points indicate coefficient 
point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals.
SECONDARY OUTCOME 3.4: DIVERSIFICATION AWAY FROM AID: LOCAL REVENUE INDEX
The team’s second secondary outcome under financial resiliency is the share of revenue coming 
from local sources. Research on financial diversification among non-profits has generally focused 
on NGOs operating in developed countries and emphasized funding across three aggregated 
revenue streams: donations, earned income, and investment revenue. However, aid-driven NGO 
sectors in countries like Cambodia face unique challenges including aid volatility, a lack of 
connection and accountability to (and legitimacy with) domestic constituencies, and vulnerability 
to government attacks on NGOs as “foreign agents.” For this reason, ResiliencyCambodia aims 
to help CSOs diversify their revenues away from foreign aid and pursue new, more sustainable 
revenue streams. The team hypothesizes that the share of revenue from local sources will 
increase more for the treatment group than for the control group. To measure the share of 
revenues from local sources at baseline and endline, the team sums the amount of revenue 
coming from the local sources (listed below) and divides it by the total amount of revenue received 
in the most recent fiscal year. The team uses this variable as an outcome variable in Equation 1. 
Share of revenue from:
● Earned income: 
− Services rendered to the government. 
− Services rendered to another NGO/community-based organization. 
− Fundraisers or other special events. 
− Fees paid by recipients of services rendered by the organization. 
− Sale of goods and other commercial activities. 
− Membership fees and dues. 
− Fundraisers or other events. 
− Income from renting property or equipment. 
− Other. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 53
● Donations: 
− Cambodian individuals. 
− Cambodian businesses. 
− Domestic NGOs and foundations. 
● Awards: 
− Domestic NGOs and foundations. 
− RGC. 
Figure 44 compares the total revenue from local sources and the total revenue from local sources 
as a share of total revenue for each CSO between the baseline and endline periods. The red line 
indicates zero changes between baseline and endline. The black text indicates the change in the 
mean value for each treatment group between the baseline and endline periods. Figure 45 
presents coefficients estimating the impact of the treatment on the change in these outcomes 
between baseline and endline. 
There is no evidence of a positive impact of the intervention on the share of revenue from local 
sources. There is some evidence for an increase in the total volume of local revenue for the 
treatment group. The team estimates that the average increase for the treatment group was about 
$19,000 larger than the increase for the control group, corresponding with a 0.23 standard 
deviation increase. However, this result is only apparent after removing one control CSO due to 
extremely large changes in the reported volume of donations between baseline and endline. It is 
important to note that some CSOs received small grants from the ResiliencyCambodia program 
(funding from an international source), which may artificially decrease the share of funding coming 
from local sources. However, these small grants would not artificially reduce the total local 
revenue.
Figure 45. Comparison of the total revenue from local sources and the total revenue from local sources as a share of total 
revenue for each CSO between the baseline and endline periods
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 54
Figure 46. Plot coefficients estimating the impact of the treatment on local revenue outcomes—points indicate coefficient point
estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals
SECONDARY OUTCOME 3.5: SHARE OF REVENUE FROM FOREIGN SOURCES
The team’s third secondary outcome under financial resiliency is the share of revenue coming 
from foreign sources. Foreign aid is the largest overall source of revenue for the NGO sector in 
Cambodia and for CSOs in the sample. However, a reliance on these sources may decrease local 
embeddedness and expose NGOs to high levels of financial vulnerability. The team 
hypothesizes that the share of revenue from foreign sources will decrease more for the 
treatment group than for the control group. To measure the share of revenues from foreign 
sources at baseline and endline, the team sums the amount of revenue coming from foreign 
sources (listed below) and divides this by the total amount of revenue received in the most recent 
fiscal year. The team uses this variable as an outcome variable in Equation 1. 
● Share of revenue from foreign government donors:
− USAID.
− China.
− Russia.
− Other foreign government aid agencies or multilateral organizations.
● Share of revenue from international NGOs and foundations.
● Share of revenue from foreign businesses.
● Share of revenue from foreign individuals.
Figure 46 compares the total revenue from foreign sources and the total revenue from foreign 
sources as a share of total revenue for each CSO between the baseline and endline periods. The 
red line indicates zero changes between baseline and endline. The black text indicates the 
change in the mean value for each treatment group between the baseline and endline periods. 
Figure 47 presents coefficients estimating the effect of the treatment on the change in these 
outcomes between baseline and endline for the treatment group. There is no evidence that the 
total amount of foreign revenue or the reliance on foreign revenue decreased as a result of the 
intervention.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 55
Figure 47. Comparison of the total revenue from foreign sources and the total revenue from foreign sources as a share of total 
revenue for each CSO between the baseline and endline periods
Figure 48. Plot coefficients estimating the impact of the treatment on foreign revenue outcomes—points indicate coefficient 
point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 56
8.0 FINDINGS—RESILIENCY TO CHANGING CIVIC SPACE OF 
TREATED CSOS
Figure 49. Summary of results for Outcome Family 4
Building the capacity of CSOs to navigate Cambodia’s narrowing civic space is critical to 
ResiliencyCambodia’s objectives. Because resiliency to changing civic space is a difficult 
concept to operationalize and measure, the team draws on several approaches. 
