May 2015
SHOUHARDO II Final Quantitative
Performance Evaluation
SHOUHARDO II Multi-Year Assistance Program
Acknowledgements 
TANGO International wishes to extend its gratitude to CARE Bangladesh staff that participated in the 
Quantitative Performance Evaluation. We would particularly like to acknowledge the support of Mr. 
AKM Abdul Wadud and his M&E team. These individuals shared their time, insights and documentation 
in an effort to ensure that the QPE is a valuable learning opportunity for all stakeholders of the 
SHOUHARDO II program. While it is not possible to exhaustively identify every individual involved, the 
team is particularly grateful to CARE Bangladesh staff members for their efforts and contributions to the 
study. 
We would also like to acknowledge the excellent support to the fieldwork provided by the CARE 
Bangladesh representatives in each of the four regional offices and offer special thanks to the entire 
quantitative team fielded by Mitra and Associates for the dedication and excellent work in collecting 
information for the QPE. 
Finally, we are most indebted to the individuals and families who gave freely of their time and company 
to be interviewed by our teams. Without their generosity and openness in welcoming us into their 
homes and sharing invaluable information about their lives, this important review would not have been 
possible. We sincerely hope that the findings of the QPE enable future programs to contribute to further 
improvements in their food and livelihood security and overall wellbeing. 
TANGO International 
 February 20, 2015
i 
Table of Contents 
List of Figures ............................................................................................................................................... iii
List of Annexes .............................................................................................................................................. v 
Acronyms ..................................................................................................................................................... vi
Executive Summary .................................................................................................................................... viii 
1. Introduction .......................................................................................................................................... 1 
1.1. Country Context ............................................................................................................................ 1 
1.2. Program description...................................................................................................................... 4 
1.3. SHOUHARDO II Implementation Mechanism ............................................................................... 5 
1.4. Performance evaluation methodology ......................................................................................... 6 
1.5. Household participation in interventions ...................................................................................10
2. Results/ Program Effectiveness ..........................................................................................................11
2.1. SO1: Availability of and access to nutritious foods .....................................................................11
2.1.1. Description of project activities ..........................................................................................11
2.1.2. Progress towards key impact indicators .............................................................................11
2.1.3. IR 1.1 Improved and diversified agriculture systems ..........................................................13
2.1.4. IR 1.2 Increased household income among PEP ................................................................. 15
2.1.5. Differences in progress made for female- and male-headed households ..........................16
2.1.6. Impact and outcome activities by participation in SO1 activities .......................................17
2.1.7. Summary of progress toward SO1 ......................................................................................18
2.2. SO2: Improved health, hygiene and nutrition of children under 2 years of age ........................19
2.2.1. Description of project activities ..........................................................................................19
2.2.2. Progress towards key impact indicators .............................................................................20
2.2.3. IR 2.1 Access to and utilization of health and nutrition services improved .......................24
2.2.4. IR 2.2 Caregivers of children under 2 adopt improved health, hygiene and nutrition 
behavior and caring practices .............................................................................................................25
2.2.5. Differences in progress made for female- and male-headed households ..........................28
2.2.6. Summary of progress towards SO2.....................................................................................29
2.3. SO3: PEP women and adolescent girls empowered ...................................................................30
2.3.1. Description of project activities ..........................................................................................30
2.3.2. Progress towards key impact indicators .............................................................................30
2.3.3. IR 3.1: Increased ability of women to influence decisions ..................................................31
2.3.4. IR 3.2: Men and women working together to end violence ...............................................32
2.3.5. Differences in progress made for female- and male-headed households ..........................33
2.3.6. Summary of progress toward SO3 ......................................................................................34
ii
2.4. SO4: Responsiveness and accountability to the PEP of local elected bodies and government 
service providers increased ....................................................................................................................35
2.4.1. Description of project activities ..........................................................................................35
2.4.2. IR 4.2: PEP access to entitlements and services increased, including safety nets and 
natural resources ................................................................................................................................35
2.4.1. Differences in progress made for the PEP and non-poor ...................................................37
2.4.2. Differences in progress made for female- and male-headed households ..........................38
2.4.3. Summary of progress toward SO4 ......................................................................................38
2.5. SO5: Improved preparation, mitigation and response to disasters and adaptation to climate 
change 39
2.5.1. Description of project activities ..........................................................................................39
2.5.2. Progress toward key impact indicators ...............................................................................39
2.5.3. IR 5.1: Disaster contingency systems in place and functioning ..........................................40
2.5.4. Differences in progress made for female- and male-headed households ..........................41
2.5.5. Summary of progress toward SO5 ......................................................................................42
3. Comparative effectiveness of the MCHN/PEP and PM2A programming approaches........................43
3.1. Success in implementing the programming approaches ............................................................44
3.2. Success of randomization in equalizing MCHN/PEP and PM2A households at baseline ...........45
3.3. Comparison of changes from baseline to endline in malnutrition prevalence across MCHN/PEP 
and PM2A households ............................................................................................................................46
4. Conclusions and recommendations ....................................................................................................47
4.1. General conclusions ....................................................................................................................47
4.2. Recommendations for Qualitative Impact Evaluation of SHOUHARDO II ..................................48
4.3. Recommendation for future project design and evaluation ......................................................50
iii
List of Figures 
Figure 1. Map of SHOUHARDO II intervention areas .................................................................................... 3 
Figure 2: Number of MAHFP by region and sex of household head ...........................................................12
Figure 3: Averaged HDDS, by region and sex of household head ...............................................................12
Figure 4: Indicators of hunger in the last four weeks .................................................................................13
Figure 5: Percent of households engaged in agricultural activities ............................................................13
Figure 6: Percent of field crop producers adopting the minimum number of technologies (three), by 
region and by sex of household head .........................................................................................................14
Figure 7: Percent of vegetable producers adopting the minimum number of technologies (three), by 
region and by sex of household head .........................................................................................................14
Figure 8: Mean monthly household income per capita (in TK), by region and sex of household head .....15
Figure 9: Index of domestic assets, by region and sex of household head ................................................15
Figure 10: Indicators of access to markets and services, percent of households at endline .....................16
Figure 11: Percent difference from baseline to endline of key livelihood indicators by sex of household 
head ............................................................................................................................................................17
Figure 12: Percent of children under five stunted, wasted and underweight, by sex ................................21
Figure 13: Percent of children under-five stunted, by region ....................................................................22
Figure 14: Percent of children under-five wasted, by region .....................................................................22
Figure 15: Percent of children under-five underweight, by region ............................................................22
Figure 16: Percent of children under-two stunted, wasted and underweight ...........................................22
Figure 17: Percent of mothers of children under five years old underweight, by region ..........................23
Figure 18: Percent of children 6-23 months with diarrhea in the last two weeks, by region ....................24
Figure 19: Immunization and ORS for diarrhea among children under two ..............................................24
Figure 20: Percent of children immunized against 8 diseases by 12 months (12-23 months), by region ..25
Figure 21: Caring practices for mothers, percent of women utilizing practice or receiving service ..........25
Figure 22: Nutrition-related caring practices for children under two for all participants (percent) ..........26
Figure 23: Percent of mothers who feel it is important to wash their hands at five critical times by region
....................................................................................................................................................................27
Figure 24: Percent of children under three whose feces are disposed of safely by region........................27
Figure 25: Percent of households with access to an improved drinking water source, by region .............28
Figure 26: Percent of households with access to an improved sanitation facility, by region ....................28
Figure 27: Percent difference from baseline to endline, child anthropometric indicators, by sex of 
household head ..........................................................................................................................................29
Figure 28: Percent of women with control over economic resources, by region and sex of household 
head ............................................................................................................................................................31
iv
Figure 29: Indicators of women's empowerment ......................................................................................32
Figure 30: Index of women's decision making power by region and sex of household head ....................32
Figure 31: Percent of women who believe a man is justified in hitting/abusing the wife, by situation ....33
Figure 32: Percent of households in which female was yelled at or struck during the last year by region 
and sex of household head .........................................................................................................................33
Figure 33: Percent difference from endline to baseline of key women's empowerment indicators, by sex 
of household head ......................................................................................................................................34
Figure 34: Role of SO4 and VDC in linking with other SOs ..........................................................................35
Figure 35: Mean number of services utilized, by region, well-being category and sex of household head
....................................................................................................................................................................36
Figure 36: Average number of safety nets engaged in, by region, well-being category and sex of 
household head ..........................................................................................................................................36
Figure 37: Average number of common property resource utilized by region, well-being category, and 
sex of household head ................................................................................................................................37
Figure 38: Percent difference from baseline to endline, government and service indicators, by well-being 
category ......................................................................................................................................................38
Figure 39: Percent different from baseline to endline, government and service indicators, by sex of 
household head ..........................................................................................................................................38
Figure 40: Percent of households distress selling, by region, sex of household head ...............................40
Figure 41: 2010 Taka value of household assets sold to meet urgent household need ............................40
Figure 42: Percent of households that experienced any disaster in the last 12 months, by region, well￾being category, and sex of household head ...............................................................................................41
Figure 43: Average number of disasters experienced, by region, well-being category, and sex of 
household head ..........................................................................................................................................41
Figure 44: Percent difference from baseline to endline, experiential disaster management, by sex of 
household head ..........................................................................................................................................42
v 
List of Annexes 
Annex 1: References 
Annex 2: Scope of Work 
Annex 3: Endline Survey Instrument 
Annex 4: Sample Weights 
Annex 5: IPTT matrix and SAPQ 
Annex 6: Quantitative tables
vi
Acronyms 
ANC Antenatal Care
CAP Community Action Plan
CBO Community Based Organization
CDRM Climate and Disaster Risk Management 
CHV Community Health Volunteer
C-IMCI Community-Based Integrated Management of Childhood Illnesses
CMAM Community Based Management of Acute Malnutrition
COG Core Occupational Group
CVCA Climate Vulnerability and Capacity Analysis
DRR Disaster Risk Reduction
ECCD Early Childhood Care and Development
EKATA Empowerment Knowledge and Transformative Action
EPI Expanded Program on Immunization
FFP Food For Peace
GMP Growth Monitoring and Promotion
GOB Government of Bangladesh
IPTT Indicator Performance Tracking Table
IR Intermediate Result
IYCF Infant and Young Child Feeding
LoA Life of Award
M&E Monitoring and Evaluation
MCHN Mother Child Health and Nutrition
MDG Millennium Development Goal
MoH&FW Ministry of Health and Family Welfare
NBD Nation Building Department
NGO Non-Governmental Organization
PACC Program Advisory and Coordination Committee
PEP Poor and Extreme Poor
PM2A Prevention of Malnutrition among Children under 2 Years of Age
PNGO Partner Non-Governmental Organization
QPE Quantitative Performance Evaluation
RTC Randomized Control Trial
SBTB School-based Teen Brigades
SO Strategic Objective
TK Bangladesh Taka
UDMC Union Disaster Management Committee
UP Union Parishad
USAID United States Agency for International Development
vii 
USD United States Dollar
VAW Violence Against Women
VGF Vulnerable Group Feeding
VDC Village Development Committees
viii 
Executive Summary 
Despite progress on combating poverty and improving a number of other human development 
indicators, Bangladesh remained ranked 142nd out of 187 countries in the 2014 Human Development 
Index (HDI).1
 Similarly, it ranked 57th out of 76 countries listed in the 2014 Global Hunger Index (GHI), 
placing the country in the category of “serious” levels of food insecurity.2
 Although Bangladesh has meet 
the Millennium Development Goals (MGDs) for prevalence of undernourishment, there are still some 25 
million people undernourished and progress has been slowing since the mid-2000s due to the 
inadequate consumption of the minimum daily-recommended amount of food and/or intake of high￾value food products. In addition to extreme poverty and food insecurity, vulnerable populations 
throughout the country cope with limited access to education and health services, widespread gender 
inequity, and repeated exposure to natural disasters and climate change. Despite improvement in 
economic indicators and progress toward the MGDs over the previous decade, undernourishment 
especially among women and children remains a major impediment to further development of the 
country.3
 
It was within this context that CARE Bangladesh and Partner NGOs first designed and implemented the 
SHOUHARDO Development Assistance Program (DAP) from 2005-2010.4
 The final evaluation of the first 
phase confirmed that it was largely effective in addressing not only the availability, access and utilization 
issues that lead to food insecurity, but also the underlying factors that contribute to vulnerability such as 
a lack of social participation and disempowerment that prevent people from realizing their full potential 
to lead healthy and productive lives. 
The second phase, SHOUHARDO II, implemented from June 2010 to May 2011 in 11 of the poorest and 
most marginalized districts of the country5
, builds upon the successes of the initial SHOUHRDO program 
by maintaining a strong emphasis on improving livelihood security, food security, nutrition, poor and 
extreme poor (PEP), and women’s empowerment at the community level, while adding components 
aimed at strengthening local governance and improving adaptation to climate change. CARE Bangladesh 
implemented SHOUHARDO II with the assistance of 16 NGOs and 4 Technical Partners. 
Objectives of the Quantitative Performance Evaluation (QPE): 
The primary objectives of the QPE were to: 
 Assess the overall impact of the project including effectiveness of program strategies and 
implementation of interventions; 
 Compare results of the two programming approaches (MCHN/PEP and PM2A) implemented 
separately in randomly selected sites; 
 
