EVALUATION OF SECONDARY 
PREVENTION IN THE COMMUNITY, FAMILY 
AND YOUTH RESILIENCE PROGRAM IN ST. 
LUCIA, ST. KITTS AND NEVIS AND GUYANA
FINAL BASELINE REPORT
August 2019
This publication was prepared independently by Stephanie Gimenez Stahlberg, Alberto Díaz-Cayeros, 
Corinna Bordewieck, Daniel Sabet, Jordan Robinson, Catherine Caligan, and Julia Kresky of Social 
Impact. It was produced at the request of the United States Agency for International Development as 
part of the Democracy, Human Rights, and Governance – Learning, Evaluation, and Research activity. 
EVALUATION OF SECONDARY PREVENTION 
IN THE COMMUNITY, FAMILY AND YOUTH 
RESILIENCE PROGRAM IN ST. LUCIA, ST. 
KITTS AND NEVIS AND GUYANA 
FINAL BASELINE REPORT
August 2019 
AID-OAA-M-13-00011 
DISCLAIMER
The author’s views expressed in this publication do not necessarily reflect the views of the United States 
Agency for International Development or the United States Government.
ACKNOWLEDGMENTS
This evaluation has required extensive coordination between many different groups and individuals. First 
and foremost, we would like to thank our colleagues at the United States Agency for International 
Development (USAID) for continually pushing and advocating a rigorous evaluation design despite the 
many challenges this entailed. We are grateful to Morgan Holmes, Andrew Greer, Ted Lawrence, Ryssa 
Brathwaite, Mansfield Blackwood, Chloe Noble, and Sharon Ramsaran. The evaluation required the 
implementer Creative Associates to make numerous compromises and invest a great deal of time and 
energy into ensuring both strong programming and a rigorous evaluation. We want to thank Debra 
Wahlberg, Jason Wilks, Carlo Arze, Donald Cole, Dwynette Eversley, Waqas Mahmood, Sara Danish, 
Enrique Roig, Guillermo Cespedes, Chuck Katz of Arizona State University, and University of Southern 
California colleagues Karen Hennigan and Kathy Kolnick. The National Opinion Research Center (NORC) 
at the University of Chicago led much of the data collection, and we appreciate the efforts of Ron Wendt, 
Kareem Kysia, Alexander Rigaux, Letitia Onyango, and Jeff Telgarsky. We would also like to thank data 
collection partners in the three countries, including The Solomon Group, Specialists in Sustained Youth 
Development and Research (SSYDR), and the St. Lucia Central Statistical Office as well as Social Impact 
(SI) field coordinators Carol English, Sacha Harris, and Marecia Pemberton and former SI staff who 
supported the Democracy, Human Rights, and Governance – Learning, Evaluation and Research (DRG￾LER) mechanism, Alison Miranda and Meredith Feenstra. Finally, we are grateful to the many youth, 
parents, and caregivers who participated in the Youth Services Eligibility Tool (YSET) assessment and 
caregiver survey. 
CONTENTS
LITERATURE REVIEW – REDUCING YOUTH RISK LEVELS FOR ENGAGING IN 
CRIME AND VIOLENCE 3
THE STATE OF YOUTH VIOLENCE IN THE CARIBBEAN 3
THE CAUSES OF YOUTH VIOLENCE IN THE REGION 3
WHAT WORKS IN PREVENTING YOUTH VIOLENCE 4
EVALUATION PURPOSE, QUESTION, AND METHODOLOGY 10
EVALUATION PURPOSE 10
EVALUATION QUESTIONS 10
EVALUATION DESIGN 10
THE YSET AND SELECTION OF YOUTH 10
CAREGIVER SURVEY 13
COMMUNITY SURVEY 14
HUMAN SUBJECTS PROTECTIONS 14
ASSUMPTIONS AND RISKS 14
COMMUNITY CONTEXT: RELEVANT FINDINGS FROM THE CFYR COMMUNITY 
SURVEY 17
CRIME VICTIMIZATION 18
PERCEPTION OF CRIME PREVALENCE 21
OVERALL SOCIAL CAPITAL INDEX 23
YOUTH SOCIAL CAPITAL 24
ATTITUDES TOWARD THE LAW 26
VIEW OF THE POLICE 27
RESOURCES FOR YOUTH 28
SELF-EFFICACY 29
INDEX SUMMARY 30
FINDINGS BASED ON THE YSET ASSESSMENT 36
QUESTION 1: DO SECONDARY PREVENTION EFFORTS IMPACT OVERALL YOUTH 
RISK, CRIME AND VIOLENCE RELATED BEHAVIOR, AND YOUTH RISK AND 
RESILIENCE FACTORS? WHY OR WHY NOT? 36
UNDERSTANDING RISK ACROSS THE FULL YOUTH POPULATION 36
SCALE B: WEAK PARENTAL MONITORING 40
SCALE C: CRITICAL LIFE EVENTS 42
SCALE DE: IMPULSIVE RISK TAKING 43
SCALE F: GUILT NEUTRALIZATION 44
SCALE G: NEGATIVE PEER INFLUENCE 46
SCALE H: PEER DELINQUENCY 47
SCALE T: FAMILY ANTI-SOCIAL INFLUENCE 48
SCALE IJ: SELF-REPORTED DELINQUENCY 50
SCALE FRD: RISKY GROUP BEHAVIORS 52
COMPARISON BETWEEN HONDURAS AND CARIBBEAN YOUTH IN 
TARGETED COMMUNITIES 54
BASELINE FINDINGS FROM THE CAREGIVER SURVEY 56
Q2: DO SECONDARY PREVENTION EFFORTS IMPROVE THE FAMILY 
ENVIRONMENT? WHY OR WHY NOT? 56
INTRODUCTION 56
CAREGIVER SURVEY RESPONDENTS AND FAMILY STRUCTURE 56
FAMILY CHARACTERISTICS 59
FAMILY COHESION 64
FAMILY DECISION-MAKING 66
PARENTING ATTITUDES: LOCUS OF CONTROL 68
CAREGIVER PRESENCE 70
NEGATIVE FAMILY INFLUENCES 76
KNOWLEDGE OF YOUTH BEHAVIOR 77
CONCLUSIONS 80
ANNEX I: ELIGIBILITY DETERMINATION AND RANDOMIZATION 82
DETERMINATION OF ELIGIBILITY 82
MATCHING AND RANDOMIZATION 82
ANNEX II: DATA COLLECTION INSTRUMENTS 84
YSET TOOL 85
CAREGIVER SURVEY 110
COMMUNITY SURVEY 123
ANNEX III: REGRESSION RESULTS 156
THRESHOLDS 156
ANNEX IV: SUPPORTING TABLES AND FIGURES 159
MEASURES OF INCOME 159
RELIGION 160
ANNEX V: INDEX CONSTRUCTION 165
ANNEX VI: GENDER DISAGGREGATIONS 167
YSET 167
CAREGIVER SURVEY 176
REFERENCES 187
TABLES Table 1: Sample sizes in various surveys used in this report............................................................... 13
Table 2: Assumptions and risk mitigation strategies.............................................................................. 15
Table 3: Index Summary............................................................................................................................... 31
Table 4: Percent of assessed youth at risk in the overall YSET and on each scale (n=5,109) ..... 37
Table 5: Odds ratios with Guyana, females, in school, ages 10 to 14, and ethnicity-Black/Afro￾Guyanese as the base (n=5,101) ................................................................................................................ 38
Table 6: Differences in predicted probabilities for each group of respondents scoring at 
risk/eligible (n=5,101) ................................................................................................................................... 39
Table 7: Percent of assessed youth at risk for weak parental monitoring (n=5,109).................... 41
Table 8: Percent of assessed youth at risk for critical life events (n=5,109).................................... 42
Table 9: Percent of assessed youth at risk for impulsive risk taking (n=5,109)............................... 43
Table 10: Percent of assessed youth at risk for guilt neutralization (n=5,109) ............................... 45
Table 11: Percent of assessed youth at risk for negative peer influence (n=5,109)....................... 46
Table 12: Percent of assessed youth at risk for peer delinquency (n=5,109).................................. 47
Table 13: Percent at risk for family antisocial influence (n=5,109)..................................................... 49
Table 14: Percent of assessed youth at risk for self-reported delinquency (n=5,109).................. 50
Table 15: Percent at risk for risky group behaviors (n=5,109)........................................................... 53
Table 16: T-tests of means with entire samples (eligible and ineligible)............................................ 55
Table 17: T-tests of means with eligible youth only .............................................................................. 55
Table 18: Caregivers: primary (by gender) and additional household members (n=941)............. 57
Table 19: Share of household members who are gainfully employed (n=942)................................ 61
Table 20: Average number of gainfully employed household members (n=947)............................ 62
Table 21: Number of youth between the ages of 16 and 18, by employment and school status 
(n=537) ............................................................................................................................................................. 62
Table 22: Youth behaviors as reported by self, caregivers, and differences between the two ... 78
Table 23: Share of matches and mismatches in reporting youth behavior, by type of primary 
caregiver (n=1,135 for youth and 1,002 for caregivers)....................................................................... 79
Table 24: Average scores for 34 highest risk youth and final thresholds across nine risk modules 
in Guyana......................................................................................................................................................... 82
Table 25: Results of stage 1 and 2 randomizations in Guyana ............................................................ 83
Table 26: YSET, caregiver survey, and implementation dates............................................................. 84
Table 27: Odds ratios with St. Lucia, females, in school, ages 10 to 14, and ethnicity-Black/Afro￾Guyanese as the base.................................................................................................................................. 156
Table 28: Odds ratios with St. Kitts and Nevis, females, in school, ages 10 to 14, and ethnicity￾Black/Afro-Guyanese as the base............................................................................................................. 156
Table 29: Comparing shares of risk on each YSET scale by country, by setting the risk 
thresholds to be the same......................................................................................................................... 157
Table 30: Logistic regressions comparing shares of risk on each YSET scale by country, by 
setting the risk thresholds to be the same ............................................................................................ 158
Table 31: PCA with polychoric correlations......................................................................................... 166
FIGURES
Figure 1: Number of youth in communities, by eligibility status, and interviewed for each survey
........................................................................................................................................................................... 12
Figure 2: Crime victimization reported by respondents in the last 12 months by country and 
community....................................................................................................................................................... 18
Figure 3: Crime victimization among other members of the household in the last 12 months by 
community and country ............................................................................................................................... 19
Figure 4: Crime victimization: Count of reported crimes of different type by community.......... 20
Figure 5: Crime victimization: Count of reported crimes of different types by country.............. 21
Figure 6: Average Perception of Crime Prevalence Index scores by community and country ... 22
Figure 7: Rankings: Perceptions of crime prevalence versus reported crime victimization by 
community....................................................................................................................................................... 23
Figure 8: Average Overall Social Capital Index scores by community and country....................... 24
Figure 9: Average Youth Social Capital Index scores by community and country ......................... 25
Figure 10: Rankings: Average Overall Social Capital Index and Youth Social Capital Index scores 
by community ................................................................................................................................................. 26
Figure 11: Average Attitudes Towards the Law Index scores by community and country ......... 27
Figure 12: Average View of the Police Index scores by community and country........................... 28
Figure 13: Average Resources for Youth Index scores by community and country...................... 29
Figure 14: Average Self-Efficacy Index scores by community and country....................................... 30
Figure 15: St. Lucia community risk factor visualizations with captioned occurrences 
represented by a red dot and non-occurrences by a black dot.......................................................... 33
Figure 16: St. Kitts and Nevis community risk factor visualizations with captioned occurrences 
represented by a red dot and non-occurrences by a black dot.......................................................... 34
Figure 17: Georgetown, Guyana community risk factor visualizations with captioned 
occurrences represented by a red dot and non-occurrences by a black dot.................................. 35
Figure 18: Average standardized YSET scores from 0 to 1 on each scale for eligible and ineligible 
youth (n=5,109) ............................................................................................................................................. 40
Figure 19: Responses to select questions on parental monitoring among eligible youth (n=1,138)
........................................................................................................................................................................... 41
Figure 20: Responses to select questions on parental monitoring among eligible youth (n=1,138)
........................................................................................................................................................................... 42
Figure 21: Responses to questions on critical life events in the last year among eligible youth 
(n=1,139) ......................................................................................................................................................... 43
Figure 22: Responses to questions on impulsive risk taking among eligible youth (n=1,134)...... 44
Figure 23: Responses on select questions on guilt neutralization among eligible youth (n=1,137)
........................................................................................................................................................................... 45
Figure 24: Responses on select questions on guilt neutralization among eligible youth (n=1,135)
........................................................................................................................................................................... 46
Figure 25: Responses on negative peer influence among eligible youth (n=1,133)......................... 47
Figure 26: Reported involvement of peers in delinquent behavior by eligible youth (n=1,135).. 48
Figure 27: Responses on family antisocial influence questions among eligible youth (n=1,132).. 49
Figure 28: Response on family antisocial influence question among eligible youth (n=1,133) ..... 50
Figure 29: Self-reported delinquency among eligible youth in the last six months (n=1,136)...... 51
Figure 30: Self-reported delinquency by eligible youth with a group of friends 
(n=44;1,017;160;520;203;12;847;460;810;726;513;124) ....................................................................... 52
Figure 31: Eligible youth reporting on behaviors their friends have engaged in during the last six 
months (n=1,008).......................................................................................................................................... 53
Figure 32: Family structures in St. Lucia, St. Kitts and Nevis and Guyana in caregiver survey 
(n=941) ............................................................................................................................................................. 58
Figure 33: Number of assets in households by country as reported in the caregiver survey 
(n=947) ............................................................................................................................................................. 59
Figure 34: Age distribution of mothers of eligible youth (regardless of caretaking role or family 
structure .......................................................................................................................................................... 60
Figure 35: Age distribution of fathers of eligible youth (regardless of caretaking role or family 
structure .......................................................................................................................................................... 60
Figure 36: Levels of educational achievement by country and relationship to program youth 
(n=941) ............................................................................................................................................................. 61
Figure 37: Share of youth ages 16-18 that are neither employed nor in school (n=537)............. 63
Figure 38: Average frequency of religious service attendance, by country (n=944)...................... 64
Figure 39: Responses to questions on vertical family relations (n=1,000) ....................................... 65
Figure 40: Distribution of scores in the Family Cohesion Index, by country (n=943) .................. 66
Figure 41: Responses to questions on family decision-making among households with more than 
one caregiver, n=872 .................................................................................................................................... 67
Figure 42: Caregiver responses to questions on family decision-making, by country (n=872).... 67
Figure 43: Caregiver responses to questions on parenting locus of control (n=947)................... 68
Figure 44: Caregiver responses to questions on parenting locus of control (cont) (n=947)....... 69
Figure 45: Caregiver responses to questions on parenting locus of control, by country (n=947)
........................................................................................................................................................................... 70
Figure 46: Caregiver responses to questions on parenting locus of control (cont.), by country 
(n=947) ............................................................................................................................................................. 70
Figure 47: Number of days in a typical week caregivers report select behavior for each eligible 
youth................................................................................................................................................................. 71
Figure 48: Number of days in a typical week caregivers report select behavior for each eligible 
youth (n=996)................................................................................................................................................. 72
Figure 49: Caregiver responses to questions on caregiver presence for each eligible youth 
(n=998) ............................................................................................................................................................. 73
Figure 50: Caregiver responses to questions on caregiver presence for each eligible youth 
(n=1,001) ......................................................................................................................................................... 74
Figure 51: Caregiver responses to questions on communication with each eligible youth 
(n=1,001) ......................................................................................................................................................... 74
Figure 52: Caregiver responses to questions on caregiver presence, by caregiver’s gender 
(n=1,001) ......................................................................................................................................................... 75
Figure 53: Responses to select questions on caregiver presence (n=1,001) ................................... 76
Figure 54: Share of matches and mismatches between youth and caregiver answers, for each 
behavior (n=1,135 for youth and 1,002 for caregivers)........................................................................ 79
Figure 55: Household socioeconomic ranking according to assets reported in caregiver survey 
(n=947) ........................................................................................................................................................... 159
Figure 56: Share of caregiver self-reported religious affiliation, in St. Lucia (n=413)................... 160
Figure 57: Share of caregiver self-reported religious affiliation, in St. Kitts and Nevis (n=155) 161
Figure 58: Share of caregiver self-reported religious affiliation, in Guyana (n=377) .................... 161
Figure 59: Caregiver reported crimes that household members have been charged with in the 
last ten years (n=76)................................................................................................................................... 162
Figure 60: Caregiver reported household crime victimization in the last 12 months (n=42).... 163
Figure 61: Shares of matches and mismatches of youth behavior, by type of primary caregiver
......................................................................................................................................................................... 164
Figure 62: Responses to select questions on parental monitoring, by gender.............................. 167
Figure 63: Responses to select questions on parental monitoring, by gender.............................. 167
Figure 64: Responses to questions on critical life events, by gender............................................... 168
Figure 65: Responses to questions on impulsive risk taking, by gender.......................................... 169
Figure 66: Responses to questions on impulsive risk taking, by gender.......................................... 169
Figure 67: Responses on select questions on guilt neutralization, by gender................................ 170
Figure 68: Responses on select questions on guilt neutralization, by gender................................ 170
Figure 69: Responses on negative peer influence, by gender ............................................................ 171
Figure 70: Responses on negative peer influence, by gender............................................................ 171
Figure 71: Responses to questions on peer delinquency, by gender............................................... 172
Figure 72: Responses on family antisocial influence questions, by gender...................................... 173
Figure 73: Responses on family antisocial influence question, by gender....................................... 173
Figure 74: Responses on self-reported delinquency, by gender........................................................ 174
Figure 75: Responses on risky group behavior, by gender................................................................. 175
Figure 76: Responses to questions on family leadership, by caregiver’s gender, n=104 Male, 767 
Female ............................................................................................................................................................ 176
Figure 77: Responses to questions on parenting, by caregiver’s gender......................................... 176
Figure 78: Responses to questions on parenting, by caregiver’s gender......................................... 177
Figure 79: Number of days in a typical week caregivers report select behavior, by caregiver’s 
gender ............................................................................................................................................................. 177
Figure 80: Number of days in a typical week caregivers report select behavior, by caregiver’s 
gender ............................................................................................................................................................. 178
Figure 81: Responses to select questions on caregiver presence, by caregiver’s gender........... 178
Figure 82: Responses to select questions on caregiver presence, by caregiver’s gender........... 179
Figure 83: Responses to select questions on caregiver presence, by caregiver’s gender........... 179
Figure 84: Responses to select questions on caregiver presence, by caregiver’s gender........... 180
Figure 85: Responses to select questions on caregiver presence, by caregiver’s gender........... 180
Figure 86: Responses to select questions on household drug problems, by caregiver’s gender
......................................................................................................................................................................... 181
Figure 87: Responses to question on crime in the last 12 months, by caregiver’s gender......... 181
Figure 88: Responses to question on crime in the last 12 months, by caregiver’s gender......... 182
Figure 89: Responses to question on crime in the last 12 months, by caregiver’s gender......... 183
Figure 90: Responses to question on crime in the last 12 months, by caregiver’s gender......... 184
Figure 91: Responses to question on gang membership, by caregiver’s gender ........................... 185
Figure 92: Responses to select question on community, by caregiver’s gender........................... 185
Figure 93: Responses to select question on community, by caregiver’s gender........................... 186
Figure 94: Responses to questions on police perception, by caregiver’s gender........................ 186
ACRONYMS
Acronym Definition
CBT Cognitive Behavioral Therapy
CFYR Community, Family and Youth Resilience
DRG-LER Democracy, Human Rights, and Governance – Learning, Evaluation, and 
Research
ESC Eastern and Southern Caribbean 
FACES Family Adaptability and Cohesion Scale
GUY Guyana
GRYD Gang Reduction and Youth Development
IDB Inter-American Development Bank
LAC Latin American and Caribbean
LST LifeSkills Training
MDES Minimum Detectable Effect Size
MICS Multiple Indicator Cluster Survey
NEET Not in Education, Employment, or Training
NORC National Opinion Research Center
PCA Principal Component Analysis
RCT Randomized Control Trial
SI Social Impact
SKN St. Kitts and Nevis
SLU St. Lucia
SSYDR Specialists in Sustained Youth Development and Research
UNICEF United Nations Children’s Fund
USAID United States Agency for International Development
USC University of Southern California
YES Youth Empowerment Services
YSET Youth Services Eligibility Tool
i 
EXECUTIVE SUMMARY
In response to crime and violence in the Eastern and Southern Caribbean (ESC), in 2016 the United States 
Agency for International Development (USAID) launched the Youth Empowerment Services (YES) 
project, which includes the Community, Family and Youth Resilience (CFYR) initiative. Implemented by 
Creative Associates, CFYR employs a public-health informed, place-based strategy in 15 communities in 
Guyana, St. Lucia, and St. Kitts and Nevis. This impact evaluation focuses on CFYR’s Family Matters 
secondary prevention efforts, a key component of the CFYR intervention. The Prevention and 
Intervention Family Systems Model (PIFSM) used by the Family Matters program is comprised of two 
interrelated components: 1) the Youth Services Eligibility Tool (YSET) diagnostic process which identifies 
nine risk factors predictive of crime and violence behaviors and 2) the seven phases of family systems 
intervention which reduces the nine risk factors.
A strict, randomized selection process was established in which the YSET was applied to a large percentage 
of the population of youth between the ages of 10 and 17 in the 15 communities. CFYR then used the 
YSET to determine which youth were at higher risk and in need of secondary prevention efforts. Eligible 
youth were randomized into a treatment group, which would be eligible for one year of the PIFSM right 
away, and a control group, which would be eligible for the program in the following year1. There was no 
ethical dilemma presented by the delayed treatment, given that the size of the treatment group
corresponds with the capacity constraints of the implementing party. 
This randomization strategy allows us to compare similar youth to estimate the effect of the intervention 
on treatment youth and their families vis a vis control youth and their families after one year. The tools 
used for the evaluation include a community survey conducted by CFYR, the YSET for youth, which is 
applied at baseline, after six months, and after one year, a caregiver survey of the youth’s primary caregiver, 
which is conducted at baseline and at the end of the one-year program, and qualitative fieldwork. The 
evaluation of the intervention itself will not be available until a year from now, but this report presents 
findings for the baseline community survey, YSET, and caregivers survey. Key findings include the following:
1. While not as well evaluated as some other approaches, a family-based approach to countering 
youth violence is supported by the existing literature on the causes of youth violence in the 
Caribbean and Latin America. 
2. The 15 communities included in the study are suffering from varying degrees of crime 
victimization with variation in the composition of the crimes suffered by each community. 
Many residents have feelings of insecurity, and both victimization and perceptions of crime 
prevalence are particularly strong in Sophia in Georgetown, Guyana.
3. We find that overall social capital and a sense of self-efficacy is high in all 15 communities, 
suggesting that there are important opportunities to leverage connections and community 
1 It should be noted that in Creative Associate’s previous implementation of the PIFSM, there was no separation between applying 
the YSET and commencing with treatment. In the Family Matters program, there was a time delay between the YSET and when 
the family first received a visit from their assigned family counselor. Also, the application of the YSET for Family Matters was done 
by enumerators instead of the family counselors that would begin working with the family. Although hiring enumerators for data 
collection was essential for a robust impact evaluation, this change in the PIFSM probably made it harder for family counselors to 
engage the family in their initial encounter.
ii
resources in supporting youth resilience programs. Residents are generally not satisfied with 
services for youth, although respondents are somewhat more satisfied in St. Kitts and Nevis
(particularly in Basseterre) than the other two countries.
4. Attitudes towards the rule of law are relatively positive and consistent across communities 
and countries. On average, respondents also have a generally trusting view of the police, 
although there is variation across communities. 
5. Although there are differences between countries, there is typically greater variation across 
communities than between countries. This is a positive finding for the evaluation design, as it 
suggests comparability between the three countries on key indicators. 
6. The YSET tool was adapted by the University of Southern California (USC) team for the three 
countries, allowing for the identification of higher risk youth eligible to be included in the 
program. One-fifth of youth surveyed scored above the eligibility cutoff. 
7. Single risk factors are highly correlated with the YSET composite scoring of eligibility, 
suggesting a relative robustness of the tool. These single risk factors identify a similar share of 
youth engaged in risky behavior or exposed to risk factors in all communities. 
8. Males are at significantly higher risk than females in the overall YSET and in each scale, except for 
family antisocial influence, where the difference between males and females is not statistically 
significant. Older youth (ages 15-17) are at higher risk than younger youth (ages 10-14) in all YSET 
scales.