There is mixed evidence for the impact of the treatment on CSO resiliency to changing civic space. 
Counter to expectations, reporting of external challenges increased more for members of the 
treatment group. Furthermore, the effect size is relatively large at 0.3 standard deviations and 
statistically significant. These unexpected findings may suggest that the treatment increased 
CSOs’ awareness of civic space issues, leading to higher reporting of these challenges. 
Alternatively, these findings could suggest that participation in ResiliencyCambodia or the 
differential increase in time spent on advocacy attracted more government attention to and greater 
repression of treatment CSOs. 
There is also evidence for a greater increase in the share of time treatment CSOs spend on 
political advocacy compared to treatment CSOs. Increased time spent on advocacy may be 
driving increases in the external challenges CSOs face in a heavily restrictive environment like 
Cambodia. Alternatively, the intervention’s emphasis on civic space issues may have increased 
the awareness or salience of these issues for treatment CSOs, leading to higher levels of 
reporting.
PRIMARY OUTCOME 4.1: EXTERNAL CHALLENGES INDEX
The first primary outcome is drawn from a question about the characteristics of the organization’s 
external environment inhibiting its ability to achieve its goals or fulfill its Mission. Specifically, the 
team selects six external challenges that ResiliencyCambodia training is designed to help CSOs 
navigate. If ResiliencyCambodia is successful, the team expects CSOs to see fewer external 
factors as inhibiting their ability to fulfill the organization’s Mission. The team hypothesizes that 
the number of external challenges that inhibit CSOs will decrease more for the treatment 
group than for the control group. The team combines the following variables into a single count 
variable to be used as an outcome variable in Equation 1:
● Lack of cooperation within the civil society sector. 
● Harassment or direct attacks by the government on the civil society sector. 
● Lack of public trust in CSOs. 
● Restrictive or politicized legal environment. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 57
● Restrictive or politicized media environment. 
● Restrictions on the speech or activities of CSOs. 
Figure 49 presents the share of CSOs reporting that each of these external challenges interferes 
with their ability to fulfill their Mission. The share of organizations reporting being affected by these 
external challenges decreases in both the treatment and control groups but decreases more for 
controls. Figure 50 plots the values of an averaged z-score combining these components into a 
single summary variable. Figure 51 presents coefficients estimating the effect of the treatment on 
the change in the reporting of external challenges between baseline and endline. Surprisingly, 
there is strong evidence that treatment CSOs are more likely to report that these external 
challenges are a problem. Looking at the combined index, the treatment group reports 0.3 
standard deviation increase in these challenges. This result is substantively meaningful and 
statistically significant.
There are two possible interpretations for this finding. The first is that participation in the 
ResiliencyCambodia program has made organizations more aware of the challenges facing civil 
society and their own organization in particular, which makes them more likely to report 
experiencing these challenges. The second interpretation is that participation in the program has 
placed organizations more at risk for these types of challenges, particularly government 
harassment and a lack of civil society cooperation. Though it seems unlikely that the 3 percent 
increase in time spent on advocacy reported under Outcome Family 1 would be enough to make 
an organization a target, USAID should consider the ethics involved in working closely with CSOs 
in restrictive environments like Cambodia. 
Figure 50. Share of CSOs reporting that each of these external challenges interferes with their ability to fulfill their Mission
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 58
Figure 51. Values of an averaged z-score combining these components into a single summary variable
Figure 52. Plot coefficients estimating the impact of the treatment on external challenges outcomes—points indicate coefficient 
point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence intervals.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 59
PRIMARY OUTCOME 4.2: CSO NETWORK DIVERSIFICATION: HHI
The second primary outcome measure is the change in the diversification of partnerships by 
sector reported between January and December 2019 (12-month pre-treatment period before the 
call for applications) and between April 2021 and March 2022 (12-month post-treatment period 
before endline). Networks facilitate the flow of information and resources across CSOs. Forging 
connections across sectors ensures that CSOs have access to a greater diversity of information 
and resources, making cross-sector connections especially important for resiliency in changing 
spaces. The team hypothesizes that the diversification of partners will increase more for 
the treatment group than for the control group. The team measures partnership diversification 
at baseline and endline using an HHI on three aggregated revenue streams (Hung and Hager, 
2018). Specifically, the team calculates the share of partners focusing on each of the following 
activities. The team combines the following variables into a single HHI to use as an outcome 
variable in Equation 1. 
● Share of partners focusing on:
− Democracy and governance.
− Environment.
− Food security.
− Public health.
− Education.
− Economic development.
− Other.
Figure 52 compares the concentration of partners across sectors for each CSO between the 
baseline and endline periods. The red line indicates zero changes between baseline and endline. 
The black text indicates the change in the mean value for each treatment group between the 
baseline and endline periods. Figure 53 plots coefficients estimating the effect of the treatment 
on the concentration of partners by sector. There is no evidence that the concentration of partners 
by sector decreased in the treatment group.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 60
Figure 53. Comparison of the concentration of partners across sectors for each CSO between the baseline and endline periods. 