1
 UNDP. 2014. Human Development Report 2014: Sustaining human progress: Reducing vulnerabilities and building resilience. 
New York, UNDP. 
2
 IFPRI, Concern and Welthungerhilfe. 2014. Global Hunger Index: The challenge of hidden hunger. Washington, D.C., IFPRI. 
3
 FAO, WFP and IFAD. 2013. The State of Food Insecurity in the World 2013. The multiple dimensions of food security. Rome, 
FAO. 
4
 SHOUHARDO is an acronym that stands for Strengthening Household Ability to Respond to Development Opportunities. 
SHOUHARDO is also a Bengali word for “amity.”
5
Kurigram, Nilphamari, Rangpur, Dinajpur, Bogra, Sirajganj, Pabna, Mymensingh, Jamalpur, Sunamganj, Cox’s Bazar. 
ix
 Evaluate progress of SHOUHARDO II towards its objectives.
Methodology 
In order to meet these objectives, the QPE team measured the program’s “population-based” 
performance on key indicators with comparisons to baseline and midterm survey results. To ensure 
comparability and the validity of findings in the baseline and midterm, the quantitative methodology 
utilized the same household survey instrument and sampling frame. In addition, two levels of 
stratification were employed. The first was the division of SHOUHARDO II impact population into four 
regions – Coast (Cox’s Bazar), Haor (Mymensingh), Mid Char (Sirajganj) and North Char(Rangpur) –
reflecting the distinct geographic areas where the program is operational. The second level of 
stratification was into the different service delivery strategies termed MCHN/PEP and PM2A. MCHN 
services across the entire program were provided under either one or the other of these two 
approaches. 
The findings and analysis contained in the full QPE report are intended to draw lessons learned for the 
selection, design and implementation of future programs in Bangladesh. CARE will also make extensive 
use of the findings from the survey to document and disseminate best practices and lessons learned. 
The discussion that follows is a brief summary of QPE results that directly corresponds to the objectives 
of the review as laid out in the Terms of Reference. 
A. Effectiveness of strategies and implementation of interventions 
The overall goal of the SHOUHARDO II program is to: 
“Transform the lives of 370,000 Poor and Extreme Poor (PEP) households (HH) in 11 of the poorest 
and marginalized districts in Bangladesh by reducing their vulnerability to food insecurity.”
In order to achieve this goal, CARE-Bangladesh and partner collaborations have established the 
following Strategic Objectives (SO) and Intermediate Results (IR):
SO1: "Availability of" and "access to" nutritious foods enhanced and protected for 370,000 PEP 
households.
IR1.1: Improved and diversified agriculture systems developed and linked with private and public 
services. 
IR1.2: Increased household income among PEP in the target communities. 
SO2: Improved health, hygiene and nutrition status of 281,000 children under 2 years of age.
IR2.1: "Access to" and "utilization of" health and nutrition services improved to care givers of 
children under 2 years of age.
IR2.2: Care givers of children under 2 adopt improved health, hygiene and nutrition behavior and 
caring practices.
SO3: PEP women and adolescent girls empowered in their families, communities and Union Parishad.
IR3.1: Influence of PEP women and adolescent girls in decision making increased. 
IR3.2: Local support systems strengthened to reduce Violence Against Women (VAW).
SO4: Local elected bodies and government service providers responsiveness and accountability to the 
x 
PEP increased.
IR4.1: Nation Building Departments (NBD) and Union Parishads proactively work to address the 
needs of the PEP, especially women.
IR 4.2: PEP access to entitlements and services increased, including safety nets and natural resources.
SO5: Targeted community members and government institutions are better prepared for, mitigate, 
and respond to disasters and adapt to climate change.
IR5.1: Disaster contingency systems in place and functioning. 
IR5.2: Influence local and national humanitarian assistance initiatives.
SHOUHARDO II prioritized community-based interventions and placed considerable emphasis on the 
capacity of Village Development Committees (VDCs) to not only assess local factors constraining food 
security but to also oversee program efforts to address them. In particular VDC, with program staff, 
were charged with coordinating the efforts of four thematic groups in target communities. They 
included: 
 Core Occupational Groups aimed at increases in food production, and/or income (targeting both 
men and women); 
 Growth Monitoring and Promotion (GMP) Groups aimed at improved nutrition and health status 
of children under 5 years of age; 
 Early Childhood Community Development (ECCD) Groups for mothers and children between 0 
and 8 years of age); and 
 Empowerment, Knowledge and Transformative Action (EKATA) Groups focused on educating 
women and girls on empowerment and protection. 
B. Comparison between implemented program designs 
A critical new element of SHOUHARDO II’s efforts to improve child nutrition was the trial of the 
Prevention of Malnutrition in Children under 2 Approach – (PM2A). Under PM2A all pregnant women, 
mothers of children 0-23 months and children under 2, regardless of nutritional and economic status 
received a monthly food ration, in addition to a package of health promotion and preventative and 
curative health services. Meanwhile, SHOUHARDO II continued to implement a more targeted approach 
to improving maternal and child health and nutrition (MCHN), under which food rations are only 
provided to pregnant women, mothers and children in the PEP well-being categories. The comparison of 
nutrition outcomes under these approaches was intended to provide insight into the relative costs and 
benefits of each in the context of Bangladesh. 
C. Progress of the SHOUHARDO II Program towards its objectives 
Overall, SHOUHARDO II has made significant progress toward each of its SOs. However, a number of 
targets were not met by the close of the program despite gains made between baseline and endline. 
The table below provides comparable quantitative information on key impact and outcome indicators 
collected during the baseline, midterm, and endline surveys. It shows improvements across a number of 
indicators, especially for household food security and child nutrition under SO1 and SO2. Yet, despite 
xi
substantial gains, indicators for hygiene and sanitation, child immunizations, and women’s 
empowerment fell significantly short of meeting performance targets. Note that there were no 
impact/outcome indicators in the SHOUHARDO II Indicator Performance Tracking Table (IPTT) under 
SO4. Also, the change in taka value of distress sales from baseline to endline shown in SO5 was not 
statistically significant. 
In the following table and throughout the full report, all statistical significance compares endline values 
to baseline values. If an endline value in a figure has an asterisk next to it, this indicates that endline 
value is statistically significant compared to the baseline value of that indicator (not to the Life of Award 
(LoA) target).6
Progress toward impact/outcome indicators7
Indicator Target
Base￾line
Mid￾term 
End￾line Difference 
(Endline-Baseline)
SO 1: "Availability of" and "access to" nutritious foods enhanced and protected for 370,000 poor & extreme poor (PEP) 
households
Average household dietary diversity score 6.0 4.8 5.9 8.7 3.9 *
Average number of months of adequate household 
food provisioning
8.0 5.9 9.9 11.0 5.1 *
SO 2: Improved health, hygiene and nutrition status of 176,706 children under 2 years of age
Percent of underweight children under five years of 
age (0-59 months) 38.4 42.2 34.4 35.9 -6.3 *
 Females 34.6 37.2 34.9 36.4 -0.8
 Males 42.2 46.7 33.4 34.3 -12.4 *
Percent of stunted children under five years of age (6-
59 months) 55.1 61.7 52.7 50.4 -11.3 *
 Females 51.2 56.5 52.2 47.8 -8.7 *
 Males 58.9 66.1 53.2 49.7 -16.4 *
Percent of children under age two who had diarrhea 
in the prior two weeks 10.3 13.8 11.6 8.1 -5.7
Percent of children underweight under the age of 2 
years 31.1 33.2 32.1 29.5 -3.7
Percent of children immunized against 8 diseases by 
12 months of age 82.0 58.6 74.9 73.0 14.4 *
Percent of pregnant and lactating women taking iron 
supplements in last 7 days 45.0 15.2 29.8 52.0 36.8 *
Prevalence of exclusive breast feeding of children 
under 6 months 66.2 64.1 66.4 62.2 -1.9
Percent of mothers who feel it is important to wash 
hands at five critical times£
35.0 9.8 10.9 29.8 20.0 *
Number of people in target areas with improved 
access to sanitation facilities 60.0 20.3 32.4 52.8 32.5 *
 
6
 Refer to Annex 5 for a comparison of endline results to LoA targets and for 95% confidence intervals of indicators . 
7
 Baseline and midterm numbers in this table were computed by TANGO during endline analysis. Some of these numbers differ 
from the baseline and midterm numbers in IPTT indicator table. The IPTT indicator is included as an annex to this report, with 
original and revised baseline numbers included.
xii 
Percent of children 6–23 months of age who receive a 
minimum acceptable diet 20.0 8.7 17.5 47.9 39.2 *
Number of people in target areas with improved 
access to drinking water supply 80.0 59.5 76.3 77.6 30.3 *
SO 3: PEP women and adolescent girls empowered in their families, communities and Union Parishad
Percent of women control over economic resources 25.0 5.7 11.3 19.8 247.4 *
SO 5: Targeted community members and government institutions are better prepared for, mitigate, and respond to 
disasters and adapt to climate change
Percent of PEP households distress selling 7.0 9.6 9.5 9.8 2.1
Taka value of distress sale (2010) 15000 19979 21304 16,321 46.0
Note: Stars (*) indicate difference is statistically significant at the 10% (*) level 
£ The five critical times are: Before eating, before breastfeeding or feeding a child, before cooking or preparing food, after 
defecation/urination, and after cleaning a child that has defecated/changing a child's diaper. 
Conclusions 
1. In SO1, SO2, and SO3, there have been strong improvements in all impact/outcome indicators 
from baseline to endline. However, it should be noted that for some indicators, improvements 
were largely seen from baseline to midterm, and stabilized from midterm to endline, especially 
those under SO2. However most indicators showed a steady trend from baseline to midterm to 
endline. 
2. These improvements may be explained, at least in part, by changes in behaviors of households 
(adoption of improved agricultural practices, improved access to markets and agricultural 
services, adoption of child feeding practices, changes in attitudes about violence to women, 
etc.). 
3. The household-level indicators for SO5 showed no statistically significant change from baseline 
to endline survey rounds. However, these indicators cannot capture the main short-term 
impacts of the interventions of these SOs, which are directed toward improving the 
performance of community and local government stakeholders. 
4. Direct comparisons of results between the two SO2 programming approaches indicate that the 
PM2A programming approach had greater impact on nutritional status of children than the 
targeted MCHN/PEP approach. However, several factors other than the different programming 
approaches may explain the differences between these two groups. In particular, there 
significant differences in characteristics of households between the two groups that may explain 
the differences in nutritional outcomes. In addition, there is evidence that households in the 
PM2A group had greater access to all project services, not just access to food rations. In 
particular, a greater proportion of households in the PM2A group participated in EKATA groups, 
and their overall higher proportion of non-poor households than compared to MCHN/PEP 
villages. Given these circumstances, no strong conclusions can be drawn regarding the relative 
impacts of the PM2A and MCHN/PEP approaches on nutritional status. It must be emphasized 
that although there are no clear differences in nutritional benefits, the relative cost of the PM2A 
is substantially higher than the targeted MCHN/PEP approach. 
xiii 
Recommendations for further research 
The findings from the quantitative performance evaluation survey raised a number of issues that should 
be further pursued in qualitative evaluation of SHOUHARDO II. Generally, these questions are around 
understanding why project outcome and impact indicators did or did not change, and how project 
interventions may have affected the changes in these indicators. Some more specific questions that 
were raised from the quantitative findings include the following: 
 The Haor Region stands out compared to other regions because of smaller improvements or 
negative outcomes. Qualitative inquiry should follow up to find explanations for the differences 
in patterns of changes across project regions. 
 Many indicators – especially impact-level measures – progressed more between baseline and 
midterm than from midterm to endline. Can this be explained by program implementation, 
household factors, or other factors? 
 Was there a significant difference in impact-level indicators between participants and non￾participants in project activities? Note that CARE has commissioned both qualitative and 
quantitative impact evaluation studies to address this question. 
 Adoption of improved technologies increased substantially for both SO1 participants and non￾participants. Further research is needed to understand the causes of this adoption, and the role 
that SHOUHARDO II interventions have played in promoting the observed increases in adoption. 
1 
1. Introduction 
The Strengthening Household Ability to Respond to Development Opportunities (SHOUHARDO) II 
Program was implemented by CARE Bangladesh from June 2010—September 2015. The project is 
funded by the United States Agency for International Development (USAID) and the Government of 
Bangladesh (GOB) at nearly US$130 million,8
 making SHOUHARDO II one of the largest non-emergency 
food security programs in the world. The program extended across four regions (North Char, Mid-Char, 
Haor and Coastal), 11 districts, 30 Upazilas and 172 Unions to reach 370,000 households. The five-year 
Food for Peace (FFP) Title II Multi-Year Assistance Program (MYAP) advanced the model from the 
previous SHOUHARDO Program (2005-09) that successfully reduced child malnutrition, and 
strengthened livelihoods and women’s empowerment. 
Technical Assistance to Non-Governmental Organizations (TANGO) International, Inc. in collaboration 
with Dhaka-based Mitra and Associates conducted the population-based quantitative performance 
evaluation survey that is the topic of this report from 17 November through 12 December 2014. The 
primary purpose of the endline survey was to collect and compare data on key project indicators with 
the baseline (conducted in December 2010) and midterm (conducted in December 2012) data to 
measure progress towards SHOUHARDO II’s strategic objectives and intermediate results. The primary 
audiences for the findings are CARE, USAID and program partners, who will draw on the lessons learned 
to inform future programming. The Government of Bangladesh (GOB) is a secondary user of the 
findings. 
This report describes the survey methodology and process, assesses progress made since the baseline 
and midterm studies, examines key successes and challenges, and provides recommendations for 
improved performance. A compendium of annexes is included with this report. 
1.1. Country Context 
Bangladesh has made important social gains in recent years, yet as a low-income country, poverty and 
food insecurity remain major challenges. Bangladesh is one of the world’s most densely populated 
countries, with a total population of approximately 159 million. Forty-three percent of the population is 
poor, living on less than US$ 1.25/day, and about one in five (21 percent) is severely poor. The 19 
percent of the population that is above the poverty line nevertheless hovers close to impoverishment 
(2011).9
 Bangladesh will reach middle-income country (MIC) status in coming years, though inequality in 
development persists. Gross Domestic Product (US$ 150 billion in 2013) steadily increased over the past 
decade.10 Bangladesh is ranked 142 out of 187 countries on the 2013 Human Development Index (HDI), 
just reaching the medium development category (>0.55) with a score of 0.558; however, when the HDI 
is adjusted for the loss of human development due to inequality in income, education and life 
expectancy, the value is 0.396. The sex-disaggregated HDI is 0.528 for females compared to 0.582 for 
males of Bangladesh.11
 
 
8
 Plus 287,420 MT worth of commodities for both direct distribution and monetization (80 percent). 
9
 UNDP. 2013. Note: The Millennium Development Goal (MDG) target by 2015 is 35.1% below US$1 (PPP) per day. 
10 World Bank. 2014. 
11 UNDP. 2013. 
2 
The Government’s long-term development strategy (2010-2021) is implemented through two medium￾term development plans. This strategy focusses on poverty and income, education, nutrition, health and 
women’s empowerment, among others,12 which have been key areas of emphasis of the SHOUHARDO 
programs. Bangladesh has seen considerable improvements in universal primary education with a 
special focus on female education, achieving the Millennium Development Goal (MDG) in this area. 
School attendance for all age groups is on the rise and is currently on track to meet the MDG 2015 
target. Attendance is highest at age 10 for boys and at age 11 for girls. Incomplete primary education 
from drop-out is a critical issue particularly for poor and ultra-poor households.13 Bangladesh has also 
met the MDG targets for under-five underweight and mortality rates and is on track to meet the goal for 
reducing infant and maternal mortality.14 Yet food insecurity continues to be a challenge for many 
households due to natural and political/economic shocks. Bangladesh is a low-lying country and 
increasingly vulnerable to flooding, erosion and cyclones as sea levels rise with climate change.15 Strikes 
and blockades, political violence, market and food price volatility have regularly occurred, compounding 
these risk factors.16
 
The four operational regions of SHOUHARDO II shown in Figure 1—Coast, Haor, Mid Char and North 
Char—share common risk factors of high food insecurity and malnutrition rates, high natural disaster 
risk, remoteness, high presence of the country’s most marginalized groups, low literacy, and high 
poverty rates. The regions are nevertheless distinct in their topography and determinants of risk. The 
North (Rangpur) and Mid (Sirajganj) Chars are riverine islands surrounded by water most of the year. 
They are prone to dramatic erosion and floods, which result in crop loss, isolation, and poor access to 
markets and services. Also highly flood-prone and with similar food insecurity issues to the Chars is the 
northeastern Haor area (Mymensingh), characterized by vast expanses of depressed wetlands with 
scattered, elevated mounds that become largely inhabitable islands during the wet season. The 
waterlogged areas have fish, but poor households often lack access to this source of income and/or 
protein, and all transport is by boat. In the dry season, the water subsides to harsh terrain that must be 
crossed by foot. According to the baseline and midterm studies, the most common livelihoods in these 
northern regions are farming and agricultural contract labor. In addition, many households reside on 
khas land, which by law gives first priority to the poor to access, utilize, and reside, but often sources of 
conflict arise with elites and organized groups. The delta-like Coast region (Cox’s Bazar) is in the deep 
southeast of the country where fishing and non-agricultural labor are common occupations. Food 
security is threatened by regular storm surges and slow-onset disasters such as water-logging and land 
salinization, and the impacts of climate change. 
 