9. Other personal characteristics correlate with CFYR’s risk determination. Out of school youth 
are at higher risk on most scales, and East Indian/Indo-Guyanese youth are at significantly lower 
levels of risk than other ethnic groups in Guyana. Country differences are not significant for 
determining eligibility, but the level of risk by country does vary across the different YSET 
scales. Youth in the islands are more likely to score at risk in the weak parental monitoring 
and negative peer influence scales, even when the scale thresholds are set to be the same for 
the three countries (see full table in Annex III). 
10. A surprisingly large amount of eligible youth have had some critical life event happen in the last 
year. Some of the events included in the critical life events scale can be stressful or even 
traumatic, such as hearing about deaths or injuries caused by violence in their neighborhoods 
(which happened to 71 percent of eligible youth), having a fight or problem with a friend (67 
percent), or having someone close to them die, get seriously injured, very sick or leave the 
country (64 percent). 
11. Youth seemed to be relatively forthcoming in reporting behaviors. They were asked if they had 
engaged in a set of twelve negative behaviors, such as fighting or stealing, in the last six months. 
The caregivers of eligible youth were also asked if they thought the youth had engaged in those 
same behaviors in the last six months. The youth answered yes much more often than their 
caregivers did in the caregiver survey. Some of the percentage point differences in the reporting 
are as high as 36 points, and no single type of caregiver (e.g. parent, grandparent, etc.) seemed
to be misreporting more often. Either caregivers are not very aware of how often youth are 
engaging in these negative behaviors, and which behaviors they engage in, or they are not willing 
to report these behaviors, presumably due to embarrassment or fear that the youth would get 
into trouble.
12. The caregiver survey allowed us to create a more comprehensive picture of the family 
composition and family life of the youth that are engaged in the program (eligible families in 
iii
the treatment and control groups). The majority of primary caregivers are mothers (72 
percent), while 11 percent are fathers and 17 percent are grandmothers, aunts, sisters and 
other family members. The most typical eligible household is one with a single mother and no 
father present in the household (52 percent). On average, around two additional household 
members, besides the primary caregiver, participate in care for the youth. This family 
composition is similar across countries, although Guyanese households tend to have more 
children.
13. On average, eligible households have 1.63 members who are gainfully employed. Employment 
levels are much higher in St. Kitts and Nevis (SKN) compared to the other two countries.
14. About 24 percent of youth ages 16-18 in eligible households are neither in school nor 
employed. There is significant variation across countries, with only four percent of treatment 
youth in St. Kitts and Nevis being out of school and unemployed, compared to 31 percent in 
Guyana. 
15. Nearly half of respondents (47 percent) believe that community conditions are making it more 
difficult for them to keep their children out of trouble, but the vast majority (91 percent) think 
that parents can have a major influence on their kids. Caregivers already appear to have a 
relatively strong internal locus of control when it comes to raising their children. 
16. 57 percent of caregivers report always knowing where their child is when not at home or in 
school, and 49 percent report always knowing who they are with. About 10 percent never or 
rarely know where their child is when outside the home or school. Only half report being able 
to always discuss problems with their child. This suggests that there is room for improvement 
in the quality and frequency of communication between youth and their caregivers. 
17. A higher percentage of male caregivers report being listened to than female caregivers. Males 
are also more likely to report that they are able to talk to their child and solve their problems 
and less likely to report saying mean things or pushing or hitting their child when angry. It is 
not clear if these differences reflect actual behavior differences or gender differences in 
reporting. 
18. In seven percent of eligible households, caregivers report that drug use has created a problem in 
the household. In these households, marijuana was mentioned 67 percent of the time and alcohol 
36 percent. 
1 
INTRODUCTION
Social Impact (SI) is contracted with the United States Agency for International Development (USAID) to 
evaluate the Community, Family and Youth Resilience (CFYR)’s Family Matters program. CFYR includes 15
communities with relatively high levels of crime and violence, distributed evenly across the countries of 
Guyana (GUY), St. Kitts and Nevis (SKN), and St. Lucia (SLU). The Family Matters secondary prevention 
program is targeted towards high-risk youth, and it is one of four components of CFYR’s programing. 
Crime and violence are growing problems in the Caribbean region. With increased globalization, the 
region has experienced a proliferation of trans-national crime and drug-trafficking that is unprecedented 
in scope and scale (CARICOM, p. 16). A CARICOM Commission on Youth Development found that 
crime and violence are primary concerns for youth in the Caribbean, as young people speak of “fear, 
perceptions of lack of safety and concern for their general well-being as a result of the increased crime 
and violence; of self-imposed curfews, diminished participation in community activities, restriction of night￾time activities, and changes in social practice as a consequence” (CARICOM, pp. xvi-xvii). Constant 
exposure to crime and violence can deeply affect a person’s well-being, through increases in stress levels, 
emotional scarring, grief, and creating a sense of loss (CARICOM, p. xvi). 
Although the level of social and economic development varies in the three countries included in this 
project, all three have marginalized communities where youth face significant hardships at the personal, 
family and community levels. Each of the countries exhibits profound dislocations in their family structures, 
characterized by single-mother-headed households, missing male role models, and a relatively large 
transient population due to migration.2 Local gangs that control drug sales and instigate violence afflict 
some of these communities. The communities are all urban and some are low-income, although they show 
some variation in educational, health and public services provided. 
The main goal of the CFYR intervention is to reduce youth involvement in crime and violence. CFYR’s 
Family Matters program is a secondary prevention intervention consisting of counseling sessions for youth 
aged 10-17 and their families. Family counselors who have been trained on the PIFSM engage with the 
families to identify the problems behind the youth’s potential anti-social behavior and work on ways to 
improve the family environment, parental involvement, and ultimately the youth’s behavior. The population 
of youth in each community was identified through schools and community outreach, with enumerators 
knocking on doors. All youth who had the proper consent were then interviewed using the Youth Services 
Eligibility Tool (YSET), which was first designed by researchers at the University of Southern California 
(USC) to identify youth who were at the highest risk for violence associated with gangs in Los Angeles
and which has been adapted to the Caribbean context.3 The intervention targeted youth considered to 
be at a higher risk level. 
2 Census and other national data for three countries was analyzed as background for this evaluation.
3 The YSET tool has been validated as a gang risk assessment tool. See Hennigan, Karen M., et al. "Identifying high-risk youth for 
secondary gang prevention." Journal of Crime and Justice 37.1 (2014): 104-128; Hennigan, Karen M., et al. "Targeting youth at 
risk for gang involvement: Validation of a gang risk assessment to support individualized secondary prevention." Children and
youth services review 56 (2015): 86-96.
2 
The opportunities afforded by these countries to understand engagement with criminal activity and youth 
violence are unique. While we know much about how these processes unfold in cities and neighborhoods 
in the United States (and increasingly more knowledge is being generated in Central America), the 
opportunity to study prevention strategies in these small, tightly knit communities may shed light on how 
to create successful programs that can impact local social dynamics in many other developing countries, 
including the rest of the Caribbean. In the three CFYR countries, at-risk youth will be offered counseling 
services in very supportive social environments before serious escalations of violence or gang activity have 
taken place. In that sense, this study can shed light on youth services that can be effective when early 
warning signs first appear, before organized crime and youth recruitment to gangs become so daunting 
that it may be too late to intervene. 
Two of the countries chosen for the intervention are very small, rendering it unfeasible (financially and in 
terms of the statistical power of a quantitative approach) to randomize treatment across communities. 
However, the evaluation team designed a multi-site, randomized controlled trial (RCT), whereby youth 
were randomized into control or treatment groups within 15 pre-selected, high risk communities.4 As 
such, the impact evaluation allows programs to be encompassing and comprehensive in the sense that 
they cover 15 high risk communities, rather than a random selection of treated and untreated communities 
within hotspot territorial areas.
This baseline report begins with a brief literature review on the state of youth violence and violence 
prevention policies and programs. Next, we cover the evaluation purpose and evaluation questions, 
background on the CFYR project, evaluation design, methods and limitations, and, finally, the data analysis 
findings. The results include background information on the communities studied from a community 
survey, findings on youth behavior and beliefs from YSET analysis, and information regarding the home 
environment and parenting style of caregivers from the caregiver survey. It is important to note that this 
is only a baseline report and we are not yet able to report on the impact of Family Matters program. 
4 In Guyana, the communities selected for the intervention were not necessarily the ones at highest risk, as CFYR excluded from 
its selection process several high risk communities with programming implemented by the Inter-American Development Bank 
(IDB). In St. Lucia, due to small population sizes and limited disaggregated data to inform community selection, some of the 
“communities” include areas from different parishes/counties.
3 
LITERATURE REVIEW: 
REDUCING YOUTH RISK LEVELS FOR 
ENGAGING IN CRIME AND VIOLENCE
THE STATE OF YOUTH VIOLENCE IN THE CARIBBEAN
Citizen security has become a greater concern for citizens of the Caribbean, where the level of violent 
crime is particularly high and on the rise. Violent crime, including homicides, assaults and threats, appears 
to be more common in the Caribbean than in all other regions covered by the International Crime 
Victimization Survey database. Nearly one in three Caribbean residents reported witnessing a violent act 
in which another person got injured or died. Youth (18-24) and young adults (25-30) are disproportionally 
likely to be victims or perpetrators of violent crime, in consideration of their share of the population 
overall (Sutton and Ruprah, 2017).
A 2013 Caribbean youth assessment prepared by SI noted among respondents a perception of heightened 
crime and violence involving youth in the region. Thefts and house break-ins, which are considered to be 
the predominant youth crimes, were on the rise and becoming more sophisticated. Key informants in St.
Vincent and the Grenadines, Dominica and St. Lucia reported that stolen articles are handed off to 
collaborators on boats and then transported and sold in other islands. This level of organization within
local and inter-island networks also suggests that a gang problem is on the rise (Social Impact, 2013). 
This perception of high levels of youth involvement in crime and violence in the Caribbean appears to be 
based on fact. A 2016 report titled “Prevalence and Patterns of Troublesome Youth Groups in the 
Caribbean” found that youth themselves report high levels of involvement in these activities. Two-thirds of 
surveyed students reported having committed criminal and/or violent offenses in the previous year, and this 
number was even higher for students who acknowledged being gang members (Katz and Nuño, 2016). This 
report suggests that gang involvement starts at a very young age, as early as nine years old. 
THE CAUSES OF YOUTH VIOLENCE IN THE REGION
Set against a backdrop of drug trafficking, organized crime, weak judicial and law enforcement systems, 
and a lack of opportunities and support for youth who live in deprived communities, Laura Chioda (2017), 
Social Impact (2013), and Katz and Nuño (2017) identify several factors that drive youth participation in 
crime and violence in the Latin American and Caribbean (LAC) region. Laura Chioda (2017) finds that 
LAC youth are at a higher risk of committing and falling victim to violence, which bears important 
consequences for their life trajectories and for society as a whole. In her study using a panel of LAC 
countries, Chioda (2017) also finds that although the general level of employment was not related to crime 
levels, youth unemployment was.5 
5 Micro-level data for Mexico and Brazil also revealed that it was not just the employment status that matters for the relationship 
between labor markets and criminal offending, but also the quality of employment.
4 
For the Caribbean region, in particular, youth development workers interviewed for the 2013 Social 
Impact youth assessment reported that besides economic pressures, “deteriorating social support 
systems, lack of positive male role models, and the growing number of unattached males in an age group 
that tends to be less risk-averse combine to create opportunities for disorder and crime” (Social Impact, 
2013, p. vii). One reason cited for increased gang membership in Guyana was the high number of parents 
who live elsewhere due to better economic opportunities, thus leaving their children with a single 
parent/caretaker or even with an older sibling. As household or family ties fail, youth seek to form other 
ties, including by engaging in promiscuous behaviors and joining gangs. Also, in Guyana, almost half of 
interview respondents characterized crime as incidents related to getting “fast cash” to fulfill basic needs 
and pursued primarily by young women in the form of transactional sex. In terms of organized crime, a 
third of interview respondents said that youth are often used by adult perpetrators as front-line workers 
in drug trafficking and prostitution rings (Social Impact, 2013).
By interviewing youth in the study countries, Katz and Nuño (2017) were able to identify the risk factors 
and reasons for joining what they term troublesome youth groups (TYGs). Among the risk factors found 
to be significant, they list low self-control (all countries), low commitment to school (St. Lucia), negative 
peer commitment (St. Lucia and Guyana), and lack of parental supervision (Guyana). The most commonly 
cited reasons for joining TYGs were to make friends and because someone in the family was a group 
member. In Guyana, another reason was to participate in group activities. One stakeholder explained it 
this way: “Generally speaking . . . persons do not ‘join’ a gang, per se. It is simply a continuation of the 
community socialization process. Gangs are found in communities where many of their members reside; 
therefore, they are known to most persons in the communities and [have] socialized with them at some 
point. Being part of the group simply requires one to ‘lime’ with the gang in their customary location” 
(Katz and Nuño, pp. 12-3). 
WHAT WORKS IN PREVENTING YOUTH VIOLENCE 
According to a 2015 Inter-American Development Bank (IDB) study entitled Closing Knowledge Gaps, youth 
violence prevention strategies can be classified by their social ecology: “treatment-specific (i.e. cognitive 
behavioral therapy (CBT), counseling and social skills training), family-based (i.e. behavioral parent training 
and home visitation), and school- or community-based (i.e. mentoring and after-school programs, as well 
as social capital building outside of the family)” (Jaitman and Guerrero Compeán, 2015, pp. 16-7). The 
report identifies the most successful youth violence prevention programming to date, most of which were 
implemented in the United States, by reviewing various studies and evaluations. 
Cognitive-behavioral interventions, such as the Big Brother/Big Sister program, which is a long-running 
mentoring match between youth at risk and adults, were the most successful at reducing anti-social 
behavior over time, more than double the measured effect of counseling therapy (Lösen and Beelmann, 
2003). CBT and family therapeutic interventions are usually also identified as successful at preventing 
violence. However, one study that reviewed eight randomized controlled trials in the United States found 
that multisystemic therapy, a type of intensive family and community-based treatment, had no statistical 
effect on arrests and convictions (Littell, Popa, and Forsythe, 2005), suggesting that some of the optimism 
regarding these programs should be tempered. A more recent study that did a randomized controlled 
trial of a Functional Family Therapy (FFT) program in Philadelphia found that the effectiveness of such a 
program on recidivism varied by levels of risk of gang membership. Among youth that were at a high risk 
5 
for joining a gang, the study found significantly lower recidivism rates, compared to the control group that 
was receiving the “treatment as usual.” For youth that were at a lower risk for gang membership, the 
study did not identify consistent differences between the treatment and control groups (Thornberry et al, 
2018). 
In a very thorough review of the literature on violence in Latin America and policies targeting children 
and youth, Chioda (2017) found that adolescence and young adulthood are critical ages for policy 
intervention “because of loosening of parental control, the increased pull of peers, the transition from 
youth to adult roles, and the development of three different regions of the brain, which are responsible 
for the regulation of automatic or instinctive reactions, risk-taking behavior, self-control, and reflective 
reasoning” (p. 34). The most effective behavioral interventions identified by Chioda’s review are CBT, life 
skills training and family-based interventions.6
School-based interventions can be successful, particularly when they are designed to alter behavior, 
thinking patterns, attitudes and beliefs. Those that discourage students from dropping out during their 
final years of secondary education yield benefits even decades later, in the form of reducing participation 
in violent and property crime (Chioda, 2017). One such school-based intervention is the LifeSkills Training 
(LST) program which was implemented in schools in Jamaica and which received the first evaluation for a 
program of this kind in a developing country context. LST is a 3-year intervention for middle and junior 
high school students, targeting violence as well as tobacco, alcohol and marijuana use. Youth receive daily 
supervision at school from 8am until 2pm, receiving instruction, socialization and “training in social skills, 
self-management, and social resistance skills to address a number of cognitive, psychological, attitudinal, 
and social factors related to risky behaviors (including violence)” (Guerra 20137 as cited in Chioda, 2017, 
pp. 239-40). LST was extensively evaluated and has shown positive results in the short and long-term, 
acquiring the “effective” program label by the Blueprints Initiative. Although participants in the treatment 
group did not differ from their counterparts in the control groups in terms of “triggers” of aggression, as 
measured by a propensity scale, they showed better self-control by being less likely to act on those triggers 
(Chioda, 2017). Policies and programs that target the early years of life have also shown to be effective in 
preventing violence later in life.8 
Finally, family-based interventions constitute a type of violence prevention programming that has also 
resulted in some successes. Developmental research findings suggest that the family environment includes 
key risk factors for involvement in crime and violence, namely household structure (single-parent 
households, siblings as caregivers, etc.), parent’s age at first birth, and the stability of the home 
6 Cognitive behavioral therapy (CBT) is a short-duration intervention that helps participants recognize and mitigate unhelpful 
thoughts and behaviors, or in a way, gets them to be “thinking about thinking” (Chioda, 2017). Participants learn to recognize 
situations where automatic, intuitive decision-making kicks in and results in negative outcomes, as a means of prevention. The 
CBT-based becoming a Man program for minority male youth in Chicago received considerable attention from the Obama 
administration and the press after an impact evaluation showed very positive results. Participation in the program raised an index 
of school engagement and performance by 0.14 and 0.19 standard deviations in the program and follow-up years, with a resulting 
10-23 percent increase in the graduation rate relative to the control group. Program participants also experienced a 44 percent 
reduction in violent crime arrests and a 36 percent decline in arrests related to vandalism and weapons crimes during the program 
(Chioda, 2017). CBT has been successfully implemented in other cities in the United States. 7 The original Guerra 2013 study could not be located.
8 Among the US programs, the most well-known are probably the perinatal home visitation programs, such as the Nurse-Family 
Partnership, and the HighScope Perry Preschool Program for children living in poverty in Chicago. Program that target the early 
years help build human capital by investing in the health, and cognitive and psychosocial skills needed for adequate development, 
which determines a person’s life outcomes for many years to come (Chioda, 2017).
6 
environment (Chioda, 2017). Family history and criminal background, particularly of the parents, are 
among the strongest predictors of involvement with criminality, even stronger than income or 
employment status (Chioda, 2017). Gang membership has also been shown to be a significant risk factor 
for offending and involvement in delinquency (see discussion in Thornberry et al, pp. 955-6). Improving 
parent-adolescent communication can lead to better mental health, better social functioning and more 
positive behavioral outcomes (UNICEF, 2017).
The most common types of interventions that seek to help families achieve better functioning include functional 
family therapy, multisystemic therapy, and brief strategic family therapy. Through therapy, these interventions 
seek to replace dysfunctional family behaviors with healthier ones to improve antisocial behavior in youth. 
Participating families will have one or more youth who are struggling with serious antisocial behavior or are 
even already in trouble with the law and will work with a trained therapist over the course of three months 
or longer. Each of these three types of family-based therapies have been evaluated multiple times and were 
found to reduce incidences of anti-social behavior, arrests and recidivism (Chioda, 2017). However, Chioda’s 
review does not include the Littell, Popa and Forsythe (2005) study mentioned earlier, which found no 
statistically significant effect of multisystemic therapy on arrests and convictions. 
Some programs include more than one strategy and component. The Gang Reduction and Youth 
Development (GRYD) program in Los Angeles includes six interrelated approaches, which were: 1) 
primary prevention (for instance, a Gun Buy-Back program and community forums where GYRD staff 
present information on gang risk factors), 2) secondary prevention (targeting youth ages 10-15 who are 
assessed as high risk), 3) intervention – family case management (youth ages 14-25 who are engaged in 
gang activity), 4) intervention – incident response (immediate response to gang-related violent incidents, 
to control rumors and mitigate tensions that might lead to further issues such as retaliatory violence), 5) 
community engagement (engage the community and police in a community policing capacity, and 
incorporating community engagement into all of GYRD’s work), 6) suppression (the GYRD office 
continually communicates with law enforcement). Therefore, this program includes family-based 
programming but also goes beyond that to include five other pillars seen as essential to violence prevention 
(Cahill et al, 2015). 
An evaluation of this program found that youth clients and parents improved family functioning and problem￾solving over the course of the study. The process evaluation clearly showed very important shifts in behavior 
and perception. However, when assessing changes in outcomes that could be attributed to the intervention, 
the study had mixed findings. While there were large reductions in the incidence of crime and violence and 
a shrinking of crime hot spots, these general trends were not limited to the treatment communities. The 
evaluation found no statistically significant difference suggesting that treatment communities did better than 
the control group (Cahill et al, 2015). Hence, although the evaluation generally showed encouraging findings 
in terms of reported improvements by the participant youth and the community, it is not altogether clear 
that crime levels changed as a consequence of the program. 
The Proponte Más program in Honduras, which was also designed by Creative Associates, was based on 
the family systems secondary prevention model implemented in Los Angeles. By strengthening family 
relationships and youth support and supervision, the program sought to build resistance to gang 
recruitment. Individual and group activities were carried out through seven phases to promote change in 
youth high-risk behaviors inside and outside the family environment. The group activities were designed 
7 
to observe and intervene in high-risk behaviors by the youth in an environment outside of the family. 
Although the Proponte Más program is similar to Family Matters in terms of the family counseling provided 
to participants, there are important differences. For instance, Proponte Más worked with referrals instead 
of blanketing the communities and visiting schools to implement YSETs, as was done for Family Matters. 
In addition, with Proponte Más family counselors applied the YSETs instead of enumerators.
A total of 300 youth were referred to Proponte Más, and 108 were found to be at high risk and eligible 
for treatment using the risk assessment tool. Out of these 108, 81 went into the treatment group receiving 
the intervention, and 27 in the control group. An evaluation showed a reduction in the percentage of 
youth scoring at-risk for different risk factors, before and after the intervention (at 6 months and 1 year). 
The risk factors that had the greatest reduction were: inadequate supervision, critical life events, the 
tendency to take more impulsive risks, and neutralization of guilt (Creative Associates International, 2015).
Despite these positive findings, the evaluation was not an experimental study with a control group and 
estimation of the counterfactual. 
While many interventions have been studied in isolation, a multi-year, multi-method and multi-country 
evaluation measured the effect of USAID’s efforts in combatting crime and violence in Central America. 
The list of activities and programs evaluated included planning by municipal-level committees, crime 
observatories and data collection, better environmental design (such as street lighting and cleaned up 
public spaces), programs for at-risk youth, and community policing. Both quantitative and qualitative data 
pointed to a significant improvement in treatment communities with the program. By comparing treatment 
and control communities, the most significant effects identified were a difference in the reported cases of 
murder (51 percent), extortion (51 percent), people avoiding walking through dangerous areas (35
percent), sales of illegal drugs (25 percent), robberies (19 percent), community organization to prevent 
crime (18 percent), and report of young people in gangs (14 percent). These incidents were measured as 
reported to interviewers, not to the police (Berk-Seligson et al., 2014).
Another comprehensive study, a meta-review that examined thirty different strategies, also provided 
important insights and lessons on what works in reducing violence (Abt and Winship, 2016). The authors 
identified elements of effectiveness, which seemed to be common among successful interventions: focusing 
on the people, places and behaviors that are most at risk for violence, being proactive in preventing 
violence, working together with partners and creating a legitimate loop between formal (police) and 
informal social control (communities), good capacity in terms of resources and good implementation, and 
a clear theory of change (Abt and Winship, 2016). Of the 30 different strategies examined in their review, 
two stood out as particularly successful: focused deterrence and CBT.9 
9 Focused deterrence is also known as the “Pulling Levers” strategy, and consists of: 1) selection of a crime problem, usually youth 
or gun homicide; 2) identify and analyze they key group of offenders and their behaviors; 3) assemble a multi-sector task force 
that includes the police, service providers and community representatives; 4) target these groups for special enforcement 
operations, aiming to alter their behavior by any legal means necessary; 5) supplement the law enforcement work with legitimate 
offers of assistance to these groups, and engage the “moral voice of the community:; 6) communicating clearly, often, and face￾to-face with these groups to let them that they are under scrutiny, and their actions will have either positive or negative 
consequences, individually and for the whole group. Therefore, this strategy reaches out to offending groups directly, tracks and 
communicates with them clearly and often, and by “pulling every lever” with credible threats of enforcement and assistance, it 
seeks to alter their behavior. This intervention also has a clear focus, for instance on reducing the number of shootings, rather 
than seeking to stop all crime or convince members to leave their gangs. In Chicago, participants were 23% less likely to be 
8 
USAID CFYR PROJECT BACKGROUND AND 
THEORY OF CHANGE
In response to crime and violence in the Eastern and Southern Caribbean (ESC), in 2016 USAID launched 
the Youth Empowerment Services (YES) project, which includes the CFYR initiative. Implemented by 
Creative Associates, CFYR employs a public health informed, place-based strategy in fifteen communities 
in Guyana, St. Lucia, and St. Kitts and Nevis. Communities were selected by CFYR through a multistep 
process that involved ranking communities using available quantitative crime and demographic indicators 
followed by a qualitative analysis entailing focus groups, interviews, and asset mapping in potential 
communities, and concluding with validation of results with stakeholders.10
Consistent with the public health approach, CFYR programming includes primary prevention, focused on 
the general youth population, secondary prevention, focused on youth at a higher risk level, and tertiary
prevention, aimed at those who have already been involved in crime and violence. This impact evaluation 
focuses on CFYR’s secondary prevention efforts, rather than CFYR as a whole. The secondary prevention 
family model that CFYR is based on was first developed by the Los Angeles Mayor’s Office of GRYD
between 2009 and 2011. As mentioned above, it was one component of that city’s comprehensive strategy 
to reduce gang activity. In 2012, the city of Los Angeles and USAID signed an agreement to share best 
practices in violence prevention used in Los Angeles to inform programming in Central America, Mexico 
and the Caribbean. The Northern Triangle of Central America began in 2013 with a pilot in five high risk 
communities in Tegucigalpa, Honduras. Later, USAID funded Proponte Más, also in Honduras, to 
implement six-month and one-year cycles of treatment with families in the secondary and tertiary levels 
of risk (CFYR Family Counseling Training Manual, 2017).