Figure 54. Plot coefficients estimating the impact of the treatment on network diversification outcomes—points indicate 
coefficient point estimates, thick lines indicate 95 percent confidence intervals, and thin lines indicate 90 percent confidence 
intervals.
PRIMARY OUTCOME 4.3: SHARE OF TIME ENGAGING IN ADVOCACY
The first secondary outcome measure is drawn from a question about the share of staff time spent 
on a series of activities during a typical month. Specifically, the team selects the amount of time 
CSOs spent on advocacy. Although relatively few CSOs in the sample focus primarily on 
advocacy, the majority report spending at least some of their time on advocacy work. Because 
interviews suggest that many CSOs see advocacy as important for their Mission but refrain from 
advocacy to avoid inviting government scrutiny, there is an increased willingness to engage in 
advocacy as an important indicator of resiliency. The team hypothesizes that the share of time 
spent on advocacy activities will increase more for the treatment group than for the control 
group. Although the number of CSOs in the sample that focus primarily on advocacy is relatively 
low, 80 percent of CSOs report devoting at least some time to advocacy (the median CSO reports 
spending 5 percent of their time on advocacy). If CSOs feel more resilient to changing civic space, 
the team expects that they will spend more time advocating for the communities that they serve 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 61
than for the causes that they care about. The team uses this variable as an outcome variable in 
Equation 1:
● Share of staff and management time spent on advocacy or raising awareness in a typical month.
Figure 54 plots the distribution of CSOs’ reports of how much time they spent on political 
advocacy. Figure 55 reports coefficients estimating the impact of the treatment on changes in the 
share of time spent on political advocacy by treatment organizations. There is evidence for a 
differential increase in time spent on advocacy for the treatment group. Specifically, the treatment 
CSOs increased their time dedicated to advocacy by 2-2 percent. This magnitude is relatively 
small but approaches statistical significance at the 0.1 level. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 62
Figure 55. Distribution of time spent on political advocacy
Figure 56. Coefficients estimating the impact of the treatment on changes in the share of time spent on political advocacy by 
treatment organizations
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 63
PRIMARY OUTCOME 4.4: SHARE OF TIME ENGAGING IN ADVOCACY
The second secondary outcome measures the extent to which intervention by local governments 
impacts the operation decision of CSOs. In Cambodia, local governments are often charged with 
enforcing government regulations of the non-profit sector. These local governments often have 
different approaches to handling NGOs, with some facilitating the work of NGOs and others 
obstructing it. Evidence from the baseline analysis suggests that CSOs in Cambodia actively 
avoid working in localities with more repressive local governments. 
If ResiliencyCambodia makes CSOs more resilient to closing civic space or more confident in 
their ability to navigate government scrutiny, the team expects that treatment CSOs should be 
less likely to avoid working in communities with repressive local governments. This conjoint 
survey experiment is part of a broader study involving surveys of NGOs in Uganda and Serbia. 
The details of the experiment and analysis are pre-registered separately (EGAP Registration ID: 
20220421AA). For this analysis, the team looks at heterogeneous effects for treatment CSOs and 
control CSOs using the procedures described in the PAP. Although a detailed discussion of the 
results is beyond the purview of this report, the team summarizes the main findings and their 
bearing on the other results under the outcome family.
In this conjoint experiment, the team varies the attributes of two communities and ask respondents 
about which community their NGO would be more likely to select for activities. Specifically, the 
team varies the community’s level of need, its remoteness, the level of operational intervention 
by local government, the level of rhetorical intervention by the local government, the circulation of 
anti-CSO narratives on social media, the local government (LG) benefits to CSOs that allow 
government influence of their activities, and the amount of funds available for projects being 
implemented in the community. 
After selecting which community their CSO would prefer to work in based on the joint, the team 
also asked respondents to choose the community in which CSOs would prefer to engage in 
various project activities, including involving members of the public in the design or 
implementation of project activities, organizing public action, partnering with other NGOs, 
partnering with local community-based organizations, partnering with religious or traditional 
leaders, or partnering with local government officials. Each respondent completes seven of these 
community and activity choice tasks.
Overall, there are no significant differences between the treatment and control groups. However, 
this is likely due to the very small sample size and limited statistical power. Consistent with the 
other findings under this outcome family, and contrary to the team’s expectations, there is
suggestive evidence that treatment CSOs were more discouraged by anti-CSO rhetoric by LG 
and on social media. Specifically, treatment CSOs were even less likely to report a preference to 
work in communities or pursue partnerships with other CSOs or community-based organizations
in communities where anti-CSO rhetoric was present. This contributes additional evidence that 
the treatment caused CSOs to be more sensitive to the challenges of closing civic spaces. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 64
Figure 57. Coefficient estimates for the effect of community, LG, and grant characteristics on CSOs’ preference for working in 
a given community—panels present results for control, treatment, and the difference between them (third panel), points to 
the left of the grey line indicate a negative causal effect of the attribute on grant selection relative to the baseline category (on 
average).