12 Government of the People’s Republic of Bangladesh (Ministry of Planning, General Economics Division). 2012. 
13 NIPORT, Mitra and Associates, and ICF International. 2013. DHS 2011. (P.11, 23-25) 
14 Government of the People’s Republic of Bangladesh and UNDP Bangladesh. 2014.
15 Karmalkar, A., C. McSweeney, M. New and G. Lizcano. 2010. 
16 BBC. 2014. 
3 
Figure 1. Map of SHOUHARDO II intervention areas 
4 
1.2. Program description 
CARE has been working in Bangladesh (previously East Pakistan) since 1949 and has implemented Title II 
programs in the country for four decades in collaboration with the GOB and local partners. The 
integrated design of the SHOUHARDO II program has drawn from CARE’s longstanding presence, 
experience and commitment to Bangladesh and its mission to address both humanitarian and 
developmental needs with a strong focus on the most vulnerable populations. The program operated 
from 1 June 2010 to 31 May 2015 in the districts of Rangpur, Kurigram, Nilphamari and Dinajpur in the 
North Char region; Sirajganj, Bogra and Pabna districts in Mid Char; Mymensingh, Jamalpur and 
Sunamganj districts in Haor; and Cox's Bazar in the Coastal region. 
The overall goal of the SHOUHARDO II program was to: 
“Transform the lives of 370,000 Poor and Extreme Poor (PEP) households (HH) in 11 of the poorest 
and marginalized districts in Bangladesh by reducing their vulnerability to food insecurity.”
In order to achieve SHOUHARDO II’s program goal, CARE partnered with 16 NGOs and in close 
collaboration with a dozen government ministries to achieve the following Strategic Objectives (SO) and 
Intermediate Results (IR):17
SO1: "Availability of" and "access to" nutritious foods enhanced and protected for 370,000 PEP households. 
IR1.1: Improved and diversified agriculture systems developed and linked with private and public 
services. 
IR1.2: Increased household income among PEP in the target communities. 
SO2: Improved health, hygiene and nutrition status of 176,706 children under 2 years of age.18
IR2.1: "Access to" and "utilization of" health and nutrition services improved to care givers of 
children under 2 years of age. 
IR2.2: Care givers of children under 2 adopt improved health, hygiene and nutrition behavior and 
caring practices. 
SO3: PEP women and adolescent girls empowered in their families, communities and Union Parishad. 
IR3.1: Influence of PEP women and adolescent girls in decision making increased. 
IR3.2: Local support systems strengthened to reduce Violence Against Women (VAW). 
SO4: Local elected bodies and government service providers’ responsiveness and accountability to the 
PEP increased. 
IR4.1: Nation Building Departments (NBD) and Union Parishads proactively work to address the 
needs of the PEP, especially women. 
IR 4.2: PEP access to entitlements and services increased, including safety nets and natural 
resources. 
SO5: Targeted community members and government institutions are better prepared for, mitigate, 
and respond to disasters and adapt to climate change. 
 
17 The internal logic will be described in more detail in each of the SO-specific sections of the report. 
18 Note: Footnote in Endline Evaluation TOR states revision to 176,706. 
5 
IR5.1: Disaster contingency systems in place and functioning. 
IR5.2: Influence local and national humanitarian assistance initiatives. 
1.3. SHOUHARDO II Implementation Mechanism 
The program follows a standard set of procedures and guidelines to interact with participating 
communities and identify ways and means of transferring services and resources to beneficiaries. As 
with any program of this scale, the implementation mechanism is fairly consistent. 
The targeted governance structure at the community level (and entry point) is the establishment of a 
Village Development Committee (VDC) – with elected members, including both men and women – that 
jointly develop a Community Action Plan (CAP) through a participatory process. Prior to developing the 
action plan the VDC helps facilitate the process of identifying community problems, conducting Climate 
Vulnerability and Capacity Analysis (CVCA) and gender analysis, and prioritization of community needs. 
An action plan is then drafted to address identified needs that fall within the scope of the program (e.g. 
shortage of food, ending violence against women, ending dowry, water, sanitation, and climate and 
disaster risk management (DRM) – with construction of infrastructure such as embankments, bridges, 
etc.). Field facilitators then work with VDCs and other specific groups in the community to develop 
strategies for addressing individual items on the CAP. Each VDC reviews its progress on an annual basis 
using results from the CVCA, gender analysis and action planning process to hold itself accountable. 
Once established, the VDC also helps monitor progress made by four thematic groups in the 
communities. These are: 
 Core Occupational Groups (COGS) aimed at increases in food production, and/or income 
(targeting both men and women); 
 Growth Monitoring and Promotion (GMP) Groups aimed at improved nutrition and health status 
of children under 5 years of age; 
 Early Childhood Community Development (ECCD) Groups for mothers and children between 0 
and 8 years of age); and 
 Empowerment, Knowledge and Transformative Action (EKATA) Groups focused on educating 
women and girls on empowerment and protection. 
Each group is facilitated by a paid volunteer recruited from within the community.19 As such there are a 
total of 8,000 paid volunteers across the 1,573 SHOUHARDO II target villages, divided into four 
categories; 1) Community Health Volunteer (CHV), 2) EKATA, 3) ECCD volunteer, 4) Agriculture 
volunteer. Following training provided by SHOUHARDO II, volunteers are tasked with facilitating group 
sessions, counseling during courtyard sessions and home visits and referral to specific government 
services such as health clinics and hospitals. Each volunteer is responsible for servicing a large number of 
households (>250 households). Full-time employed “field facilitators” oversee a specific geographic area 
and provide technical support and oversight to volunteers from the four different groups. Field 
 
19 As an incentive volunteers receive 2,000 TK per month. 
6 
facilitators are employed by private non-governmental organizations (PNGOs) and, in areas of direct 
delivery, by CARE. They receive technical support from technical managers within these organizations. 
 
The MCHN/PEP and PM2A approaches to service delivery 
Project activities falling under SO2, “Improved health, hygiene and nutrition status of 176,706 children 
under 2 years of age” were delivered following two different approaches in order to test which is most 
effective at reducing child malnutrition. The first, the MCHN/PEP (Maternal and Child Health and 
Nutrition/PEP) approach, follows the traditional SHOUHARDO model of including only poor and 
extreme-poor (PEP) households as participants in MCHN activities, including food ration receipts, 
educational activities, and growth monitoring and promotion. Food rations are made available to 
pregnant PEP women, PEP women with children under two, and children under two living in PEP 
households. The second approach, the Preventing Malnutrition in Children Under 2 (PM2A) approach, 
includes as participants all pregnant women and children under two years of age in project villages, 
regardless of socio-economic status. Both approaches are based on the basic PM2A ideas that (a) the 
first “1000 days”, from conception to age 24 months is the window of opportunity for preventing 
malnutrition; (b) efforts should be focused on the prevention of malnutrition rather than its cure; and (c) 
such prevention should take place by addressing malnutrition’s three underlying causes (access to 
adequate food, the quality of caring practices for children and mothers, and access to health services 
and a healthy environment).i
 SHOUHARDO II project villages were randomly selected into the 
MCHN/PEP and PM2A groups, with roughly 17 percent of villages chosen to follow the PM2A approach. 
1.4. Performance evaluation methodology 
As with the earlier survey rounds, the purpose of the endline QPE survey is to provide statistically 
representative estimates of project impact and outcome level indicators that are measured at the 
household level. To provide statistically valid estimates of these indicators, the survey was conducted on 
a sufficiently large, randomly selected sample of households from within the SHOUHARDO II 
implementation area. 
Sampling design: The sampling approach for the SHOUHARDO II household surveys was first formulated 
in August of 2010, prior to the baseline study. It was then reviewed by TANGO and finalized with 
feedback and suggestions from FANTA II and CARE Bangladesh. This agreed upon sampling approach 
was applied during the baseline and midterm studies. To ensure comparability, this sampling approach 
was also applied for the endline survey. 
The methodology utilizes two levels of stratification. The first was a division of the SHOUHARDO II 
operational area into its four geographical regions – Coast (Cox’s Bazar), Haor (Mymensingh), Mid Char 
(Sirajganj), and North Char (Rangpur) – reflecting the distinct geographic areas where the program is 
operational. The baseline and midterm reports had shown that stratification by region captures 
significant variation in results since these regions have distinct and different physical features and 
resources were spread out over large distances. 
7 
The second level of stratification is by programming approach (MCHN/PEP and PM2A) discussed above. 
The design was based on sampling an equal number of villages (clusters) and households for both PM2A 
and MCHN/PEP areas and across the four regions, that is the four geographic strata were then further 
stratified by programming approach to provide a total of eight stratat. For analysis, cases were weighted 
to account for the different proportions of sampled households drawn for the survey in relation to the 
total number of households in each stratum. Application of these sampling weights allowed for 
statistically valid comparisons between PM2A and MCHN/PEP target areas and across the four regions. 
The sampling frame for the survey is a comprehensive list of all households in villages where 
SHOUHARDO II is operational. At the beginning of implementation, the project conducted a census of 
households in the project implementations area to obtain these household listings. Because the 
households have been randomly drawn from the sample frame, the results of the study are
generalizable to all households in the SHOUHARDO II operational area. Note that villages included in the 
project were selected by CARE Bangladesh based on their vulnerability characteristics and therefore 
were not necessarily representative of the regions as a whole. 
The baseline sample size was computed to be able to detect a 10 percentage point difference in 
stunting prevalence across sample strata. This request for statistically representative comparisons 
across geographic sub-groups was from the SHOUHARDO II project, and resulted in an overall sample 
size much larger than required by FFP. An initial assumed value of stunting prevalence of 50% was 
used, along with a 95 percent confidence level, 80 percent power, and a design effect of 2.0 to estimate 
the minimum required sample of children 6-59 months old. Using these parameters yielded a minimum 
sample size of 606 children 6-59 months old per stratum. Following the FANTA Sampling Guide, the 
required number of households to be interviewed to ensure a minimum sample of 606 6-59 month old 
children was estimated on the basis of proportion of the population in Bangladesh that is aged 6-59 
months and the average household size. The national data (Population Census, 2001) revealed that 
children aged 6-59 months comprise 11.9 percent of the overall Bangladesh population and the average 
household size is five persons. Applying t these factors resulted in a minimum required sample 
[606/(0.119*5)] or 1,017 households per stratum. An additional 10 percent was added to this sample 
size in order to account for non-response, thus resulting in a final target sample size of 1,119 households 
per stratum. Applying this sample requirement across the eight cells (four regions by two areas) yielded 
a total target sample size of 8,952 households. 
In order to achieve this target sample size, the sample design employed a cluster design of 200 X 45. A 
total of 50 clusters (villages) from each of the four regions (25 villages from PM2A areas and 25 villages 
from MCHN/PEP areas within each region) were selected using probability proportionate to size (PPS) 
methods. In each of the selected villages, 45 households were randomly selected, without replacement 
from the village sampling frames developed by the SHOUHARDO II program for all villages in the 
intervetion areas. Table 1 provides details of the sample by program strata. 
8 
Table 1. Target sample size by program strata
Program strata Target # 
Households
Number of 
villages
Sample HH 
Region Approach per Village
Coast
MCHN/PEP 1125 25 45
PM2A 1125 25 45
Haor
MCHN/PEP 1125 25 45
PM2A 1125 25 45
Mid Char
MCHN/PEP 1125 25 45
PM2A 1125 25 45
North Char
MCHN/PEP 1125 25 45
PM2A 1125 25 45
Total Sample 9000 200 9000
This same sample size was used for the endline survey in order to have comparable levels of precision of 
estimates between the survey rounds, and to allow comparisons across the survey rounds. 
This sample size is much larger than the required minimum size to be able to detect a change of 10 
percentage points in the prevalence of stunting from baseline to endline at the level of the program as a 
whole. The minimum sample size required per survey round is computed using the following formula: 
n = [(Zα + Zβ)
2
 * {P1(1-P1) + P2(1-P2)}/(P2-P1)
2
] * D * Nf
where: 
n = required minimum sample size per survey round or comparison group (strata) 
P1 = stunting rate at baseline, 61.7% = 0.617
P2 = the expected level of stunting at endline for the program area such that the quantity (P2 - 
P1) is the size of the magnitude of change it is desired to be able to detect, SHOUHARDO II life of 
award (LOA) target reduction of 10 percentage points , 51.7% = 0.517
Zα = the Z-score corresponding to the degree of confidence with which it is desired to be able to 
conclude that an observed change of magnitude (P2-P1) would not have occurred by chance (α - 
the level of statistical significance for one-tailed test), 95% = 1.645 
Zβ = the z-score corresponding to the degree of confidence with which it is desired to be certain 
of detecting a change of magnitude (P2-P1) if one actually occurred (β - statistical power), 80% = 
0.840.
D = Actual SHOUHARDO baseline design effect for stunting = 2.43 
Nf
 = Non-response factor (assuming a 10% non-response rate) = 1.10
Based on these parameter values, the estimated minimum sample size (n) is 729 children 6-59 months 
old. Considering that not all households have children 6-59 months, the sample size of households must 
be adjusted according to the Addendum to Fanta Sampling Guide to ensure that a sufficient number of 
9 
children 6-59 months old are included in the sample.20. Assuming that the average number of children 
6059 months old in the population is 11.5 percent and the average households size is 4.621, the total 
number of households required to be interviewed to reach 729 U5s is 1,639 households.22 This sample 
size is adequate to detect a reduction of 10 percentage points ( from 61.7% to 51.7%) in the stunting 
rate of children 6-59 months old at the program level (LOA target in IPTT). The actual target sample size 
of the endline survey of 9,000 households is substantially larger than this minimum required sample 
size. 
The anthropometric data were collected following FFP guidance that data be collected from all children 
under five in a household rather than only one randomly-selected index child as had been the method 
followed for the baseline and midterm surveys. However, for comparability to the baseline and 
midterm results, anthropometric measures for the index child in the endline survey were used in this 
analysis. Indicators computed from all under-5 children in the sampled households are reported in 
Annex 6. 
The analytic sample (n=7,148) did not meet the target number of households estimated for 
MCHN/PEP.23 Nevertheless, significant differences in the anthropometric indicators were detected in 
the ex post analysis. Because the sample size calculation was based on the required number of under￾five children for the main stunting indicator, the number of under-two children was inadequate to be 
able to detect threshold differences for this age group by region for some of the MCHN indicators. 
Data treatment and quality assurance: Editing and processing of the completed paper questionnaires 
was done by 17 editors to verify that they were correctly filled in and completed, the correct numbers of 
households per village were interviewed, and the items of information recorded were consistent in each 
questionnaire. A senior professional staff member with Mitra then checked 10 percent of edited 
questionnaires after verification. This system ensured high quality data processing during the editing 
work. 
For questions with pre-coded categorical responses, a possible response of ‘other’ was also permitted, 
and interviewers wrote in specific responses when the ‘other’ code was selected. These written
responses for ‘other’ were translated from Bangla to English and then post- coded by creating additional 
code values that were added sequentially to the pre-coded responses. Translation, post-coding and 
editing were simultaneous tasks, beginning two weeks after the start of data collection. 
Data entry and cleaning: Data processing was conducted by trained staff of Mitra and Associates. The 
team consisted of one computer programmer, one data entry supervisor and ten data entre operators. 
Completed questionnaires were brought to the Mitra office where a registration officer (questionnaire 
administrator) logged them prior to entry by the data-entry team. All data were entered using CSPro 
with double entries, resolving discrepancies (100 percent verification) through referral to the original 
 