The secondary prevention approach is grounded on family system theory and practice, including Structural, 
Multigenerational, Communications and Strategic approaches. The theory of change postulates that all the 
risk factors and associated behaviors are reinforced by relations within the family. If family member 
behaviors and practices associated with risk factors can be identified and modified, family dynamics and 
relations should improve and risk factors and associated negative behaviors should be reduced (CFYR 
Family Counseling Training Manual, 2017). Other factors that have been identified as being important for 
improving anti-social behavior include increasing youth’s resilience factors, such as a more stable home 
environment, the presence of positive male role models, higher self-esteem, peaceful conflict resolutions 
skills, and decreasing overall risk factors, such as low literacy levels, unemployment, domestic violence and 
child abuse.
The risk assessment tool that is used by these programs is the above-mentioned YSET, and it measures 
the accumulation of risk factors at individual, peer and family level domains (Hennigan et al., 2013). Youth 
are eligible for programming if they have four or more of the nine risk factors, all of which are weighted 
equally. The nine risk factors included in the Caribbean YSET tool are: Weak parental monitoring, Critical 
life events, Impulsive risk taking, Guilt neutralization, Negative peer influence, Peer delinquency, Family 
involved in shootings and 32% less likely to become a gunshot victim in the year after treatment. Focused deterrence has been 
successfully implemented in other cities in the United States as well (Abt and Winship, 2016). 10 CFYR (March 2017). Draft CFYR Community Selection: St. Lucia.
9 
antisocial influence, Self-reported delinquency, and Risky group behaviors.
After the initial YSET assessment, the intervention is carried out for a year in a collaborative relationship 
between the family, their counselor, and a strategy team that provides on-going consultation. After six 
months, the youth is reassessed using the YSET Retest (YSET-R), and after another six months of 
treatment the youth receives the final YSET-1 year assessment. These two six-month cycles are each 
composed of six phases, which last one month each. During each phase, the counselor meets with the 
family at least twice and meets at least once individually with the eligible youth. After each family meeting, 
family members are given “homework” that is designed to identify emotional assets, build cohesion at the 
multi-generational vertical family level, and improve problem-solving capacity at the horizontal family level. 
The seven phases of intervention are: Referral and collaboration, Building agreements, Redefining, 
Celebrating changes, Integrating, Next level agreements, and Re-evaluation (CFYR Family Counseling 
Training Manual, 2017). 
10
EVALUATION PURPOSE, QUESTION, AND 
METHODOLOGY
EVALUATION PURPOSE
The purpose of this evaluation is to measure the impact of the CFYR secondary prevention intervention 
on participating youth on overall risk and across nine risk factors. In addition, the evaluation team designed 
a caregiver survey to gather more background information on the families, which allows us to see if the 
intervention impacts caregiver behaviors and attitudes, providing variables that can be used as control in 
regressions measuring the impact of the program on youth.
EVALUATION QUESTIONS
This evaluation seeks to answer two core questions: 1) Do secondary prevention efforts impact the 
participating youth’s overall risk, crime and violence-related behavior, and their individual risk and 
resilience factors? Why or why not? and 2) Do secondary prevention efforts improve the family 
environment? Why or why not? The first question will be answered primarily with YSET data analysis, and 
the second primarily with information collected from the caregiver survey.
EVALUATION DESIGN
The evaluation team considered many options when designing an impact evaluation of the CFYR program, 
and, in collaboration with colleagues at CFYR and USAID, ultimately decided to compare youth eligible 
for secondary prevention efforts inside 15 treatment communities selected by CFYR. Given the small size 
of St. Lucia and St. Kitts and Nevis and other donor interventions in Guyana, it was not possible to identify 
an adequate number of eligible high-risk communities to randomize treatment at the community level or 
select comparison communities. However, it was possible to identify youth eligible for treatment and 
randomize those youth into treatment and comparison groups. The evaluation is, therefore, a multi-site, 
RCT without comparison communities but with comparison youth within treatment communities. Youth 
randomized into the control group are not denied treatment but will be able to obtain treatment in the 
second year of CFYR programming. CFYR and its partners lacked the capacity to treat all eligible youth 
at the same time, and this delayed treatment approach minimizes the ethical considerations to withholding 
treatment for some youth. 
THE YSET AND SELECTION OF YOUTH
In the ESC, the YSET was used as an assessment tool to determine eligibility for treatment and by the 
evaluation team as an evaluation tool to measure changes on key outcomes over time.11 Prior to the 
intervention, CFYR and the National Opinion Research Center at (NORC) at the University of Chicago, 
working in collaboration with the evaluation team and data collection partners, conducted an extensive 
11 In other contexts, Creative Associates also uses the YSET as a means by which to engage with families, begin a relationship 
with the family, and enter into the “family unit.” This trust building was undermined by emphasizing the evaluative aspect of the 
YSET and having enumerators rather than counselors administer the YSET. 
11
outreach campaign in all 15 communities to reach the population of youth aged 10 to 17. This included 
outreach through schools both inside and outside of targeted communities and door-to-door canvassing
to target youth out of school or that were previously missed for any reason.12 After all youth YSET 
assessments had been conducted, CFYR’s partners at USC established cut-points for eligibility on a 
country-by-country basis and determined which youth were eligible for treatment (see Annex I for greater 
detail). SI stratified households with eligible youth across communities and randomized households and 
their eligible youth into treatment and control groups, ensuring balance across gender, age, and risk 
assessment scores.13 
Figure 1 shows the flow of steps in the identification youth that have been collected for both the 
intervention itself and the evaluation baseline (highlighted in color). CFYR undertook a detailed effort to 
identify the population of youth age 10-17 in the schools and identified 5,845 youth (3,162 in St. Lucia,
1,362 in Guyana, and 1,321 in St. Kitts and Nevis). Other out of school youth were identified through 
household canvassing and made up 6.3 percent of the final sample. While we do not know the population 
of youth out of school, if we assume that they make up at least 6.3 percent of the total population of 
youth age 10-17, then this population was at least 6,236. Assessments required informed consent by the 
youth and by a parent or guardian, and in total, 5,109 youth were assessed, representing as much as 82 
percent of the potential population. Some additional youth may have been assessed through referrals from 
police, schools, or other sources. These were excluded from the evaluation as randomizing these youth 
into the control group might have undermined CFYR’s relationships with its referral networks. Of the 
5,109 youth, 2,086 are in Guyana, 2,076 in St. Lucia and 921 in St. Kitts and Nevis (see Table 1 below.)
Divided by gender, the sample is 51.5 percent female and 48.5 percent male. Youth range in age from 10 
to 17, with 69 percent between the ages of 10 and 14, and 31 percent between the ages of 15 and 17. 
After the YSET, 1,133 (22 percent) were considered eligible for treatment and randomized into control and 
treatment groups. This includes 437 in Guyana, 468 in St. Lucia, and 208 in St. Kitts and Nevis (see Table 1). 
12 Baseline YSET assessments occurred through various iterations. Originally, CFYR was responsible for applying the YSET in all 
three countries in collaboration with data collection partners, including Specialists in Sustained Youth Development and Research 
(SSYDR) (Guyana), the Central Statistical Office (St. Lucia), and other partners in St. Kitts and Nevis. For a variety of reasons, 
DRG-LER co-contractor NORC was later brought on to take over data collection oversight from CFYR in Guyana and St. Kitts 
and Nevis. SYDDR continued to conduct the assessments in Guyana and NORC contracted the Solomon Group (TSG) to 
complete the YSETs in St. Kitts and Nevis. In St. Lucia, CFYR and its partner the Central Statistical Office continued to lead the 
YSET process. All partners participated in the same trainings and -- with some minor, agreed to deviations across countries --
followed the same protocols.
Originally, the evaluation was expected to last two years and as such CFYR followed a careful sampling approach to ensure 
replicability in the second year. This entailed a mixed strategy of randomly sampling youth from schools that served CFYR 
communities, including schools outside of community boundaries, and randomly selecting households in hot spot communities. 
The school sampling had the advantage of accessing numerous youth, and the household sampling had the advantage of identifying 
youth that were not enrolled in school. After it became clear that the number of youths were fewer than expected, the evaluation 
was shortened to one year, and it no longer became necessary to follow a strict random sampling approach that could be 
replicated the following year. As a result, the approach shifted to identifying and YSET assessing the population of youth between 
ages 10 and 17 in the selected communities. Data collection teams targeted the population of youth in the identified schools, 
door to door sampling was increased, and even referral networks were tapped into to help identify youth. We do not know the 
population of youth in the CFYR communities, and as such we cannot know for sure what percent of the population was reached, 
but we estimate that it was not greater than 82 percent. Without information on the full population, we do not know if there 
was sampling bias in the youth and families that were reached or agreed to participate. It would be reasonable to suggest that 
certain high-risk families would be more likely to not be identified or refuse to provide consent. Nonetheless, we feel confident 
that the vast majority of youth were identified and YSET assessed.
13 As the intervention occurs at the household level, it was necessary to randomize youth in the same household to the same 
treatment arm. The randomization process is described in greater detail in Annex I. 
12
We should note that an error was discovered in the final report revision involving the erroneous 
inclusion of 26 invalid observations (not included in the figures above), which were to have been 
dropped for various reasons including duplicates, missing documentation, youth out of the targeted age 
range or targeted community. This affects only some tables and figures that contain observations that 
were erroneously included in the YSET analysis. The error involves 15 observations in St. Lucia, 11 in St. 
Kitts and Nevis, and none in Guyana. While we strive to present the most accurate data, given that this 
error (a) only affects two percent of the eligible youth, (b) does not impact the conclusions below, and 
(c) the final baseline figures will shift slightly in the endline report, we have opted not to revise all the 
YSET tables and figures. 
Figure 1: Number of youth in communities, by eligibility status, and interviewed for each survey
Using the YSET, the evaluation will measure changes in overall risk and changes across all nine risk factors
(Weak parental monitoring, Critical life events, Impulsive risk taking, Guilt neutralization, Negative peer 
influence, Peer delinquency, Family antisocial influence, Self-reported delinquency, and Risky group 
behaviors). The youth in both treatment and control (delayed treatment) groups are to be assessed again 
after six months of treatment and again after 12 months of treatment.
13
CAREGIVER SURVEY
Another key source of data for the evaluation is a survey of caregivers for eligible treatment and control 
youth. This survey is used to (1) collect caregiver-level outcome data to answer the second evaluation 
question, (2) collect additional family-level data that will inform the youth analysis and help contextualize
the family environment in which youth live, and (3) validate youth responses to the YSET.14 In addition, at 
endline this survey will (4) measure additional caregiver output indicators focused on participation in the 
intervention. The survey includes question batteries on the family environment, including family structure, 
family characteristics, and potentially negative family influences (e.g., family involvement in crime and 
violence), community engagement, and family attitudes and behaviors that might change as a result of the 
intervention, including family cohesion, family decision-making attitudes (locus of control in raising 
children), caregiver presence, and knowledge of youth behavior. 
The caregiver survey targeted those with primary caregiver responsibility for the youth. While this was 
most often the mother, it was at times fathers, siblings, aunts, grandparents, and guardians. As seen in 
Table 1, out of the 1,133 eligible youth, we interviewed caregivers for 1,002, or 88 percent (12 percent 
non-response). This included interviews with 946 caregivers; in households with more than one eligible 
youth, caregivers responded to questions about each eligible youth within that household. The remaining 
caregivers either refused to be interviewed or could not be located. The caregiver survey includes 404 
interviews from Guyana, 434 from St. Lucia and 164 from St. Kitts and Nevis. Caregiver non-response 
rate was highest in St. Kitts and Nevis (21 percent). 
Table 1: Sample sizes in various surveys used in this report 
Country Community 
survey
YSET 
youth
Eligible 
youth
Eligible youth with 
caregiver survey
Percent without 
caregiver survey
Guyana 1,602 2,086 437 404 8%
St. Lucia 1,500 2,076 468 435 7%
SKN 1,412 921 208 164 21%
Total 4,514 5,109 1,133 1,002 12%
Data collection for the caregiver survey occurred in Guyana between April and May 2018 and in St. Lucia
and St. Kitts and Nevis between June and August 2018. Interviews typically took place in the household 
and caregivers provided informed consent. Challenges in conducting the survey included caregivers that 
could not be reached, creole language barriers, respondents with serious disabilities, noisy environments, 
and the presence of drugs and alcohol.
14 The caregiver survey was designed to validate youth responses to questions about negative behaviors (e.g., theft, fighting) that 
youth might be dishonest about in their YSET responses. As will be discussed in the findings section of the report, this goal was 
not achieved, as caregivers generally underestimated youth participation in these behaviors. 
14
COMMUNITY SURVEY
A final source of data that informs this report is a survey of community members conducted by CFYR. 
The CFYR community survey was collected between April 2017 and February 2018 with around 300 
surveys conducted in each community, including 1,602 respondents from Guyana, 1,500 from St. Lucia
and 1,412 from St. Kitts and Nevis (see Table 1). Community members were stratified into youth and 
adults and selected randomly, in order to ensure a representative sample of the community. This survey 
is not part of the evaluation design, per se; however, it provides important information on how the 
countries and the communities compare to one another. 
HUMAN SUBJECTS PROTECTIONS
All survey participants provided informed consent to participate in the YSET and caregiver survey and 
youth also required the consent of a parent or guardian. Data was captured electronically on encrypted 
tablets and securely stored. The evaluation team conducted the analysis with de-identified datasets to limit 
opportunities for a breach of confidentiality, while still linking caregivers to eligible youth. 
ASSUMPTIONS AND RISKS
A few assumptions must hold to minimize the risk of arriving at erroneous conclusions about the impact 
of secondary prevention efforts. These are summarized along with mitigation strategies in Table 2. 
15
Table 2: Assumptions and risk mitigation strategies
Assumption and risk Risk mitigation strategy
Attrition will be low and control youth will participate in the follow￾up YSET interviews at the same rate as those in treatment. There 
is a risk that control youth will have less incentive to participate in endline 
data collection. It might also be harder to obtain parental consent for follow￾up interviews with control youth. This could be problematic if those that 
decline to participate are different than those that participate and if 
differences between the control and treatment group are introduced. 
The initial YSET outreach includes detailed contact information including name, 
cellphone, address, parents/guardians’ names, and their cellphone numbers to 
allow the evaluation team to follow-up. Second, by conducting a 6-month and 
12-month follow-up, we hope that regular contact will increase contact rates 
and thereby reduce attrition. Third, the evaluation team will offer control youth 
a small in-kind incentive (e.g. cellphone minutes) to participate in endline. 
Fourth, we will conduct an attrition analysis to detect any bias potentially 
introduced by variable attrition. 
CFYR, NORC and data collection partners will systematically apply 
the YSET according to established protocols at baseline, midline
and endline. The evaluation is not adequately powered to control for 
differences across different data collection units contracted by CFYR and 
NORC across the three countries. 
NORC, CFYR, and their data collection partners have used the same protocols 
and attended one another’s trainings to minimize variation across data 
collection teams. 
CFYR and its partners will respect the YSET scoring and the 
randomization. Failure to do so (e.g., control youth receive the 
intervention) will reduce the comparability between the treatment and 
control groups. 
CFYR has committed to this. The evaluation team will be able to monitor this 
at midterm through CFYR’s monitoring data. 
Social desirability bias and the testing effect of using the YSET at 
baseline, midline and endline can be minimized. We assume that 
measurement bias in key outcome variables will be minimal and that any bias 
will be the same across treatment and comparison youth. This is a risk to the 
evaluation, as youth receiving treatment might feel a stronger social 
desirability bias than those not receiving treatment. 
Social desirability bias was discussed as part of enumerator trainings. The 
evaluation team had hoped to use data from police/schools and from the 
caregiver survey to validate youth responses; however, police and school data 
are inconsistently available, and the baseline shows that caregivers tend to 
underestimate undesirable youth behavior. 
The program is implemented consistently across countries and 
locations. Another risk for the evaluation is implementation problems in 
one or more of the 15 sites. Implementation quality could be compromised 
by problems with a site, with a partner, or with individual counselors. If there 
CFYR trains counselors consistently across the three countries and uses a 
common counselor manual. CFYR has also implemented a monitoring system 
16
are problems with a site or with a partner that result in the elimination of 
one or two sites from the study, then the minimum detectable effect size 
(MDES) will jump considerably (meaning less power). While minor 
adaptations are expected to be made to accommodate different realities 
across countries, we assume no substantial differences in the nature or 
intensity of the CFYR intervention between countries or sites. The greater 
the variation by site, the lower the statistical power. 
across the fifteen sites. The evaluation team will review monitoring data and 
collect qualitative data at the midpoint in the intervention. 
The evaluation will be adequately powered. The evaluation design 
assumed a greater number of youth in the fifteen communities and treatment 
youth than occurred in practice. The design assumed 800 treatment youth; 
however, for a variety of reasons only 450 are expected to receive 
treatment. This will reduce the study’s power and undermine the evaluation’s 
ability to speak confidently about heterogeneous effects across gender and 
age group. This number will be further eroded by attrition – both from the 
program and from the survey – which could reach 20 percent. 
As discussed above, SI, CFYR, and NORC have taken actions to reduce 
attrition. There is little that can be done beyond this. 
Spillover is minimal and can be understood through additional data 
analysis. Spillovers represent a potential challenge for the evaluation. 
Because treatment and control youth live in the same communities and might 
have overlapping social networks, it is possible that a treatment youth will 
(a) be positively affected by the program, (b) interact with control youth, and 
in so doing (c) have a positive influence on those control youth. This is 
expected to be a greater problem in smaller communities. 
It will be necessary to explore the potential of spillovers through alternative 
means. As noted above, first, we will ask control youth at end-line if they have 
noticed positive or negative behavior changes in their friends across a few 
different types of behaviors. Then we will attempt to link any positive changes 
to the program. In the analysis, we will then compare those who are more 
likely to experience spillover effects to those who are less likely to 
experience spillover effects. Second, we will explore the issue of spillover 
qualitatively in focus groups with youth receiving treatment and those not 
receiving treatment.
17
COMMUNITY CONTEXT: RELEVANT 
FINDINGS FROM THE CFYR COMMUNITY 
SURVEY 
CFYR’s community survey provides important information about the local environment, information 
which could not be obtained from available national-level data. This survey is particularly helpful to the 
evaluation because it allows us to understand differences between communities and countries on key 
indicators. If communities were found to differ mostly between countries, rather than within them, it 
would be difficult to justify an evaluation design that examines all three countries together. In order to 
provide the context of the communities in which the CFYR project is working, only some select data from 
the community survey is presented here. Relevant information about crime, policing, the rule of law, and 
community support is given, using summary statistics, indices and some maps. 
Our main conclusion is that although there are differences between countries, there is typically greater 
variation across communities within countries. Except for higher perceptions of crime in Guyana, greater 
resources for youth in St. Kitts and Nevis, and some demographic differences (e.g., larger family sizes in 
Guyana), practically all the metrics estimated from the baseline survey show that the three countries are 
quite comparable on the variables of interest. 
In the discussion that follows, we first begin with reported crime victimization. We then create and 
examine a series of indices measuring different aspects of the community context. The indices provide a 
single metric based on several questions in the community survey that measure characteristics of the 
social fabric, attitudes towards crime and law enforcement and agency within those communities. The 
following indices are calculated:
• Perception of crime prevalence
• Social capital, subdivided into two groups 
• Attitudes towards the law
• View of the police
• Resources for youth
• Self-efficacy
These indices were constructed using polychoric principal component analysis (PCA), a method for 
extracting a single indicator from a group of correlated variables. This single indicator, presented in the 
form of a unitless score, simplifies analysis and interpretation of the larger group of variables, while 
retaining a reasonable representation of the information contained in the original variables. Details on this 
methodology, including the principal component factor loadings (the correlation of each underlying 
variable to the resulting index), are given in Annex V. 
18
CRIME VICTIMIZATION
One of the most critical aspects of risk that must be assessed in order to understand the CFYR 
communities is to know how unsafe residents feel in these communities. The instrument included several 
questions on crime victimization. In particular, respondents were asked whether they had been a victim 
of any type of crime in the past 12 months or whether any other person living in the same household had 
been a victim in the same period. When a crime was reported, respondents were asked to specify the 
type of crime according to the categories. In most countries where many crimes go unreported, such 
victimization surveys tend to offer better measures of crime than official statistics. Nonetheless, 
victimization surveys suffer from recall error, and certain crimes with a social stigma, such as rape, sexual 
assault, and domestic violence, are frequently underreported. 
Figure 2 shows that the average rate of self-reported crime victimization is similar in each of the 
three countries. Responses are given for each community and for each country in aggregate; 
the percent of respondents who did not answer the question is captured on the right side of 
the figure in purple. Victimization ranges from six percent in St. Kitts and Nevis to nine percent 
in Guyana, with greater variation across communities, ranging from three percent in Sandy 
Point, St. Kitts and Nevis to 14 percent in Sophia, Guyana. 
Figure 2: Crime victimization reported by respondents in the last 12 months by country and 
community
19
We find a similar pattern in reported victimization of respondents’ family members in the last 12 months. 
As shown in Figure 3, the average rate of reported household crime victimization ranges from four percent 
in St. Kitts and Nevis to seven percent in Guyana, and from three percent (Corriverton) to ten percent
(Sophia). In sum, victimization figures suggest relative similarities between the target countries with greater 
variation across communities. 
Figure 3: Crime victimization among other members of the household in the last 12 months by 
community and country
It should be noted that reliability of self-reported victimization is difficult to ascertain in many contexts. 
For this reason, the proportion of “missing” responses are also included in the figures. This percentage 
averaged around one percent for self-reported crime victimization and around four percent for reports 
of household crime victimization; however, the proportion of respondents who declined to respond at all 
was substantially highest in the St George community in Nevis (16 percent) than any other community. 
Figure 4 shows what specific crimes were reported by victims, including 1) unarmed robbery, no assault 
or physical threats; 2) Unarmed robbery with assault or physical threats; 3) Armed robbery; 4) Assault 
but not robbery; 5) Rape or sexual assault; 6) Domestic violence; 7) Kidnapping; 8) Vandalism; 9) Burglary 
of your home while you were not at home; 10) Burglary of your home while you were at home; 11) 
20
Extortion/blackmail; 12) Bullying; and 13) Other crimes. Only two percent of respondents reported being 
the victim of two crimes, and fewer than one percent reported being the victim of three or more crimes, 
so most observations correspond to one individual. Zero respondents stated that they had been a victim 
of kidnapping; therefore, this crime is not presented on the chart. 
Given that only a small percentage of respondents report victimization, Figure 4 provides a count of the 
number of incidences of each crime rather than a percentage of crimes. Given that there is some variation 
in the number of respondents in each community, the (broken) x-axis shows the total respondents in each 
community. For example, while we observe a large incidence of reported crime in Sophia, it is important 
to recognize that there were more respondents in this community. The most common crimes are 
unarmed robbery and burglary. While the graph illustrates the heterogeneity in type of crimes across 
communities, it should be clear that the incidence overall is quite low.
Figure 4: Crime victimization: Count of reported crimes of different type by community
Figure 5 provides a similar count (rather than percentage) of reported crime incidence by country. 
Although there is some variance across countries, the general composition of crime is quite similar. As 
noted above, a large part of the reported crime is comprised of robbery and burglary. Armed robbery is 
reported more frequently in Guyana; St. Lucia stands out for a greater share of bullying; and in St. Kitts 
and Nevis there is a larger share of domestic violence. There appears to be a greater number of reports
of rape and domestic violence cases in St. Lucia than the other two countries, although it is not clear if 
this is the result of a higher incidence rate or a greater willingness to report these stigmatized crimes. 