9.0 RECOMMENDATIONS AND LESSONS LEARNED
There is no evidence that ResiliencyCambodia caused meaningful improvements in CSO 
capacity, networks, finances, or resiliency to closing civic space. The team’s analyses estimate 
extremely small differences in how outcomes for treatment and control CSOs changed between 
baseline and endline, suggesting that these disappointing results are not attributable to small 
sample sizes or attrition from the evaluation sample. Furthermore, the prevalence of these null 
results across both objective (data collected directly from financial records or Facebook pages) 
and self-reported measures (data collected from survey questions asking about behavior or 
perceptions) of key outcomes strengthens the team’s conclusion that the program largely failed 
to achieve its main objectives. 
For the small number of outcomes where there are meaningful improvements that are larger for 
the treatment group, these increases are limited to individual components of the broader index 
variables that the team pre-registered as the primary measures of impact (see the PAP referenced 
in Appendix C). The one exception is an increase in the share of time CSOs dedicate to political 
advocacy. However, this 2–3 percent increase is relatively small when compared to the size and 
length of the intervention. Furthermore, this outcome is a self-reported rather than objective 
measure of behavior.
It is important to remember that the ResiliencyCambodia program was designed in part as a tool 
to test a refocused, less intensive, and scaled-up model of the traditional R+ program, in hopes 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 65
of developing a more far-reaching version of the program. The null results of this model offer 
important lessons for what components of a civil society capacity-building program are essential 
for success.
It is also important to note that the period during which ResiliencyCambodia occurred was one 
of unprecedented difficulties facing the civil society sector. COVID-19 posed unique challenges 
for CSOs, including being ordered to close temporarily, being unable to implement or deliver 
programs, having grant funding delayed or canceled, and adapting to remote work. Data 
presented in Section 7 show that both the treatment and control groups experienced large 
decreases in revenues during this period across almost all funding sources. This retrenchment 
likely made it a particularly difficult time for CSOs to pursue new revenue sources. Similarly, much 
of the programming that was designed to take place in person had to be moved online, which 
potentially reduced engagement and limited any opportunities for networking. Although these 
challenges did not limit the team’s ability to evaluate the impact of this implementation of 
ResiliencyCambodia, they do limit the team’s ability to make inferences about the expected 
effectiveness of similar programming implemented under normal conditions. 
RECOMMENDATIONS 
Drawing on the IE data, the team poses the following recommendations for future USAID 
activities, particularly those aimed at improving local organizational capacity. 
RECOMMENDATION 1: INVEST IN PROGRAMS TO HELP ORGANIZATIONS ACROSS ALL SECTORS 
COMBAT CLOSING CIVIC SPACE 
In Cambodia and countries across the world, local and international CSOs play a critical role as 
both service providers and sources of political accountability. Across every technical area of its 
programming, USAID relies heavily on these organizations to achieve its development objectives. 
However, efforts to constrain the activities of CSOs have increased dramatically over the last 15 
years (Young and Echague, 2017), and shrinking civic space threatens the ability of civil society 
to do this important work. 
In Cambodia, recently published work using baseline data from this evaluation demonstrates that 
government repression targets CSOs across both the advocacy and service delivery sectors, 
reduces their fiscal viability, and limits their willingness and ability to implement certain activities 
(Springman et al., 2022). The results from this IE reinforce this point. Specifically, the diverse 
sample of CSOs participating in ResiliencyCambodia reported high levels of external challenges 
related to closing civic space but were also willing and able to increase the share of time spent
on political advocacy.
USAID’s efforts to conduct civic space programming in technical areas beyond DRG are rare 
(Wibbels et al., 2022). The team believes that the LO-MTSR activity represented an important 
effort to design programming that directs civic space programming across technical areas to 
bolster organizations working not just on DRG issues like political advocacy on human rights but 
also, ostensibly, on apolitical sectors like health, education, and agriculture. The team 
recommends that USAID/Cambodia continue their support for such work and encourage other 
Missions to build on these efforts.
RECOMMENDATION 2: IES ARE A SMART INVESTMENT TO ENSURE TAXPAYER DOLLARS MAXIMIZE 
PROGRAM IMPACTS
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 66
The LO-MTSR IE of the ResiliencyCambodia program fulfilled the vision of the SBAR pilot by 
successfully applying cutting-edge scientific research methods to an applied research question. 
Furthermore, this research provided a first attempt for USAID at conducting an IE on CSO 
programming, contributing the first systematic evidence on the impact of CSO capacity-building 
and providing a model for future evaluations in this sector. 
Although disappointing, null results can be helpful for providing concrete feedback on program 
successes and failures, potential unexpected consequences, and the need to adapt or redesign 
activities. Without additional support for an RCT, the evaluation of ResiliencyCambodia may 
have relied on feedback from a select group of participants that reported satisfaction with the 
program and recommended that funding for ResiliencyCambodia be scaled up in Cambodia or 
expanded to other countries. Alternatively, a less sophisticated evaluation may have relied on 
pre-/post- pmeasures of changes in the treatment group showing that revenues decrease 
dramatically over the treatment period; this would have led to an erroneous conclusion that 
ResiliencyCambodia had caused great harm to participating organizations.8 Instead, the IE 
successfully identified the program’s failure to benefit treatment CSOs, allowing for informed 
decisions about future allocations of funding and revisions to the ResiliencyCambodia model. 