20 Stukel, Diana & Deitchler, Megan. Addendum to FANTA Sampling Guide by Robert Magnani (1999): Correction to Section 
3.3.1 Determining the Number of Households that need to be Contacted. March 2012. 
21 From SHOUHARDO II Baseline.
22 All U5s in a selected household were measured for anthropometric indicators. The estimate for the proportion of children U5 
per household is consistent with the baseline sample and data from the most recent Demographic and Health Survey (DHS). 
23 Target: 1,017 households; MCHN/PEP Analytic Sample: North Char=794, Mid Char=609, Haor=679, Coast=863 
10
questionnaire. Data editing included a set of checks for internal consistency. The clean dataset was 
converted into SPSS database for tabulation and analysis. 
Data analysis: TANGO staff conducted the analysis using STATA and SPSS statistical software. Sampling 
weights were applied in the calculation of all statistics in order to compensate for unequal probabilities 
of selection and for non-response, both of which varied across the eight survey strata (Annex 4).
Weights were needed at two levels of analysis: the household level (for data collected in Parts I and II of 
the questionnaire) and the child level (for that collected in Part III). Following the procedures applied in 
the baseline and midterm surveys, the analytic sample included eligible program participants by 
excluding households in the MCHN/PEP component that were not poor or extremely poor. The analytic 
sample was thus reduced from 8,415 to 7,148 households. This was a reduction of 1,267 households. 
Throughout the full report, all statistical significance compares endline values to baseline values. If an 
endline value in a figure has an asterisk next to it, this indicates that endline value is statistically 
significant compared to the baseline value of that indicator (not to the Life of Award (LoA) target).24
Sampling weights were used to correct for the fact that different proportions of households were 
selected from different sampling strata. The weights were computed as the inverse of the proportion of 
the sampled households to the total number of households within the strata. The weights were 
normalized to a provide an overall weighted sample size equal to the actual sample size. The weighting 
was based actual responses, so accounts for non-response. 
Limitations: The analytic sample (n=7,148) did not meet the target number of households estimated for 
MCHN/PEP.25 While the target number of households needed to adequately detect specified threshold 
differences in under-five children across comparison groups was not met, significant differences in the 
anthropometric indicators were detected in the ex post analysis. Additionally, because the sample size 
calculation was based on the required number of under-five children for the main stunting indicator, the 
number of under-two children was inadequate to be able to detect threshold differences for this age 
group by region for some of the MCHN indicators. Furthermore, for certain indicators like exclusive 
breastfeeding, the sample size was found to be too small to detect the difference between the baseline 
results and the LoA target. To detect a change from 62.2 percent at baseline to 66.2 percent for the 
target would require a sample size of 3,542 mothers. 
1.5. Household participation in interventions 
Household participation in SHOUHARDO II interventions focused heavily on SO1 and SO2 (Table 2). 
About two-thirds of households took part in SO1. The encompassing livelihoods focus with activities 
targeted at both men and women accounts for this. Coast and Haor households participated at a 
marginally higher rate than those in Mid Char and North Char. SO2, which included some non-poor 
households, had the greatest participation (See Section 1.3). The proportion of households participating 
ranged from 70 to 75 percent, regionally. 
 
24 Refer to Annex 5 for a comparison of endline results to LoA targets. 
25 Target: 1017 households; MCHN/PEP Analytic Sample: North Char=794, Mid Char=609, Haor=679, Coast=863 
11
SO3 also included a notable proportion of households, though the drop-off is clear. Just over one-fifth of 
households in Coast, Mid Char and North Char were included in SO3, while just 13 percent in Haor 
participated. 
Table 2: Percent of households participating in SHOUHARDO II interventions under SO1, SO2, 
SO3, by region
All Coast Haor
Mid 
Char
North 
Char
SO1: Availability of and access to nutritious foods 60.9 67.4 64.6 58.6 57.5
SO2: Improved health, hygiene and nutrition of children under 
2 years of age 72.8 70.2 72.6 70.6 74.6
SO3: PEP women and adolescent girls empowered 17.5 22.4 13.4 19.4 20.5
SO4 and SO5 were both generally more focused on community-level or higher interventions, and 
household-level information does not provide a complete picture of the degree of participation in these 
SOs by other key stakeholders. 
2. Results/ Program Effectiveness 
This section presents summary results for several indicators. Each SO is introduced with a brief 
description of activities. This is followed by discussion and bar chart figures comparing baseline, 
midterm, and endline analysis for impact, outcome and output indicators. Each SO also includes a 
discussion and break down of differences between male-headed households and female-headed 
households. SO4 and SO5 include discussion comparing PEP and non-poor households. A short 
summary of main findings concludes each segment. More detailed quantitative tables are in Annex 6. 
2.1. SO1: Availability of and access to nutritious foods 
2.1.1. Description of project activities 
SO1 sought to enhance 370,000 PEP households’ access to and the availability of nutritious foods by 
improving and diversifying household agriculture systems and increasing household income. The SO1 
program component delivered various sizes of packages of inputs for various types of livelihood 
activities (agriculture, fisheries, homestead development, other income-generating activities) and 
training to COGs and target households. Non-poor community members and PEP households received 
the same training and technical assistance) but only PEP households received input packages, which 
varied by PEP status. The SO1 also included self-help savings groups. 
2.1.2. Progress towards key impact indicators
The months of adequate household food provisioning (MAHFP)26 and household dietary diversity score 
(HDDS)27 are the two main indicators for measuring household food access.28 The number of MAHFP and 
 
26 MAHFP is a measure of a household’s ability to manage vulnerability in such a way as to ensure that food is available above a
minimum level for up to one year. The indicator is a self-reported number of months by the person in the household who is 
responsible for preparing the food. Source: Bilinsky, P. and A. Swindale. 2010.
27 HDDS is measured on a scaled of 0-12. It is a measure of the diversity of the diets of members of a household over a given 
time. Twelve food groups are included in the measure: cereals, roots and tubers, vegetables, fruits, meat/poultry/ offal, eggs, 
12
the average HDDS increased significantly across the baseline, midterm and endline surveys in all four 
regions and for both male- and female-headed households (Figure 2 and Figure 3). On average, 
households had enough food to meet their family’s needs for 11 months out of the year, up from about 
10 months at midterm and six months at baseline; little variation was observed between regions (Figure 
2). 
The HDDS nearly doubled from baseline to endline for the project participant group as a whole (Figure 
3). By endline, the measure ranged from 8.1 in the Coast region to 9.0 in Haor (on a scale of 1-12). This is 
an improvement from an average score just under five at baseline. Interestingly, MAHFP saw the 
greatest increase between baseline and midterm, while HDDS showed larger improvement from 
midterm to endline. Notably, both HDDS and MAHFP endline values are well above the LoA targets of 
six and eight, respectively. 
Figure 2: Number of MAHFP by region and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Figure 3: Averaged HDDS, by region and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Consistent with the measured increases in MAHFP and HDDS, the prevalence of households 
experiencing hunger, as well as each of the three experiential indicators used to measure hunger (no 
 
fish and seafood, pulses/legumes/ nuts, milk and milk products, oils/fats, sugar/honey, and miscellaneous. Source: Swindale, A., 
and Bilinsky, P. 2006. 
28 Bilinsky, P. and A. Swindale. 2010. FANTA III 
13
food in the house, go to sleep hungry, any household member went entire day without eating), 
decreased significantly (Figure 4). The percentage of households with hunger dropped to less than half 
(42 percent), down from 91 percent at baseline. Likewise, 17 percent of households reported having no 
food in the house while more than three-quarters (78 percent) reported this at baseline. The percentage 
of households with a member going to sleep hungry dropped from 61 to 11, and that with a member 
who skipped an entire day of eating fell from 43 to 7. Overall, these indicators point to a marked 
improvement in household food security conditions within the SHOUHARDO program area from 
baseline to endline. 
Figure 4: Indicators of hunger in the last four weeks
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.1.3. IR 1.1 Improved and diversified agriculture systems 
Though some fluctuation exists between midterm and endline, several descriptive indicators providing 
background information on agricultural production improved significantly from baseline (Figure 5). In 
particular, the percentage of households growing vegetables increased to 63 percent from 17 percent at 
baseline. Field crop and livestock production also gained popularity (44 percent and 86 percent, 
respectively), while fish production increased to nine percent of households. 
Figure 5: Percent of households engaged in agricultural activities
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
14
Looking at progress made in key SO1 outcome indicators regarding technology adoption, the proportion 
of field crop producers adopting at least three new technologies more than doubled between the 
baseline and endline, rising to 93 percent (Figure 6). The percentage of Coast producers adopting new 
technologies more than quadrupled to 84 percent of producers, with a large part of that increase 
between baseline and midterm. All other regions increased adoption more than 100 percent. Both male￾and female-headed households experienced significant increases in technology adoption, although 
increases were greater for male-headed households. Among the four practices for which data were 
collected, adoption rates for every one increased significantly. Regarding specific practices, 
intercropping increased the most from three percent at baseline to 30 percent, followed by using 
recommended seed storage practices, which increased from two percent at baseline to 21 percent at 
endline (Annex 6). Other practices for which marked increased was observed include: maintain proper 
spacing (331 percent increase), use of integrated pest management (244 percent increase), and use of 2-
3 seedlings per hill for rice (215 percent increase). 
Figure 6: Percent of field crop producers adopting the minimum number of technologies (three), by 
region and by sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Vegetable producers saw an almost identical significant increase in adoption of at least three new 
technologies to more than 90 percent of all vegetable producing households at the endline. Figure 7 
shows that in contrast to field crops, female-headed households increased adoption of technologies at a 
higher rate compared to male-headed households.29
Figure 7: Percent of vegetable producers adopting the minimum number of technologies (three), by 
region and by sex of household head
 
29 This study did not collect information on yield. Annex 6 includes tables on the percentage of vegetable producers that 
adopted at least three improved technologies and the average number of vegetable crops produced in the last year. 
15
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.1.4. IR 1.2 Increased household income among PEP
Strong and significant gains in income were observed in every group (Figure 8). Overall, monthly per 
capita household income grew more than 50 percent to 981TK. Coast households reported the highest 
monthly income per capita (1,083TK) and the largest increase from baseline (87 percent). Conversely, 
Haor households experienced the smallest increase, 42 percent. Male- and female-headed households 
saw income increases of 39 percent and 62 percent, respectively. 
Figure 8: Mean monthly household income per capita (in TK), by region and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
The pattern for ownership of domestic assets mirrors that seen for income (Figure 9). Overall, survey 
households’ domestic asset index30 climbed 56 percent. All groups saw significant increases in the 
domestic asset index. Regionally, Haor experienced the largest increase (62 percent), but maintained the 
lowest overall value at endline (37.4). 
Figure 9: Index of domestic assets, by region and sex of household head
 
