Abstract
We investigate the effects of a positive income shock on mental health among adolescent girls using evidence from a cash transfer experiment in Malawi. Offers of cash transfers strongly reduced psychological distress among baseline schoolgirls. However, these large beneficial effects declined with increases in the transfer amount offered to the parents conditional on regular school attendance by the adolescent girls. Improved physical health, increased school attendance, personal consumption, and leisure contributed to the effects. There was also strong evidence of increased psychological distress among untreated baseline schoolgirls in treatment areas. All of these effects dissipated soon after the program ended.
I. Introduction
Mental health disorders developed during adolescence are not only of immediate intrinsic importance as the leading contributor to the global disease and injury burden among adolescent females (World Health Organization 2004),1 but they also can have negative long-run health consequences through increased risky decision making (Di Clemente et al. 2001; Fishbein et al. 2006) and future mental and physical health problems (Evans et al. 2007; McLoyd et al. 2009).2 Moreover, since youth is the period of life when most people are completing school, establishing friendships and romantic relationships, and seeking jobs, mental disorders during adolescence can have effects that extend into adulthood by reducing the likelihood of successfully completing these tasks (Patel et al. 2007). For instance, mental health problems are known to be linked to lower educational achievement (see, for example, Currie and Stabile 2006; Eisenberg and Golberstein 2009; Fletcher 2008; Fletcher 2010; Fletcher and Wolfe 2008; Kessler et al. 1995; Stein and Kean 2000). Self-esteem and self-efficacy, which are highly correlated with psychological distress among adolescents (Tait, French, and Hulse 2003), also have been shown to strongly influence schooling decisions, wages, and a host of behaviors that determine social and economic success (Heckman, Stixrud, and Urzua 2006; Krishnan and Krutikova 2012).
Poverty also can precipitate poor mental health (Dercon and Krishnan 2009; Lund et al. 2011; Patel et al. 2007), resulting in a vicious circle of poor mental health and low socioeconomic status (Patel and Kleinman 2003). Despite this circular relationship, there exists very little evidence on what types of interventions can break this cycle—especially in low- and middle-income countries.3 A recent review suggests that while interventions to improve mental health, such as residential drug treatment programs or psychotherapy, were associated with improved economic outcomes, the mental health effects of poverty alleviation programs were inconclusive (Lund et al. 2011). The authors further discuss the difficulty of identifying the channels of impact for interventions that influenced mental health and suggest that future “evaluations that include an analysis of separate components of the interventions might contribute to a clearer picture,” such as the conditionality of cash payments or the amount of transfers.
In this paper, we use a randomized cash transfer experiment to identify the causal effects of a positive income shock on the mental health of school-age girls in Malawi. The unique design of the intervention allows us to compare the impact of unconditional cash transfers (pure income shocks) with the impact of cash transfers made to families conditional on regular school attendance by the school-age girl in the household. Multiple overlapping randomized treatment layers allow us to further look into channels of impact by investigating the elasticity of mental health with respect to transfer size, examining whether the identity of the recipient within the household matters, and identifying effects on untreated girls in treatment communities.
Using the General Health Questionnaire 12 (GHQ-12), a screening instrument widely used in clinical settings, we show that a positive income shock can cause a substantial reduction in psychological distress. While the intervention was ongoing, baseline schoolgirls who were offered unconditional cash transfers (UCT arm) were approximately 14 percentage points, or 38 percent, less likely to be suffering from psychological distress than the control group. Those who had been offered transfers conditional on regular school attendance (CCT arm) also experienced a statistically significant reduction in psychological distress, but at six percentage points (17 percent) this impact was significantly smaller than that in the UCT arm. Furthermore, CCTs had no effect on psychological distress among those who had already dropped out of school at baseline.
Using nonexperimental methods, we find that the beneficial effects of the program among baseline schoolgirls were due to improvements in physical health, increased schooling and family support for education, as well as higher levels of individual consumption and leisure. These factors account for one-third to one-half of the program impact in the UCT and the CCT arms respectively. Furthermore, exploiting the exogenous variation in the size of the transfers made separately to parents and the school-age girls, we find that the muted effect of cash transfers in the CCT arm is largely due to the fact that psychological distress among baseline schoolgirls increased by approximately three percentage points with each additional dollar offered to her parents (over and above the minimum monthly transfer amount of $4) conditional on her school attendance. We interpret this finding as suggesting that when the transfers become an important source of income for the entire family and depend on the adolescent girl’s actions each month, they might become a heavy burden to shoulder and increase her stress.
The intervention also resulted in a substantial increase in the prevalence of psychological distress among schoolgirls who lived in treatment areas but were not invited to participate in the program, and a decrease among female siblings of schoolgirls who had been offered a transfer. The detrimental effects on the mental well-being of those randomly excluded from the program in intervention areas is consistent with the idea that an individual’s utility depends on her relative consumption (or income or status) within her peer group in addition to absolute consumption. Finally, the impact of the intervention on psychological distress does not appear to be permanent: The measured differences between the treatment and control groups dissipated to a large extent soon after the intervention ended, especially in the UCT arm and among untreated girls in treatment areas.4
The remainder of this paper is as follows. Section II surveys the previous literature on the relationship between poverty, income shocks, and mental health and discusses the potential mechanisms through which a cash transfer intervention may improve mental health. Section III discusses the instrument we use to measure mental health. Section IV describes the study setting, the research design and the intervention. Section V describes our estimation strategy while Section VI presents the results. Section VII concludes.