Ultimately, IE results can help USAID create more effective development solutions that best utilize 
US taxpayer dollars.
The data collected in this IE will provide many opportunities for additional learning about civil 
society in Cambodia and has already resulted in one peer-reviewed publication in the September 
2022 volume of International Studies Quarterly. The research team is also exploring future 
publications and conference presentations to widely share the results with the academic and 
practitioner communities, as well as opportunities to present findings to USAID’s New Partnership 
Initiative and the USAID/DRG/Bureau for Development, Democracy, and Innovation’s civil society 
team. More broadly, the findings from this IE will help USAID/Cambodia and USAID to adapt 
existing programming, design new programs, and understand the development challenges facing 
civil society in Cambodia.
RECOMMENDATION 3: BUILD THE EVIDENCE BASE AROUND EACH COMPONENT OF THE PROGRAM 
THEORY OF CHANGE
The original theory that motivated this project states the expectation that increases in 
organizational capacity, financial diversification, and networks would enhance their resilience to 
closing civic space through. However, the team finds that the intervention failed to yield 
improvements in these three preliminary outcomes through which gains in resiliency were 
expected to develop. This suggests that securing improvements in these preliminary outcomes is 
more difficult than USAID/Cambodia anticipated. The team recommends that before developing 
more complex theories about how capacity, finances, and networks may be related to resilience 
to closing civic space, future research should invest more heavily in developing and testing more 
simple theories about how to help CSOs improve each of these important preliminary outcomes.
In other words, each of these outcomes requires its own researched theory of change to show 
that these outcomes are backed by evidence to promote organizational capacity. The team 
recommends that before designing any future capacity-building program with a similar theory of 
change, USAID should conduct a review of the evidence to support this theory of change and 
8 As is clear from the plots of the raw data, the financial situation for most CSOs in the sample got worse between 
baseline and endline. In this situation, smaller declines among the treatment group relative to control would indicate a 
successful intervention.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 67
make adjustments as needed. If there is no evidence to support one or more of these outcomes 
as a pathway to change, the theory of change should be re-examined. 
If program designers at the Mission lack the time or capacity to conduct a review of the evidence, 
they should utilize resources at USAID, such as the Evidence and Learning Team at 
USAID/DDI/DRG in Washington. For example, through the Evidence and Learning Team at the 
DRG Center, at a cost of approximately $20,000, a Mission can commission an evidence review 
conducted by leading academics to review experimental and quasi-experimental literature and 
compile recommendations on what types of programming can be most successful in a particular 
context, often within a four-month period.9 
LESSONS LEARNED 
Going beyond the IE data, the team drew on discussions with organizations throughout the 
intervention, initial findings from a qualitative assessment of similar R+ programming, and 
numerous pause-and-reflect sessions to provide the following programmatic lessons learned for 
future USAID activities in hopes of improving the capacity of local organizations. 
LESSON 1: FLEXIBLE, NEEDS ASSESSMENT-DRIVEN AWARDS ARE BEST FOR CUSTOMIZED CAPACITY￾BUILDING 
The initial problem statement in the SBAR solicitation identified a lack of organizational and 
technical capacity among organizations to effectively manage funding and develop strategies to 
become self-reliant. The solicitation identified financial diversification, particularly a reduction of 
funding from USAID and from malign actors, as the primary program objectives, with an increase 
in network connections as a secondary objective. These contract objectives mandated a top-down 
approach to determining what factors CSOs would aim to strengthen through the 
ResiliencyCambodia program. To be compliant with the contract objectives and monitoring, 
evaluation, and learning targets, organizations developed Resiliency Roadmaps with three pre￾defined priorities, with a fourth priority that could be customized for each organization, and a 
prescriptive number of paths for fulfilling each priority. However, feedback from organizations 
during the Peer Learning Event suggests that financial diversification, particularly shifting away 
from traditional donors and toward social enterprise and donations, was not the top priority for 
organizations.
The traditional R+ model, on the other hand, relies on a bottom-up approach where organizations 
can create a fully customized roadmap based on the seven different resiliency areas. This level 
of flexibility ensures the Resiliency Roadmaps address the top priorities of the organization as 
defined by the organization, which makes it more valuable both for increasing organizational 
capacity and for promoting buy-in from organizations. A more flexible model would also have 
allowed the ResiliencyCambodia program to pivot when USAID moved away from the Journey 
to Self-Reliance toward localization, by supporting organizations to become stronger potential 
partners for USAID/Cambodia and large implementing partners. Instead, ResiliencyCambodia
had to continue its mandate to move organizations away from USAID funding, which was likely 
not in the best interests of the organizations or USAID/Cambodia.