21
Figure 5: Crime victimization: Count of reported crimes of different types by country
PERCEPTION OF CRIME PREVALENCE
We created a Perception of Crime Prevalence Index from the questions listed below, with responses 
solicited on a Likert scale of Never to Always (with one exception, for which the scale was reversed). A 
higher score indicates a perception of higher crime prevalence and a lower score indicates a perception 
of lower crime prevalence. 
How often do the following situations occur within your community?
• Vandalism or property damage
• People using or selling drugs
• People getting robbed 
• People getting sexually assaulted or raped
• During the last year, do you think that crime in your community has 
[increased/decreased/stayed the same]? (scale reversed)
• How safe do you feel walking in your community during the daytime?
• How safe do you feel walking in your community after dark?
Figure 6 shows each community’s and each country’s average score on this index. Individual scores ranged 
from 2.60 to 14.76. Error bars indicate a five percent error margin for the value of each average. In other 
words, assuming no bias, we are 95 percent confident that the true population average falls within the 
range shown by the error bands. It is important to note that the index values do not have a substantive 
22
interpretation. The scores are standardized according to a normal distribution, such that the mean is the 
underlying central tendency of all the respondent attitudes, according to an underlying latent variable 
(assumed to be a normal distribution too) captured in their responses to the individual items on the 
categorical scales. Nonetheless, the index is still helpful to compare differences in crime perceptions across 
communities and countries. As can be seen in the figure, crime perceptions are somewhat higher in Guyana 
than the other two countries, with the worst crime perceptions in Sophia, Guyana, followed by Basseterre, 
St. Kitts and Nevis, followed by Castries, St. Lucia. As with crime victimization, there is more homogeneity 
across countries than individual communities. 
Figure 6: Average Perception of Crime Prevalence Index scores by community and country 
Interestingly, the communities with the highest perception of crime prevalence do not always correspond 
to those with the highest reported victimization. For example, as shown in Figure 7, Anse-La-Ray in St. 
Lucia has a low average victimization rate but a high average perception of crime score. In St. Kitts Nevis, 
the communities of St. John (Nevis) and Basseterre (St. Kitts) found themselves on opposite ends of the 
spectrum from one ranking to another. In Guyana, community rankings for both metrics were more 
closely aligned for all five communities. 
23
Figure 7: Rankings: Perceptions of crime prevalence versus reported crime victimization by 
community
OVERALL SOCIAL CAPITAL INDEX
There are several questions in the survey that serve as a proxy for “social capital,” a concept with varying 
definitions, but frequently understood as social networks and their associated norms of trust and 
reciprocity.15 We divide social capital into an Overall Social Capital Index and a Youth Social Capital Index. 
While there are numerous approaches to measure social capital in surveys, questions usually hinge on 
issues of interaction, trust, and reciprocity. The first index is derived from the following questions: 
• How long have you lived in this community?
• Would you say that people in this community are very trustworthy, somewhat trustworthy, not 
very trustworthy, or untrustworthy? (scale reversed)
• What is the likelihood that people from your community would try to stop a fight if one broke 
out in front of their house, with someone being beaten? (Very Unlikely to Very Likely)
A high score corresponded to high social capital, and a low score to low social capital. The resulting index 
showed a high level of uniformity across both countries and individual communities. Individual scores 
ranged from 1.95 to 9.23, but country and community averages ranged only from 5.69 to 6.62, indicating 
both homogeneity and a relatively high feeling of community trust and cohesion. 
15 See for example Robert D. Putnam (2000) Bowling Alone: The Collapse and Revival of American Community. New York: 
Simon and Schuster.
24
Figure 8: Average Overall Social Capital Index scores by community and country
YOUTH SOCIAL CAPITAL 
The Youth Social Capital index was constructed from a different set of questions and a different subset of 
respondents. Only responses from interviewees aged 10 to 17 years were used to create this metric, and 
the index compiled responses to the following set of input questions given on a Likert scale from Strongly 
Disagree to Strongly Agree. As with the Overall Social Capital Index, a higher score corresponds to 
stronger social capital: 
• I feel like a member of this neighborhood 
• I belong in this neighborhood
• I have a say about what goes on in my neighborhood
• People in this neighborhood are good at influencing each another
• I feel connected to this neighborhood
• I have a good bond with others in this neighborhood
As seen in Figure 9, scores on this index were somewhat less uniform than above. It is important to 
note that sample sizes were smaller and therefore the margin of error for each estimate was greater, 
particularly in Guyana. Individual scores range from 2.39 to 16.75, and country and community scores 
25
ranged from 9.77 to 13.20. We observe both similarity across communities and countries and relatively 
strong feelings of trust, belonging, and respect in their neighborhoods.
Figure 9: Average Youth Social Capital Index scores by community and country
Figure 10 compares how communities ranked in terms of the Overall Social Capital Index versus the youth 
index. Although each index was constructed from different data, the second index was youth-specific and 
can be thought of as a proxy for how well young people feel supported and nurtured in their communities.
Making this assumption, a comparison of the two can draw out disparities in terms of youth perception 
of their communities versus the overall population, and whether they feel exceptionally supported or 
exceptionally isolated. 
Community rankings for both indices were relatively similar, with the notable disparities of Corriverton, 
Guyana and Soufriere, St. Lucia; both communities had high rankings on the first index, and low rankings 
on the second, indicating that there may be some rift in how young people in these communities feel 
integrated and supported. It is possible that some of the inconsistencies between the two sets of averages 
are explained by the large margins of error in the youth averages. 
26
Figure 10: Rankings: Average Overall Social Capital Index and Youth Social Capital Index scores by 
community
ATTITUDES TOWARD THE LAW
The Attitudes Toward the Law Index was constructed from responses to the following set of questions. 
Responses were solicited to several statements on a Likert scale of Strongly Disagree to Strongly Agree. 
This scale was reversed (with one exception) so that a higher score indicates positive attitudes toward 
the law and a lower score indicates negative attitudes toward the law: 
• The law does not protect my interests (scale reversed)
• If the police can break the law, then so can I (scale reversed)
• I try to obey the law, even if it goes against what I think is right
• I do not trust the law to protect me (scale reversed)
• Most people would commit crime if they know they would not get caught (scale reversed)
• In order to catch criminals, police should be able to occasionally bend the rules (scale reversed)
The Attitudes Toward the Law index showed uniformity across both countries and communities. 
Individual scores ranged from 2.10 to 15.07, but country and community scores clustered very closely 
around a value of 8.91 with very little variation. In fact, when considering the error bands, the communities 
are barely distinguishable. This means that most of the variance in this perception comes from individual 
respondents, regardless of the community they live in. 
27
Figure 11: Average Attitudes Towards the Law Index scores by community and country
VIEW OF THE POLICE 
We derive the View of the Police Index from responses to the following set of input questions in the form 
of a Likert scale of Very Dissatisfied to Very Satisfied. A high score indicates a positive view of the police, 
and a low score indicates a negative view of the police: 
• Responsiveness of the police to incidents of crime and violence
• Police patrols of the community
• Approachableness of the police if you require assistance from them
• People in my neighborhood are willing to call the police to help their neighbors who are in 
trouble (scale reversed)
As seen in Figure 12, the View of the Police index shows general uniformity across countries. Individual 
scores ranged from 1.94 to 13.03, and community scores fall under a much narrower range from just 6.41 
to 8.35. 
28
Figure 12: Average View of the Police Index scores by community and country
RESOURCES FOR YOUTH
The Resources for Youth Index was constructed from several questions asking respondents to rate their 
satisfaction with youth services provided in their communities on a scale from Strongly Dissatisfied to 
Strongly Satisfied. A high score indicates strong satisfaction with services, and a low score indicates strong 
dissatisfaction with services.
• Programs for youth not enrolled in school
• School programs meeting the needs of students in the community
• Care and support for residents injured in crime and violence by hospitals/clinics
• Programs to help youth recently released from prison
• Victim support services for residents hurt by crime and violence
The Resources for Youth Index also showed general uniformity across both countries and communities, 
with a somewhat higher greater satisfaction in St. Kitts and Nevis. Individual scores ranged from 2.22 to 
15.42 and community averages range between 5.00 and 7.71. This indicates a neutral or low satisfaction 
for resources for youth across all communities and countries. 
29
Figure 13: Average Resources for Youth Index scores by community and country
SELF-EFFICACY
Last, we constructed a Self-Efficacy Index from the responses to a series of questions about one’s own 
sense of agency (or lack thereof). Interviewees were asked to respond on a scale of Strongly Disagree to 
Strongly Agree. A high score on the index corresponds with high self-efficacy, and a low score corresponds 
with low self- efficacy. 
• I have little control of things that happen in my life (scale reversed)
• I do not have any solutions for some of the problems I am facing (scale reversed)
• I cannot do much to change things in my life (scale reversed)
• When I face life difficulties, I feel helpless (scale reversed)
• I feel my life is determined by others and fate (scale reversed)
• I believe things happening in my life are mostly determined by me
• I can finish almost everything that I am determined to do
This scale results in similar average scores across both communities and countries, and relatively high
scores overall. The minimum individual score was 2.61, and the maximum individual score was 18.29, with 
30
a country/community average of 13.64. Average scores suggest that respondents are generally able to 
determine the course of their lives and make changes if they desired. 
Figure 14: Average Self-Efficacy Index scores by community and country
INDEX SUMMARY
Although there are differences between countries, there is typically greater variation across communities than 
between countries. This is a positive finding for the evaluation design, as it suggests comparability between the 
three countries on key indicators. In addition, while there are differences across communities in some of the 
indicators, the similarities are more common (see Table 3).16 This is a promising feature of the study, in that it 
suggests most variation comes from the individual youth and not effects between communities. 
16 We define “similar” as the majority (two or more for countries and four or more for communities) of error bars 
overlapping, and we define "divergent" as the majority of error bars failing to overlap.
31
Table 3: Index Summary
INDEX COUNTRIES COMMUNITIES
Crime Victimization (percentages) Similar Divergent
Perception of Crime Prevalence Similar Divergent
Overall Social Capital Similar Similar
Youth Social Capital Similar Divergent
Attitudes Towards the Law Similar Similar
View of the Police Similar Similar
Resources for Youth Similar Similar
Self-Efficacy Similar Similar
VISUALIZATION OF COMMUNITIES
The community survey provides important contextual information about the intervention sites. In addition, 
it is possible to use the geocoding of the survey respondents to understand that, notwithstanding some 
territorial dispersion, these are relatively homogenous territorial units with characteristics that make them 
comparable with each other. In particular, it is possible to map some characteristics related to stability and 
permanence of the residents, as measured by the years they have lived in the community; the trust towards 
neighbors; how prevalent crime may be as measured by reported victimization; whether respondents feel 
unsafe when walking in their community after dark; and some basic socio-demographic patterns that may be 
associated with risky youth behavior or less stable family environments, specifically the presence of single 
mothers who are unemployed, or households characterized by employed married couples. 
As seen in Figure 15, in St. Lucia, the communities can be clearly visualized throughout the island, with a 
higher territorial concentration of households in Castries, where people live in very close contiguity, while 
in Vieux Fort there is a greater distance between households. The communities are very stable, with very 
few respondents reporting having lived in the community for three years or less. The neighbors in the 
communities are perceived as trustworthy, there is very little victimization and, in general, residents feel 
safe walking after dark. The family composition shows a high prevalence of single mothers that are 
unemployed, and a relatively low presence of married employed households. There is no particular 
territorial distinctiveness in these patterns between communities, which confirms the previous findings 
suggesting that the communities are somewhat similar in their demographics and feelings of insecurity and 
social trust. 
Figure 16 presents similar images for St. Kitts and Nevis. The figure shows that communities in St. Kitts 
and Nevis are also very stable in terms of the years of residence, but they exhibit less trust across 
communities, particularly in Basseterre and Sandy Point. There is very little victimization, but in contrast 
with St. Lucia, a lot of the residents feel unsafe when walking in their neighborhoods after dark. In St. Kitts 
and Nevis there are more stable family structures; only a few scattered households report having single 
unemployed parents, while there is greater prevalence of employed married households.
In Guyana, the visualization in Figure 17 only shows Georgetown, although the patterns are similar in 
Corriverton. There is a marked difference in terms of residential stability, in the sense that there are many 
32
households that have only been living in the community for three years or less. Sophia is somewhat 
different in that most respondents report living there for a longer time. There is a strong feeling of 
insecurity in terms of the possibility of walking in the community after dark; however, the communities 
are relatively similar to those in St Kitts and Nevis and St. Lucia, in terms of the actual victimization, with 
no particular concentration in specific places, perhaps with the exception of La Penitence. Finally, the 
composition of the households in terms of single unemployed versus married employed is similar to what 
was found in the other countries, with no particular distinctiveness for any specific community. 
The territorial visualizations suggest that, in general, while there are differences between communities, 
these do not imply distinctive dynamics of insecurity or risk for youth in particular locations, nor is there 
a suggestion that specific streets or blocks within the communities are particularly problematic in terms 
of their public safety environments. There is no particular concentration of victimization in specific 
territories within the communities, nor is there any particular demographic pattern that may be found 
within a few blocks inside a community. 
33
Figure 15: St. Lucia community risk factor visualizations with captioned occurrences represented by 
a red dot and non-occurrences by a black dot
3 years or less living in the Community People in this community are untrustworthy
Victimization last 12 months Feel unsafe walking in community after dark
Single unemployed households Married employed households
34
Figure 16: St. Kitts and Nevis community risk factor visualizations with captioned occurrences 
represented by a red dot and non-occurrences by a black dot
3 years or less living in the Community People in this community are untrustworthy
Victimization last 12 months Feel unsafe walking in community after dark
Single unemployed households Married employed households
35
Figure 17: Georgetown, Guyana community risk factor visualizations with captioned occurrences 
represented by a red dot and non-occurrences by a black dot
3 years or less living in the Community People in this community are untrustworthy
Victimization last 12 months Feel unsafe walking in community after dark
Single unemployed households Married employed households
36
FINDINGS BASED ON THE YSET
ASSESSMENT
QUESTION 1: 
DO SECONDARY PREVENTION EFFORTS IMPACT OVERALL 
YOUTH RISK, CRIME AND VIOLENCE RELATED BEHAVIOR, AND 
YOUTH RISK AND RESILIENCE FACTORS? WHY OR WHY NOT?
Our analysis of the YSET data included disaggregation based on country, sex and age. According to the 
United Nations Children’s Fund (UNICEF), results of research done on adolescents should be 
disaggregated for reporting, by sex, age and locality. We have followed their age divisions and separate 
youth into those 10-14 years of age and 15-17. The importance of age disaggregation stems from the fact 
that some interventions may be appropriate for younger adolescents but less so for older adolescents and 
young adults, or vice-versa (UNICEF, 2017). In the analysis that follows, YSET scales for the total 
population of youth are reported with sex and age disaggregation. Individual questions are typically 
reported in aggregate, and only when there are significant country differences on a certain set of questions 
are those questions disaggregated by country. 
UNDERSTANDING RISK ACROSS THE FULL YOUTH POPULATION
As noted in the methodology section, the method applied by CFYR and their colleagues at USC 
categorizes a YSET respondent to be at risk or not at risk across nine risk scales, including weak parental 
monitoring, critical life events, impulsive risk taking, guilt neutralization, negative peer influence, peer 
delinquency, family antisocial influence, self-reported delinquency, and risky group behaviors. Youth that 
are determined to be at risk for four or more of the nine scales are considered to be at a higher level of 
risk and thus eligible for treatment. Table 4 reports the percentage of youth that were found to be at risk 
by country, sex, and age. The findings show that males and older youth (ages 15 to 17) are at a higher risk 
on every single scale, and there is country-level variation across scales. The scales with the highest 
numbers of youth scoring at risk are guilt neutralization and impulsive risk taking. In the following sections 
we examine each of these risk factors in turn. 
37
Table 4: Percent of assessed youth at risk in the overall YSET and on each scale (n=5,109)
Total Country
GUY
Country
SLU
Country
SKN
Gender
Female
Gender
Male
Age
10 to 14
Age
15 to 17
Eligible: at-risk in four 
or more scales 22% 21% 23% 24% 17% 27% 18% 32% 
Weak parental 
monitoring 19% 10% 29% 20% 11% 28% 16% 26% 
Critical life events 24% 25% 20% 29% 22% 26% 20% 33% 
Impulsive risk taking 35% 36% 34% 34% 33% 36% 32% 40% 
Guilt neutralization 40% 45% 41% 24% 36% 43% 35% 50% 
Negative peer 
influence 18% 11% 19% 32% 16% 20% 16% 23% 
Peer delinquency 16% 18% 16% 13% 14% 19% 13% 24% 
Family antisocial 
influence 19% 17% 19% 20% 18% 19% 16% 25% 
Self-reported 
delinquency 23% 20% 24% 26% 18% 28% 19% 31% 
Risky group behaviors 12% 13% 11% 13% 8% 16% 11% 15% 
It is important to note that a higher percentage of at-risk youth in a given country is not necessarily an 
indication of higher risk levels in that country, as the thresholds vary by country. For instance, in the weak 
parental monitoring scale, CFYR and USC determined that youth would be classified as at risk if they 
scored an average of at least 2.5 in Guyana, 2 in St. Lucia and 2.3 in St. Kitts and Nevis. Therefore, a higher 
share of youth scoring at risk in St. Lucia in weak parental monitoring does not necessarily mean that a 
higher share of youth in that country selected “risky” answers. 
Table 5 presents the results of a logistic regression testing the impact of country, sex, age group, school 
status, and ethnicity on risk levels. This analysis allows us to see if there are specific factors that make it 
more likely for certain groups to score at risk, meaning that they are more vulnerable. The table presents 
odds ratios, which are the difference in likelihood of being eligible or scoring at risk on each of the scales, 
by group. For instance, the odds ratio of 4.106*** on the independent variable St. Lucia and the dependent 
variable weak parental monitoring can be interpreted as the odds that a youth in St. Lucia will be 
considered at risk for weak parental monitoring are 4.1 times greater than the odds that a youth in Guyana 
will be considered at risk. In this case, Guyana is not included in the equation and is considered the base 
country by which St. Lucia and St. Kitts and Nevis are compared. The three stars indicate that this
difference is statistically significant at the 99 percent confidence level. 
The results in this table that are most striking are: 1) youth in the islands are much more likely to score 
at risk in the weak parental monitoring and negative peer influence scales compared to Guyana. Because 
of the different risk thresholds set for each country, higher average scores in the islands could stem from 
the lower risk thresholds rather than from real country differences. We tested for this and found that 
when setting the same threshold for the three countries, the differences in average scores are still 
statistically significant but the magnitude is smaller. 2) In other scales, such as impulsive risk taking, guilt 
38
neutralization and peer delinquency, Guyanese youth are more likely to be at risk. When the thresholds 
were set to be the same for the impulsive risk taking and guilt neutralization scales, the differences 
remained similar between Guyana and St. Lucia, but were reversed for Guyana and St. Kitts and Nevis
(with the same thresholds, St. Kitts and Nevis youth are now at higher risk than Guyanese youth in three
out of four cases). For the peer delinquency scale, the threshold is the same for all three countries so 
different average scores reflect real differences in risk levels. 3) Males are more likely than females to be 
at risk on all scales, except for family antisocial influence, where the difference is not statistically significant. 
4) Older youth are more likely to score at risk on all scales. 5) Youth who are out of school are more 
likely to be at risk on most scales. 6) While Amerindians are not statistically significantly more at risk than
Black/Afro-Guyanese youth, respondents who are East Indian/Indo-Guyanese are much less likely to be at 
risk on all scales. For instance, East Indian/Indo-Guyanese youth are half as likely to have had a critical life 
event occur in the last year compared to Black/Afro-Guyanese youth.
Table 5: Odds ratios with Guyana, females, in school, ages 10 to 14, and ethnicity-Black/Afro￾Guyanese as the base (n=5,101) 
SLU SKN Male Ages 15 
to 17
Out of 
school
Ethnicity - 
Amerindian
Ethnicity -
East 
Indian/Indo 
Guyanese
Eligibility 1.006 1.048 1.804*** 2.037*** 1.828*** 0.817 0.503***
Weak parental 
monitoring
4.106*** 2.551*** 3.370*** 1.773*** 2.104*** 1.589* 0.652**
Critical life events 0.620*** 1.034 1.331*** 1.824*** 1.655*** 0.678* 0.500***
Impulsive risk 
taking
0.834** 0.832** 1.126** 1.334*** 1.521*** 0.985 0.655***
Guilt neutralization 0.717*** 0.327*** 1.300*** 1.781*** 1.677*** 0.774 0.597***
Negative peer 
influence
1.771*** 3.565*** 1.296*** 1.618*** 1.279 1.099 0.720**
Peer delinquency 0.777*** 0.611*** 1.491*** 2.066*** 1.220 0.798 0.577***
Family antisocial 
influence
0.981 1.037 1.092 1.665*** 1.519*** 0.899 0.523***
Self-reported 
delinquency
1.192** 1.356*** 1.701*** 1.768*** 1.976*** 0.850 0.561***
Risky group 
behaviors
0.685*** 0.858 2.160*** 1.249** 1.548*** 0.730 0.518***
Note: significance is * p<.1, ** p<.05, ***p<.01. 
When St. Lucia is the base country (reference group) instead of Guyana, we can see that St. Lucian 
youth are more likely to score at risk on a few scales (weak parental monitoring, guilt neutralization, and 
peer delinquency). St. Kitts and Nevis youth are more likely to be at risk on a few others (critical life 
events, negative peer influence, and group risky behaviors).
Given the importance of age and sex, Table 6 explores differences in risk in age and sex for each of the 
three countries using differences in predicted probabilities rather than odds ratios. In this table, we look 
at the differences in risk/eligibility considering the differences in location (country), gender, age group, 
39
school status and ethnicity that exist in the samples. The first column suggests that all else being equal, 
males are ten percentage points more at overall risk than females. This difference is consistent in Guyana, 
St. Lucia, and St. Kitts and Nevis. In terms of individual risk factors, the impact of sex is generally consistent 
across countries once all other factors are considered, with the strong exception of parental monitoring. 
For this risk factor, males are 17 percentage points more at risk than females across all countries; however, 
there is only a ten-percentage point difference in Guyana and a 23-percentage point difference in St. Lucia. 
Predicted differences in risk across age groups are also fairly consistent across the three countries. The
predictive impact of age matters somewhat less for negative peer influence in Guyana (ten percentage 
points) than it does in St. Lucia (11 percentage points). As with the community survey, these generally 
consistent findings are encouraging for the evaluation, as they suggest that -- at least in terms of sex and 
age -- risk dynamics are similar across the three countries. 
Table 6: Differences in predicted probabilities for each group of respondents scoring at risk/eligible
(n=5,101)
Male - 
female
Male - 
female 
in GUY
Male - 
female 
in SKN
Male - 
female 
in SLU
Ages 15 
to 17 - 
ages 10 
to 14 
overall
Ages 15 
to 17 - 
ages 10 
to 14 in 
GUY
Ages 15 
to 17 - 
ages 10 
to 14 in 
SKN
Ages 15 
to 17 - 
ages 10 
to 14 in 
SLU
Eligibility 0.10 0.10 0.10 0.10 0.13 0.13 0.13 0.13
Weak parental 
monitoring
0.17 0.10 0.18 0.23 0.08 0.05 0.09 0.11
Critical life events 0.05 0.06 0.06 0.04 0.11 0.12 0.12 0.10
Impulsive risk 
taking 0.03 0.03 0.03 0.03 0.07 0.07 0.07 0.07
Guilt neutralization 0.06 0.06 0.05 0.06 0.14 0.14 0.11 0.14
Negative peer 
influence 0.04 0.03 0.06 0.04 0.07 0.05 0.11 0.08
Peer delinquency 0.05 0.06 0.04 0.05 0.10 0.12 0.09 0.10
Family antisocial 
influence 0.01 0.01 0.01 0.01 0.08 0.08 0.08 0.08
Self-reported 
delinquency 0.09 0.08 0.10 0.09 0.10 0.10 0.11 0.10
Risky group 
behaviors 0.08 0.09 0.08 0.07 0.02 0.03 0.03 0.02
Figure 18 compares standardized scores of eligible and ineligible youth across all nine risk factors.17 Eligible 
youth have higher average scores on every scale, and in the case of three scales, in particular, they have 
17 Because the scales had different numbers of questions and maximum scores, we standardize the scales to allow for comparison 
in one chart. For example, in the scale Critical Life Events, eligible youth had a mean of four events, compared to a mean of two
events for ineligible youth. Because the number of questions in that scale is eight, with a maximum total of eight events, these 
means translate in the chart to a 0.5 (or 50% of 8) for eligible youth and 0.25 (or 25% of 8) for ineligible youth.
40
much higher average scores. In the self-reported delinquency scale, eligible youth report that they have 
done certain delinquent behaviors in the last six months with much higher frequency than ineligible youth. 