II. Poverty, Income Shocks, and Mental Health
A. Previous Research
Poverty can precipitate or maintain poor mental health through higher levels of stress and social exclusion, among other channels (Lund et al. 2011). These pathways are thought to be particularly applicable to common mental disorders, such as depression, anxiety, and somatoform disorders.5 That said, empirical evidence on the link between poverty and mental health is mixed. Das et al. (2008) are unable to confirm a relationship between poverty and mental health using survey data from five developing countries while Lund et al. (2011), in a systematic review, find that the associations of common mental disorders with income, consumption, and employment are more equivocal than those with education, food insecurity, and financial stress. Moreover, although Kling, Liebman, and Katz (2007) find a consistent relationship between exogenous reductions in neighborhood poverty rates and improved mental health, they attribute this improvement to a reduction in stress associated with lower levels of random violence in better-off neighborhoods rather than increased earnings or economic self-sufficiency at the individual level.
Although evidence on the relationship between poverty levels and prevalence of mental disorders is mixed, the evidence on the effect of income shocks (either positive or negative) on mental health paints a more consistent picture (Das et al. 2007, 2008). The Indonesian financial crisis of the 1990s, for instance, had a detrimental effect on the psychological well-being of the Indonesian people (Friedman and Thomas 2008), while individuals who were selected by lottery to migrate from Tonga to New Zealand (which has higher living standards) exhibited significantly improved mental health outcomes (Stillman, McKenzie, and Gibson 2009). However, since both of these events also led to many other covariate changes, it is impossible to directly identify the effect of income shocks on mental health from these studies. In fact, Stillman, McKenzie, and Gibson (2009) argue that changes in income contributed little to the improvement in mental health from migration. More direct evidence comes from studies of lottery winners, which indicate positive effects of winning lottery prizes on mental health among British and Swedish lottery players (Gardner and Oswald 2007; Lindahl 2005).6
B. Potential Mechanisms
Given the dearth of evidence on economic interventions that can improve mental health, it is natural to ask how a cash transfer program might improve psychological well-being among school-age girls. The extant literature on the subject is thin, but it nonetheless suggests several potential mechanisms. First, schooling raises self-esteem and self-efficacy (Heckman, Stixrud, and Urzua 2006), which are strong predictors of mental health among adolescents (Tait, French, and Hulse 2003). Cash transfers, especially those conditional on regular school attendance, can hence improve mental health by increasing school attendance, attainment, and achievement.7 Second, physical illness is associated with poor mental health (Patel et al. 2007), thus cash transfers can reduce psychological distress by decreasing the likelihood of suffering from other illnesses—either through improved nutritional intake or other protective behaviors. Third, cash transfer programs can improve self-esteem through increased personal consumption, such as wearing shoes and clean clothing or being able to buy food/snacks at school. Qualitative interviews from our study suggest that girls are ashamed to go out without shoes or in dirty/tattered clothing. Fourth, cash transfers can increase opportunities for leisure and foster more connectedness with friends and community, both of which are cited as protective factors for mental well-being (Patel et al. 2007). Finally, cash transfers can reduce depressive symptoms among mothers (Ozer et al. 2011) by reducing the frequency of stressful situations and increasing their sense of control, which, in turn, may reduce stress among their children (Fernald and Gunnar 2009).8 Similarly, increased household income can improve mental health among adolescents as reduced financial stress among parents can improve childrearing behavior and decrease adverse moods (Conger et al. 1994).
III. Mental Health Instrument
A. The GHQ-12 and What It Measures
Our main instrument to assess mental health among the study population is the GHQ-12. The GHQ-12 was developed as a screening instrument to detect individuals who are likely to have mental disorders (or are at risk of developing them) and provides a measure of the common mental health problems of anxiety, depression, and social withdrawal (Jackson 2007). Validation studies, using face-to-face clinical interviews for comparison, show that it produces measures of psychological distress that are highly associated with mental disorder and is able to discriminate DSM-IV and ICD-10 cases from noncases with satisfying levels of reliability (Jackson 2007; Kessler et al. 2002).9 In this paper, we refer to identified cases from the GHQ-12 as individuals who are “suffering from psychological distress.”10
Common mental health problems, such as anxiety and depression, cover large domains of highly prevalent mental health disorders (ICD-10; WHO 2004), so it is important to mention certain aspects of these mental health problems that are pertinent here. First, anxiety and depression are regarded to be the result of a combination of biological and environmental factors. Second, anxiety and depressive disorders can vary greatly in duration: Some conditions are temporary and last for a period of months, or even weeks, while others are chronic and can last for years. The impact of the cash transfer intervention that we examine in this paper should be expected to lie at the intersection of those anxiety and depressive disorders that are affected by environmental factors and are temporary in nature. Chronic disorders and disorders that are primarily genetic or biological in nature are unlikely to be affected.
B. The GHQ-12 in This Study
The GHQ-12 was originally designed for use in clinical settings, but it is regularly used to screen for psychological distress in nonclinical settings (Weich and Lewis 1998; Weich, Lewis and Jenkins 2001; Wiggins et al. 2004). Originally drafted in English by Goldberg and Williams (1988), it has been shown to have high sensitivity and specificity in different settings.11 It has also been shown to be a reliable instrument among adolescents (Tait, French, and Hulse 2003; French and Tait 2004; Montazeri et al. 2003).