LESSON 2: COACHING IS A VALUABLE TOOL FOR BUILDING CSO CAPACITY, BUT INTENSITY MATTERS
9 See an example evidence review on women’s political participation conducted for USAID Kosovo on the Development 
Experience Clearinghouse: https://pdf.usaid.gov/pdf_docs/PA00ZF2R.pdf
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 68
In interviews with CSOs, the support and mentoring of the Resiliency Coaches were consistently 
praised as the most valuable part of the program. Coaching provided organizations with 
personalized support on the topics that mattered most to them and provided an opportunity to 
apply the knowledge passed on through trainings and toolkits. However, qualitative evidence from 
the traditional R+ model suggests that the intensity of the coaching that is needed to have a 
significant impact is higher than in the ResiliencyCambodia program design.
The traditional R+ model relies on intensive coaching by senior civil society executives who have 
frequent training and support from the PartnersGlobal staff. In the traditional R+ model, a pair of 
coaches support three organizations over 18 months for a total of 80 days of coaching support 
per organization. In the ResiliencyCambodia model, this support was reduced to 12 days of 
support per organization in the first year, and just four days of support in the second year, with 
each coach supporting ten CSOs.10 The Illuminating New Solutions and Programmatic 
Innovations for Resilient Spaces coaches also received more in-person training and virtual 
support from the PartnersGlobal team than the ResiliencyCambodia coaches. Coaches from the 
traditional R+ model were also recruited from executive directors in leading CSOs, as opposed to 
the mid-level civil society experts recruited for ResiliencyCambodia. Future USAID activities that 
involve coaches should plan for more hours of coach support per organization, fewer 
organizations per coach, and more training and support from the implementing partner.
LESSON 3: TRADITIONAL TRAININGS SHOULD BE COUPLED WITH SPACES FOR PEER LEARNING AND 
SHARING
According to discussions with CSOs, traditional trainings and toolkits were less valuable to 
organizations than the coaching and mentoring. Some organizations shared that the trainings 
were too high-level without appropriate opportunities for building concrete skills. Others found the 
content helpful but would have preferred in-person training to virtual. Offering virtual trainings was 
an essential way for the ResiliencyCambodia program to provide support to CSOs during the 
COVID-19 lockdowns and restrictions on international travel. However, the team recommends 
future USAID activities utilize in-person trainings as much as possible, even though virtual 
trainings are easier to scale.
Whether in person or virtual, future USAID capacity-building activities should learn from the 
success of the traditional R+ approach and build in peer learning spaces for participating 
organizations to informally come together to discuss both the training content and other 
challenges or issues they are currently facing at their organization. This approach was used at 
the end of the ResiliencyCambodia activity with the Social Lab component, and initial feedback 
from organizations has been positive.
10 This slimmed-down approach was proposed intentionally in order to test a more cost-effective and scalable model of 
the R+ model. The full approach would have been too resource-intensive to implement under the SBAR pilot. 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 69
APPENDIX A. DESIGN REPORT 
The LO-MTSR ResiliencyCambodia IE Design Report can be found at the following URL: 
https://pdf.usaid.gov/pdf_docs/PA00WCZ5.pdf
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 70
APPENDIX B. BASELINE REPORT 
The LO-MTSR ResiliencyCambodia Baseline Report can be found at the following URL: 
https://pdf.usaid.gov/pdf_docs/PA00X47C.pdf
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 71
APPENDIX C. PRE-ANALYSIS PLAN 
The LO-MTSR ResiliencyCambodia PAP can be found at the following URL:
https://osf.io/9wejn
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 72
APPENDIX D. DEPARTURES FROM PRE-ANALYSIS PLAN
This section records any aspects of the analysis that do not adhere to the team’s registered PAP. 
Changes to the statistical model or the presentation of results:
● The PAP states that the team will only present results using ordinary least squares because of 
balanced attrition. Although respondents are assigned to treatment arms with the same probability 
across blocks, attrition is not perfectly balanced between the treatment and control arms, 
introducing the possibility of some bias in estimation. Furthermore, the potential for heterogeneous 
effects across treatment blocks also creates the potential for bias in the estimates. To address these 
concerns, the team also estimates all models and presents results using the IWE suggested by 
Gibbons et al. (2018).
● Due to attrition, one block was left with two treatment and zero control NGOs. In order to include 
block fixed-effects in the models, the team merged this block with its most similar block according 
to the distance score used for block assignment.
● To estimate spillover effects, the team intended to include a covariate indicating whether each CSO 
reported a partnership with a treatment CSO and use inverse probability weights to account for 
how network location impacts the probability of being connected to a treated unit (Aronow and 
Samii, 2015). However, an analysis of the baseline level of connection across assignment groups 
indicated that these connections are rare, reducing concerns about spillovers. However, the team 
will add this to the final analysis prior to submission to the Development Experience Clearinghouse.
Changes to the team’s definition of key variables:
Primary Outcome 1.3: Administrative Capacity
● In the PAP, the team registered an intent to use small sample methods to calculate each CSO’s 
conformity to the Benford distribution and use this continuous measure as an outcome variable in 
Equation 1 to test for larger decreases in levels of non-conformity in the treatment group (Wheeler, 
2015). However, these small-sample models rely on simulation and were far too computationally 
intensive to run for nearly all CSO budgets in the sample. For this reason, the team uses the KS test 
from the 2021 Benford analysis report.