In the scale critical life events, eligible youth are much more likely to have had important and serious (and 
in some cases, traumatic) events happened to them in the last year. And in the scale risky group behaviors, 
eligible youth report with much higher frequency that their peers engage in negative behaviors.
Figure 18: Average standardized YSET scores from 0 to 1 on each scale for eligible and ineligible 
youth (n=5,109)
SCALE B: WEAK PARENTAL MONITORING
In the following sections we examine each scale in detail. Each scale section begins with a table that shows 
the percentage of youth that scored “at risk” for that scale, including all eligible and ineligible youth (this 
is the same information presented in Table 4). In addition to looking at the scales as a whole, we also 
report results for many of the individual questions. As noted in the chart titles, when examining specific 
questions, we use only data on eligible youth, because that is the population of interest for the evaluation. 
When there are interesting or striking results by disaggregating country or gender, these are included. 
The remaining charts containing gender disaggregation can be found in Annex VI. 
The first scale includes six questions on the level of parental monitoring and awareness of youth’s 
whereabouts and who they are with (see Annex II for the full questionnaire). For each of the six questions 
(e.g., how often they return home on time), the youth respondents were asked to select from one of five 
options: always=1, often=2, half the time=3, rarely=4 or never=5. Responses were then averaged across 
the six questions. Youth that score an average of at least 2.5 in Guyana, 2 in St. Lucia and 2.3 in St. Kitts 
and Nevis were classified as at risk in this scale. Table 7 repeats the percentage of youth that were found 
41
to be at risk for weak parental monitoring across countries, gender and age group, with 19 percent of 
respondents overall scoring at risk and considerable variation across the different disaggregated results. 
Females and younger youth report higher levels of parental monitoring of their whereabouts and company.
Table 7: Percent of assessed youth at risk for weak parental monitoring (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Weak parental 
monitoring
19% 10% 29% 20% 11% 28% 16% 26%
When we limit the sample to just eligible youth, we find that a minority of these respondents report that 
they rarely or never tell their caregivers where they are going (18 percent), nor return home at the time 
the caregivers asked (20 percent), and that caregivers do not know where they are when outside of school 
(13 percent) (see Figure 19). Conversely, over 60 percent of eligible youth report that they always or 
often tell their parents or guardians where they are going and that their parents always or often know 
where they are. 
Figure 19: Responses to select questions on parental monitoring among eligible youth (n=1,138)
Similarly, more than half of eligible youth report that their caregivers often or always know who they are 
with (56 percent) and they tell their caregivers where they are going even if not asked (60 percent). The 
vast majority of these respondents said that their caregivers care about where they go and what they do. 
42
Figure 20: Responses to select questions on parental monitoring among eligible youth (n=1,138)
SCALE C: CRITICAL LIFE EVENTS
Scale C includes eight questions about difficult things that happen in young people’s lives and asks whether 
any of these happened to them in the last year. A count of at least four events in Guyana and St. Lucia, 
and three in St. Kitts and Nevis classified youth as at risk in this scale. About a quarter of respondents 
scored at risk, with a slightly higher percentage (29 percent) of youth in St. Kitts and Nevis. 
Table 8: Percent of assessed youth at risk for critical life events (n=5,109)
 Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Critical 
life 
events
24% 25% 20% 29% 22% 26% 20% 33% 
Among eligible youth, the least often reported critical life events regard school, namely failing or dropping 
out of a class or failing to go on to the next grade (25 percent) or getting suspended/expelled/thrown 
out/transferred to another school (25 percent). The most common critical event in the lives of eligible youth 
respondents in the last year was hearing about deaths or injuries caused by violence in their neighborhoods, 
which occurred with 71 percent of respondents. 67 percent of respondents also said that they had gotten 
into a fight or had a problem with a friend. These important and serious events can have significant and 
negative impacts on youth’s wellbeing and behaviors. Two of the events included, namely going on a date for 
the first time or starting to hang out with a new group of friends, are not as clearly negative, and may in fact 
be positive events. The questionnaire does not dig deeper into these so we cannot confirm whether or not 
all of these had negative impacts.
43
Figure 21: Responses to questions on critical life events in the last year among eligible youth
(n=1,139)
SCALE DE: IMPULSIVE RISK TAKING
The impulsive risk-taking scale asks about respondents’ level of agreement with four statements about 
doing dangerous actions or things that might get them in trouble. Responses include strongly agree=5, 
agree=4, neither agree nor disagree=3, disagree=2 or strongly disagree=1. Youth that had an average 
score of at least 3 in Guyana and St. Lucia and 3.3 in St. Kitts and Nevis are considered to be at risk. A 
total of 35 percent of respondents scored at risk in this scale; one of the most common risk factors. There 
was not as much variation across countries and gender as in other scales.
Table 9: Percent of assessed youth at risk for impulsive risk taking (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Impulsive risk 
taking 35% 36% 34% 34% 33% 36% 32% 40% 
When we limit the sample to just eligible youth, we find that about 60 percent of these respondents 
agreed or strongly agreed with the statement that they like to do something dangerous for the fun of it 
44
and to do things without thinking about whether they will get in trouble. Slightly under half (48 percent) 
agreed or strongly agreed that they find it exciting to do things that might get them into trouble, and 71
percent agreed or disagreed that they like to have fun when they can, even if they will get in trouble for 
it later. This reflects a higher level of agreement than other items in the YSET. 
Figure 22: Responses to questions on impulsive risk taking among eligible youth (n=1,134)
SCALE F: GUILT NEUTRALIZATION
This scale lists six statements that attempt to measure rationalization (or guilt neutralization) of lying, 
stealing, and beating others. Youth are asked if they strongly agree=5, agree=4, neither agree nor 
disagree=3, disagree=2 or strongly disagree=1 with these behaviors. Youth are considered at risk if they 
averaged 2.5 or above in Guyana and St. Lucia and 3 in St. Kitts and Nevis. This scale had the highest 
number of respondents scoring at risk, at 40 percent. Respondents in St. Kitts and Nevis were much less 
likely to be scored at risk; however, as demonstrated in Annex III where we set the risk thresholds to be 
the same for all three countries, this difference is due to the higher threshold – when the threshold is set 
to be the same, St. Kitts and Nevis youth’s responses are at a higher risk level in three out of the four
cases tested.
45
Table 10: Percent of assessed youth at risk for guilt neutralization (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Guilt 
neutralization 40% 45% 41% 24% 36% 43% 35% 50% 
Turning to the eligible youth, we see that half of these respondents agree or strongly agree with the 
statement that it is okay to lie to keep themselves or their friends from getting in trouble (Figure 23). 
Similarly, a majority of eligible youth are also able to rationalize causing harm through fighting, as 70
percent believe it is okay to beat people up if they were hit first or were standing up for themselves. By 
contrast, these respondents are less willing to rationalize stealing; only eight percent agree or strongly 
agree that is okay to steal something from somebody who is rich (Figure 23). 
Figure 23: Responses on select questions on guilt neutralization among eligible youth (n=1,137)
46
Figure 24: Responses on select questions on guilt neutralization among eligible youth (n=1,135)
SCALE G: NEGATIVE PEER INFLUENCE
Scale G has four questions asking respondents how likely they are to go along with their friends in certain 
situations. Responses include yes, definitely=5, probably=4, not sure=3, probably not=2, no, definitely 
not=1. The threshold for risk in this scale is an average score of at least 2.3 in Guyana and St. Lucia and 2 
in St. Kitts and Nevis. A total of 18 percent of respondents scored at risk in this scale, with a higher share 
of youth in St Kitts and Nevis at 32 percent. 
Table 11: Percent of assessed youth at risk for negative peer influence (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Negative 
peer 
influence
18% 11% 19% 32% 16% 20% 16% 23% 
Figure 25 provides responses to the four questions in this scale for eligible youth. About 60 percent of 
these higher-risk youth report that they would definitely or probably not go along with their friends if 
they were doing something wrong or bad. Responses are similar when eligible youth are asked about 
doing something that would get them in trouble at home or in school. A much larger percentage (85 
percent) of eligible youth say that they would definitely not or probably not go along with their friends if 
they were getting in trouble with the police. 
47
Figure 25: Responses on negative peer influence among eligible youth (n=1,133)
SCALE H: PEER DELINQUENCY
The peer delinquency scale lists a series of six actions and asks respondents to count how many of their 
friends did each action during the last six months. The module also asks how many friends their caregivers 
would consider to be “bad company.” Respondents selected from the options all=5, most=4, half=3, a 
few=2 or none=1. An average score of 1.9 or higher in any of the three countries classifies the respondent 
as at risk, and a total of 16 percent of respondents scored at risk in this scale. The largest difference in 
the disaggregation was based on age; older youth are much more likely to have more friends engaging in 
these negative behaviors or considered to be bad company.
Table 12: Percent of assessed youth at risk for peer delinquency (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Peer delinquency 16% 18% 16% 13% 14% 19% 13% 24% 
Figure 26 lists eligible youth responses to the six actions in order from least common to most common. 
The most serious actions were also the least common, including breaking into a building to steal something, 
participating in crime and violence, using or selling illegal drugs, and attacking someone with a weapon. 
For example, 78 percent of youth reported that none of their friends had broken into a building to steal 
something whereas only 26 percent of eligible youth reported that none of their friends had skipped 
48
school. Four-fifths of eligible respondents said that at least some of their friends would be considered “bad 
company” by their caregivers.
Figure 26: Reported involvement of peers in delinquent behavior by eligible youth (n=1,135)
SCALE T: FAMILY ANTI-SOCIAL INFLUENCE
The family anti-social influence scale contains three questions about the respondents’ and their family 
members’ involvement with crime and violence. Respondents were asked how many of their family 
members think the respondent will be involved in crime in violence someday, how many family members 
have been involved in crime and violence, and if they think they will be involved with a group engaging in 
crime and violence A respondent who answered zero for the first two questions and no on the third 
scored a three, going up by increments of one if the answer was higher than zero or no. The threshold 
for at risk is a six for all three countries. A total of 19 percent of respondents scored at risk for this scale, 
and there was not a significant variation across countries and sex. 
49
Table 13: Percent at risk for family antisocial influence (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Family 
antisocial 
influence
19% 17% 19% 20% 18% 19% 16% 25% 
Just over half (53 percent) of eligible youth have no family members who believe that they (the 
respondents) will one day become involved in crime and violence (see Figure 27). Most respondents (66
percent) report having at least one family member who has been involved in crime and violence, and 18
percent have four or more. Three-quarters of respondents answered no when asked if they thought that 
they would one day become involved in crime and violence, or if they are involved now. Only seven 
percent answered probably or yes (see Figure 28). 
Figure 27: Responses on family antisocial influence questions among eligible youth (n=1,132)
50
Figure 28: Response on family antisocial influence question among eligible youth (n=1,133)
SCALE IJ: SELF-REPORTED DELINQUENCY
The self-reported delinquency section is divided into two parts: first, the YSET asks youth if they did or did 
not engage in a series of behaviors in the last six months; second, if they answered yes to any of those 
behaviors, they are asked if they engaged in that behavior with the group of friends they hang out with. Every 
time the respondent says yes, he/she gets one added to the total score for this section. CFYR considers a 
youth to be at risk if they score a sum of six or higher in any of the countries. A total of 23 percent of 
interviewed youth scored at risk, with some variation across countries, sex and age groups.
Table 14: Percent of assessed youth at risk for self-reported delinquency (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Self￾reported 
delinquency
23% 20% 24% 26% 18% 28% 19% 31% 
Figure 29 ranks delinquent behaviors from least common to most common among eligible youth. Hitting 
others and getting into fights were the most commonly reported behaviors, followed by skipping classes 
without an excuse or permission. Breaking into a building, selling drugs, and stealing were the least 
commonly reported behaviors by eligible youth. 
51
Figure 29: Self-reported delinquency among eligible youth in the last six months (n=1,136)
As noted above, if youth reported engaging in a behavior, they were then asked if they engaged in the 
behavior with the group of friends they hang out with. As shown in Figure 30, out of the behaviors that 
youth reported doing in the last six months, the most commonly done with a group of people they hang 
out with were: selling and inhaling drugs, skipping school, and getting into serious group fights. 
52
Figure 30: Self-reported delinquency by eligible youth with a group of friends
(n=44;1,017;160;520;203;12;847;460;810;726;513;124)
Males are more likely than females to report engaging in the various behaviors. There are, however, two 
exceptions. Females are more likely than males to report hitting someone with the purpose of hurting 
him/her and skipping classes without an excuse or permission (see Figure 74 in Annex VI). 
SCALE FRD: RISKY GROUP BEHAVIORS
The final scale concerns the behaviors of the group of friends the respondent hangs out with. The 
respondent is asked to answer yes or no to a series of questions regarding risky, illegal and belligerent 
group behavior. Each yes answer equals one, and a count of three or higher classifies the respondent as 
at risk in this scale. A total of 12 percent of youth scored as at risk in this scale. Risk does not vary much 
by country but the share of males at risk is twice as high as females. This means that the group of friends 
that females hang out with is less likely to engage in risky behaviors. 
53
Table 15: Percent at risk for risky group behaviors (n=5,109)
Total GUY SLU SKN Female Male Ages 10 
to 14
Ages 15 
to 17
Risky 
group 
behaviors
12% 13% 11% 13% 8% 16% 11% 15% 
The most commonly reported risky behaviors among eligible respondents’ group of friends were getting 
into serious fights, using weapons in fights, claiming or defending a territory as their own, and selling things 
that do not belong to them (see Figure 31). A total of 29 percent of eligible respondents said that their 
group of friends is known by others in the neighborhood as being violent (compared with 18 percent for 
the total youth sample). 
Figure 31: Eligible youth reporting on behaviors their friends have engaged in during the last six 
months (n=1,008)
54
COMPARISON BETWEEN HONDURAS AND CARIBBEAN YOUTH IN TARGETED 
COMMUNITIES
Given that a very similar YSET tool was used in both Honduras and the Caribbean,18 the evaluation team 
thought it would be informative to compare youth answers and test for differences by module. While 
Honduras experiences higher levels of violence than any Caribbean country, the samples of youth included 
in the two programs might or might not reflect these differences, and the samples can be compared using 
their YSET answers. For this exercise, we included only questions that were identical or nearly identical 
across the two YSETs, and we used the same method for both countries/regions when calculating average 
scores. The two samples are, of course, different in many regards, and we were able to test for gender 
and age differences. We found that the Honduran sample is more male (68 percent compared with 49
percent in the Caribbean) and the Caribbean sample is older (average age is 13.14 years compared with 
12.14 for Honduras, where the minimum age for YSET interviews was eight years old). These gender and 
age imbalances could account for some of the differences in average scores between Honduras and the 
Caribbean youth in targeted communities. 
The following table contains the mean scores for Honduran and Caribbean youth, by module, and the p￾value for the two-sample t-tests. A low p-value (typically, smaller than 0.05) tells us that the means for the 
two samples are statistically different. When comparing the entire two samples, with both eligible and 
ineligible youth, we see that Honduran youth have higher average scores in all modules but self-reported
delinquency (where their means are not statistically different), and family antisocial influence (where 
Caribbean youth have higher average scores). The family antisocial influence module here includes two 
questions only, and their wording is slightly different in the two YSETs: in Honduras, youth were asked 
how many members of their family thought that they would one day become a gang member, and how 
many of their family members are in a gang; in the Caribbean, the words gang and gang member were
replaced by being involved or engaged with crime and violence. The more general terms used in the 
Caribbean YSET might explain at least some of the difference in average scores in the family antisocial 
influence module.
18 By Caribbean we mean the three countries included in this study, and not the overall Caribbean region.
55
Table 16: T-tests of means with entire samples (eligible and ineligible)
Honduras
(n=2,181)
Caribbean
(n=5,109) p-value
Weak parental monitoring 1.80 1.58 0.000
Critical life events 2.83 1.88 0.000
Impulsive risk taking 2.74 2.59 0.000
Guilt neutralization 2.58 2.36 0.000
Negative peer influence 2.01 1.50 0.000
Self-reported delinquency 1.05 1.08 0.429
Family antisocial influence 0.79 1.08 0.000
Risky group behaviors 0.82
(n=2,062)
0.76
(n=3,950) 0.043
When comparing only eligible youth, we found that Honduran youth had higher average scores in all 
modules except for self-reported delinquency, family antisocial influence and risky group behaviors. The 
risky group behaviors about risky behavior among the respondent’s group of friends, and the higher 
average age in the Caribbean sample might help explain the higher average scores in that module. 
Table 17: T-tests of means with eligible youth only
Honduras
(n=665)
Caribbean
(n=1,128) p-value
Weak parental monitoring 2.35 2.11 0.000
Critical life events 3.63 3.17 0.000
Impulsive risk taking 3.46 3.36 0.007
Guilt neutralization 3.10 2.93 0.000
Negative peer influence 2.22 1.95 0.000
Self-reported delinquency 2.25 2.82 0.000
Family antisocial influence 1.65 2.56 0.000
Risky group behaviors 1.54
(n=646)
1.69
(n=985) 0.041
56
BASELINE FINDINGS FROM THE CAREGIVER 
SURVEY 
QUESTION 2:
DO SECONDARY PREVENTION EFFORTS IMPROVE THE FAMILY 
ENVIRONMENT? WHY OR WHY NOT?
INTRODUCTION
Family structures and the environment at home can impact the behavior of youth, particularly when 
parents are either ill-equipped or absent. A survey of caregivers, parents or those with primary caretaker 
responsibility, was designed to understand at the baseline some of the family dynamics and the 
characteristics of the parents or others who may be in charge of youth at such a critical time of their lives. 
It also aims to measure parenting, family attitudes, and behaviors targeted by the intervention and test if 
the intervention leads to positive changes in these attitudes and behaviors over time. 
This section first introduces the caregivers participating in the survey and looks at the overall family 
structure. We then look at diverse family characteristics, including income and age, and examine potential 
negative family influences, such as drug and alcohol use and past engagement in criminal activity. These 
will serve as potentially important variables in exploring variation in risk and programmatic impact at 
endline. We then transition into items that might improve as a result of the intervention, including family 
cohesion, family decision-making, parenting attitudes (e.g., locus of control in raising children), caregiver 
presence, and knowledge of youth behaviors. We conclude by looking at families’ community engagement. 
Surveys were conducted with the primary caregiver of eligible youth in both the treatment and control 
groups. Out of the 1,139 eligible youth, we interviewed caregivers for 1,002 recipients, or 88 percent of 
eligible youth. This includes interviews with 947 caregivers; in households with more than one eligible 
youth, caregivers responded to relevant questions separately for each eligible youth. The remaining 
caregivers either refused to be interviewed or could not be located. The caregiver survey includes 404 
interviews from Guyana, 434 from St. Lucia and 164 from St. Kitts and Nevis. As noted in the methodology 
section, the caregiver non-response rate was highest in St. Kitts and Nevis (21 percent). 
CAREGIVER SURVEY RESPONDENTS AND FAMILY STRUCTURE
Variations in family structure might help explain risk levels and the success of a family-based intervention. 
The majority of the primary caregivers in Guyana, St. Kitts Nevis and St. Lucia (72 percent) are mothers; 
however, there is a small minority of fathers (11 percent) who responded to the survey as being the 
primary caregiver. These men are, according to the survey, primarily single parents, unlikely to be married 
or in common law unions. Mothers who take primary responsibility for the youth usually do so without 
the help of a father (an important finding further discussed below). Another 17 percent of the primary 
caregivers are grandmothers, aunts, sisters, and other guardians. This group of non-parental caregivers is 
overwhelmingly female. 
57
While the survey focuses on a primary caregiver, there are usually additional household members who 
contribute to the care of the eligible youth. We asked respondents to list out members of the household 
and to note those with caregiver responsibility for the eligible youth. In addition to the primary caregiver, 
there are on average around two additional household members who participate in care. These additional 
caregivers mostly include fathers (and some mothers), grandparents, aunts and uncles, and a large number 
of older siblings. 
Table 18 provides the breakdown of who contributes, according to the survey, to the care of the youth 
eligible to participate in CFYR. The first two sets of columns show the reported primary caregiver, broken 
down by gender. The additional providers of care are reported in the third set of columns. This includes 
primarily spouses and grandparents but also a large number of siblings, aunts, uncles and other members 
of the extended family. When families have either fathers or grandparents living in the household, they 
are almost always involved in the care of the youth. Aunts and uncles may live in a household and not be
involved in the care of the youth, most likely because they are taking care of a child of their own.
Table 18: Caregivers: primary (by gender) and additional household members (n=941)
Male
Primary
Caregiver
freq.
% of 
primary 
caregivers 
Female 
Primary
Caregiver 
freq.
% of 
primary 
caregiver
Add’l 
Caregiver
Freq.
% of 
add’l
Total
caregivers
Freq. 
% of 
total
Father/Mother 102 11% 677 72% 419 23% 1198 43%
Brother/Sister 3 0% 24 3% 713 38% 740 26%
Grandparent 2 0% 65 7% 207 11% 274 10%
Uncle/Aunt 4 0% 27 3% 236 13% 267 10%
Self 0 0% 1 0% 28 2% 29 1%
Others 9 1% 27 3% 257 14% 293 10%
Total 120 13% 821 87% 1860 100% 2801 100%
Overall, the most typical household in the caregiver survey is one with a single mother with no father (52 
percent). Only a quarter of the households are nuclear families where both a father and a mother live 
together in the household and participate in the care (26 percent). There is a significant number of 
households where neither father nor mother appear as primary caregivers (12 percent); and a number of 
homes where the father is the single caregiver, with no mother present in the household (nine percent).
It is important to note that the family structure in households with eligible youth is rather similar across 
the three countries. Families do, however, differ across countries in terms of size. The family size of eligible 
households in Guyana is larger, which had an average 2.46 household members, than in St. Kitts and Nevis
(2.06) and St. Lucia (1.91). 
In order to visualize these profiles more clearly, Figure 32 provides a breakdown of the family structure 
in each of the three countries. The axes in the figure are a ratio. A 1:1 ratio would be one where there is 
a particular family member for each youth. For example, in the case of St. Lucia, approximately 80 percent 
of youth have a female parent while only roughly 40 percent of youth have a father living in the same 
household. Over 60 percent have both a male and female sibling, and roughly 20 percent have a female 
58
grandparent in the household. This standardization is done in order to make the data comparable across 
countries and so that gender gaps and other family structure differences are not the consequences of 
differences in the way the graphs are scaled. The figures show a disproportionate number of mothers, a 
relative absence of fathers, similar numbers of male and female siblings (although many more young 
children in Guyana) and a somewhat greater presence of grandmothers in St. Kitts and Nevis. 
Figure 32: Family structures in St. Lucia, St. Kitts and Nevis and Guyana in caregiver survey (n=941) 
In St. Lucia and Guyana, we are able to draw on Multiple Indicator Cluster Survey (MICS)19 to compare 
eligible households with the rest of the nation. These health surveys are designed to understand various 
aspects of caregiving, with detailed information regarding mothers and children, as well as data on 
household characteristics. There are some limits to comparability, as MICS4 surveys were carried out in 
Guyana in 2015 and St. Lucia in 2012 and there are some differences in question wording. Nonetheless, 
the MICS provide a useful point of comparison. 
In the case of St. Lucia, we find that nationally the share of single mothers identified in the MICS is virtually 
identical to the one found in the caregiver survey: 52.3 percent of the households had a mother with no 
father present. In addition, for St. Lucia, in 4.3 percent of the households neither a father or a mother was 
living in the household. And only 1.6 percent of the households had a father with no mother living in the 
household. The MICS survey includes some information on caregiving responsibilities, but only for primary 
care. The largest difference with respect to our caregiver survey is that the MICS survey in St. Lucia found 
41.7 percent of the households had nuclear families compared with 28 percent in the caregivers survey. 
Keeping in mind that there are some limitations to comparability, single mother patterns found among 
eligible households are not atypical of a country like St. Lucia; however, the share of nuclear families in 
our sample of eligible youth seems to be much lower than for the country as a whole.
In the case of the MICS for Guyana, when the whole country is considered, the share of single mother 
households is 32.7 percent, with most homes being nuclear families. However, when only the urban area 
19 MICS is a UNICEF initiative to collect data on the situation of women and children, with governments typically carrying out 
the data collection with technical and financial support from UNICEF and partners. Data are collected in nationally or sub￾nationally representative samples of households. 
59
of Georgetown is analyzed, the share of single mother household increases to 54.1 percent, again virtually 
the same as what was found in the caregiver survey.
FAMILY CHARACTERISTICS
In addition to the family structure, there are other family characteristics that might be important to 
explaining variation in risk or variation in programmatic impact. For example, income levels, caregivers’ 
age, and their education levels might influence risk and resilience to violence in the larger community. 