The GHQ-12 contains twelve Likert items, which are listed in Appendix 1A. We use a version of the GHQ-12 where each question has a five-level Likert item response format. The responses to the five-level Likert item are: “much more than usual,” “more than usual,” “same as usual,” “less than usual,” and “much less than usual.” According to conventional practice, respondents who report an item as applying to them “much more than usual,” or “more than usual” (“much less than usual,” or “less than usual” for positively phrased items) score 1 on that item, and 0 otherwise. The summed scores, ranging from 0 to 12, provide an indication of the respondent’s mental health. Summed scores with a value of three or higher are classified as having high psychological distress—or diagnosable with the common mental disorders described above (Jackson 2007). We follow this practice and use a binary indicator, which is equal to unity if the summed GHQ-12 score is equal to three or higher, as our main indicator of psychological distress.
We did not validate the widely used threshold score of 3, using in-depth psychiatric interviews.12 It is possible that this threshold score is not optimal among our study population. We therefore examine the robustness of our findings using the summed GHQ-12 scores instead of a binary indicator of psychological distress. We also use another screening instrument called the Mental Health Inventory 5 (MHI-5) to further confirm robustness of our findings. The MHI-5 (described in Appendix 1B) is a more focused instrument that contains five Likert items aimed specifically at depression and anxiety (see, for example Cuijpers et al. 2009).
IV. Research Setting and Design13
A. Location
Malawi, the setting for this research project, is a small and poor country in southern Africa. 81 percent of its population of 15.3 million lived in rural areas in 2009, with most people relying on subsistence farming. The country is poor even by African standards: Malawi’s 2008 GNI per capita figure of $760 (PPP, current international $) is less than 40 percent of the sub-Saharan African average of $1,973 (World Development Indicators Database 2010). According to the same data source, net secondary school enrollment is very low at 24 percent.
B. Sampling
Zomba District in the Southern region was chosen as the site for this study. Zomba District is divided into 550 enumeration areas (EAs), which are defined by the National Statistical Office of Malawi and contain an average of 250 households spanning several villages. Fifty of these EAs lie in Zomba city, while the rest are in seven traditional authorities. Prior to the start of the experiment, 176 EAs were selected from three different strata: Zomba city (urban, 29 EAs), near rural (within a 16 KM radius of Zomba city, 119 EAs), and far rural (28 EAs).
In these 176 EAs, each dwelling was visited to obtain a full listing of never-married females, aged 13–22. The target population was then stratified into two main groups: Those who were out of school at baseline (baseline dropouts) and those who were in school at baseline (baseline schoolgirls). In each selected EA, all eligible baseline dropouts and a percentage of all eligible baseline schoolgirls were sampled for inclusion in the study. This procedure resulted in a sample size of 3,796 young women with an average of 5.1 dropouts and 16.5 schoolgirls per EA.
C. Research Design and Intervention
Treatment status was assigned at the EA level and the sample of 176 EAs was randomly divided into two equally sized groups: treatment and control. In the 88 treatment EAs, all baseline dropouts were offered conditional cash transfers. The 88 treatment EAs were then randomly assigned to one of three groups to determine the treatment status of baseline schoolgirls: in 46 EAs baseline schoolgirls received transfer offers conditional on regular school attendance (CCT arm), while in 27 EAs they received offers for unconditional cash transfers (UCT arm).14 In the remaining 15 EAs no baseline schoolgirls received any transfer offers. To measure potential spillover effects of the program, a randomly selected percentage of baseline schoolgirls in each treatment EA were randomly chosen to participate in the cash transfer program. In the 15 treatment EAs, where no baseline schoolgirls were offered cash transfers, this percentage was equal to 0. In these 15 EAs, the only effects on untreated baseline schoolgirls in treatment EAs would be due to the baseline dropouts who were offered CCTs. No EA in the sample had a similar cash transfer program before or during the study. Figure 1 provides a flowchart that further illustrates treatment assignment.
Study sample and attrition
CCT Arm—After the random selection of EAs and individuals into the treatment group, the local NGO retained to implement the cash transfers held meetings in each treatment EA between December 2007 and early January 2008 to invite the selected individuals to participate in the program. At these meetings, the program beneficiary and her parents / guardians were made an offer that specified the monthly transfer amounts being offered to the beneficiary and to her parents, the condition to regularly attend school, and the duration of the program. It was possible for more than one eligible girl from a household to be invited to participate in the program.
The offer to participate in the program consisted of a transfer to the parents, a transfer directly to the girl, and payment of school fees for girls attending secondary school. Transfer amounts to the parents were varied randomly across EAs between $4, $6, $8, and $10 per month, so that each parent within an EA received the same offer. Within each treatment EA, a lottery was held publicly to determine the transfer amount to the school-age program beneficiaries, which was equal to $1, $2, $3, $4, or $5 per month. Secondary school fees were paid in full directly to the schools upon confirmation of enrollment for each term (primary school is free in Malawi).