Primary Outcome 3.2: Revenue Diversification Index
● The PAP registered a definition of this variable to include the three largest grants from major 
sources. Due to heterogeneity in the granularity of budget data provided by CSOs, the team does 
not observe this value for all CSOs. Instead, the team uses the aggregate value from each major 
source and then supplements this with a separate analysis taking the fractionalization of all revenue 
line items at the lowest level of granularity available for each CSO.
Primary Outcome 3.3: Financial Health
● The PAP registered the inclusion of assets and divestment in the measure of financial health. 
However, most CSOs did not report any line items in these categories. To avoid bias from 
differential detail in reporting, the team excludes these values from the index.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 73
Primary Outcome 3.4: Diversification Away from Aid: Local Revenue Index
● The PAP registered a more limited range of local revenue sources to be included in this index. 
During the final instrument design, the team added a number of additional categories based on 
feedback from enumerators and pilot studies. These additional local sources are all included in the
index variable.
Primary Outcome 3.1: Revenue Generation Index
● The PAP registered the inclusion of a question asking about the value of new awards to be applied 
for in the next 12 months. This question was not asked on the survey and is not included in the 
index.
● The PAP registered the inclusion of questions asking about both fundraisers and membership fees. 
To increase the clarity of this question, the team asked only about membership fees. 
● The PAP did not register the inclusion of questions asking about whether CSOs started charging 
fees for services rendered to other CSOs, began charging fees for services rendered to the 
government, or made specific plans to begin doing so. The team added these questions to the 
endline survey and included them in the index.
Secondary Outcome 2.4: Partnership-Seeking Behavior Index
● The PAP did not register questions asking about new partners and donors CSOs are seeking awards 
within the next 12 months. These questions were added to the endline survey and included in the 
index. 
● The PAP registered a question asking about the number of new donors listed as having received 
awards or donations within the last 12 months. This question was not asked on the endline survey 
and was not included in the index.
Outcomes excluded from the analysis:
Secondary Outcome 2.4: Social Media Capacity Index: Number and Quality of Social Media Posts
● This information was difficult to extract from scraped Facebook data and was excluded from the 
analysis.
Secondary Outcome 3.6: Share of Revenue From Malign Sources
● No CSOs reported receiving any revenue from Russia or China at baseline or endline.
Primary Outcome 4.3: CSO Network Diversification
● Data cleaning for this outcome is ongoing and will be included prior to the submission of this report 
to the Development Experience Clearinghouse.
Primary Outcome 4.4: Civic Space Petition Signatures
● Cambodia’s increasingly restrictive civic space environment made this task more difficult to 
implement. After endline scoping activities, the team decided to drop plans to collect data through 
a petition experiment.
 
 
 
 
 
 
 
 
 
APPENDIX E. BUDGET CATEGORIES 
BUDGET CATEGORY SUBCATEGORIES 
Awards 
The team defines an award as a sum of money received 
by an organization to facilitate the completion of a 
specific project or specific activities. This includes direct 
grants and subgrants, contracts and subcontracts, and 
other awards or subawards. 
● United States 
● Russia 
● China 
● Other foreign government donors or multilateral 
organizations 
● International NGOs and foundations 
● Domestic NGOs and foundations 
● RGC 
● Other 
Donations 
The team defines donations as gifts from a private individual, organization, foundation, or group that may 
take the form of money, goods, or services. Donations 
are given without expectations for actions in return and 
are not intended to fund a specific project. If donations 
include goods or services, the value should report the 
cost of those goods/services. 
● Foreign individuals 
● Foreign business 
● Cambodian individuals 
● Cambodian business 
● Anonymous/Other 
● International NGOs and foundations 
● Domestic NGOs and foundations 
Earned income 
The team defines earned income as income earned 
through commercial activity such as the sale of goods or 
services that is used to fund the organization's 
operations or activities. 
● Membership fees and dues 
● Fees paid by recipients of services rendered by the 
organization 
● Income from services rendered to the government 
● Income from services rendered to another 
NGO/community-based organization 
● Income from fundraisers or other special events 
● Income from the sale of goods and other 
commercial activities 
● Income from renting out property, vehicles, or 
equipment owned by the organization 
● Income from bank interest paid 
● Other 
Expenditures 
The team defines expenditures as any money spent
to fund an organization's operations. 
● Employee costs (includes wages, benefits, and 
other payments) 
● Allowances and per diems for program 
beneficiaries 
● Project materials 
● Transportation costs 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 74 
 
 
 
 
BUDGET CATEGORY SUBCATEGORIES 
● Vehicle purchases 
● Rent for use of land or buildings 
● Purchase or mortgage of land or buildings 
● Payment for services rendered (translations, 
auditing, design) 
● Utilities (electricity, water, internet, cellular data, 
etc.) 