INCOME LEVELS AND ASSET INDEX
Income levels and socioeconomic status are notoriously difficult to measure even in relatively 
sophisticated survey instruments. There are many reasons why respondents may find it difficult to estimate 
their own household income, which may be driven by seasonality, complex income streams coming from 
different family members, as well as imperfect recall and other sources of imprecision. In addition, there 
are social desirability pressures that may prevent respondents from truthfully revealing their own income 
or socioeconomic levels. Annex IV provides information on three distinct methods for estimating 
household socio-economic status. 
In Figure 33, we provide a simple count of assets. The variables included are whether the household has 
a flush toilet, television, computer, internet connection, refrigerator, washing machine, stove, microwave, 
telephone, cellular phone, water pump, hair dryer, heater, satellite or cable television, vehicle and a boat. 
Many of these household goods are widespread, while others are rare within the caregiver sample. Only 
three households had all 16 items, and only two of them reported not having any of them at all. The 
median household has eight of these items. There are some differences between countries, which reflect 
the relative development of each of them, but as can be seen in the histograms reported in Figure 33, the 
most important feature is that there is a great deal of overlap in the asset indexes across countries. The
average for St Kitts Nevis is highest at 9.8 assets, Guyana has the lowest average at 7.3, while St. Lucia
falls in between with an average of eight. 
Figure 33: Number of assets in households by country as reported in the caregiver survey (n=947)
60
AGE
Mothers who are the primary caregivers of the eligible youth are no younger or less educated than the 
fathers of the youth (living in the same household) (see Figure 34 and Figure 35). This is the case for all 
the countries, even though there are some differences, particularly in education, across the countries. It 
is not possible to compare the age profiles to the non-eligible youth, but comparisons across gender and 
country reveal some salient features. Fathers who participate as caregivers are slightly older than the 
mothers. The distributions are relatively dispersed in St. Lucia and much tighter around the mode in St.
Kitts Nevis. In the case of the mothers, the distribution looks very similar across countries, with a slightly 
lower age in St. Lucia. 
Figure 34: Age distribution of mothers of 
eligible youth (regardless of caretaking role or 
family structure
Figure 35: Age distribution of fathers of eligible 
youth (regardless of caretaking role or family 
structure
EDUCATION
As with age, it is not possible to use the caregiver survey to compare the education characteristics with non￾eligible households. However, some comparisons across family relationships and countries are very 
illustrative. Education is measured by a rank ordering, where 5 = tertiary (college/university), 4 = secondary, 
3 = primary, 2 = less than primary, and 1= no formal education. Vocational training was placed in the same 
ranking as secondary (high school) education, with a score of 4. Average education levels across countries, 
groups, and genders are presented in Figure 36. While sample sizes are small across multiple disaggregation, 
the figure suggests the following: females tend to have a slightly higher level of education than males, siblings 
tend to have a higher level than parents, and those in St. Kitts and Nevis have higher levels than those in the 
other two countries. Perhaps surprisingly, the profile of education in St. Lucia is not so different from the 
one in Guyana, even though Guyana is clearly a less developed country.20
20 In UNDP’s 2018 Human Development Index (HDI), Guyana ranks 125th out of 189 countries, St. Lucia ranks 90th and St. Kitts 
and Nevis 72th. The HDI measures progress along three dimensions: a long and healthy life, being knowledgeable, and having a 
decent standard of living. 
61
Figure 36: Levels of educational achievement by country and relationship to program youth (n=941) 
EMPLOYMENT 
Caregiver survey respondents were asked about gainful employment for each household member who is 
aged 16 and above. Although we do not know the number of hours worked or salary they receive, we 
know whether each member has a source of income through work. Of all the primary caregivers that 
answered the survey, 64 percent are gainfully employed. Among single mothers, 65 percent are working, 
and among single fathers, 84 percent are working. Grandparents who live in the household are working 
in about 31 percent of cases. In our sample, employment levels are much higher in St. Kitts and Nevis
compared to the other two countries, which is primarily driven by higher employment rates for mothers. 
Table 19: Share of household members who are gainfully employed (n=942)
All 
countries SLU SKN GUY
Primary caregiver (n=942) 64% 59% 87% 60%
All mothers (n=751) 64% 60% 92% 58%
Single mothers (n=489) 66% 59% 90% 62%
All fathers (n=332) 86% 83% 88% 88% 
Single fathers (n=83) 84% 85% 82% 85%
Grandmothers (n=159) 31% 23% 41% 36% 
Grandfathers (n=53) 32% 32% 37% 29% 
Aunts (n=119) 68% 67% 91% 60%
Uncles (n=112) 71% 55% 92% 81%
An important factor for household income level and economic stability is how many household members 
are gainfully employed. On average, households have 1.63 persons aged 16 and above employed, and this 
number is lower for households that have a single parent or no parent at all (i.e. no mother or father are 
62
reported to live in the household). With the exception of single-father households, St. Kitts and Nevis
again shows higher levels of employment. That finding is especially interesting when we consider that 
Guyanese households are larger on average. 
Table 20: Average number of gainfully employed household members (n=947)
All 
countries SLU SKN GUY
All households (n=947) 1.63 1.41 1.96 1.73
Single-mother households (n=489) 1.53 1.24 1.84 1.72
Single-father households (n=83) 1.41 1.36 1.27 1.52
Households without any parents (n=118) 1.49 1.09 2 1.59
Youth employment levels are measured for those aged 16-18, including both eligible youth and other 
youth in the household. Among this group, 24 percent are neither in school nor working. The case of not 
being in school nor working is sometimes referred to as NEET (“not in education, employment or 
training”) in the literature and is typically considered a strong risk factor. 
Table 21: Number of youth between the ages of 16 and 18, by employment and school status
(n=537)
In school Employed
Yes
Employed
No
Total
Yes 25 (5%) 327 (61%) 352 
(66%) 
No 57 (11%) 128 (24%) 185 
(34%) 
Total 82 (15%) 455 (85%) 537 
(100%) 
Figure 37 breaks down the number of NEET youth by country and compares all youth in the caregiver 
survey sample. The share of youth that are NEET is much higher in Guyana compared to the other two 
countries, with 35 percent in Guyana and only six percent in St. Kitts and Nevis. St. Lucia falls somewhere 
in between at 17 percent. If breaking down the numbers of NEET youth by community in Guyana, this 
share becomes 32 percent of all sample youth in Sophia, 35 percent in East Ruimveldt, 36 percent in East 
La Penitence, 40 percent in Lodge and 43 percent in Corriverton. Therefore, compared to the 
Georgetown communities, the town of Corriverton appears to have a larger share of youth in our sample 
that is not currently studying or working.
63
Figure 37: Share of youth ages 16-18 that are neither employed nor in school (n=537)
RELIGION AND RELIGIOUS ATTENDANCE
There is a large degree of reported religious diversity in our sample of caregivers and considerable 
differences across countries. The largest denomination among respondents in St. Lucia is Roman Catholic 
(60 percent); in St. Kitts and Nevis it is Methodist (19 percent); and in Guyana it is Pentecostal (38
percent). Religious diversity is the highest in St. Kitts and Nevis where similar percentages identify as 
Methodist, Baptist, Pentecostal, Anglican, and other. Pie charts of self-reported religious affiliation can be 
found in Annex IV. 
Out of those reporting a religion, about half attend religious services more than once a month (see Figure 38).
There are some differences in the frequency of religious service attendance by country, with respondents in 
St. Lucia attending somewhat less frequently than respondents in Guyana and St. Kitts and Nevis. 
64
Figure 38: Average frequency of religious service attendance, by country (n=944)
FAMILY COHESION 
To measure family cohesion, the caregiver survey uses a series of questions, mostly drawn from the Family 
Adaptability and Cohesion Scale (FACES), an instrument designed to measure family interactions. We 
inquire about the frequency with which a youth calls or visits older family members such as grandparents 
and uncles and whether the family shares traditions or family stories. According to caregivers, about one￾fifth of eligible youth do not ever call or visit older family members, and one-quarter do not ever or rarely 
get traditions shared with them. As noted above, if there is more than one eligible youth in the household, 
caregivers are asked separate questions about each youth. As such, sample sizes are somewhat higher 
than the number of caregivers interviewed.
65
Figure 39: Responses to questions on vertical family relations (n=1,000)
To facilitate analysis, the data from several additional family cohesion questions are reduced, through a 
factor analysis, into an index in a single dimension. The items included in the index include evaluations by 
the caregiver of whether family members are involved in each other’s lives, whether they feel very close 
to each other, if are supportive of each other in difficult times, and if they can discuss problems calmly 
with each other. The index also includes questions about the quality of communication between family 
members, the family’s ability to resolve conflicts, and the amount of time spent together as a family. The 
factor analysis shows a very strong correlation among all these variables.21 The graphs show that this index 
of family cohesion is quite similar among all three countries, with most households bunched toward the 
lower end of the family cohesion spectrum. This skew is slightly stronger in St. Kitts and Nevis and the 
distribution somewhat more normal in Guyana. 
21 The survey also asked a question related to anger in the family. The question is the following: “When angry, family members 
seldom say negative things about each other.” The factor analysis for the index showed that this question was not correlated with 
the rest of the variables measuring family cohesion. The lack of correlation with all the other family cohesion measures suggests 
that this variable triggers a very different aspect of family dynamics, probably related to anger.
66
Figure 40: Distribution of scores in the Family Cohesion Index, by country (n=943)
FAMILY DECISION-MAKING
To provide a measure of how decisions are made in families, caregivers in households with more than one 
caregiver were asked how often they disagree and argue with the youth’s other caregivers in front of the 
youth. More arguing, fighting, and bad mouthing signifies a less unified parenting front and can give youth 
mixed messages about how they are expected to behave. As shown in Figure 41, more than half (61
percent) of caregivers admitted to very often/often/sometimes disagreeing with other caregivers on how 
to discipline the youth in front of them. Talking badly about the other caregivers or arguing with them in 
front of the youth are reported to be far less common.
67
Figure 41: Responses to questions on family decision-making among households with more than one 
caregiver, n=872 
Although there is variation across countries, as shown in Figure 42, no single country stands out as having a 
higher or lower share of caregivers reporting to disagree or argue with other caregivers in front of the youth.
Figure 42: Caregiver responses to questions on family decision-making, by country (n=872)
68
PARENTING ATTITUDES: LOCUS OF CONTROL
Caregivers were also asked six questions about their parenting beliefs, particularly about whether parents 
and the community have an influence on kids’ likelihood of getting into trouble, as a measure of parental 
locus of control. As show in Figure 43, while the vast majority (91 percent) of respondents believe that 
parents can have a major influence in their kids, a similar majority (85 percent) also believe that children 
will get into trouble no matter what their parents do. Therefore, most believe that their children’s 
behavior can be influenced but not ultimately controlled, and children will still get into trouble. Nearly half 
of respondents (47 percent) believe that community conditions are making it more difficult for them to 
keep their children out of trouble, while the rest does not necessarily think so, even though they all live 
in communities deemed to be vulnerable. Figure 44 shows that the vast majority (95 percent) of caregivers 
believe it is their responsibility to keep their children out of trouble. They also report that in the last 12 
months they have improved as a parent (91 percent) and that their influence on their children has 
increased (87 percent). These findings suggest that caregivers already have a relatively strong internal locus 
of control when it comes to raising their children.
Figure 43: Caregiver responses to questions on parenting locus of control (n=947)
69
Figure 44: Caregiver responses to questions on parenting locus of control (cont) (n=947)
Respondents in St. Kitts and Nevis are more likely to agree that parents can have a major influence on 
their kids, but also that children will get into trouble no matter what they do (see Figure 45). Guyanese 
caregivers agree in higher percentages (58 percent) with the statement that the community conditions 
prevent them from keeping children out of trouble. St. Kitts and Nevis caregivers were more likely to 
strongly agree with all three questions in Figure 46, although the sum of agree and strongly agree is similar 
for all three countries on all questions.
70
Figure 45: Caregiver responses to questions on parenting locus of control, by country (n=947)
Figure 46: Caregiver responses to questions on parenting locus of control (cont.), by country
(n=947)
CAREGIVER PRESENCE
The survey asked caregivers how often they spend time with their child doing a few different specified 
activities over the course of the month prior to the survey. Caregivers were also asked about how often 
71
they know where their child is, who they are with, and when they are out of school. The survey asked 
about parent-child communication, including whether they say mean things or hit or push when angry and 
whether they have other adults to count on to help raise their child. The last question asks them whether 
they are proud of their child.
Figure 47 and Figure 48 show how many days a week caregivers report sitting down to eat, helping with 
school work, doing something fun as a family, and doing something religious as a family. The activity that 
caregivers and youth most often do together is sharing meals. 42 percent of caregivers report that they 
have meals together seven days a week and only 12 percent report not having any meals together. Other 
activities are less common, including doing homework together, doing something fun as a family, or doing
something religious as a family. Doing something fun as a family is the least likely activity to take place 
every day of the week, with only 10 percent of respondents saying they did that seven days a week last 
month, and close to a quarter (23 percent) of caregivers reporting not having done anything fun as a family 
at all.
Figure 47: Number of days in a typical week caregivers report select behavior for each eligible youth
Note: While many caregivers with youth out of school did not answer the question on helping youth with school work, 
several did. Most of these stated that they helped youth zero days a week, but some did report helping with school work, 
which could have been understood as other educational programming. Excluding out of school youth, the percentage 
stating zero would drop from 34 percent to 32 percent.
72
Figure 48: Number of days in a typical week caregivers report select behavior for each eligible youth
(n=996) 
The share of caregivers who report always knowing where their child is when not at home or school is 
57 percent, and the share who report always knowing who they are with is 49 percent (see Figure 49).
About ten percent never or rarely know where their child is when outside the home or school. These 
numbers are similar to the YSET youth responses, where 18 percent reported that they rarely or never 
tell their caregivers where they are going, and 13 percent that their caregivers do not know where they 
are when outside of school. More than half of youth (63 percent) reported that their parents or guardians 
often or always know where they are when they are outside of school, which is close to the 57 percent
of caregivers reporting the same thing.
73
Figure 49: Caregiver responses to questions on caregiver presence for each eligible youth (n=998) 
About a third of respondents said that when the youth breaks family rules, he/she only faces consequences 
half the time/rarely/never (see Figure 50). Only 55 percent of caregivers say that their child always listens 
to them when they have something important to say, and about ten percent admit they only listen to their 
child when he/she has something important to say half the time or less often (see Figure 51). Only half 
report being able to always discuss problems with their child. When angry, a quarter of caregivers (26 
percent) always or often say mean things to their child and 11 percent may push or hit them. These 
responses suggest that there is room for improvement in terms of parental monitoring and caregiver-child 
effective communication. 
74
Figure 50: Caregiver responses to questions on caregiver presence for each eligible youth (n=1,001) 
Figure 51: Caregiver responses to questions on communication with each eligible youth (n=1,001) 
75
There are gender differences in reporting on communication with eligible youth. A higher percentage of 
male caregivers report being listened to when compared to female caregivers (see Figure 52). Males are 
also more likely to report that they are able to talk to their child and solve their problems. Male caregivers 
are also less likely to report saying mean things or pushing or hitting their child when angry. It is not clear 
if these differences reflect actual behavior differences or gender differences in reporting. 
Figure 52: Caregiver responses to questions on caregiver presence, by caregiver’s gender (n=1,001)
As shown in Figure 53, about 17 percent of respondents report not having another adult to count on to 
help raise their child. Despite the challenges, parents are overwhelmingly proud of their children. Only 
three percent said that they are never or rarely proud. 
76
Figure 53: Responses to select questions on caregiver presence (n=1,001) 
NEGATIVE FAMILY INFLUENCES 
Caregivers were asked if members of the family had drug or alcohol problems or if they had been in 
conflict with the law (this was done without indicating specific family members.) In seven percent of 
households, caregivers report that drug use has created a problem in the household. In those households, 
marijuana was mentioned in 67 percent of these cases and alcohol in 36 percent. It is worth mentioning 
that enumerators noted instances where they observed signs of drug use. One enumerator noted that 
he/she “saw [a family member] smoking marijuana on the street” while doing the interview. Another noted 
that the “interview conducted outdoors and there was a smell of marijuana on respondent's person.” A 
third included a comment regarding “the uncles” who “were smoking marijuana,” and a final one reports 
that “the mother bought marijuana in front of me.” 
When disaggregating the data by caregiver’s gender, we found that three percent of male-led household 
reported that drug-use creates problems in the household, compared with seven percent of female-led 
households. Out of those who reported a problem, no male caregivers mentioned alcohol, and they all 
mentioned marijuana.
In seven percent of households, respondents reported that a member had been charged with a crime in 
the last ten years. The most common types of crime were robbery, assault without robbery, and domestic 
violence (see Figure 59 in Annex IV for a breakdown of reported crimes). In two percent of households, 
respondents reported that a member had been a member of a gang, defined as a group of people who 
engage in violent and/or criminal acts together. This question does not include relatives, such as a parent, 
uncle, cousin or older sibling who currently live somewhere else. It is difficult to know if there is 
77
underreporting due to social desirability bias, but these responses indicate a relatively low level of 
household members in conflict with the law.
The survey also asked caregivers if members of the household had been a victim of a crime in the last 12 
months. Only four percent of respondents reported victimization. This is lower than the average rate of 
household victimization found through representative samples of the fifteen communities reported above: 
nine percent in Guyana, eight percent in St. Lucia, and six percent in St. Kitts and Nevis. For those who 
said yes, the most common types of crime were assault without a robbery and home burglary (see Figure 
60 in Annex IV).
KNOWLEDGE OF YOUTH BEHAVIOR
Caregivers were asked about the youth’s behavior in the last six months across the same set of behaviors 
that youth were asked in the YSET. The main reason for employing this strategy was to compare caregiver 
and youth responses as a means to estimate youth bias when self-reporting negative behavior. 
Interestingly, youth reported engaging in these behaviors in larger numbers than their caregivers, in all 
categories. While this battery does therefore not provide a useful validation tool, it does provide a 
measure of caregiver knowledge about youth behavior. 
Table 22 provides a comparison between youth and caregiver responses. The largest differences can be 
seen in the cases of hitting someone with the purpose of hurting him/her and being involved in fights, 
where 48 percent and 36 percent of youth, respectively, said yes, but only 12 percent and less than one 
percent of caregivers, respectively, said yes. Breaking into a building was the least commonly reported 
behavior, and for this no differences were identified in reporting between youth and their caregivers. 
These results indicate two possible phenomena: either caregivers were not very aware of how often youth 
were engaging in these negative behaviors and which behaviors they engaged in, or they were not willing 
to say yes even when they thought the answer was yes, presumably due to embarrassment or fear that 
the youth would get into trouble. 
78
Table 22: Youth behaviors as reported by self, caregivers, and differences between the two
Item
Share of 
eligible youth 
that has done 
in the last 6 
months
(n=1,135)
Share of caregivers 
reporting that 
youth has done in 
the last 6 months
(n=1,002)
Percentage 
point 
difference 
between 
eligible youth 
and caregivers
Hit someone with the purpose of hurting 
him/her 48% 12% -36
Were involved in fights 36% 0% -36
Involved in serious group fights 39% 11% -28
Skipped classes at school without an 
excuse of permission 38% 18% -20
Purposely damaged or destroyed property 
that belonged to others 30% 6% -24
Carried a hidden weapon for protection 31% 6% -25
Inhaled drugs or substances that make you 
"high" 30% 9% -21
Lied about age to get into a place or to buy 
something 26% 4% -22
Attacked someone with a weapon 14% 3% -11
Stole or tried to steal something valuable 10% 3% -7
Sold marijuana or other illegal drugs 9% 2% -7
Broke into a building to steal something 1% 1% 0
Figure 54 shows where caregivers and eligible youth agree and disagree. The chart is ordered so that least 
commonly reported behaviors by youth appear first, starting with breaking into a building, and most 
commonly reported behaviors appear last, ending with hitting someone. Among the least commonly 
reported behaviors, there are larger shares of agreement between caregivers and youth. Conversely, 
among the most commonly reported behaviors, there are larger shares of disagreement, particularly with 
the youth reporting having done that behavior in the last six months, and the caregiver saying no. Overall, 
looking at the mismatches, it is far less common for the youth to say no and for the caregiver to say yes, 
than the other way around.
79
Figure 54: Share of matches and mismatches between youth and caregiver answers, for each 
behavior (n=1,135 for youth and 1,002 for caregivers)
As shown in Table 23, different types of caregivers (parents, grandparents, and others) are equally likely to say 
a youth did not commit an act when the youth reported that s/he did. Other caregivers (e.g., aunt, sibling)
were somewhat more likely to say that a youth did something when youth reported that s/he did not. 
Table 23: Share of matches and mismatches in reporting youth behavior, by type of primary 
caregiver (n=1,135 for youth and 1,002 for caregivers)
Parents Grandparents Other 
caregivers
Match 74% 75% 73% 
Youth yes, caregiver no 23% 22% 22% 
Youth no, caregiver yes 3% 3% 6% 
80
CONCLUSIONS 
The analysis of the community survey using the indices created by the evaluation team indicates the 
following general findings. (1) There are relatively low rates of reported crime, but there are
heterogeneous perceptions of crime prevalence, and perceptions of crime prevalence do not always match 
self-reported crime victimization statistics. (2) There is relatively strong community trust and cohesion, 
as measured by two different indicators of social capital. (3) Community members have somewhat neutral 
attitudes toward the law and toward police. (4) There are similar levels of resources for youth across the 
three countries, and higher variation across communities in each country. (5) There are generally high
feelings of self-efficacy and self-determination. 
Although the levels of victimization and crime in the communities in the program are not as high as in other 
countries of Latin America or the Caribbean, there are substantial numbers of youth at risk within these 
communities. The YSET identified that more than one fifth of the youth to whom the instrument was applied 
were above a threshold of risky attitudes and behaviors to warrant the family intervention. As far as what is 
learned from analyzing the YSET data, implementation of the same YSET tool in the three countries provides 
a useful snapshot of the degree of risk that youth face in each location, and if there are differences by gender 
and age. Males and older youth (ages 15-17) are at a higher risk in all scales. Males stand out in relation to 
females in one scale in particular: weak parental monitoring, implying that caregivers monitor the 
whereabouts and company of female youth more closely, and in St. Lucia especially. Youth in the islands are 
more likely to score at risk in the weak parental monitoring and negative peer influence scales, even when 
the scale thresholds are set to be the same for the three countries (see full table in Annex III). When the 
thresholds are set to be the same, St. Kitts and Nevis has a higher share of youth scoring at risk in very 
single scale except for one, scale H – peer delinquency. Predicted differences across gender and age groups 
suggest that risk dynamics are similar across the three countries (see Table 6).
Impulsive risk taking and guilt neutralization are the most common risk factors. These scales measure 
levels of agreement with statements about enjoying doing things that are dangerous or that might get them 
in trouble, and rationalizing behaviors such as lying, stealing and beating others. A surprisingly high number 
of eligible youth have had some critical life event happen in the last year. Some of the events included in 
the critical life events scale can be stressful or even traumatic, such as hearing about deaths or injuries 
caused by violence in their neighborhoods (which happened to 71 percent of eligible youth), having a fight 
or problem with a friend (67 percent), or having someone close to them die, get seriously injured, very 
sick or leave the country (64 percent). The frequency for negative behaviors as measured by the IJ scale 
was also noticeably high, with the mean for all the Caribbean youth a little higher than (but not statistically 
different from) the mean for Honduras. Eligible Caribbean youth had a higher mean score in the IJ scale 
than Honduran eligible youth.
These data also show that there is significant underreporting of youth negative behaviors by caregivers. 
We believe that either caregivers are not very aware of how often youth are engaging in these negative 
behaviors, and which behaviors they engaged in, or they are not willing to report these behaviors, 
presumably due to embarrassment or fear that the youth would get into trouble. 
Through the caregiver survey, we are able to obtain a more comprehensive picture of the family 
composition and family life of the youth that are engaged in the program. The majority of primary 
81
caregivers are mothers (72 percent) and the most typical household in the program is one with a single 
mother and no father present in the household (52 percent). 
According to caregivers, almost a quarter of household youth between the ages of 16 and 18 (24 percent)
are neither in school nor working. Employment levels in St. Kitts and Nevis are much higher than in the 
other two countries, for youth and for household members in general. In our sample, Guyanese youth in 
particular are more likely to be in the risky situation of being out of school and out of work.
Parents generally have a relatively strong internal locus of control, believing both that they have a 
responsibility to keep youth out of trouble and that they are able to influence youth. In addition, very few 
caregivers report problems of drug and alcohol problems in the family or family participation in gangs or 
criminal activities. 