Monthly school attendance of all the conditional cash transfer recipients was checked and payment for the following month was withheld for any student whose attendance was below 80 percent of the number of days school was in session for the previous month. However, participants were never removed from the program for failing to meet the monthly 80 percent attendance rate, meaning that if they subsequently had satisfactory attendance, then their payments would resume. Offers to everyone, identical to the previous one they received and regardless of their schooling status during the first year of the program in 2008, were renewed between December 2008 and January 2009 for the second and final year of the intervention, which ended at the end of 2009.
UCT Arm—In the UCT arm, the offers were identical with one crucial difference: There was no requirement to attend school to receive the monthly cash transfers. Other design aspects of the intervention were kept identical so as to be able to isolate the effect of imposing a schooling conditionality on the outcomes of interest. For households with girls eligible to attend secondary schools at baseline, the total transfer amount to the parents was adjusted upwards by an amount equal to the average annual secondary school fees paid in the conditional treatment arm. This additional amount ensured that the average transfer amounts offered in the CCT and UCT arms were identical and the only difference between the two groups was the “conditionality” of the transfers on satisfactory school attendance. Attendance was never checked for recipients in the unconditional arm and they received their payments by simply presenting at the transfer locations each month.
D. Household Survey
The data used in this paper were collected in three household survey rounds. Baseline data (Round 1) were collected between October 2007 and January 2008, before the offers to participate in the program took place. First followup data collection (Round 2) was conducted approximately 12 months later—between October 2008 and February 2009. The second followup (Round 3) data collection was conducted between February and June 2010—after the completion of the two-year intervention at the end of 2009 to examine the final impacts of the program. The intervention period coincided with the 2008 and 2009 school years in Malawi.
The annual household survey consisted of a two-part multitopic questionnaire: one that was administered to the head of the household and the other administered to the core respondent—the sampled girl from our target population. The former collected information on the household roster, dwelling characteristics, household assets and durables, shocks and consumption. The survey administered to the core respondent provides detailed information about her family background, schooling status, health, dating patterns, sexual behavior, fertility, and marriage. Our main mental health instrument, the GHQ-12, was part of the two followup surveys administered to the core respondent during Rounds 2 and 3, but not at baseline. The additional mental health instrument, the MHI-5, was only administered during Round 3.
V. Estimation Strategy
A. Attrition and Balance
Table 1 investigates whether there are differences in the sample attrition rates between the various treatment arms. The dependent variable in Table 1 is a binary indicator for being surveyed in all three rounds. Column 1 shows the results for baseline dropouts, Column 2 for baseline schoolgirls, and finally Column 3 for the spillover group (also baseline schoolgirls). The extensive tracking protocols that were put in place for the study ensured that the overall attrition rate was kept to a minimum among this highly mobile young study population: in the control group, 84.3 percent of baseline dropouts and 89.3 percent of baseline schoolgirls were successfully interviewed during all three rounds of data collection. The regression results show that there are no differences between study arms in the likelihood of having been interviewed in all three rounds, suggesting that the findings of program impact are unlikely to be biased by differential attrition.
Analysis of Attrition
In all regressions described below, we include baseline values of the following variables as controls: age dummies, a dummy that indicates whether a girl lives with her father, an indicator for having been ill during the two weeks prior to the interview, and an indicator for ever having had sex. These variables were chosen because they are predictive of mental health and, as a result, improve the precision of the impact estimates. Following Bruhn and McKenzie (2009) we also include indicators for the strata used to perform block randomization—Zomba Town, rural within 16 kilometers of the town, and rural beyond 16 kilometers.
Table 2 provides a summary of these baseline characteristics, indicators used in the analysis of mechanisms, as well as additional variables that the literature suggests may be correlated with mental health.15 Columns 1 and 3 show the mean values in the control group for the two cohorts examined in this study. We note that baseline dropouts are older, less likely to have their father living in the household, have a lower level of assets, and are much more likely to be sexually active than baseline schoolgirls. Columns 2, 4, 5, and 7 investigate whether the average baseline characteristics are significantly different in the treatment arms or the spillover group than the control group. Girls in the CCT arm are younger than those in the control group (Columns 2 and 4) and the UCT arm (Column 6). There is also a difference between the CCT and UCT arms in highest grade attended (a variable that is highly correlated with age). As one would expect under a successfully implemented randomization procedure, however, the experiment is balanced across observable baseline characteristics that are prognostic of mental health status. Appendix 2 shows that the experiment is also well balanced with respect to the randomized transfer amounts.
Baseline Summary Statistics
B. Specification
We analyze the intention-to-treat (ITT) effects of the intervention on Round 2 and Round 3 mental health indicators using cross-sectional regressions. The regression-adjusted ITT impact of the program for each Round t is estimated with Ordinary Least Squares (OLS) using (straightforward variations of) the following cross-sectional linear regression model:
(1)
where Yit is the mental health outcome for individual i in Round t (t is equal to 2 or 3). Xi1 is a vector that contains the baseline covariates discussed above and presented in Table 2. 