● Grants to other NGOs or community-based 
organizations 
● Paying interest on debt 
● Costs of personnel training 
● Costs of advertising 
● Costs for conference attendance or hosting 
● Computer software, systems, or security 
purchased 
● Other organizational expenses 
● Other program expenses 
Assets 
The team defines assets to include the value of any property owned by the organization. This includes the 
value of any buildings or land, vehicles, or equipment 
owned by the organization. 
● Value of land, buildings, or property 
● Value of equipment and machinery 
● Value of inventories 
● Value of vehicles 
● Value of financial assets 
● Other 
Divestments 
The team defines divestments as money earned from the sale of property, investments, or other durable 
assets. This includes the value of any builds or land, 
vehicles, or equipment owned by the organization. 
● Income from the sale of land or buildings 
● Income from the sale of vehicles 
● Income from the sale of equipment or machinery 
● Other 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 75 
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 76
APPENDIX F. REFERENCES
Aronow, P.M., Samii, C., & Assenova, V.A. (2017). Cluster–Robust Variance Estimation for Dyadic 
Data. Political Analysis, 23 (4), 564–77.
Beaman, L.., Yishay, A. B., Magruder, J., & Mushfiq Mobarak, A. (2018). Can Network Theory-Based 
Targeting Increase Technology Adoption? American Economic Review Forthcoming. 
Curley, M. (2018). Governing civil society in Cambodia: implications of the NGO law for the “Rule of 
Law”. Asian Studies Review, 42(2), 247-267.
Cruz, C., Labonne, J., & Querubin, P. (2020). Social Network Structures and the Politics of Public Goods 
Provision: Evidence from the Philippines. American Political Science Review, 114 (2), 486-501.
Dalaibuyan, B. (2013). A Network Approach to Ngo Development: Women’s Ngos in Mongolia. The 
International Journal of Not-for-Profit Law, 15 (86). 
Dang, C.T., Burger, R., & Owens, T. (2020). Do Better-Performing Nongovernmental Organizations 
Report More Accurately? Evidence from Financial Accounts in Uganda. Economic Development 
and Cultural Change. 
Hill, T.P. (1995). A Statistical Derivation of the Significant-Digit Law. Statistical Science, 10 (4), 354–63.
Hung, C., & Hager, M.A. (February 2019). The Impact of Revenue Diversification on Nonprofit Financial 
Health: A Meta-Analysis. Nonprofit and Voluntary Sector Quarterly, 48 (1), 5–27. 
Gibbons, C.E., Suárez Serrato, J.C., & Urbancic, M.B. (2019). Broken or Fixed Effects? Journal of 
Econometric Methods, 8 (1).
Long, S., & Ervin, L. (2000). Using heteroscedasticity consistent standard errors in the linear regression 
model. The American Statistician, 54(3), 217-224.
Marshall, J.H., & Suarez, D. (2014). The Flow of Management Practices: An Analysis of Ngo Monitoring 
and Evaluation Dynamics. Nonprofit and Voluntary Sector Quarterly, 43 (6), 1033-51.
McKenzie, D.(2012). Beyond baseline and follow-up: The case for more T in experiments. Journal of 
Development Economics, Elsevier, 99(2), 210-221.
Drake, P., & Nigrini, M. (2000). Computer assisted analytical procedures using Benford's Law. Journal of 
Accounting Education, 18(2), 127-146.
Petrova, T., & Tarrow, S. (2007). Transactional and Participatory Activism in the Emerging European 
Polity: The Puzzle of East-Central Europe. Comparative Political Studies, 40 (1), 74-94.
Qu, H., Steinberg, R., & Burger, R. (2020). Abiding by the Law? Using Benford’s Law to Examine the 
Accuracy of Nonprofit Financial Reports. Nonprofit and Voluntary Sector Quarterly, 49 (3), 548–70.
USAID.GOV LO-MTSR RESILIENCY CAMBODIA IE FINDINGS REPORT | 77
Springman, J., Malesky, E., Right, L., & Wibbels, E. (2022). The Effect of Government Repression on Civil 
Society: Evidence from Cambodia. International Studies Quarterly, 66(3), sqac028.
Springman, J.; & Wibbels, E. (2021). Mapping NGO Networks: Evidence from a Survey of NGOs in Cambodia. 
Prepared for United States Agency for International Development Cambodia.
Suarez, D.; & Marshall, J.H. (2014). Capacity in the Ngo Sector: Results from a National Survey in 
Cambodia. VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations, 25 (1), 176-
200.
Viterna, J., Clough, E., & Clarke, K. (2015). Reclaiming the ‘Third Sector’ from ‘Civil Society’: A New 
Agenda for Development Studies. Sociology of Development, l (1),173-207.
Wheeler, A. (2014). Testing Serial Crime Events for Randomness in Day‐of‐Week Patterns with Small 
Samples. Journal of Investigative Psychology and Offender Profiling, 13, 148-165.
Wibbels, E., Andrzejewski, K., Springman, J., & Schewel, K. (2022). Retrospective of USAID 
Response to Pandemic Enabled Democratic Backsliding Findings Report. USAID DRG Center.
Youngs, R. & Echague, A. (2017). Shrinking Space for Civil Society: The EU Response. Technical 
Report, European Parliament Subcommittee on Human Rights.