Questions about caregivers’ knowledge of youth whereabouts and company when not at home or in 
school show that improvements can be made in the level of parental monitoring. Many caregivers are 
either unaware or unwilling to report that their youth engages in normatively undesirable behavior even 
when their children do so. Many caregivers are also struggling with communicating with youth about 
important topics they want or need to address, with only half reporting they are always able to discuss 
problems with their child. These seem to be problem areas that the family counseling intervention may 
be apt to address.
One of the most challenging aspects of this evaluation is related to the small size of the communities (and 
even countries) involved, which limits the statistical power of the inferences that may be obtained 
regarding the effects of the intervention. After implementation began, several processes have generated a 
smaller sample size than what was originally envisaged. One is related to the willingness of families and 
youth to become beneficiaries, which can partly be attributed to the time that passed between YSET 
consent to the enumerator and the family counselor’s first visit. A second one is the consequence of 
mobility and migration of youth and families that had been participating in the program. Even though survey 
teams have gone to great lengths to find and survey all the youth and their families, some of them are not 
located. A third one is that there is likely to be some attrition in numbers due to youth and families that 
cease to participate for the full duration of the intervention. The feasibility of measuring minimally 
detectable effects will depend on ensuring that efforts continue to collect data as comprehensibly and with 
as full of a coverage as possible, at least for those aspects that are under the control of the implementer 
and the evaluator.
82
ANNEX I: ELIGIBILITY DETERMINATION AND 
RANDOMIZATION
This annex summarizes the approach used to determine eligibility and randomize youth into treatment 
and control. 
DETERMINATION OF ELIGIBILITY
Creative Associate’s project sub-contractor USC determined youth eligibility for the intervention. The 
USC team added up the scores for each YSET module, identified the highest risk youth across all modules, 
and then calculated the average module scores across this highest risk group. These averages were then 
used in conjunction with the distribution of the remaining youth to inform a purposively selected cut￾point for each module on a country-by-country basis. Using Guyana as an example, Table 24 provides the 
average risk scores for the 34 highest risk youths identified and the final thresholds. The thresholds are 
somewhat different across each country, and as such, raw scores cannot be compared across countries. 
Table 24: Average scores for 34 highest risk youth and final thresholds across nine risk modules in 
Guyana
B C DE F G H T IJ EG
Average scores 
for highest risk 
youth
2.21 3.97 3.45 2.92 2.24 2.11 6.24 5.91 4.47
Final cut-points 2.5 4.0 3.0 2.5 2.3 1.9 6.0 3 3 
This approach is similar to that used by USC in Los Angeles and Central America, although in these 
locations the scores of gang members were used to identify the thresholds. In the Caribbean, where gang 
membership is far less common, this approach could not be used. 
MATCHING AND RANDOMIZATION
Given that the intervention targets families, one of the challenges with randomization was ensuring that 
two eligible youth in the same household were randomized to the same treatment condition. After 
matching youth into households, we followed a two-stage randomization process. In the first stage all 
eligible households were stratified by community and randomized into treatment and control. We then 
conducted balance tests with the individual youth level data on several variables including sex, age, grade, 
working or not, in school or not, and a summary scores on each of the nine criteria. The randomization 
procedure was written to re-randomize until all variables balanced. In the case of Guyana, this resulted 
in 229 treatment and 228 control youth. This exceeds CFYR’s capacity to treat, which was 150 youth.
Therefore, the second step divides the treatment and control youth into primary study youth and 
replacement study youth. This step was done to maximize distribution of cases across communities and 
subsequently maximize power. In the case of Guyana, 150 youth were assigned to the primary study group 
and the remaining 79 were assigned to the replacement group. Using Guyana as an example, Table 25
83
shows the number of youth by community randomized into treatment and control at both stage one and 
stage two. 
Table 25: Results of stage 1 and 2 randomizations in Guyana
Stage I
Treatment
Stage I
Control
Stage I
Total
Stage 2
Treatment
Stage 2
Control 
Stage 2
Total
Corriverton 43 43 86 30 30 60
East La Penitence 17 14 31 17 14 31
East Ruimveldt 41 41 82 30 33 63
Lodge 30 32 62 30 30 60
Sophia 98 98 196 43 43 86
Total 229 228 457 150 150 300
84
ANNEX II: DATA COLLECTION 
INSTRUMENTS
Table 26: YSET, caregiver survey, and implementation dates
YSET application dates Caregiver survey dates Implementation 
start date
Guyana Sept 2017-March 2018 April 2018 - June 2018 Jun-18
SLU Sept 2017-April 2018 June 2018 - August 2018 Aug-18
SKN October 2017-June 2018 July 2018 - August 2018 Aug-18
85
YSET TOOL
GUYANA EXAMPLE
86
87
88
89
90
91
 
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
 
109
110
CAREGIVER SURVEY
Impact evaluation of secondary prevention: Caregiver Survey 
March 14, 2018
Contents 
A. Preload Field Control .......................................................................................................110
B. Introduction/Gatekeeper speech ......................................................................................111
C. Informed consent .............................................................................................................111
D. Roster and respondent information ..................................................................................111
E. Family cohesion (from the FACES instrument) ................................................................113
F. Parenting .........................................................................................................................114
G. Relationship with the youth .................................................. Error! Bookmark not defined.
H. Youth behavior and relationships .....................................................................................116
I. Other family members......................................................................................................118
J. Community ...................................................................................................................119
K. Household characteristics ............................................................................................120
L. Contact information ......................................................................................................121
M. End matter................................................................................................................122
A. Preload Field Control
Enumerator name
Enumerators ID number
Unique identifier of youth interviewed
Unique identifier of caregiver
Country
Community
EA Name
EA Code
Number of siblings in the study
Unique identifier of youth interviewed 
Unique identifier of youth interviewed 
Unique identifier of youth interviewed 
Unique identifier of youth interviewed 
Timestamp
111
B. Introduction/Gatekeeper speech
Hello, my name is XXX and I work for The Solomon Group a local research firm conducting a 
study of youth in (St. Lucia/Guyana/St. Kitts and Nevis). In the last few months, a youth from 
this household participated in a survey for a project funded by the United States Agency for 
International Development (USAID) called the Community, Family and Youth Resilience Project. 
We would like to conduct some follow-up from that survey. I am looking to speak with the 
primary caregiver of XXXX. By primary caregiver I mean the adult in the household that most 
looks after XXXX. This is often the mother of the household but it can also be the father, 
grandparents, or other relatives or guardians. Is he/she available?
[If Yes skip to informed consent]
[Add a note to the enumerator with the name of the person who originally consented for the 
child to participate in the YSET]
[If No] Do you know a good time for me to return so I can speak with XXX’s primary caregiver?
[Record name and time for follow-up] 
A. Informed consent
Hello, my name is XXX and I work for The Solomon Group, a local research firm conducting a 
study of youth in (St. Lucia/Guyana/St. Kitts and Nevis). XXXX recently participated in an 
interview and now we would like to conduct a follow-up survey with you as the primary caregiver 
to XXXX. This survey is funded by the United States Agency for International Development as 
part of an evaluation of a project being conducted in your community called the Community, 
Family and Youth Resilience Project. The project is implemented by a US organization in 
collaboration with the government of (St. Lucia/Guyana/St. Kitts and Nevis). The goal of this 
study is to better understand how to provide support and services to your community and 
communities like yours throughout the Caribbean. Please note that while we work in close 
coordination with the Community, Family and Youth Resilience Project, we are not part of the 
project. 
The survey should take no more than 45 minutes to complete. Your participation in this survey 
is voluntary. Deciding not to participate will not affect your child’s or ward’s ability to get 
services. If you are unable or do not want to answer a question, you do not have to. All of your 
responses will be kept strictly confidential, which means that we will never reveal your name.
If you have any questions or concerns about this study now or later you may contact XXXXX at 
XXXXXXXXXX so that we may assist you.
CONSENT. Are you willing to proceed?
1 Yes
2 No PROG: SKIP TO CLOSURE
112
D. Roster and respondent information
HOUSEHOLD ROSTER: I want you to think about your household. By ‘household’ I mean a 
group of people living in the same dwelling space and who eat meals together at least three 
times a week. 
We would like to learn a little about who lives in this household. First, list yourself. Then list 
every other member of your household. We do not need to know their names, so you can just 
give me the first letter of their name or nickname. For each person I will ask a few questions, the 
answers to which will all be kept confidential. Former household members who have been gone 
for more than a year should not be listed.
[ID] 1. [Name
] 
2. [Relation to HH] 3. [Sex] 4. [Raising
] 
5. [Age] 6. Education 7. In 
School
8. Emplo
yment
ID Please 
provide the 
first names 
of all of the 
people who 
are part of 
this 
household 
starting with 
yourself?
• Probe: 
is there 
anyone 
else? 
What is their 
relationship to XXX?
1. Father/Mother
2. Brother/Sister
3. Brother-in￾Law/Sister-In￾Law
4. Grandparent
5. Uncle/aunt
6. Spouse/Partner
7. Son/Daughter
8. Step-Child
9. Father-in￾Law/Mother-in￾Law
10. Other (specify): 
__________
What is 
[…]’s 
gender?
1. Male
2. Female 
3. Other
Does […] 
help 
raise 
XXX?22
1. Yes 
2. No
Approximat
ely how old 
was […] on 
their last 
birthday? 
(Enter 
number of 
Years) 
_______
98 - Don’t 
know
What is the 
highest level of 
education 
completed by 
[…]? 
1 No formal 
education
2 Less than 
primary
3 Primary 
4 Secondary
5 Tertiary 
(college/u
niversity)
6 Vocational
Answer only 
for those 6 
to 18 years 
old:
Did […] 
attend 
school this 
year?
1. Yes 
2. No
Answer 
only for 
those 16 
years and 
older:
Is this 
person 
currently 
employed 
in a paying 
job?
1. Yes 
2. No
RES
P1 
Interviewee
: 
_________
___
RES
P2
RES
P3
RES
P4
RES
P5
RES
P6
RES
P7
RES
P8
RES
P9
22 Guidance: Helping to raise the child means being responsible for them. Parents typically “raise” their children, but grandparents, older siblings 
or other extended family members might also help raise a youth. Raising is different than temporarily babysitting or looking after a youth. 
113
9. Does anybody else, who does not live in 
this house, play a role in raising XXX?
1-Yes, 2-No
10. [If yes] Who? 1-Another parent, 2-Another relative, 3-A 
family friend, 4-Teacher or counselor, 5-
Priest, pastor or other religious leader, 6-
Other (specify)
Thank you. Now I would like to ask you a few questions about yourself.
11. [If country = SKN or St. Lucia] What is 
your present civil status?
1-Single/never married, 2-Married, 3-
Common law, 4-Widowed, 5-Divorced, 6-
Separated, 7-Visiting at least once a week
12. [If country = Guyana] What is your 
present marital status?
1-Single/never married, 2-Married, 3-
Common law, 4-Widowed, 5-Divorced, 6-
Separated, 7-Visiting at least once a week
13. What is your religion? 1-None, 2-Anglican, 3-Baptist, 4-Hindu, 5-
Methodist, 6-Muslim, 7-Pentecostal, 8-
Rastafarian, 9-Roman Catholic, 10-Seventh 
Day Adventist, 11-Church of God, 12-
Evangelical, 13-Other
14. How often do you attend religious 
services, on average? 
1-Everyday, 2-Every week, 3-More than once 
a month, 4-Once a month, 5-Several times a 
year, 6-Once a year, 7-Less than once a 
year, 8-Never
E. Family cohesion (some questions are from the FACES instrument)
Now I want to ask you a few questions about your family.
1. How often does XXX visit any family 
members who are older than you, such 
as grandparents or great aunts and 
uncles? Visits include in-person meetings 
only, not over the phone.
1- Very often, 2-Often, 3-Sometimes, 4-
Rarely, 5-Never, -99 RA N/A
2. How often does XXX call any family 
members who are older than you, such 
as grandparents or great aunts and 
uncles? These may include phone or 
video calls.
1- Very often, 2-Often, 3-Sometimes, 4-
Rarely, 5-Never, -99 RA N/A
3. How often do family members share 
family traditions or family stories with 
XXX?
1- Very often, 2-Often, 3-Sometimes, 4-
Rarely, 5-Never, -99 RA N/A
Would you agree or disagree with the following statements about this family or household:
114
4. Family members are involved in each 
other's lives
[Potential prompt: “Involved can include a 
number of activities, including regular 
communication, attending events
together, and monitoring progress in 
school.”]
1-Strongly agree; 2-Agree, 3-Neither agree 
nor disagree, 4-Disagree, 5-Strongly disagree
5. Family members feel very close to each 
other 
1-Strongly agree; 2-Agree, 3-Neither agree 
nor disagree, 4-Disagree, 5-Strongly disagree
6. Family members are supportive of each 
other during difficult times.
1-Strongly agree; 2-Agree, 3-Neither agree 
nor disagree, 4-Disagree, 5-Strongly disagree
7. Family members can calmly discuss 
problems with each other.
1-Strongly agree; 2-Agree, 3-Neither agree 
nor disagree, 4-Disagree, 5-Strongly disagree
8. When angry, family members seldom say 
negative things about each other.
1-Strongly agree; 2-Agree, 3-Neither agree 
nor disagree, 4-Disagree, 5-Strongly disagree
How satisfied are you with…
9. The quality of communication between 
family members?
1- Extremely satisfied; 2- Very satisfied; 3-
Generally satisfied; 4- Somewhat dissatisfied; 
5-Very dissatisfied
10. Your family’s ability to resolve conflicts? - Extremely satisfied; 2- Very satisfied; 3-
Generally satisfied; 4- Somewhat dissatisfied; 
5-Very dissatisfied
11. The amount of time you spend together 
as a family?
- Extremely satisfied; 2- Very satisfied; 3-
Generally satisfied; 4- Somewhat dissatisfied; 
5-Very dissatisfied
F. Family leadership
[Section to be asked in households where another person is listed as having a role in raising 
XXX in the Roster section [Raising] or the follow-up section. If there is nobody else, this section 
should be skipped.] 
For the following questions I want you to think about ALL THE PEOPLE that play a role in 
raising XXX, who may live in this house or not.
10. How often do you disagree on how to 
raise or discipline XXX in front of XXX?
1-Very often, 2-Often, 3- Sometimes, 4-
Rarely, 5- Never
11. How often do you talk badly about these 
other people who help raise XXX in front 
of XXX? 
1-Very often, 2-Often, 3- Sometimes, 4-
Rarely, 5- Never
12. How often do you argue or fight in front of 
XXX?
1-Very often, 2-Often, 3- Sometimes, 4-
Rarely, 5- Never
115
G. Parenting
Now I would like to ask you some questions about parenting. For each question, I 
would like for you to tell me if you strongly agree, agree, neither agree nor disagree, 
disagree, or strongly disagree.
1. Children will get into trouble no matter 
what their parents do.
1-Strongly agree; 2-Agree, 3-Neither agree nor 
disagree, 4-Disagree, 5-Strongly disagree
2. The conditions in this community prevent 
me from keeping my children out of 
trouble.
1-Strongly agree; 2-Agree, 3-Neither agree nor 
disagree, 4-Disagree, 5-Strongly disagree
3. Parents can have a major influence on 
their kids.
1-Strongly agree; 2-Agree, 3-Neither agree nor 
disagree, 4-Disagree, 5-Strongly disagree
4. It is primarily my responsibility to keep my 
children out of trouble.
1-Strongly agree; 2-Agree, 3-Neither agree nor 
disagree, 4-Disagree, 5-Strongly disagree
5. In the last 12 months, I have improved as 
a parent. 
1-Strongly agree; 2-Agree, 3-Neither agree nor 
disagree, 4-Disagree, 5-Strongly disagree
6. In the last 12 months, my influence on my 
children has increased.
1-Strongly agree; 2-Agree, 3-Neither agree nor 
disagree, 4-Disagree, 5-Strongly disagree
H. Caregiver Presence
[Module repeats if multiple study youth in the household. Names of youth should auto-populate 
into the question] 
In answering the following questions, I would like you to think about your experience over the 
last month.
1. In a typical seven-day week, how many 
days do you sit down to eat together with 
XXX? [synonyms for typical: normal, usual]
Enter number from 0 to 7]
2. In a typical seven-day week, how many 
days do you help XXX with school work? 
Enter number from 0 to 7.
-97 if youth is not enrolled in school] 
3. In a typical seven-day week, how many 
days do you do something fun as a family 
with XXX, such as play a game, play sports, 
go to the beach, have a BBQ, and so forth?
Enter number from 0 to 7]
4. In a typical seven-day week, how many 
days do you do something religious as a 
family with XXX such as go to church or 
temple, pray or read the scriptures together?
Enter number from 0 to 7]
116
Now I would like to ask you about your relationship with XXX. For each question, I would like 
for you to tell me if it happens always, often, half the time, rarely, or never. 
7. When XXXX goes out, he/she tells me 
where he/she is going either in-person or 
through a text, WhatsApp, or phone-call
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
8. When XXXX is not at home or school, I 
know WHERE he/she is.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
9. When XXXX is not at home or school, I 
know WHO he/she is with.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
10. If XXXX does not let me know where 
he/she is going, he/she lets another adult 
in the household know.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
11. There are consequences if family rules are 
broken.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
12. When I have something important to say, 
XXXX listens to me.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
13. When XXXX has something important to 
say, I listen to him/her.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
14. XXXX and I are able to talk and solve our 
problems.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
15. When I’m angry, I may say mean things to 
XXXX.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
16. When I’m angry, I may push or hit XXXX. 1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
17. I have other adults I can count on to help 
me raise my child.
1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
18. I am proud of XXXX. 1-Always, 2-Often, 3-Half the time, 4-Rarely, 
5-Never
I. Youth behavior and relationships (from the YSET)
Now, we would like to talk about XXXX’s behavior. We are going to ask you about a series of 
actions and if he/she has done any of these in the last 6 months. Please remember that your 
responses are confidential. Answering yes to any of the questions will not get anyone in any 
trouble. We know some of these questions are sensitive, but we are asking them to all parents 
of kids aged 10 and above. We hope you feel confident in answering these questions honestly. 
Please let us know at any time if you have any concerns. [If respondents demonstrate concern 
during this set of questions, remind them of the material in this introduction and that their 
responses are confidential. You might also mention: “when the researchers analyze the survey 
responses they only look at all the survey respondents together. For example, they are 
117
interested in if 10% or 20% of youth committed these acts – not if a particular household 
answers yes or no.”] , 
In the last 6 months, has XXXX… [If the respondent 
answers “Yes” to any 
of the questions] How 
sure are you of this? 
1. Inhaled (sniffed, smoked or drank) drugs or 
substances that make you “high”?
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
2. Skipped classes at school without an 
excuse or permission? (skipped school = 
ditched or not turned up for school)
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
3. Lied about his/her age to get into some 
place or to buy something?
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
4. Purposely damaged or destroyed property 
that did not belong to him/her?
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
5. Carried a hidden weapon for protection? 1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
6. Stolen or tried to steal something valuable? 1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
7. Broken into a building to steal something? 1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
8. Hit someone with the purpose of hurting 
him/her? (clarify: intending to hurt 
someone, not just playing)
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
9. Attacked someone with a weapon? 
(attacked = jumped with gun, knife or other 
weapon)
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
10. Used a weapon or force to get money or 
things from people?
1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
11. Been involved in group fights? 1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
12. Been involved in gang fights? 1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
13. Sold marijuana and illegal drugs? 1-Yes, 2-No, -98 DK, 
-99 RA
1-Very sure, 2-
Somewhat sure, 3-
Not at all sure
14. Thinking about whether this has ever 
happened, at any point in time… 
1-Yes, 2-No
118
Have you ever discovered drugs in 
XXX’s room or belongings?
15. [if yes] In the last 6 months? 1-Yes, 2-No
[For the enumerator to answer] 
16. Was there any indication that the 
respondent did not answer the questions 
honestly? (Below we will ask you to 
explain any YES responses)
1-Yes, 2-No
J. Other family members
Now, I would like to ask you a few questions about the household members in general. Some of 
these questions are somewhat sensitive; however, please remember that your answers are 
confidential. [If respondents demonstrate concern during this set of questions, remind them of 
the material in this introduction and that their responses are confidential. You might also 
mention: “when the researchers analyze the survey responses they only look at all the survey 
respondents together. For example, they are interested in if 10% or 20% of households exhibit 
these traits – not if a particular household answers yes or no.”] , 
1. In the last year, has drug use among any 
household members created any 
problems in this household?
1-Yes, 2-No (skip to 46)
2. If so, which drugs?
a. Alcohol
b. Marijuana
c. Other drugs 
1-Yes, 2-No
3. Has anybody in this household been a 
victim of a crime in the last 12 months?
1-Yes, 2-No
4. If you are comfortable doing so, could 
you please tell us the type of crime or 
crimes. Please select all that apply. 
1-Unamed robbery with no assault or 
physical threat
2 – Unarmed robbery with assault of physical 
threats
3 - Armed robbery
4 – Assault but not robbery
5- Rape or sexual assault
6-Domestic violence
7-Kidnapping
8-Vandalism
9- Burglary of your home while no one was 
home
10-Burglary of your home while you were 
home
119
11-Extortion/blackmail [define extortion: 
obtain money through force or threats]
12-Bullying
13-Other
-98 Do not know
-99 Refused to answer
5. Has anybody in this household been 
charged with a crime in the last 10 
years? We do not need to know who 
specifically. 
1-Yes, 2-No
6. If you are comfortable doing so, could 
you please tell us the type of crime or 
crimes the household member has been 
charged with. Please select all that apply.
1-Unamed robbery with no assault or 
physical threat
2 – Unarmed robbery with assault of physical 
threats
3 - Armed robbery
4 – Assault but not robbery
5- Rape or sexual assault
6-Domestic violence
7-Kidnapping
8-Vandalism
9- Burglary of your home while no one was 
home
10-Burglary of your home while you were 
home
11-Extortion/blackmail [define extortion: 
obtain money through force or threats]
12-Bullying
13-Other
-98 Do not know
-99 Refused to answer
7. Has any member of this household ever 
been a member of a gang? By gang, we 
mean a group of people who engage in 
violent and/or criminal acts together. 
Again, we do not need to know who. 
1-Yes, 2-No
K. Community (from the CFYR Community Survey) 
Now, I would like to ask you a few questions about your community.
1. How long have you lived in this 
community?
1-Less than 6 months, 2-Between 6 months 
and a year, 3-Between 1-2 years; 4-Between 
2-5 years; 5-Between 5-10 years, 6-More 
than 10 years
For the next question, I am going to read 
you a statement, and I would like for you 
to answer on a 1 to 7 scale where 1 
1 Strongly satisfied, 2,3,4,5,6, 7 Strongly 
dissatisfied; -98 DK; -99 RA
120
means strongly AGREE and 7 means 
strongly DISAGREE 
2. I have a good bond with others in this 
neighborhood. 
For the next two questions I am going to 
ask you about how satisfied you are with 
police services provided in your 
community in the last 12 months. Please 
rate your level of satisfaction on a 1-7 
scale where 1 means strongly 
SATISFIED and 7 means strongly 
DISSATISFIED. 
3. Police responsiveness to incidents of 
crime and violence.
1 Strongly satisfied, 2,3,4,5,6, 7 Strongly 
dissatisfied; -98 DK; -99 RA
4. Approachableness of the police if you 
require assistance from them. 
[Clarification if needed: 
approachableness means that the 
police make it clear that you can 
come to them if you need assistance]
1 Strongly satisfied, 2,3,4,5,6, 7 Strongly 
dissatisfied; -98 DK; -99 RA
L. Household characteristics 
Now, I would like to ask you a few questions about your home. These are questions that are 
asked on the census. We ask them here to see how our study households differ from the rest of 
the country. 
1. Television 1-Yes, 2-No
2. Computer 1-Yes, 2-No
3. Internet access/connection 1-Yes, 2-No
4. Vehicle (private car, bus, or truck) 1-Yes, 2-No
5. Refrigerator/freezer 1-Yes, 2-No
6. Washing machine 1-Yes, 2-No
7. Stove (gas/electric/solar) 1-Yes, 2-No
8. Microwave 1-Yes, 2-No
9. Telephone (landline) 1-Yes, 2-No
10. Mobile/cellular phone 1-Yes, 2-No
11. Water pump 1-Yes, 2-No
12. Air condition unit 1-Yes, 2-No
13. Water heater 1-Yes, 2-No
121
14. Cable TV/satellite 1-Yes, 2-No
15. Flush toilet 1-Yes, 2-No
16. Boat (with engine or paddle) 1-Yes, 2-No
17. In thinking about the families in XXXXX, 
How does your household’s income level 
compare with other households in 
XXXX? Would you say that you are:
1-Much wealthier, 2-Wealthier, 3-Average, 4-
Lower income, 5-Much lower income
M. Contact information
That basically concludes the survey. As we noted above, we would like to contact you again in 
one year. We have your basic contact information, but we sometimes have a hard time getting a 
hold of people after one year, so if you could provide us with some additional contact 
information it would be very much appreciated. 