is a dummy variable that is equal to 1 in Rounds 2 and 3 if a girl was offered a conditional (unconditional) transfer and 0 otherwise. The error terms (εit) are clustered at the EA level to account for the design effect of our EA-level treatment and for the heteroskedasticity inherent in the linear probability model used to estimate the impact of the intervention on the binary indicator of psychological distress. Age- and stratum-specific sampling weights are used to make the results representative of the target population in the study area. To make the results comparable across survey rounds, the analysis includes respondents if and only if they were interviewed in all three survey rounds and had the GHQ-12 binary indicator of psychological distress, our main outcome of interest, defined in both Round 2 and Round 3.16
VI. Results
A. Basic Treatment Impacts on Mental Health
Average ITT Effects—In Table 3, we present the average impact of the CCT and UCT treatment arms on the GHQ-12 binary indicator of psychological distress. Panels A and B present program impacts among baseline dropouts and baseline schoolgirls, respectively. The first two columns investigate program impacts in Round 2 (measured during the first followup survey while the intervention was still in progress) and the next two columns investigate them in Round 3 (measured during the second followup survey shortly after the intervention had ended). Regression models presented in Columns 1 and 3 do not include any controls other than treatment status, while regressions presented in Columns 2 and 4 include the set of baseline characteristics described in Section VA as controls.
Average Impact of the Zomba Cash Transfer Program on GHQ-12 Binary Measure of Psychological Distress
Panel A shows that CCTs did not affect psychological distress among baseline dropouts in either round. Panel B, on the other hand, shows that CCTs did have a significant impact on the psychological distress among baseline schoolgirls in Round 2, for whom the probability of suffering from psychological distress was 6–8 percentage points (or at least 17 percent) lower. The impact on psychological distress was even more pronounced in the UCT arm in Round 2. Baseline schoolgirls in the UCT arm exhibit a 14 percentage points reduction in psychological distress (Panel B, Columns 1 and 2)—an improvement of 38 percent over the control group. The F-tests at the bottom row of the table show that the large difference between the CCT and UCT arms is statistically significant at the 5 percent level, but only when baseline controls are included.
The results presented in Columns 3 and 4 of Table 3 suggest that the large impact of the intervention on psychological distress among baseline schoolgirls dissipated shortly after the intervention ended. Psychological distress is approximately four percentage points lower in either treatment arm in Round 3, but these effects are statistically insignificant. Finally, we note that the inclusion of baseline controls had only a modest effect on the point estimates for baseline schoolgirls in the CCT arm (due to the imbalance in age). As our findings are robust to the exclusion of baseline controls for the rest of the analysis conducted in this paper, we only present impact regressions with baseline controls for brevity.
Robustness to the Chosen Threshold Score—Our binary indicator of psychological distress uses a threshold score of 3 in the overall GHQ-12 score to define psychological distress. However, as discussed in Section IIIB, this threshold score has not been validated in the context of Malawi. For example, some studies suggest a threshold score of 4 (Jackson 2007). Furthermore, program impacts on mental health can be nonlinear (Stillman, McKenzie, and Gibson 2009). For this reason, we have repeated the analysis presented in Table 3 using different threshold scores. For any cutoff score between 2 and 9, the decrease in psychological distress in the UCT arm is large and statistically significant.17 The effects also are always negative in the CCT arm, but the statistical significance varies. The effects (as a percentage of the mean in the control group) grow larger as the cutoff score is increased. The reduction in psychological distress is significantly larger in the UCT than the CCT arm for any threshold score below 6. We therefore conclude that the program impacts presented in Table 3 are robust to the choice of the cutoff score used to define our binary indicator of psychological distress.18
Table 4 investigates program impacts on mental health by using the summed GHQ-12 scores (scored bimodally with a range of 0–12; and scored as a Likert scale with a range of 0–36) and the MHI-5 score (scored as a Likert scale ranging from 0–25) as dependent variables.19 We see that using these alternative measures to a binary measure of psychological distress does not alter the main findings: CCTs do not improve mental health among baseline dropouts (Panel A) while mental health is significantly improved in Round 2 in both treatment arms among baseline schoolgirls (Panel B). The relative impact sizes among baseline schoolgirls are similar regardless of the measure used in Tables 3 and 4 with the Round 2 effect in the UCT arm always significantly higher than that in the CCT arm.20
Robustness Checks using Alternative Measures of Mental Health
B. Channels of Impact
In Section II, we briefly discussed the mechanisms through which a cash transfer program, such as the one examined in this study, can influence mental health among its adolescent female beneficiaries and identified five main channels: increased schooling and family support for education, improved physical health, higher levels of personal consumption, opportunities for leisure and social activities, and reductions in parental stress. The rich set of data collected under this study allows us to construct indicators for each of these categories except parental stress.21 We construct eight indicators for the school-age respondent among baseline schoolgirls:22 whether she was enrolled in school during Term 3 of the 2008 school year (coinciding with Round 2 surveys for most of the respondents), whether her household supports her education, whether she has been ill in the past two weeks, whether her self-reported health status is very good, whether she can walk a distance of five kilometers easily, whether she usually wears shoes, the number of days she consumed meat, fish, or eggs last week, and the number of days she got together with friends for food or drinks during the past month.23
Table 5 first examines the program impact on these indicators and shows that the CCT and UCT treatments had different impacts on these indicators, which may serve as likely channels for the differential impacts on psychological distress observed in Tables 3 and 4 above. Both regular school attendance and household support for schooling are increased in the CCT arm. Girls offered conditional transfers also are less likely to have been ill in the past two weeks, consistent with the evidence that they were more likely to sleep under bed nets (Baird, McIntosh, and Özler 2011). Finally, the number of days CCT recipients consumed proteins from meat, fish, or eggs over the past seven days is increased by approximately half a day. However, CCT recipients were neither more likely to wear shoes nor to get together with their friends more often. In contrast, school attendance did not increase significantly in the UCT arm, despite the significantly larger increase in the support for education they report receiving from their households. Nor were they likely to have improved physical health. But, personal consumption and leisure in the UCT arm increased substantially: they were 12 percentage points more likely to wear shoes and had more opportunities for leisure with friends. These findings indicate that the program had a strong effect on outcomes that might influence mental health among school-age girls. Furthermore, these effects are in the expected directions in the two treatment arms: the effects in the CCT arm are on school attendance and related outcomes (illness and food consumption), while the effects in the UCT arm are on personal consumption and opportunities for leisure.24
Program Impacts on Potential Channels among Baseline Schoolgirls (Schooling, Physical Health, Personal Consumption and Leisure)
To provide an estimate of the share of the overall treatment effect on psychological distress accounted for by these channels we depart from experimental analysis. Following Flores and Flores-Lagunes (2009), we reestimate the regressions presented in Columns 2 and 4 of Table 3, but, this time, include the Round 2 values of the potential channel indicators. Assuming that the true relationship between the channels and psychological distress is linear and common to treatment and control, the channel indicators pick up the mechanism effects, while the effects remaining in the treatment dummies are the net effects in each treatment arm. Table 6 presents the results for each of the main channels separately (Columns 2–5) and for all of them (Column 6), and also indicates the fraction of the overall treatment effect explained by the specified channel.