30 Domestic asset index is a proxy indicator for measuring livelihoods and food security status. This index generally included 
non-productive assets, such as jewelry and households goods (e.g. furniture, radios, TVs, etc.). Each asset is assigned an 
associated weight based on its relative value. Each item owned in a household is multiplied by its weight, then values for all 
assets are summed to generate the index.
16
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Figure 10 reports on four indicators of market access, including: 1) percent of households that purchase 
food for whom the nearest food vendor is close by (<30 minute walk), 2) percent of households that sell 
food for whom the nearest point of sale is close by, 3) percent of households that purchase agricultural 
inputs for whom the nearest vendor is close by, and 4) percent of households accessing agri-inputs and 
financial services. While each indicator increased significantly from baseline, growth from midterm to 
endline is less marked (and actually decreased for one indicator). By endline, nearly all farmers (98 
percent) reported access to agri-inputs or financial services, which is a key SO1 outcome indicator. 
However, this measure decreased among female-headed-households, though the result is not 
statistically significant. Also of note is a highly significant drop (18 percent) among Coast households for 
percentage of households that purchase food for whom the nearest food vendor is close by (Annex 6). 
Figure 10: Indicators of access to markets and services, percent of households at endline
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.1.5. Differences in progress made for female- and male-headed households 
Both male- and female-headed-households showed large significant gains in key indicators (Figure 11). 
Male-headed households generally saw slightly larger increases from baseline to endline. Among field 
crop producers, male-headed households increased adoption of at least three new technologies twice as 
much as females. Meanwhile, increase in adoption of new technologies among vegetables producers 
was substantially higher for female-headed households than male-headed households, primarily 
because these activities were targeted at women. Female-headed households increased MAHFP almost 
twice as much as male-headed households, with both groups reporting about 11 months at endline. 
Finally, the percent of female-headed households that accessed agri-inputs, finances services decreased 
from baseline (though the change was not statistically significant), while it increased for male-headed 
households. 
17
Figure 11: Percent difference from baseline to endline of key livelihood indicators by sex of household 
head
Blue bar: % change from baseline to endline, Male HH Red bar: % change from baseline to endline, Female HHH
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Note: For ‘Percent of households accessing agri-inputs or financial services’, female-headed-households decreased 2.8 percent 
from baseline to endline. This change was not significant.
2.1.6. Impact and outcome activities by participation in SO1 activities 
The midterm and endline QPE surveys collected information about household participation in project 
interventions. Table 3 provides information about key SO1 impact and outcome indicators broken down 
by households that did and did not participate in SO1 interventions from the midterm and endline 
rounds. In particular, participants are households that reported having received either training or 
training plus inputs for livelihood activities supported by SHOUHARDO II. The results from this table 
show that both impact and outcome indicators increased substantially for both participants and non￾participants. Interestingly, the average per-capita income of participants was just over 90 percent of 
that of non-participants at the midterm, but by the endline, the income of participants had relatively 
more than non-participants so that their income levels were actually slightly higher than those of non￾participants. 
Adoption of improved technologies by participants and non-participants shows generally similar 
patterns of change. At the time of the mid-term the rate of adoption of technologies was higher for 
participants than non-participants for field crops, vegetable production, and livestock production. 
However, the rate of increase of adoption of non-participants from med-term to endline was higher 
than for participants. There are at least two possible interpretations of these results. The first is that the 
adoption of technologies by non-participants is the result of demonstration effects from participants to 
non-participants, with some time lag. This interpretation is supported by the fact that the adoption rates 
at the mid-term are substantially lower for non-participants, but by the time of the endline adoption by 
18
non-participants is closer to that of participants. An alternative explanation is that adoption of these 
practices are the result of factors other than the SHOUHARDO interventions. A more study into the 
reasons for adoption of recommended technologies by both participants and non-participants is needed 
to more fully understand these trends. 
Table 3: S01 key indicators , by S01 program participation
Indicator Midterm Endline Percent Difference 
(Endline-Midterm)
Number of observations
Midterm Endline
MAHFP by participation in S01 activities
Participant 9.8 11.0 12.1% *** 2,909 5,062
Non-participant 10.1 11.1 10.1% *** 4,007 2,082
HDDS by participation in S01 activities
Participant 5.1 8.7 71.8% *** 2,909 5,062
Non-participant 5.1 8.4 65.8% *** 4,007 2,082
Monthly per capita income (2010 Taka) by participation in S01 activities
Participant 791.8 998.7 26.1% ***
Non-participant 866.5 978.5 12.9% ***
Adoption of three or more field crop production technologies by field crop program participation
Participant 33.4 60.3 80.8% *** 714 978
Non-participant 17.7 37.2 110.2% *** 6,202 6,168
Adoption of three or more vegetable production technologies by vegetable crop program participation
Participant 68.7 69.3 0.9% 898 2,499
Non-participant 41.9 51.4 22.6% *** 6,018 4,649
Adoption of three or more livestock production technologies by livestock program participation
Participant 12.8 41.2 222.7% *** 1,380 2,857
Non-participant 8.7 36.1 314.6% *** 5,536 4,291
Note: Stars indicate difference is statistically significant at the 10% (*), 5%(**) or 1%(***) levels.
2.1.7. Summary of progress toward SO1 
SO1: “Availability of” and “access to” nutritious foods enhanced and protected for 370,000 poor & 
extreme poor households. 
 At endline, almost all households reported nearly 11 months of adequate food provisioning, 
compared with only about six months at baseline. 
 Likewise, HDDS was improved for all households, with female-headed increasing the most of all 
sub-groups. 
 Regionally, Coast and Haor households experienced the largest increased in MAHFP, while Haor 
improved the most in HDDS. 
19
 While indicators of overall hunger decreased significantly between the baseline and endline, 
household food insecurity was still an issue that needs addressed, as more than 40 percent of 
households reported hunger at endline. 
 Regarding technology adoption31, both field crop producers and vegetable producers actively 
adopted new technologies. Regionally, North Char households showed the largest increase in 
adoption of at least three technologies in both field crops and vegetables. The remarkable 
increase in adoption among Coast field crop producers was not reflected among vegetable 
producers in the region. 
 Monthly income grew substantially, with most of the increase coming between baseline and 
midterm. 
 All indicators for access to markets and services significantly improved since baseline. Nearly all 
farmers (98 percent) reported access to agri-inputs or financial services by the time of the 
endline, though only about half (52 percent) reported access to the nearest point of sale. 
 Female-headed households were far more likely to adopt technologies related to vegetable 
production while male-headed households adopted field crop technologies at a greater rate. 
Both groups improved significantly across almost all indicators. 
 Adoption of improved technologies has increased substantially from midterm to endline by both 
participants and non-participants in SO1 interventions. 
2.2. SO2: Improved health, hygiene and nutrition of children under 2 years of age 
2.2.1. Description of project activities 
SO2 has built on the best practice of SHOUHARDO I, which delivered a comprehensive package that 
positively impacted the health, hygiene and nutrition of children under two and pregnant and lactating 
women. The impact was made through the results of increasing access to and utilization of health and 
nutrition services, and through health and nutrition behavior changes among target PEP households. 
The package included: monthly food ration; promotion of infant and young child feeding (IYCF) 
practices; linking preventative and curative health and nutrition services; building capacity for 
community based integrated management of childhood illnesses (C-IMCI); facilitating linkages with the 
Ministry of Health and Family Welfare (MoH&FW) program to provide micro-nutrient supplementation 
for pregnant and lactating mothers; screening and referral for community based management of acute 
malnutrition (CMAM); and promotion and provision of water and sanitation infrastructure (identified by 
partners and VDCs). CHVs were trained and recruited to deliver most of the interventions through 
community-based GMP sessions and court yard sessions, home visits, and referrals. PNGOs were also 
responsible for making referrals and linkages with MoH&FW programs for the Expanded Program on 
Immunization (EPI), and vitamin A and iron folate supplementation. 
 
31 The QPE used the IPTT definition of this indicator, which is defined as adoption of three improved technologies. 
20
2.2.2. Progress towards key impact indicators 
LoA targets for all malnutrition indicators by endline were met with the exception of underweight 
females. The target for underweight females under the age of five years was set at 34.6 percent and at 
endline, 36.4 percent of females were underweight. However, when the target and endline values were 
tested for significance, it was found that the difference of 5 percent was not significant. Underweight 
male children under five years of age, on the other hand, far exceeded their LoA target of 42.2 percent 
at endline by 19 percent (endline value of 34.4 percent). For stunting rates, targets set for all children 
and those disaggregated by sex were achieved (target: 55.1 percent; females: 51.2 percent; males: 58.9 
percent). The overall endline results for stunting among children under five years of age was 48.8 
percent with females at 47.8 percent and males at 49.7 percent (Figure 12). The endline underweight 
target was also met for children under two, with 39.5 percent of children underweight, an 11 percent 
decrease from baseline to endline. 
Other LoA targets met included prevalence of diarrhea for children 6-23 months, percent of pregnant 
and lactating women taking iron supplements, and children 6-23 months of age who receive a minimum 
acceptable diet. The prevalence of diarrhea for children 6-23 months of age at endline was 8 percent, 
which exceeded the target of 10.3 percent. This may be attributed to increases in hygiene beliefs of 
mothers and improved access to water and sanitation (Figure 23, Figure 24, Figure 25, and Figure 26). 
The percent of pregnant and lactating women exceeded the target with a difference of 16 percent. Most 
impressive, however, was the percent difference between target and endline results for children 
receiving a minimum acceptable diet. The target was set at 20 percent and at endline, the program 
surpassed it at 48 percent, a 140 percent difference. LoA targets were not met for the following 
indicators: percent of children immunized against 8 diseases by 12 months of age, and number of people 
with improved access to sanitation facilities and improved access to drinking water. For exclusive 
breastfeeding and mothers who feel it is important to wash hands at five critical times, the LoA target 
and endline values were not significantly different and thus were considered meeting target. 
Malnutrition among under-fives: Children under five, overall, experienced significant reductions in all 
measures of malnutrition—stunting, wasting and underweight (Figure 12). The12.9 percentage point 
decrease in the stunting prevalence from baseline to endline is impressive especially when compared to 
BDHS data. Sunting rates SHOUHARDO II showed a yearly decrease of 3.2 percent32 while the BDHS data 
suggest that the national trend was only 0.5 percent per year between 2007 and 2011.33, 34 The 
reduction in the wasting prevalence was 22 percent from baseline to endline; however, the endline 
value of 12.2 percent was greater than the midterm value at 9.0 percent. The percentage of 
underweight children also increased slightly between midterm and endline, but decreased overall, with 
 
32 TANGO calculated using baseline and endline values divided by 4 years (dates from which the baseline and endline results 
were gathered). 
33 National Institute of Population Research and Training, Mitra and Associates & MEASURE DHS. 2013. Bangladesh 
demographic and health survey 2011. Dhaka, Bangladesh, and Calverton, MD, USA: National Institute of Population Research 
and Training, Mitra and Associates, and Macro International. 
34 National trend was calculated from 2007 to 2011, over a four-year time period, where there was only a decrease from 43 to 
41 percent.
21
an annual reduction of 1.7 percent per year.35 Overall, most gains were made for all malnutrition 
indicators by the midterm. Differences in the patterns of change in child malnutrition indicators by 
programming approach are discussed in section 3 below. 
When comparing male and female children under five, males saw greater reductions in stunting and 
underweight whereas female children saw greater reductions in wasting. For both male and female 
stunting, the difference from baseline to endline was significant (25 percent and 15 percent, 
respectively). Wasting was only significant for female children with a reduction of 28 percent (15.4 
percent at baseline to 11.1 percent at endline). However, for underweight children, only males saw a 
significant reduction of 27 percent (Annex 6).
When data for child malnutrition was disaggregated by age of mother (15-25 years and 26-49 years), it 
was found that both age groups had significant reductions in stunted and underweight children from 
baseline to endline. No statistical differences were found for wasting (Annex 6). 
Figure 12: Percent of children under five stunted, wasted and underweight, by sex
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Decreases in malnutrition prevalence varied regionally depending on the malnutrition indicator. For 
stunting, all regions saw significant decreases ranging from 11 to 37 percent from baseline to endline. 
(Figure 13). In comparison, significant decreases for wasting were only found in the Coast and Hoar 
regions, with percent decreases at 44 and 48 percent, respectively (Figure 14). For underweight 
children, decreases were significant in the Coast, Haor and North Char regions. (Figure 15). Notably, the 
Coast and Hoar regions showed significant decreases in all three indicators. 
 
35 This is comparable to similar declines for the country as a whole between 2007 and 2011 with an annual percentage-point 
reduction of 1.3. Source: National Institute of Population Research and Training, Mitra and Associates & MEASURE DHS. 
22
Figure 13: Percent of children under-five stunted, by region36
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Figure 14: Percent of children under-five wasted, by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Figure 15: Percent of children under-five underweight, by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Malnutrition among under-twos: Children under two saw significant reductions from baseline to 
endline in stunting, but not for wasting or underweight (Figure 16). This reduction may be partially 
attributed to the improved nutrition, nutritional status and dietary diversity of mothers, which can 
affect children’s birth weights and mother’s breast milk quality. The prevalence of underweight among 
mothers of children under five declined from 37 to 31 percent. 
 
36 Figure 13 includes all children moderately stunted (<2SD). Haor region did experience a significant decrease (p=.001) in 
stunting among severely stunted (<3SD) children. (Annex 6) 
Figure 16: Percent of children under-two stunted, wasted and underweight
23
For wasting, although the overall reduction was not significant from baseline to endline, an increase was 
noted from midterm to endline. Likewise, the greatest reduction for wasting was seen from baseline to 
midterm at 6.5 percent. This reduction was found to be significant. 
When comparing values across children under 5 and children under 2, it was found that the only 
significant indicator for both age groups was stunting. Decreases in stunting rates were greater for 
children under 2 years of age with a percent difference of 23 percent from baseline to endline. The 
percent difference for children under 5 years of age was 21 percent. Data were further disaggregated by 
age of mother. It was found that both younger mothers (between the ages of 15-25 years) and older 
mothers (between the ages of 26-49 years) were statistically significant for reducing stunting in children 
under two from baseline to endline. Wasting and underweight were not significant for either mother 
age groups (Annex 6). 
Caregiver’s nutrition: Significant reduction in underweight women was observed for all households. 
Regionally, the Coast, Mid Char and North Char (Figure 17) also saw significant reductions. Further, the 
quality of mothers’ diets has improved as well: overall Mothers’ Dietary Diversity Score increased 82 
percent to 8.4. 
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.. 
Figure 17: Percent of mothers of children under five years old underweight, by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
24
Prevalence of Diarrhea: The percentage of children age 6-23 months with diarrhea declined significantly 
to less than 10 percent. This may be one reason why stunting has seen significant declines for children.37
However, Figure 18 shows that decreases in diarrheal incidences were only seen in two regions, Coast 
and Mid Char. 
Figure 18: Percent of children 6-23 months with diarrhea in the last two weeks, by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.2.3. IR 2.1 Access to and utilization of health and nutrition services improved 
Immunization and ORS: Figure 19 shows significant gains in the percentage of children immunized 
against eight diseases. However, these gains were slightly less than those achieved at midterm. 
Improvements have also been made in the percentage of children age 6-23 months receiving ORS for 
diarrhea since baseline. This indicates that SHOUHARDO II has been successful at linking children to the 
MoH EPI program and basic health services. 
Figure 19: Immunization and ORS for diarrhea among children under two
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Although the percent of children immunized did not meet the LoA target of 82 percent, some regions 
performed better than others. When the data was disaggregated by region immunization rates for 
children under two saw significant increases in the Mid Char and North Char (Figure 20). However, it is 
important to note that the greatest improvements were seen from baseline to midterm for all the 
 