1. Do you live in more than one place? 1-Yes, 2-No
2. Confirm Address 
3. [if 90=yes] Alternative address
4. [if 90=yes] Landline phone number at 
other address
5. Other parent or guardian cellphone
6. Other family member cellphone
7. Name of other family member
8. Email address
Thank you very much for your participation!!!
END OF SURVEY
[Section N: Questions for enumerator only] 
1. Were you able to see any drugs in the 
house?
1-Yes, 2-No
2. If yes, which drugs? (multiple answer 
possible)
a. Alcohol
b. Marijuana
c. Other drugs (specify _______)
3. Were you able to smell any drugs in the 
house?
1-Yes, 2-No
122
4. If yes, which drugs? (multiple answer 
possible)
a. Alcohol
b. Marijuana
c. Other drugs (specify _______)
5. If yes to question 50 (last question in 
section I), please elaborate on what you 
noticed about the respondent that indicates 
he/she may not have been truthful in the 
section on XXXX’s behavior.
1- Respondent paused before responding; 2-
uncomfortable body language including 
scratching their head, shrugging, or slouching 
with arms folded; 3-prior dishonesty; 4-
indicating that the enumerator should ask 
another caregiver about that question; -96 
other ________________________
N. End matter (cont)
6. Enter the enumerator ID
Numeric
7. Supervisor ID
8. Country 1-St. Lucia, 2-SKN, 3-Guyana, 
9. [if Country=1] Community 1-Anse-La-Raye, 2-Castries City, 3-Dennery, 
4-Soufriere, 5-Vieux-Fort
10. [if Country=2] Community 1-McKnight/Newtown/The Village, 2-Upper 
and Lower Cayon, 3-Sandy Point, 4-Rawlins, 
5-Bath Village 
11. [if Country=3] Community 1-Sophia, 2-East La Penitence, 3-East 
Ruimveldt, 4-Lodge, 5-Corriverton
12. Enumerator assessment of the survey 1- No concerns, 2-Some concerns but good 
data quality, 3-Some concerns that could 
affect data quality
13. [If 2,3] Concerns (select all that apply) 1-Respondent rushing, 2-Respondent 
potentially not answering honestly, 3-Poor 
survey conditions, 4-Other individuals 
listening
14. [If 2,3] Explain any concerns and if 
they pertain to specific questions
(Open ended)
15. Supervisor review 1-No concerns, 2-Corrective action required
16. [If 2] Explain any concerns and 
corrective action required
17. [If 2] Corrective action completed 1-Yes, 2-No
123
COMMUNITY SURVEY
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
ANNEX III: REGRESSION RESULTS
Table 27: Odds ratios with St. Lucia, females, in school, ages 10 to 14, and ethnicity-Black/Afro￾Guyanese as the base
GUY SKN
Eligibility 0.994 1.042
Weak parental monitoring 0.244*** 0.621***
Critical life events 1.613*** 1.668***
Impulsive risk taking 1.200** 0.999
Guilt neutralization 1.394*** 0.456***
Negative peer influence 0.565*** 2.013***
Peer delinquency 1.286*** 0.786**
Family antisocial influence 1.020 1.058
Self-reported delinquency 0.839** 1.138
Risky group behaviors 1.459*** 1.252*
 Note: significance is * p<.1, ** p<.05, ***p<.01.
Table 28: Odds ratios with St. Kitts and Nevis, females, in school, ages 10 to 14, and ethnicity￾Black/Afro-Guyanese as the base
GUY SLU
Eligibility 0.954 0.960
Weak parental monitoring 0.392*** 1.610***
Critical life events 0.967 0.600***
Impulsive risk taking 1.201** 1.001
Guilt neutralization 3.055*** 2.191***
Negative peer influence 0.281*** 0.497***
Peer delinquency 1.636*** 1.272**
Family antisocial influence 0.964 0.946
Self-reported delinquency 0.737*** 0.879
Risky group behaviors 1.166 0.799*
 Note: significance is * p<.1, ** p<.05, ***p<.01.
THRESHOLDS
In order to understand whether differences in average scores by country stem from different thresholds 
or from real country differences, the evaluation team set the same threshold for all three countries on 
each of the scales. The next two tables include the results.
157
Table 29: Comparing shares of risk on each YSET scale by country, by setting the risk thresholds to 
be the same
Scale Mean for all 
countries 
together
GUY
mean
SLU mean SKN mean
Weak Parental Monitoring (B) 0.19 0.10 0.29 0.20
B_flag_1 0.27 0.22 0.29 0.33
B_flag_2 0.17 0.13 0.19 0.20
B_flag_3 0.13 0.10 0.15 0.15
Critical Life Events (C) 0.24 0.25 0.20 0.29
C_flag_1 0.45 0.45 0.41 0.54
C_flag_2 0.24 0.25 0.20 0.29
Impulsive risk taking 0.35 0.36 0.34 0.34
DE_flag_1 0.37 0.36 0.34 0.47
DE_flag_2 0.26 0.25 0.24 0.34
Guilt Neutralization 0.40 0.45 0.41 0.24
F_flag_1 0.45 0.45 0.41 0.54
F_flag_2 0.21 0.22 0.18 0.24
Negative Peer Influence (G) 0.18 0.11 0.19 0.32
G_flag_1 0.23 0.16 0.26 0.32
G_flag_2 0.13 0.09 0.14 0.18
Threshold calculation: For scale B, for example, the original Guyana threshold (labeled B_flag_1), the original St. Lucia
threshold, (B_flag_2), and the original SKN threshold (B_flag_3) are used uniformly for all the countries. The other scales 
had only two different thresholds so only two alterative models are estimated for each of them. Country differences 
remained the same in terms of statistical significance, but the magnitude changed in some cases for the B, C and G scales. 
For scales DE and F, the differences remained similar between Guyana and St. Lucia, but are reversed for Guyana and SKN 
(with the same thresholds, SKN youth are now at higher risk than Guyanese youth in 3 out of 4 cases). Those reversals are
noted in bold typeface.
158
Table 30: Logistic regressions comparing shares of risk on each YSET scale by country, by setting the risk thresholds to be the same
VARIABLE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
B_flag_1 B_flag_2 B_flag_3 C_flag_1 C_flag_2 DEflag_1 DEflag_2 F_flag_1 F_flag_2 G_flag_1 G_flag_2
2.country 4.106*** 1.436*** 1.489*** 1.613*** 0.620*** 0.681*** 0.620*** 0.834** 0.826** 0.919 0.717*** 0.713*** 0.706*** 1.771*** 1.815*** 1.708***
(0.446) (0.125) (0.151) (0.187) (0.0527) (0.0499) (0.0527) (0.0627) (0.0621) (0.0767) (0.0530) (0.0526) (0.0624) (0.185) (0.166) (0.198)
3.country 2.551*** 1.833*** 1.679*** 1.731*** 1.034 1.179* 1.034 0.832** 1.428*** 1.501*** 0.327*** 1.212** 0.997 3.565*** 2.385*** 2.132***
(0.326) (0.186) (0.200) (0.235) (0.100) (0.103) (0.100) (0.0758) (0.126) (0.143) (0.0317) (0.107) (0.102) (0.401) (0.248) (0.278)
Gender 3.370*** 3.059*** 3.186*** 3.513*** 1.331*** 1.185*** 1.331*** 1.126** 1.176*** 1.087 1.300*** 1.299*** 1.177** 1.296*** 1.291*** 1.323***
(0.269) (0.210) (0.263) (0.338) (0.0896) (0.0681) (0.0896) (0.0669) (0.0692) (0.0703) (0.0770) (0.0751) (0.0822) (0.0972) (0.0881) (0.113)
Age15 to 
17 1.773*** 1.736*** 1.615*** 1.515*** 1.824*** 1.612*** 1.824*** 1.334*** 1.295*** 1.396*** 1.781*** 1.791*** 1.761*** 1.618*** 1.555*** 1.615***
(0.146) (0.127) (0.137) (0.145) (0.133) (0.104) (0.133) (0.0886) (0.0854) (0.0997) (0.118) (0.116) (0.133) (0.131) (0.116) (0.148)
Out of
School 2.104*** 2.088*** 2.236*** 2.525*** 1.655*** 1.277* 1.655*** 1.521*** 1.533*** 1.524*** 1.677*** 1.688*** 1.672*** 1.279 1.373** 1.316
(0.317) (0.277) (0.322) (0.394) (0.214) (0.161) (0.214) (0.189) (0.191) (0.198) (0.215) (0.218) (0.220) (0.197) (0.193) (0.222)
1.ethnicity 1.589* 1.340 1.538* 1.587* 0.678* 0.769 0.678* 0.985 0.979 1.053 0.774 0.769 0.919 1.099 1.025 1.092
(0.425) (0.283) (0.367) (0.427) (0.146) (0.139) (0.146) (0.181) (0.180) (0.212) (0.141) (0.140) (0.195) (0.295) (0.244) (0.327)
2.ethnicity 0.652** 0.704*** 0.648*** 0.651** 0.500*** 0.522*** 0.500*** 0.655*** 0.646*** 0.718*** 0.597*** 0.593*** 0.608*** 0.720** 0.702** 0.806
(0.121) (0.0929) (0.106) (0.124) (0.0641) (0.0547) (0.0641) (0.0712) (0.0701) (0.0876) (0.0627) (0.0620) (0.0792) (0.120) (0.101) (0.146)
3.ethnicity 1.074 0.894 0.936 0.981 0.852 0.880 0.852 1.017 0.986 1.063 0.799* 0.774** 0.817 1.061 0.973 1.020
(0.200) (0.134) (0.166) (0.199) (0.118) (0.105) (0.118) (0.124) (0.120) (0.142) (0.0984) (0.0937) (0.120) (0.179) (0.149) (0.197)
Note: significance is * p<.1, ** p<.05, ***p<.01.
Weak Parental Monitoring (Scale B) covers variables (1)-(4) | Critical Life Events (Scale C) covers variables (5)-(7)
Impulsive Risk Taking (Scale DE) covers variables (8)-(10) | Guilt Neutralization (Scale F) cover variables (11)-(13)
Negative Peer Influence (Scale G) covers variables (14)-(16)
159
ANNEX IV: SUPPORTING TABLES AND 
FIGURES
MEASURES OF INCOME 
An asset index was calculated from the caregiver survey through two methods, one based on a simple 
count of a dichotomic measure of how many household goods or services are found in the home, and a 
factor index generated through a polychoric analysis. It turns out that both methods generate a relatively 
similar rank ordering of the socioeconomic status of households in the three countries. The variables 
included are whether the household has a flush toilet, television, computer, internet connection, 
refrigerator, washing machine, stove, microwave, telephone, cellular phone, water pump, hair dryer, 
heater, satellite or cable television, vehicle and a boat. Many of these household goods are widespread, 
while others are rare within the caregiver sample. Only three households had all 16 items, and only two
of them reported not having any of them at all. The median household has eight of these items. In the 
polychoric index four of these items had to be dropped (boat, vehicle, heater, hair dryer) because their 
infrequent occurrence hindered the calculation of a correlation matrix. Figure 55 shows a scatterplot 
suggesting that the index calculated either with the count of items of the polychoric method yields very 
similar rankings across households (the correlation is 0.94).
Figure 55: Household socioeconomic ranking according to assets reported in caregiver survey 
(n=947)
160
This index cannot be readily compared with the socioeconomic status of non-eligible households in the 
countries, but it provides a metric that may rank households at the baseline according to the degree of 
deprivation and disadvantage that youth may experience relative to other youth. This ranking is relatively 
objective (since a home either has or does not have an asset, and it is unlikely that respondents will 
misreport their existence) and very different from subjective measures of well-being. The caregiver survey 
in fact included a question regarding whether the respondent believed his or her income to be higher or 
lower than other members of their community. While the respondents that reported much lower income 
also showed a much lower number of assets (5.7), and those with lower income also reported slightly 
lower assets than the median (7.4), it turns out that those saying they had average income reported nine 
assets on average, while those that perceived themselves to be wealthier or much wealthier than their 
neighbors reported 8.1 and 8.4 average assets. Hence the subjective measure of income is not a very 
reliable indicator, particularly for those reporting higher incomes. 
RELIGION 
The following figures provide a breakdown in reported religious affiliation among caregivers across the 
three countries. 
Figure 56: Share of caregiver self-reported religious affiliation, in St. Lucia (n=413)
Note: no respondents selected the options Hindu, Methodist or Muslim.
161
Figure 57: Share of caregiver self-reported religious affiliation, in St. Kitts and Nevis (n=155)
Note: no respondents selected the options of Muslim or Evangelical.
Figure 58: Share of caregiver self-reported religious affiliation, in Guyana (n=377)
162
Figure 59: Caregiver reported crimes that household members have been charged with in the last 
ten years (n=76)
163
Figure 60: Caregiver reported household crime victimization in the last 12 months (n=42)
164
Figure 61: Shares of matches and mismatches of youth behavior, by type of primary caregiver
165
ANNEX V: INDEX CONSTRUCTION
Principal component analysis (PCA) is a data reduction method: from a group of correlated variables, it 
produces a smaller set of components that give a good representation of the main characteristics of the 
original data. These components are a linear combination of the original variables that represent the 
maximum possible variance between them but are uncorrelated to each other (whereas there may be a 
high degree of correlation among the original variables). 
Described mathematically, given an original set of n correlated variables, PCA produces uncorrelated, 
orthogonal components, where each component is a linear combination of the initial variables. For 
example, from a set of variables X1 through to Xn, 
1st Principal Component = a11X1 + a12X2 + … + a1nXn
2nd Principal Component= a21X1 + a22X2 + … + a2nXn 
… 
Zth Principal Component = aZ1X1 + aZ2X2 + … + aZnXn
where aZn is the coefficient, or weight, for the Zth principal component and the nth variable. Each 
component has both magnitude and direction
The first component accounts for maximum possible variance in the data. Subsequent components 
account for information not captured by the first and are uncorrelated with the first component (and each 
other), with diminishing explanatory power of variance among the original variables. Therefore, the first 
component often can reasonably be used on its own as an index score that gives information on the
underlying variables. 
PCA does not assume the existence of a group of common factors driving the variation in the data
(whereas factor analysis does). It does, however, assume that the underlying data is normally distributed 
and calculates associations between variables using Pearson correlations. This means that PCA on its own 
is best suited for continuous data. Conversely, the data from which the indices in the Community Context
section of this report are built are ordinal, categorical data. 
PCA with polychoric correlations solves this problem by assuming that each ordinal variable is a 
representation of an underlying, inferred “latent variable” that is both continuous and normally-distributed, 
and produces principal components under this assumption. When polychoric correlations are used in 
variables with many categories, the method assumes that a discrete amount is actually a continuous 
variable. In the case of the Likert scales or categorical dichotomic variables, the assumption is more likely 
to follow when the realizations are not particularly skewed or lopsided in their distribution. 
The table in this Annex shows the variables from the Community Survey from which each index in the 
Community Context section was formed, with the factor loadings of the first polychoric principal 
component for each of them. The table shows the Eigenvalue of the first component (by convention, only 
166
components with Eigenvalues higher than 1 are usually retained). The Eigenvalue provides a measure of 
the variance explained by the first component compared to the variance that would be explained by a 
single variable in the estimation. The “variance explained” column gives the actual percentage of variance 
explained by the component. The reported loadings are correlations, providing a sense of how closely 
associated any given variable is with the estimated factor: 
Table 31: PCA with polychoric correlations
Index Name 1st Principal 
Component
N Eigenvalue Variance 
explained
# Variable Factor 
Loadings
PERCEPTION OF 
CRIME PREVALENCE
crimeprev 3336 3.07066 0.4387 B20 vand 0.3849
B21 drugs 0.3063
 B22 robbed 0.4576
 B23 sex_assault 0.3842
 B24 (recode) rc_lev_crime 0.2348
 B25 safe 0.4083
 B26 safe_dark 0.4234
SOCIAL CAPITAL I soccap1 4084 1.30642 0.4355 B1 sblive_n_comm 0.3023
 B2 (recode) rc_com_trust 0.6737
 B4 fight 0.6744
SOCIAL CAPITAL II soccap2 331 3.14447 0.5241 B36 needs3 0.4673
B37 belong 0.4726
 B38 say 0.2846
 B39 ninflu 0.2773
 B40 connect 0.4856
 B41 bond 0.4057
ATTITUDES 
TOWARDS THE LAW
lawatt 3998 1.59103 0.2652 B5 (recode) rc_law 0.5118
B6 (recode) rc_lawbreak 0.4629
 B7 lawobey -0.0085
 B8 (recode) rc_lawtrust 0.5657
 B9 (recode) rc_comcrome 0.3946
 B10 (recode) rc_bendlaw 0.2190
VIEW OF THE POLICE police 3788 2.13604 0.534 B11 polresp 0.5595
B13 ppatrols 0.5303
 B14 appolice 0.5748
 B27 (recode) rc_police_call 0.2746
RESOURCES FOR 
YOUTH
resources 2394 2.58237 0.5165 B12 prog_unenroll 0.4567
B15 scprog 0.4188
 B16 rescare 0.3546
 B18 prosonprog 0.4894
 B19 vicsupport 0.5007
SELF-EFFICACY selfeff 4494 3.30377 0.4720 D1 (recode) rc_control_life 0.3938
 D2 (recode) rc_solut_prob 0.4209
 D3 (recode) rc_change_life 0.4500
 D4 (recode) rc_feel_helpless 0.4076
 D5 (recode) rc_fate_other 0.3659
 D6 determ_life 0.2827
 D7 determ_comp 0.2919
167
ANNEX VI: GENDER DISAGGREGATIONS 
YSET
SCALE B: WEAK PARENTAL MONITORING
Figure 62: Responses to select questions on parental monitoring, by gender
Figure 63: Responses to select questions on parental monitoring, by gender
168
SCALE C: CRITICAL LIFE EVENTS
Figure 64: Responses to questions on critical life events, by gender
169
SCALE DE: IMPULSIVE RISK TAKING
Figure 65: Responses to questions on impulsive risk taking, by gender
Figure 66: Responses to questions on impulsive risk taking, by gender
170
SCALE F: GUILT NEUTRALIZATION
Figure 67: Responses on select questions on guilt neutralization, by gender
Figure 68: Responses on select questions on guilt neutralization, by gender
171
SCALE G: NEGATIVE PEER INFLUENCE
Figure 69: Responses on negative peer influence, by gender
Figure 70: Responses on negative peer influence, by gender
172
SCALE H: PEER DELIQUENCY
Figure 71: Responses to questions on peer delinquency, by gender
173
SCALE T: FAMILY ANTI-SOCIAL INFLUENCE 
Figure 72: Responses on family antisocial influence questions, by gender
Figure 73: Responses on family antisocial influence question, by gender
174
SCALE IJ: SELF-REPORTED DELIQUENCY
Figure 74: Responses on self-reported delinquency, by gender
175
SCALE FRD: GROUP RISKY BEHAVIORS
Figure 75: Responses on risky group behavior, by gender
176
CAREGIVER SURVEY
Figure 76: Responses to questions on family leadership, by caregiver’s gender
n=104 Male, 767 Female
Figure 77: Responses to questions on parenting, by caregiver’s gender
177
Figure 78: Responses to questions on parenting, by caregiver’s gender
Figure 79: Number of days in a typical week caregivers report select behavior, by caregiver’s gender
178
Figure 80: Number of days in a typical week caregivers report select behavior, by caregiver’s gender
Figure 81: Responses to select questions on caregiver presence, by caregiver’s gender
179
Figure 82: Responses to select questions on caregiver presence, by caregiver’s gender
Figure 83: Responses to select questions on caregiver presence, by caregiver’s gender
180
Figure 84: Responses to select questions on caregiver presence, by caregiver’s gender
Figure 85: Responses to select questions on caregiver presence, by caregiver’s gender
181
Figure 86: Responses to select questions on household drug problems, by caregiver’s gender
Figure 87: Responses to question on crime in the last 12 months, by caregiver’s gender
182
Figure 88: Responses to question on crime in the last 12 months, by caregiver’s gender
183
Figure 89: Responses to question on crime in the last 12 months, by caregiver’s gender
184
Figure 90: Responses to question on crime in the last 12 months, by caregiver’s gender
185
Figure 91: Responses to question on gang membership, by caregiver’s gender
Figure 92: Responses to select question on community, by caregiver’s gender
186
Figure 93: Responses to select question on community, by caregiver’s gender
Figure 94: Responses to questions on police perception, by caregiver’s gender
187
REFERENCES 
Abt, Thomas and Christopher Winship. 2016. What Works in Reducing Community Violence: A Meta￾Review and Field-Study for the Northern Triangle.
Berk-Seligson, Susan, Diana Orcés, Georgina Pizzolitto, Mitchell A. Seligson and Carole J. Wilson. 2014. 
Impact Evaluation of USAID’s Community-Based Crime and Violence Prevention Approach in 
Central America: Regional Report for El Salvador, Guatemala, Honduras and Panama. 
Cahill, M., Jannetta, J. Tiry, E., Lowry, S., Becker-Cohen, M. Paddock, E., Serakos, M., Park, L., Hennigan, 
K. 2015. Evaluation of the Los Angeles Gang Reduction and Youth Development. Program Year 4 
Evaluation Report. Washington: Urban Institute. 
https://www.urban.org/sites/default/files/publication/77956/2000622-Evaluation-of-the-Los￾Angeles-Gang-Reduction-and-Youth-Development-Program-Year-4-Evaluation-Report.pdf
Caribbean Development Bank. 2018. 2017 Caribbean Economic Review and 2018 Economic Outlook. 
Barbados: Caribbean Development Bank. 
https://issuu.com/caribank/docs/cdb_2017_caribbean_economic_review_?e=21431045/5817262
5 
CARICOM Commission on Youth Development. 2010. Eye on the Future: Investing in Youth Now for 
Tomorrow’s Community. 
CFYR. 2017. Family Counselor Training Manual – Draft. Family Counselor Training - St. Lucia. November 
27 – December 1, 2017.
Chioda, Laura. 2017. Stop the Violence in Latin America: A Look at Prevention From Cradle to Adulthood. 
Latin American Development Forum, World Bank Group.
https://openknowledge.worldbank.org/bitstream/handle/10986/25920/210664ov.pdf 
Creative Associates International. 2015. Sistematización de la Experiencia Piloto del Programa de 
Prevención y Oportunidades: Trabajando para el Entendimiento Familiar (PROPONTE).
Hennigan, K. M., Maxson, C., Sloane, D. C., Kolnick, K. A., and Vindel, F. (2013). “Identifying high-risk 
youth for secondary gang prevention,” Journal of Crime and Justice, 37 (1): 104–28. 
Henggeler, Scott W. and Ashli J. Sheidow. 2012. “Empirically Supported Family-Based Treatments for 
Conduct Disorder and Delinquency in Adolescents.” Journal of Marital Family Theory 38(1). 
Jaitman, Laura and Roberto Guerrero Compeán. 2015. Closing Knowledge Gaps: Toward Evidence-Based 
Crime Prevention Policies in Latin America and the Caribbean. Inter-American Development Bank. 
Katz, Charles and Lidia Nuño. 2016. Prevalence and Patterns of Troublesome Youth Groups in the 
Caribbean: Implications for Policy and Practice. Final Report. USAID.
188
Littell, J.H., M. Popa, and B. Forsythe. 2005. “Multisystemic therapy for social, emotional, and behavioral 
problems in youth aged 10-17.” Cochrane Database of Systematic Reviews 4: 1–51. 
Lösel, F. and A. Beelmann. 2003. “Effects of child skills training in preventing antisocial behavior: a 
systematic review of randomized evaluations.” Annals of the American Academy of Political and Social 
Science 587, Assessing Systematic Evidence in Crime and Justice: Methodological Concerns and 
Empirical Outcomes: 84–109. 
Social Impact. 2013. Eastern and Southern Caribbean Youth Assessment (ESCYA). Final Report. 
September.
Sutton, Heather and Inder Ruprah. 2017. Restoring Paradise in the Caribbean: Combatting Violence with 
Numbers. Inter-American Development Bank.
Thornberry, T., B. Kearley, D. C. Gottfredson, M. P. Slothower, D. N. Devlin, and J. J. Fader. 2018. 
“Reducing Crime Among Youth at Risk for Gang Involvement: A Randomized Trial.” Criminology 
& Public Policy 17(4). 
UNICEF. 2017. Innocenti Research Briefs on Adolescent Wellbeing. Link: https://www.unicef￾irc.org/adolescent-research-methods/
U.S. Agency for International Development
1300 Pennsylvania Avenue NW
Washington, DC 20523