Average Impact of the Zomba Cash Transfer Program on GHQ-12 Binary Measure of Psychological Distress, including Endogenous Channels as Controls
In Columns 2 and 4, we see that the schooling channel and the consumption channel each account for approximately 20 percent of the overall treatment effect in either treatment arm. Column 5 shows that increased leisure accounts for little, if any, of the impact. The physical health channel also accounts for more than 20 percent of the overall effect in the CCT arm, but none in the UCT arm, a result consistent with the strong program impact on illness that is only observed in the CCT arm. When all the potential channels are included in the regression analysis, they account for close to half of the overall effect in the CCT arm, rendering the net effect remaining in the CCT dummy statistically insignificant (Column 6). The same channels account for a little more than one-third of the effect on psychological distress in the UCT arm and the net effect that remains in the UCT dummy is still large and statistically significant at the 95 percent confidence level.
We conclude that cash transfer programs can reduce psychological distress among school-age girls through a variety of channels hypothesized in the literature. However, the size of the overall effects and the channels are likely to differ depending on whether the cash transfers are conditional on school attendance or not. Under a CCT, the improvement in mental health is through increased school attendance and related behaviors (reduced illness from sleeping under bed nets and increased food consumption), while it is more likely to be due to increased household support and personal consumption under a UCT. There is also a larger net effect that remains unexplained in this latter group.
C. Heterogeneity of Program Impacts
While the analysis above sheds light on the possible causal mechanisms responsible for the treatment effects observed in Tables 3 and 4, it does not explain why the overall effect is significantly larger in the UCT arm than the CCT arm. In this subsection, we examine the elasticity of psychological distress with respect to the randomized benefit levels offered separately to the adolescent girl and her parents / guardians, which provides a partial explanation.
As described in Section IVC, the transfers to the household were split between the adolescent girl and her parents / guardians. The transfer size to the girl was randomized at the individual level to take an integer value between $1 and $5 per month, while the transfer size to the parents was randomized at the EA level and took a value of $4, $6, $8, or $10 per month. In this subsection, we reestimate the treatment effect of the program by including additional regressors for the individual and household amounts in our analysis as follows:
(2)
where 
and 
are again dummy variables for the CCT and UCT offers, 
and 
give the individual and household transfer amounts in each treatment arm respectively, defined in differences from the lowest amount offered in the treatment arms ($1 for the individual transfer, $4 for the household transfer). The estimates of δ and ϕ thus give the marginal effect of increasing individual and household amounts by $1 under each treatment arm, while the estimates of γ measure the impact of each treatment at the lowest total transfer amount.
The results of the estimation of Equation 2 are displayed in Table 7. The coefficients on the conditional and unconditional treatment indicators in Column 3 indicate that when the amount transferred to households in which baseline schoolgirls reside is at its minimum total value of $5 / month, the beneficial impact of CCTs on the mental health of adolescent girls is large and roughly equal to that of UCTs. However, each additional dollar transferred to the parents of a baseline schoolgirl conditional on her school attendance increases the likelihood of her suffering from psychological distress during the intervention by approximately three percentage points (Column 3). This detrimental effect of additional conditional dollars to the parents disappears completely after the end of the intervention (Column 4). In contrast, we detect no relationship between the parental transfer size and the mental health of the adolescent baseline schoolgirls if the transfers were offered unconditionally. Interestingly, Column 1 shows that increased conditional cash transfers to the parents had a similarly detrimental effect on the mental health of baseline dropouts during the intervention (p value=0.122), which also disappeared by Round 3 (Column 2). In contrast to the transfers given to the parents, transfers given to the girls seem to have mildly beneficial effects on their mental health, especially in Round 3 (Column 4).