37 A 2008 meta-study of data collected from nine countries, including Bangladesh, confirms that a higher cumulative burden 
(i.e., a greater number of episodes) of diarrhea increases the risk of stunting in this age group (see Checkley et. al. 2008 Int J 
Epidemiol 37(4): 816–830). 
25
regions with the exception of Mid Char. The Coast, Hoar, and North Char saw decreases from midterm 
to endline. 
Figure 20: Percent of children immunized against 8 diseases by 12 months (12-23 months), by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater
Caring practices for mothers: Significant improvements were achieved in caring practices for mothers. A 
substantially higher percentage of women received antenatal care, additional food and rest during 
pregnancy, and iron/folic acid38 at endline than at baseline (Figure 21). This shows that key messages 
communicated by CHVs during GMP and courtyard sessions led to behavior change. Despite these gains, 
the percentage of women receiving care in a medical facility showed no significant gains from baseline 
to endline. When disaggregated by region, however, data showed significant gains for mothers receiving 
antenatal care in a medical facility in the Coast, Hoar, and Mid Char regions. Other data showed 
significant improvements across all four regions for mothers receiving antenatal care, taking in more 
food and rest during pregnancy, and taking iron/folic acid (Annex 6). 
Figure 21: Caring practices for mothers, percent of women utilizing practice or receiving service
† Pregnant and lactating women
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.2.4. IR 2.2 Caregivers of children under 2 adopt improved health, hygiene and nutrition behavior and 
caring practices 
Figure 22 shows that the quality of caring practices for children under two has improved for five of the 
six indicators of nutrition-related caring practices for children from baseline to endline. The strongest 
 
38 Pregnant and lactating women.
26
improvements were found in the area of diet and food consumption. The percentage of children 
receiving monomix or other vitamin sprinkles increased from 3 percent at baseline to 32 percent at 
endline. Likewise, the minimum dietary diversity39 and minimum acceptable diet40 indicators also 
showed similar large increases. The percentage of breastfed children with a minimum acceptable diet 
increased from 9 percent to 48 percent. However, there is no evidence of gains in exclusive 
breastfeeding for children under six months; rather, there was a slight, statistically insignificant decrease 
from baseline to endline. Additional information is needed to understand why exclusive breastfeeding 
rates did not see an increase similar to other indicators under child care practices. In the midterm 
evaluation, this was also identified as an area that needed further evaluation with only small, 
insignificant gains in exclusive breastfeeding percentages. With an endline value of 62 percent of 
children under 6 months exclusively breastfed, SHOUHARDO II did not meet its end target of 66.2 
percent. However, it is important to note that the sample size used in this analysis for breastfeeding was 
too small to detect the difference between baseline and target. To detect a change from 62.2 percent 
(baseline) to 66.2 percent (target), this would require a sample size of 3,542 mothers. Furthermore, 
when the target and the endline values were evaluated for this indicator, it was found that they were 
not significantly different. 
Figure 22: Nutrition-related caring practices for children under two for all participants (percent)
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Hygiene practices of mothers: Although SHOUHARDO II did not achieve its overall target of 35 percent 
of mothers who feel it is important to wash their hands at the five critical times,41 positive indications 
for each of the individual five critical times for hand washing saw significant increases. Improved hand 
washing has been correlated with lower rates of diarrhea, and this result is consistent with the declines 
in diarrhea prevalence documented over the same period (Figure 18). Note that the consistency does 
not hold at the regional level. Interestingly, those regions that had the smallest improvements in beliefs 
about hand washing (Coast and Mid Char) had the greatest decrease in diarrhea rates. 
 
39 The minimum dietary diversity indicator identifies whether a child has consumed at least four foods from the following seven 
food groups in the last 24 hours: grains, roots and tubers, legumes and nuts dairy products (milk, yogurt and cheese), flesh 
foods (meat, fish, poultry and liver/organ meats), eggs, vitamin-A rich fruits and vegetables, other fruits and vegetables. 
40 Proportion of children 6–23 months of age who receive a minimum acceptable diet (apart from breast milk). 
http://www.unicef.org/nutrition/files/IYCF_Indicators_part_III_country_profiles.pdf . 
41 The five critical times are: before eating, before breastfeeding or feeding a child, before cooking or preparing food, after 
defecation/urination, and after cleaning a child that has defecated/changing a child’s diaper.
27
Figure 23: Percent of mothers who feel it is important to wash their hands at five critical times by 
region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Similarly to hand washing, Figure 24 shows strong gains for the population as a whole and across all 
regions in the percent of children under three whose feces are disposed of safely. However, these values 
decreased between midterm and endline for all regions except the Coast. Across regions, the Coast saw 
the greatest increase (72 percent). This may have contributed to the significant decrease in diarrheal 
rates of children 6-23 months in the region. 
Figure 24: Percent of children under three whose feces are disposed of safely by region 
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Access to water and sanitation: Significant gains have also been achieved in access to clean water and 
sanitation facilities. Access to improved drinking water (Figure 25) has increased in all four regions. 
Although at baseline Coast had the highest percentage of households with access to improved drinking 
water, it saw the least improvement over time. North Char saw the greatest improvement, from 58 to 
81 percent. 
28
Figure 25: Percent of households with access to an improved drinking water source, by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Access to improved sanitation facilities remained lower than access to improved drinking water sources, 
but this indicator saw greater increases overall (Figure 26). Additionally, improvement in this indicator 
was strong across all regions. Mid Char saw the least improvement but had a higher percentage of 
households with improved sanitation facilities at baseline compared to the other regions. 
Figure 26: Percent of households with access to an improved sanitation facility, by region
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.2.5. Differences in progress made for female- and male-headed households 
Both male- and female-headed households saw significant reduction in underweight for children under 
five years of age. However, stunting was only significant for female-headed households and wasting was 
only significant for male headed households for this child age group. For children under two years of 
age, stunting had significantly reduced in both male and female headed households. Interestingly, 
wasting and underweight for under-two children increased among female-headed-households, though 
the difference is not significant and may be reflection of the small sample size. 
29
When comparing underweight mothers of children under five years of age, it was found that male￾headed households saw significant reductions from 37.0 percent at baseline to 29.9 percent at endline. 
However, female headed households, an increase in underweight was recorded but was not significant. 
As for the data available for child malnutrition indicators and for underweight mothers, it is important to 
note that the sample sizes for female-headed households were much smaller in comparison to male￾headed households and thus, harder to detect significant changes within this population. 
2.2.6. Summary of progress towards SO2 
 Generally the design of SO2 interventions was comprehensive and in line with global practice. 
LoA targets were met for most indicators. Those indicators for which targets were not met 
include improved access to water and sanitation, and immunization of children against eight 
diseases by 12 months of age. Significant regional differences were noted for these indicators, 
showing where improvements and additional targeting is needed. Nevertheless, endline values 
of these indicators were moving in the intended direction. Overall, when comparing baseline to 
endline, significant improvements in SO2 indicators were noted. 
 The midterm evaluation emphasized the need to focus on lagging indicators including hand 
washing, prevalence of diarrhea, and exclusive breastfeeding. Improvements were seen across 
most with the exception of exclusive breastfeeding. Further outreach to pregnant and lactating 
mothers was recommended at midterm; however, it is unclear if additional program efforts 
were made to address this. It should be noted that there was already a high percentage of 
children at baseline who were exclusively breastfed (64 percent) and in order to detect 
significant changes and to meet the LoA target for this indicator, a much larger sample size 
would need to be analyzed. 
Figure 27: Percent difference from baseline to endline, child anthropometric indicators, by sex of 
household head
Blue bar: % change from baseline to endline, Male HH Red bar: % change from baseline to endline, Female HHH
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
30
 For child malnutrition indicators, significant reductions in stunting and underweight were seen 
for children under five, whereas stunting was the only significant indicator for children under 
two. For children under five, it should be noted that the greatest gains for these same indicators 
were seen from baseline to midterm and leveled out or increased slightly between midterm and 
endline. 
 Proper disposal of feces and improved access to water supply and sanitation facilities saw 
significant improvements from baseline to endline across the four regions. It would be expected 
that diarrheal rates would show similar trends; however, only the Coast and Mid Char saw 
significant decreases in diarrheal rates. 
2.3. SO3: PEP women and adolescent girls empowered 
2.3.1. Description of project activities 
The objective of SO3 was for PEP women and adolescent girls to be empowered in decision making, to 
engage in activities to reduce food insecurity at various levels, among their families, communities and 
Union Parishad (UP), and to strengthen local support systems that reduce VAW. The interventions 
involved engaging village development committees (VDCs)42 to conduct a gender analysis and integrate 
gender-based issues into the community action plans. Additional community groups were formed in half 
of the 1,509 target villages at the midterm to address early education needs (e.g., ECCD centers) and 
promotion of girls’ enrollment in school. About one-third of the target villages also established EKATA 
groups at the midterm where women and adolescent girls with support of other community members 
defined empowerment and identified an action plan to address issues such as women’s mobility, 
domestic violence and early marriage 
2.3.2. Progress towards key impact indicators 
The percent of women with control over economic resources rose sharply in all groups (Figure 28). 
Across all survey households, the proportion more than tripled, a trend that was also seen by region. 
More than one-quarter of women in North Char reported control over resources, the highest of any 
region. The Coast region experienced the largest increase (339 percent), with one-fifth of women 
reporting control of resources at endline. Despite the significant gains, the indicator fell short of its LoA 
goal of 25 percent of women overall. The only groups to meet or exceed the goal were women in North 
Char and female-headed households (32 percent). 
 
42 According to the SHOUHARDO II Proposal: “VDCs are an umbrella for all different groups working on behalf of the PEP in the 
community, including EKATA, agriculture, commodity distribution and safety nets. They are also the primary link between the 
community and the Ward Savas, providing opportunities to institutionalize the role of VDCs in the governance process and to 
establish work practices that are inclusive and transparent.”
31
Figure 28: Percent of women with control over economic resources, by region and sex of household 
head
Note: Stars (*) indicate that difference between baseline and endline values are statistically significant at the 10% level. 
2.3.3. IR 3.1: Increased ability of women to influence decisions 
Following recommendations made in the midterm report, strong gains were made overall in indicators 
of women’s empowerment (Figure 29). Except for participation in a group, all key indicators improved 
steadily and significantly from baseline. The largest proportional improvement was in the percent of 
women earning cash income; this measure increased in every region by at least 200 percent and in 
Coast households by more than 300 percent (Annex 6).
Interestingly, the percent of women participating in any group43 was almost equal at baseline and 
endline, with large fluctuations in magnitude of variation among different groups: participation 
increased significantly in Coast households (82 percent) but dropped significantly in Mid Char (28 
percent); similarly, participation grew 37 percent in extreme poor households, but fell 40 percent or 
more in all non-poor groups. For this indicator, almost every group increased from baseline to midterm 
and decreased from midterm to endline (Annex 6). 
 
43 “Any Group” includes the following: savings or credit group, community agriculture or homestead garden or IGA group, 
community health group, parent-teacher association or school management committee, mother’s group, women’s support 
group, UP Special Committee, UP Standing Committee, UP disaster committee, or other. 
32
Figure 29: Indicators of women's empowerment
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Women’s decision making power was computed on as the average score of the following responses: 
4=woman can decide alone, 3=can decide with husband or male HH member, 2=husband makes 
decision after discussion with wife, 1 = woman not involved in decision; over 12 different types of 
decision. An overall score of 4 would mean that the woman can decide alone over all types of decision, 
while a score of 1 would mean that the woman is not involved in any type of decision. This indicator 
improved by small yet significant amounts across all categories (Figure 30). Gains from midterm to 
endline were substantially larger compared to the small increases – or decrease in the case of Mid Char 
– from baseline to midterm. 
Figure 30: Index of women's decision making power by region and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.3.4. IR 3.2: Men and women working together to end violence 
The trend in decreasing domestic violence seen at midterm continued through endline (Figure 31). At 
endline, the percent of women who believed a man was justified in hitting his wife dropped to less than 
15 percent for all situations except for one. Refusing to have sex saw the largest decrease and the 
lowest value at endline in women who believe it was justification for violence. Almost one-quarter of 
women still reported that when the wife does not obey elders, the man is justified in using violence, 
though this is down significantly from 61 percent of women at baseline; the decrease of 63 percent 
represents the smallest proportional decline, also, of any category. 
33
Figure 31: Percent of women who believe a man is justified in hitting/abusing the wife, by situation
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Likewise, actual incidents of domestic violence in the year preceding the survey fell significantly, with 
the majority of the reduction coming between baseline and midterm (Figure 32). Regionally, North Char 
households reported the largest decrease (88 percent), as well as the lowest percentage (four percent), 
while Mid Char saw a slight increase from midterm to 15 percent. Male-headed households also saw 
substantial improvement, falling from nearly one-third of households to just eight percent. In addition, 
the percent of households for which assistance was sought following an incident of domestic violence 
fell 44 percent from 14 percent at baseline to eight percent at endline (Annex 6). 
Figure 32: Percent of households in which female was yelled at or struck during the last year by region 
and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.3.5. Differences in progress made for female- and male-headed households 
Gains in women’s empowerment among all survey households were significant during the program 
period (Figure 33). Of note, the percentage of women earning cash income in male-headed households 
more than quadrupled from the baseline value; at endline, though, just 18 percent of these households 
reported this. Male-headed households saw greater increases than female-headed households in all 
other indicators except the percent of women participating in a group: among male-headed households, 
the proportion decreased from baseline, while households led by women increased 42 percent. 
34
Figure 33: Percent difference from endline to baseline of key women's empowerment indicators, by 
sex of household head
Blue bar: % change from baseline to endline, Male HH Red bar: % change from baseline to endline, Female HHH
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.3.6. Summary of progress toward SO3 
SO3: PEP women and adolescent girls empowered in their families, communities and Union Parishad 
 The percentage of women with control over economic resources increased significantly overall 
and in every sub-group but did not meet the LoA target. The largest increases came in the Coast 
region and in male-headed households, though at endline the highest percentages of women 
with control over resources were in female-headed households and in North Char region. 
 Women’s empowerment increased substantially in terms of decision-making power, freedom of 
movement and earning cash income. The largest improvement came in women earning cash 
income, particularly in male-headed households. Participation in a group fell off after midterm 
reverting to baseline values. 
 Women’s decision making power improved significantly in all categories. As expected, decision 
making power in female-headed households was highest at baseline and endline. 
 Beliefs about situations that justify domestic violence changed significantly, with fewer women 
believing a man is justified in all situations presented. At both baseline and endline “she does 
not obey elders” was the most likely reason women suggested men would be justified violence. 
 Related to beliefs, incidents of domestic violence decreased in every category, with less than 
one in ten households (seven percent) citing abuse in the previous year. North Char households, 
in particular, experienced a substantial decrease; this region reported the highest percent at 
baseline and lowest percent at endline. Conversely, Mid Char experienced a very slight increase 
from midterm and with the highest reported incidence at endline than other regions. 
 Nearly equal progress was made among both PEP and non-poor households in the women’s 
empowerment indicators. The notable difference related to seeking assistance after an incident 
of violence, in which PEP households were much less likely to seek help at endline. 
35
2.4. SO4: Responsiveness and accountability to the PEP of local elected bodies 
and government service providers 
increased 
2.4.1. Description of project activities 
As an innovative governance component, SO4 sought to 
increase the responsiveness and accountability of 
government to the needs of PEP households, particularly 
women. To this end, the program established Project 
Advisory and Coordinating Committees (PACCs) and 
VDCs, which linked SO4 with each of the other SOs 
(Figure 34). PACCs included ministry representatives and 
elected officials at various levels (e.g., national, 
divisional, district and upazila), joined by CARE program 
staff, functioning as a coordinating body for the 
government institutions involved in the program, as a 
platform for capacity building and trainings, and as a 
forum for policy discussion. The VDCs functioned to 
empower local communities to develop and implement 
community-driven action plans and to liaise between 
PEP households and government service providers or other representative bodies. 
2.4.2. IR 4.2: PEP access to entitlements and services increased, including safety nets and natural 
resources 
SO4 activities were largely community-based, however, certain indicators in the QPE were calculated for 
household participation in program activities; specifically those where resources were accessed, safety 
nets engaged, and common property resources utilized., Program participation in SO4 activities was 
extremely high for all surveyed households over the life of the project, from 98.5 percent of households 
in baseline to 99.9 at endline. 
All surveyed households utilized more services on average than at baseline (Figure 35).44 Regionally, 
Coast household saw the smallest growth, utilizing the least number of services at both baseline and 
endline, with little increase during the second half of the program period. In contrast, Mid Char and 
North Char used the most services of any region. Except for Coast households, the magnitude of change 
varied little for any category. At endline, the most common services included: primary health care (93 
percent of households), Union Parishad (87 percent), government family planning (76 percent), family 
planning services (76 percent), and Grammo Shalish (72 percent).45 Of the services included in the 
 