Impact of Transferred Amounts on GHQ-12 Binary Measure of Psychological Distress
These findings indicate that, at low levels of transfers to the parents, CCTs and UCTs are equally effective in improving mental health among baseline schoolgirls. However, simply doubling the amount offered to the parents (from $4 to $8 / month) is sufficient to entirely wipe out this beneficial effect in the CCT arm.25 Cash transfers tied to an adolescent girl make her a breadwinner for her household regardless of the conditionality. Our results suggest that when the transfers become an important source of income for the entire family and are conditional on her actions each month, they might turn into a heavy burden for her to shoulder and become detrimental to her mental health.
D. Effects on Untreated Girls in Treatment EAs
Thus far, we focused on the effects of the program on individuals assigned to one of the treatment arms, but our study sample contains a group of untreated individuals in treatment EAs. Interventions to improve health, to increase schooling, or to reduce poverty are commonly targeted to communities in need, and households or individuals within such communities. The expected effect is a Pareto-improvement in welfare: Treated households in treatment communities benefit, while those that are untreated in the same areas (or elsewhere) are no worse off. In fact, program spillovers that have been shown to exist are generally positive (Angelucci and De Giorgi 2009; Miguel and Kremer 2004).
However, it is possible that targeting interventions within communities also can have some detrimental effects among the untreated. There is a large literature on the “relative deprivation” hypothesis, which shows that one’s health can be adversely affected by his / her relative economic status within a reference group. Recent examples finding empirical support for this hypothesis include Eibner and Evans (2005), who suggest that people in the United States respond to the stress and low self-esteem caused by relative deprivation by engaging in health compromising behaviors; Mangyo and Park (2011), who find that increases in mean income in one’s township in China leads to declines in one’s own psychosocial health measured by the Center for Epidemiologic Studies Depression (CES-D) scale; and Luttmer (2005), who shows that an increase in neighbors’ earnings in the United States is associated with a decrease in one’s self-reported happiness and with an upward movement at the bottom of the CES-D scale.
A randomized cash transfer program exogenously creates an inequality in income within a reference group if only a subset of that group is chosen as program beneficiaries.26 In the context of this experiment, if adolescent girls care about their relative income or status within their peer group, the mental health of eligible nonbeneficiaries may deteriorate due to the stress and anxiety caused by the increased inequality in income and educational attainment within treatment EAs. The effects on mental health among untreated schoolgirls could also run in the other direction. If the program has positive spillover effects in income or schooling, mental health could improve among this group. Similarly, siblings of treated girls may indirectly benefit from the additional income coming into their households or simply benefit from living with sisters with improved mental health.
Participants in the Zomba cash transfer program were randomly chosen from the study population of never-married females aged 13–22 within each treatment EA. Intervention effects on untreated girls in treatment EAs can be identified by comparing their outcomes to those in control EAs. Furthermore, because the program was assigned at the individual (and not the household) level, it is possible to examine these spillover effects separately for those who lived in the same household with a program beneficiary and for those who lived in households with no program beneficiaries. Of the 564 baseline schoolgirls in our study sample who were untreated in treatment villages and remain in our panel data set, 69 lived in the same household with at least one treatment girl at baseline.27
The first column of Table 8 shows that untreated girls in treatment EAs are significantly more likely to suffer from psychological distress than the control group during the intervention period. Psychological distress among this group is 6.4 percentage points higher than in the control group, statistically significant at the 95 percent confidence level. Columns 3 and 5 of Table 8 separately examine the spillover effects during the intervention for untreated girls who did not live in a treatment household at baseline and for those who did, respectively.28 The results are striking: among baseline schoolgirls who did not live in a household with an eligible sibling at baseline, the prevalence of psychological distress in Round 2 is 9.9 percentage points higher in treatment EAs than control EAs, statistically significant at the 99 percent confidence level (Column 3). Siblings of program beneficiaries, on the other hand, seem to have reaped substantial benefits from the program: the prevalence of psychological distress among this group in Round 2 is 8.6 percentage points lower than the control group (p value = 0.103). All of these effects disappeared quickly after the intervention ended (Columns 2, 4, and 6).
Spillover Effects on GHQ-12 Binary Measure of Psychological Distress of Baseline Schoolgirls
Given that we are unable to find any effects on other outcomes that might act as causal mechanisms for the detrimental effects on mental health observed among the nonbeneficiaries—such as individual consumption, school enrollment, learning achievement, or household support—we conclude that our findings are consistent with the idea that the utility of school-age girls depend on their relative income in addition to their own income.29 The fact that the detrimental effects on mental health entirely disappeared quickly after the cash transfer program ended provides further support for this hypothesis.30
VII. Concluding Discussion
Given the large toll that mental health problems take during adolescence, as well as their potential long-run health and developmental effects, protecting adolescent mental health is an important yet underappreciated area in the public policy spheres. This study took advantage of a cash transfer experiment to assess the effects of randomly varied positive income shocks on mental health among school-age girls. There are three main findings worth discussing here.
First, unconditional cash transfers had a striking impact on the mental health of baseline schoolgirls in Malawi. The average effect size in the UCT arm, a 38 percent decline in the likelihood of scoring above the GHQ-12 screening threshold score of 3, is smaller than the effect of migration from Tonga to New Zealand (Stillman, McKenzie, Gibson 2009), but similar to the effect of moving to richer neighborhoods under the Moving to Opportunity program (Kling, Liebman, and Katz 2007), the latter of which, according to the authors, is comparable to “some of the most effective clinical and pharmacological mental health interventions.” Given that migration to richer countries, clinical interventions, and drug treatment programs are not realistic options for young people in many poor countries, it is encouraging to find that small, regular monthly cash transfers can reduce psychological distress among such populations.