44 The three response categories for service utilization were 'frequently', 'sometimes', and 'never.' In the baseline and midterm 
calculations values were mis-coded: 'sometimes' and 'never' were coded as utilized, instead of 'frequently' and 'sometimes.' 
45 Government family planning and family planning services were presented as separate choices on the questionnaire. See 
Annex 6 
Figure 34: Role of SO4 and VDC in linking with 
other SOs
Community Mobilization and Facilitation
(Technical support, coordination of services, allocation of resources)
VDCs
(with Field 
Facilitators)
ECCD
COG
GMP/ 
Courtyard
EKATA
EVAW
DVs
Upazila
PACC UDMC
Capacity Building and Advocacy
(Inclusion of PEP, open budgeting, CDRM)
NBD/UP
/WS
36
survey, each one was used by at least one-third of households, with the Department of Fisheries and the 
Department of Women’s Affairs experiencing the largest increases in use (210 percent and 188 percent, 
respectively). 
Figure 35: Mean number of services utilized, by region, well-being category and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Despite remaining generally low, on average less than one type of safety net utilized (0.63) out of a total 
of 12 possible at the endline, the average number of safety nets increased significantly from baseline to 
endline (Figure 36). The largest regional increases came in Coast and Haor households. Female-headed 
households engaged in the most safety nets of any group, both at baseline and endline. As at midterm, 
the largest increase in any safety net was for government Vulnerable Group Feeding (VGF), which 
increased 880 percent to 40 percent of households. The most-used safety net, however, was 
SHOUHARDO II, reported by 76 percent of households.46
 
Figure 36: Average number of safety nets engaged in, by region, well-being category and sex of 
household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Note: Baseline and midterm have been revised for this data.
 
46 The endline survey included two safety nets, SHOUHARDO II and self-help groups, neither of which was in the baseline or 
midterm surveys.
37
Across all survey households, the average number of common property resources utilized (resources 
such as ponds, forests, embankments that are not owned by individuals but are managed and use rights 
are provided by communities or local authorities) more than doubled, though no group used more than 
two resources (Figure 37). The most notable increase was in Haor, which saw a nearly 400 percent 
increase in use. In contrast, Mid Char increased by 36 percent from baseline, and actually decreased 
slightly between baseline and midterm. PEP and male-headed households used the most common 
property resources. At endline, the most commonly used resources included river/canal (68 percent of 
households), beel/haor (63 percent), forest land (62 percent), and grazing land (60 percent). The largest 
increases in use came for Community-based Organization (CBO) Water Body, which grew in use 271 
percent, and railway grounds, which increased 149 percent. 
Figure 37: Average number of common property resource utilized by region, well-being category, and 
sex of household head47
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.4.1. Differences in progress made for the PEP and non-poor 
While both groups experienced increases in all areas, PEP households saw larger and more significant 
improvements for SO4 measures than did non-poor households (Figure 38). PEP households equaled the 
overall average gains for average number of services utilized and common property services utilized. 
The PEP household increase in safety net usage was below the overall average, though PEP households 
used more safety nets than average at both baseline and endline. 
 
38
Figure 38: Percent difference from baseline to endline, government and service indicators, by well￾being category
Blue bar: % change from baseline to endline, PEP Red bar: % change from baseline to endline, non-PEP
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.4.2. Differences in progress made for female- and male-headed households 
Male- and female-headed households experienced nearly identical significant increases in measures of 
access and use of services (Figure 39). Regarding use of safety nets and common property resources, 
slightly larger gains were observed in households led by men. The opposite is true for the average 
number of service utilized, though the difference between male-headed households and female-headed 
households in this category is minimal. 
Figure 39: Percent different from baseline to endline, government and service indicators, by sex of 
household head
Blue bar: % change from baseline to endline, Male HH Red bar: % change from baseline to endline, Female HHH
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.4.3. Summary of progress toward SO4 
SO4: Local elected bodies and government service providers’ responsiveness and accountability to the 
PEP increased.
 Because the quantitative analysis was primarily focused at the household level, discussion 
regarding progress toward SO4 is limited to indicators measured in IR4.2. This household-level 
39
indicator represents only a limited lens on the SO4 interventions, which are mostly directed at 
the level local level governments. 
 Overall survey households, regions, and both male- and female-headed households, significant 
increases were observed in the average number of services utilized, the average number of 
safety nets engaged in, and the average number of common property resources utilized. 
 No clear trends were observed for any one region – only that all regions increased substantially 
in every area. 
 The same is true for sex of household head: considerable increases were seen for both male￾and female-headed households. Male-headed households used accessed more services and 
utilized more common property resources at both baseline and endline. Meanwhile, female￾headed households on average engaged in more safety nets at both baseline and endline. 
2.5. SO5: Improved preparation, mitigation and response to disasters and 
adaptation to climate change 
2.5.1. Description of project activities 
SO5 was designed to achieve progress in disaster preparation and response and in climate change 
adaptations with the aim to build longer-term resilience among PEP households. This objective included 
the following interventions: reactivating Union Disaster Management Committees (UDMC, linked to 
SO4) and their oversight of community disaster volunteers and contingency planning; conducting 
participatory Climate Vulnerability and Capacity Analysis (CVCA) that guided adaption plans, and which 
applied to agricultural technologies promoted in SO1; improving early warning and forecasting systems; 
creating and maintaining disaster resistant infrastructure such as water and sanitation (linked to SO2), 
embankment, wave protection walls, cyclone and flood shelters, drainage culverts, transportation 
systems, etc.; strengthening humanitarian response capacity through pre-positioned goods; and 
engaging youth in Climate and Disaster Risk Management (CDRM) through school-based teen brigades 
(SBTB). 
2.5.2. Progress toward key impact indicators 
No significant change was observed in the percentage of all households or the percentage of PEP 
household distress selling (Figure 40). Both categories included 10 percent or less of households, with 
minimal variation regionally or by sex of household head. The largest fluctuation came in Haor district, 
with a 46 percent increase in households selling, while North Char experienced the only significant 
change – a 37 percent decrease in distress selling. The LoA goal of seven percent of PEP households 
distress selling was not met, however when the endline result was compared to the LoA target, no 
significant difference was found. 
40
Figure 40: Percent of households distress selling, by region, sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
The Taka value of distress sales decreased 19 percent overall to 16,231TK and PEP households 
decreased distress selling by 17 percent to 12,500TK, but no significance was observed (Annex 6). The 
only significant change was in Haor region, where the value of sales decreased 56 percent to 13,400TK. 
Additionally, the Hoar was the only region to exceed the LoA target. However, when comparing the Taka 
value of distress sells overall, it was found that the endline result was not significantly different from the 
LoA target of 15,000TK (Figure 41). 
Figure 41: 2010 Taka value of household assets sold to meet urgent household need
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Note: Two extreme cases (more than 4 times greater than the next largest value) of value of assets sold were excluded from the 
analysis as outliers.
2.5.3. IR 5.1: Disaster contingency systems in place and functioning 
Disaster experiences varied widely across categories (Figure 42). Overall, the percentage of households 
that experience a disaster in the year preceding the survey increased. This trend was true for three of 
the four regions and both male- and female-headed households. The most notable point was a drop in 
Coast households experiencing a disaster from 95 percent at baseline to just 16 percent at endline. 
41
More than half of households reported experiencing a flood and one-quarter of households experienced 
no disaster. Other disasters mentioned include heavy rains (24 percent) and wind storms (23 percent). 
Figure 42: Percent of households that experienced any disaster in the last 12 months, by region, well￾being category, and sex of household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
Little significance was observed in the average number of disasters experienced (Figure 43). On average, 
households reported to have experienced just over one disaster. As with the previous indicator, this 
measure was most notable in Coast households, which fell from two disasters to 0.2. Program activities 
led to greater knowledge of disaster mitigation strategies that could reduce impact of future disasters. 
The most frequently mentioned measures were structural improvements to the home (48 percent of 
households), food stocks (39 percent), and water stocks (18 percent). At both baseline and midterm, 
more than 40 percent of houses said no measure existed to reduce impact in the future. At endline, just 
seven percent of houses gave this response (Annex 6). 
Figure 43: Average number of disasters experienced, by region, well-being category, and sex of 
household head
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.5.4. Differences in progress made for female- and male-headed households 
Changes in disaster-related measures during the program period varied for male- and female-headed 
households. The significant changes included (1) a 17 percent increase for female-headed households 
42
that experienced a disaster and (2) a 13 percent increase in the number of disasters experienced for 
male-headed households. This latter measure, though significant, increased from 1.1 to 1.2 disasters. 
Figure 44: Percent difference from baseline to endline, experiential disaster management, by sex of 
household head
Blue bar: % change from baseline to endline, Male HH Red bar: % change from baseline to endline, Female HHH
Note: Stars (*) indicate that difference between endline and baseline value is statistically significant at the 10% level or greater.
2.5.5. Summary of progress toward SO5 
SO5: Targeted community members and government institutions are better prepared for, mitigate, 
and respond to disasters and adapt to climate change. 
 As in SO4, the household-level information relating to SO5 interventions offers only a very 
limited perspective on the overall impacts of the interventions in this area. It should also be 
emphasized that the household-level indicators are very sensitive to the degree to which 
households were exposed to shocks prior to the time of the survey. The quantitative analysis 
alone cannot offer much discussion of progress toward achieving SO5 without community-level 
indicators that demonstrate the progress of communities’ or local government institutions’ 
progress. 
 No change was seen in the impact indicators that measured the percentage of households 
distress selling or the value of these sales. Both measures varied widely across regions, with no 
clear trend. 
 The percent of households that experienced a disaster and the number of disasters experienced 
increased or remained about the same overall and for both male- and female-headed 
households. The only notable change was in the Coast region, where both indicators dropped 
sharply. 
 While the latter indicators describe experiences rather than behavior change, there was no 
association between experiencing disasters and distress selling. The percentage of Coast 
households distress selling increased while the same region saw a sharp drop in disasters. This 
suggested that those households that experienced disasters may use other coping strategies or 
that households distress sell for other reasons beyond experiencing disasters. 
43
3. Comparative effectiveness of the MCHN/PEP and PM2A programming approaches 
As noted in the introduction to this report, the SHOUHARDO II project was designed to incorporate a 
community randomized controlled trial (RCT) to evaluate the relative impact of the traditional 
SHOUHARDO MCHN/PEP approach and the PM2A approach. In review, both approaches follow the 
basic tenets of the PM2A approach that (a) the first “1000 days” is the window of opportunity for 
preventing malnutrition; (b) efforts should be focused on preventing rather than curing malnutrition; 
and (c) the entire food-care-health triad should address underlying causes of malnutrition. The only 
difference between the two approaches as applied in the context of SHOUHARDO II was that the 
MCHN/PEP approach employs socio-economic targeting while the PM2A approach did not. Additionally, 
because PM2A provided supplementary nutrition rations to both PEP and non-poor women and 
children, PM2A cost more to implement than MCHN/PEP strategy. 
Table 4 summarizes the design of the RCT. All PEP households, regardless of whether they were 
randomly selected into a MCHN/PEP or PM2A village, were eligible to participate in SHOUHARDO II
project activities, including MCHN activities (for those with a pregnant woman and/or child under-two) 
and activities falling under the other SOs. In the case of the non-poor, those in PM2A villages were only 
eligible to participate in the MCHN activities, which involved the receipt of supplementary nutritional 
rations, while those in MCHN/PEP villages were not eligible to participate in any project activity. 
Table 4: Design of the randomized controlled trial to evaluate the relative effectiveness of 
the MCHN/PEP and PM2A approaches
MCHN/PEP PM2A
PEP
 Eligible to participate in MCHN 
(SO2) and all other project 
interventions (S01, S03, S04, SO5) 
 Received supplementary nutritional 
rations
 Eligible to participate in MCHN 
(SO2) and all other project 
interventions (S01, S03, S04, SO5) 
 Received supplementary 
nutritional rations
Non-poor
 Not eligible to participate in any 
project interventions.
 Did not receive supplementary 
nutritional rations
 Eligible to participate in MCHN 
(SO2) interventions only
 Received supplementary 
nutritional rations
It is important to note that the idea behind the MCHN/PEP targeting approach is that if limited 
resources are directed towards those who need them most—the poorest—then greater improvements 
in measures of well-being such as children’s nutritional status can be expected. This basic principle 
cannot be tested in the present evaluation because the same amount of resources was not made 
available to both MCHN/PEP and PM2A villages. In fact we can hypothesize that the PM2A approach 
was more successful at reducing child malnutrition given that additional resources, such as food 
assistance, were extended to a larger percentage of households in PM2A villages.48
 
 
48 The non-poor make up 30 percent of households in the average SHOUHARDO II village. 
44
The randomized allocation of villages at the start of the project across the two programming 
approaches, in conjunction with the survey sampling design (which stratified by programming 
approach), should in principal allow us to evaluate their relative impacts through simple comparisons of 
changes from baseline to endline in malnutrition prevalence of households in MCHN/PEP and PM2A 
villages. To judge whether this comparison was valid, it was important to determine whether the 
programming approaches were successfully implemented as laid out in Table 4 and whether the 
randomization process actually worked to equalize the groups at baseline. These tasks were carried out 
in the next two sections. Note that all analyses were undertaken only for households for which 
anthropometric data are available, including households with children under-five years with valid data 
on age, height and weight of children.
3.1. Success in implementing the programming approaches 
Table 5 shows the percent of households participating in SHOUHARDO II interventions for MCHN/PEP 
and PM2A villages. Given the design of the RCT, we expected that participation prevalence was roughly 
equal for all SOs except S02, the latter for which participation should be higher in PM2A villages where 
eligibility was universal. Indeed participation in MCHN activities was higher in PM2A villages, 
particularly for receipts of food rations. Participation in SO1 interventions and savings groups and EDDC 
were roughly equal across the two sets of villages. However, participation by children’s mothers in 
EKATA groups was significantly higher in PM2A villages, and in fact there was a very large difference in 
the percent of villages with an EKATA group. Only 38 percent of MCHN/PEP villages had an EKATA group 
while almost 80 percent of PM2A villages did, which was surprising considering that EKATA groups were 
expected to be present in all villages. These findings suggest that PM2A villages were favored in the 
establishment of SHOUHARDO II EKATA groups. If these groups were having a positive impact on 
children’s nutritional status as they did in the SHOUHARDO I project, then this unequal allocation of 
project interventions would bias the results of the RCT towards PM2A villages. 
Table 5: Percent of households participating in SHOUHARDO II interventions, by programming approach
MCHN/PEP PM2A Difference
SO1 Availability of and access to nutritious foods
 Agriculture/income generation 66.2 60.3 -5.9
SO2 Health, hygiene and nutrition of children <2
 MCHN (any SO2 intervention) 72.9 88.2 15.3 ***
 Ever received a food ration 57.2 77.6 20.3 ***
SO3 Women's empowerment
 EKATA group (mother is member) 6.2 9.5 3.3 **
 Percent of villages with an EKATA group 38.3 78.6 40.3 ***
 Savings group (mother is member) 10.0 10.6 0.6
 ECCD 12.4 14.4 2.0
Note: Stars indicate that the difference is statistically significant at the 1(***) or 5(**) percent levels.
 
45
Table 6 gives a breakdown of participation prevalence for the PEP and non-PEP within the two 
programming approach groups. In the case of SO1 and SO3 interventions, the participation of the non￾poor in both groups was quite low with the exception of ECCD, which had near 10 percent participation 
in both. Consistent with the evaluation design, the participation in MCHN interventions of the non-poor 
in PM2A villages was quite high and roughly on par with that of the PEP. The participation of the non￾poor in the MCHN/PEP villages was low, yet a substantial percentage—30%—did participate in MCHN 
activities. This, too, was a departure from the design of the RCT and would favor MCHN/PEP villages in 
terms of recorded reductions in child malnutrition. 
Table 6: Percent of households participating in SHOUHARDO II interventions, 
by well-being category and programming approach
MCHN/PEP PM2A
PEP Non-PEP PEP Non-PEP
SO1 Availability of and access to nutritious foods
Agriculture/income generation 87.1 8.8 85.3 5.8
SO2 Health, hygiene and nutrition of children <2
 MCHN (Any SO2 intervention) 88.4 30.3 90.3 83.7
 Ever received a food ration 75.6 7.1 79.4 73.6
SO3 Women's empowerment
 EKATA group (mother is member) 8.1 1.1 12.3 3.6
 Savings group (mother is member) 12.9 2.1 13.4 4.6
 ECCD 13.7 8.9 16.1 10.8
SHOUHARDO II project
95.1 39.6 95.9 87.1
3.2. Success of randomization in equalizing MCHN/PEP and PM2A households at 
baseline 
Another factor that could bias the results of the impact evaluation was the presence of inequality in 
socio-economic or physical well-being of the MCHN/PEP and PM2A groups of households at baseline. 
The randomization process should in principal render these two groups equal in all respects. 
Table 7 presents data on household and village characteristics that either would not be expected to 
change over a four-year period or, in most cases, were measured at baseline. In terms of household 
characteristics, a substantially higher percentage of households in the PM2A than the MCHN/PEP group 
reside in the Coast region, and a lower percent in North Char. Households in the MCHN/PEP group were 
also disproportionately more likely to fall into the “extreme poor” socio-economic status group and less 
likely to fall into the “lower middle” group, that is, they were more likely to be poor. This difference was 
also reflected in the fact that 55.2 percent of MCHN villages were classified at baseline as “extremely 
vulnerable” while only 47.3 percent of PM2A villages were. Finally, although one indicator of 
remoteness (the percent of villages for which the nearest town was more than a 1 hour walk) was 
equalized across the groups, another (the percent of villages for which the nearest district headquarters 
46
was greater than a 2 hour walk) indicates that PM2A villages as a group were somewhat more remote 
than MCHN/PEP villages. 
Overall, the fact that households in MCHN/PEP villages tended to be poorer than those in PM2A villages 
means that recorded reductions over time in malnutrition may be relatively lower for them than if 
poverty were equalized between the two groups. This disadvantage to the MCHN/PEP group may be 
partially offset by the fact that PM2A villages had a slight tendency to be more remote. 
Table 7: Baseline household and village characteristics, by programming approach
MCHN/PEP PM2A Difference
Household characteristics
 Age of household head 41.1 41.6 0.5
 Percent female-headed households 6.9 7.7 0.9
 Region of residence
 Coast 4.1 11.6 7.5 ***
 Haor 44.6 45.1 0.5
 Mid Char 16.1 18.4 2.3
 North Char 35.2 24.9 -10.3 ***
 Well-being category (baseline)
 Extreme poor 15.5 11.2 -4.2 *
 Poor 55.4 56.9 1.5
 Lower middle 12.4 15.9 3.4 *
 Middle 9.9 9.7 -0.2
 Rich 6.8 6.4 -0.5
Village characteristics
 Extremely vulnerable 55.2 47.3 -7.9
 Size (number of households) 512.2 532.7 20.4
 CARE is implementing agency 7.8 9.7 1.9
 Nearest town >1 hr walk 39.9 40.8 1.0
 Upazila headquarters >2 h walk 43.0 51.9 8.9 *
Note: Stars indicate that the difference is statistically significant at the 1(***) or 5(**) percent levels.
3.3. Comparison of changes from baseline to endline in malnutrition prevalence 
across MCHN/PEP and PM2A households 
As seen in Table, for both children under five and children under two the reduction between the 
baseline and endline surveys in malnutrition prevalence—the percent of stunted, wasted and 
underweight children—were greater for the PM2A group. 
47
Table 6. Change in child malnutrition prevalence from baseline to endline, by programming approach
MCHN/PEP PM2A Difference in 
difference 
(PM2A￾Baseline Endline Change Baseline Endline Change MCHN/PEP)
Children under five
 Stunting 58.4 49.9 -8.5 59.9 46.8 -13.1 -4.6
 Wasting 14.3 11.5 -2.8 15.9 10.2 -5.7 -2.9
 Underweight 40.7 36.0 -4.7 42.7 33.5 -9.2 -4.5
Children under two
 Stunting 53.5 43.9 -9.6 49.4 39.0 -10.4 -0.8
 Wasting 16.0 12.7 -3.3 17.1 9.4 -7.7 -4.4
 Underweight 36.7 31.0 -5.7 32.8 26.4 -6.4 -0.7
Note: Stars indicate that the difference is statistically significant at the 1(***) or 5(**) percent levels.
Several factors explain the greater recorded reductions in child malnutrition among PM2A households. 
First, additional resources were extended to a larger percentage of households in PM2A villages. 
Second, the EKATA groups that successfully helped reduced malnutrition prevalence in the SHOUHARDO 
I project were established in PM2A villages at a much higher rate than in MCHN/PEP villages. Third, 
households in MCHN/PEP villages tended to be poorer than those in PM2A villages at the start of the 
project. Given these circumstances regarding project implementation and findings from the data 
analysis—and taking into account the fact that the basic PM2A principals were applied throughout—we 
can come to no conclusions regarding the relative merits of the PM2A and MCHN/PEP approaches as 
applied in the SHOUHARDO II project. 
4. Conclusions and recommendations 
4.1. General conclusions 
The findings from the final quantitative performance survey support the following general conclusions 
about the trends in project outcome and impact indicators: 
 In SO1 through SO3, substantial improvements in all impact indicators are observed, comparing 
baseline to final results. In some cases, improvements happened largely from baseline to 
midterm, and stabilized from midterm to final. Specifically in SO2, stunting for children under￾five and under two showed greatest gains from baseline to midterm. Likewise, the percent of 
households with access to improved drinking water and percent of children whose feces were 
disposed of properly saw similar trends. However, but most indicators across the SOs showed a 
steady improvement from baseline through midterm to endline. 
 These improvements in impact indicators are consistent with the reported changes in behaviors 
of households (adoption of improved agricultural practices, improved access to markets and 
48
agricultural services, adoption of child feeding practices, changes in attitudes about violence to 
women, etc.) 
 Changes in the household-level indicators for SO4 improved substantially (number of services 
utilized by households increased by almost 60 percent), while the changes in household level 
indicators for SO5 (percent of households with distress sales and average value of distress sales) 
were statistically unchanged from baseline to endline. However, it is important to stress that 
these indicators do not capture the main impacts of the interventions under these two SOs, 
which have been directed primarily toward community and local-government stakeholders. The 
effects of these interventions can only be expected to show significant impact at household 
level over the course of a longer period of time than the span of these project surveys. 
 Direct comparison of malnutrition indicators between the two SO2 programming approaches 
show better results for the PM2A programming approach compared with the targeted 
MCHN/PEP approach. However, several confounding factors may explain the greater recorded 
reductions in child malnutrition prevalence among PM2A households including PM2A having 
access to additional resources, their higher rates of establishing EKATA groups, and their overall 
higher proportion of non-poor households than compared to MCHN/PEP villages. Given these 
circumstances, no strong conclusions can be drawn regarding the differences of the two 
approaches on nutritional outcomes. However, while there are not clear differences with regard 
to nutritional outcomes, it is important to keep in mind the higher implementation cost 
associated with the non-targeted PM2A approach as compared with the MCHN/PEP approaches 
that targets resource distribution only to PEP households. 
4.2. Recommendations for Qualitative Impact Evaluation of SHOUHARDO II 
Following are several issues raised from the QPE and questions that require further qualitative 
investigation.. 
General Issues and Questions 
 The Haor Region stands out compared to other regions because of smaller improvements or 
negative outcomes (e.g. Stunting among children under five, spike in disasters and large 
increase in distress selling, lower participation in SO3). Few specific trends are seen within any 
one SO, though qualitative analysis across SOs may help describe some underlying causes. 
 Many indicators – especially impact-level measures – progressed more between baseline and 
midterm than from midterm to endline. Can this be explained by program implementation, 
household factors, or something else? 
SO1 
 Explore the different types of technology adopted by female-headed households and male￾headed households. Did these groups adopt similar technologies? What caused each group to 
adopt certain technologies? 
 Regarding household income, what differences exist between COGs? 
 What is the effect of age of household head on the pattern of adopted livelihood strategies, and 
on household income level? 
49
 What is the effect of a household receiving remittances on the pattern of adopted livelihood 
strategies? 
 How do measures of household food insecurity correlate with flooding that occurred in North 
Char, Mid Char and Haor Regions? 
 What are the reasons for adoption of recommended technologies by households that did not 
participate directly in SO1 interventions? Was this trend a result of demonstration effects from 
project participants, or are there factors other than SHOUHARDO II interventions that explain 
the adoption of these practices in the SHOUHARDO intervention areas. 
SO2 
 Significant reductions were observed in stunting, wasting, and underweight in children under 
five years old, but in children under two years old, only stunting decreased significantly. What 
household or program factors could explain this? 
 Among children under five years old, both males and females experienced an increase in 
wasting from midterm to endline. To what extent could this reflect an attitude of care givers 
giving up on children who are seen as unable to survive? What other conditions could explain 
this increase? 
 What variables affected mothers’ perceptions of hygiene practices (hand washing, in particular) 
and how did these variables affect perception? How can this information be leveraged to 
increase the perception that hygiene is important? 
SO3 
 The overall rate of women participating in a group increased from 21 percent at baseline to 
more than 25 percent at midterm, and then it dropped back to 21 percent at endline. Wide 
variations in the magnitude of change were observed at baseline, midterm and endline and 
across regions. (Annex 6). What could explain the overall drop in participation and the 
fluctuations over time and geography? 
 What are the causes of violence in female-headed households? Is the offender a non-spouse 
partner, a stranger, or someone else? Does age or presence of EKATA group have an effect on 
incidence of violence? 
SO4 
 Did PEP households have sufficient access to or knowledge of safety net services? What factors 
may have limited their use? 
 Do people believe that government services have improved? Has access to government services 
improved? What additional deficiencies exist and how can these be addressed? 
SO5 
 To what extent is the increase in distress selling in Haor related to the spike in disaster 
experienced in this region? Is distress selling in Haor connected in any way to malnutrition in the 
region? 
50
 North Char, Mid Char and Haor are susceptible to annual, large-scale flooding. These regions 
also experienced a notable cold wave from late 2012 through early 2013. The Coast did not 
experience these events. How have these experiences factored into the disaster planning and 
management of the regions? How are communities managing weather-related events compared 
to before the project? 
 Do people perceive themselves to be less vulnerable to shocks than they were before the 
project? Why or why not? Do female heads of households view themselves as more vulnerable 
to shocks than male heads of households? Similarly, do males perceive themselves to be less 
vulnerable than females? Why or why not? 
4.3. Recommendation for future project design and evaluation 
 In future project performance evaluations, it is highly recommended to integrate quantitative 
and qualitative components into a single integrated study. A major shortcoming of this 
quantitative evaluation is that the team did not have the opportunity to collect qualitative 
information from community members, organizations, local governments, and other 
stakeholders that would have provided a richer context in which to interpret the quantitative 
results, understand the reasons for why changes did or did not occur over the life of the project, 
and understand the ways in which project interventions supported the observed patterns of 
change in the project outcome and impact variables. 
 
ii
 FANTA-2. Title II Technical Reference Materials. TRM-01: Preventing Malnutrition in Children Under 2 Approach (PM2A): A 
Food-Assisted Approach. Revised November 2010. Washington, D.C.: Food and Nutrition Technical Assistance II Project 
(FANTA-2). FHI 360, 2010.