Second, baseline schoolgirls who were offered cash transfers conditional on regular school attendance also experienced a significant improvement in mental health, but, at 17 percent, the decrease in psychological distress was significantly smaller than that for girls in the UCT arm. Given that a cash transfer program (conditional or not) alters many facets of a young girl’s life, the study design was not ideally suited to identify causal mechanisms underlying the observed impacts. However, using nonexperimental methods, we find that improved physical health, increased school enrollment and household support for education, and higher levels of personal consumption all contribute to the effects in both treatment arms, but do not explain why the overall effects are substantially smaller in the CCT arm. By further exploiting the randomized variation in transfer size and the identity of the recipient, we are able to provide a potential answer to this question: Our findings suggest that when an important source of income for the family is made conditional on the actions of a school-age girl, it might place a heavy burden on her and affect her mental health adversely—in a manner UCTs to the household would not.
The differing effects of UCTs and CCTs on adolescent mental health should be considered in the context of all the other program effects, including schooling. As the effects on school enrollment, attendance, and achievement were significantly higher in the CCT than the UCT arm (Baird, McIntosh, and Özler 2011), the relative effectiveness of these two treatments in terms of long-term welfare are not yet clear and future research is needed. However, given that increased CCT amounts do not improve schooling outcomes while increasing stress, policymakers may prefer smaller transfers to a larger set of beneficiaries.
Third, the intervention increased psychological distress among untreated girls in treatment villages, while it had beneficial effects on the female siblings of program beneficiaries. To our knowledge, this is the first study to find negative externalities in the context of a cash transfer program. The findings are consistent with increased stress and lower self-esteem among untreated girls in treatment EAs in response to the sudden inequality in income and school participation within their reference group. While this inequality was a result of a randomized program design in this instance, our findings may have broader implications for welfare assessments of poverty-alleviation programs that employ commonly used methods of targeting individuals or households within communities, especially around the eligibility threshold.
We found no effects of the same cash transfer program on the cohort of baseline dropouts—those who had already dropped out of school by the time the program was starting. Given that there were large increases in school attendance, household support for schooling, and food consumption, it is somewhat surprising that there were no beneficial effects on their mental health. A similar analysis of mechanisms suggests that none of the outcomes that are strongly associated with psychological distress among baseline schoolgirls influence mental health in this cohort. Unlike baseline schoolgirls, the program decreased the hours of sleep and leisure enjoyed by these girls, so perhaps there were offsetting effects associated with returning to school.
The reader may be concerned that we are detecting a change in the reference point for mental health among program beneficiaries in Round 2 rather than an actual improvement. We find this explanation unlikely. A change in the reference point would imply that the reference points, or the “usual” state, for the questions contained in the GHQ-12 became lower for program beneficiaries (relative to those in control EAs) between Rounds 1 and 2. If anything, we would expect the opposite as the cash transfers improved the lives of the study participants. A second concern may be that while the cash transfer intervention was in progress, program beneficiaries found it desirable to answer the GHQ-12 questions more optimistically than the control group. However, if this effect was responsible for our findings, then we would expect to find large reductions in psychological distress among baseline dropouts in treatment EAs.
Overall, the results presented in this paper indicate that the psychological wellbeing of adolescent girls can substantially improve when they experience positive income shocks. However, if these income shocks are administered as part of a cash transfer intervention, then small changes in program design parameters can make large differences in program impacts. Policymakers would be wise to take into consideration the potential externalities, as well as the heterogeneous program effects with regards to the conditionality, the transfer size, and the recipient of the transfers within the household while designing cash transfer programs.
Appendix 1 The GHQ-12 And The MHI-5
A. The GHQ-12
The GHQ-12 items, ordered as they appear in the questionnaires, are listed below.
During the past two weeks, have you been able to concentrate on whatever you are doing?
During the past two weeks, have you lost much sleep over worry?
During the past two weeks, have you felt that you were playing a useful part in things?
During the past two weeks, have you felt capable about making decisions?
During the past two weeks, have you felt constantly under strain?
During the past two weeks, have you felt that you couldn’t overcome your difficulties?
During the past two weeks, have you been able to enjoy your normal day-to-day activities?
During the past two weeks, have you been able to face up to your problems?
During the past two weeks, have you been feeling unhappy and depressed?
During the past two weeks, have you been losing confidence in yourself?
During the past two weeks, have you been thinking of yourself as a worthless person?
During the past two weeks, have you been feeling reasonably happy, all things considered?
Each question has five possible answers: (i) much more than usual, (ii) more than usual, (iii) same as usual, (iv) less than usual, and (v) much less than usual. When scored bimodally, respondents who report an item as applying to them “more than usual” or “much more than usual” (“less than usual” or “much less than usual” for positively phrased items) score 1 on this item, whereas others score 0. These binary scores are summed up and summed scores of 3 or higher are classified as cases suffering from psychological distress.
B. The MHI-5
The MHI-5 consists of the following five questions:
How much of the time during the past month have you been a very nervous person?
How much of the time during the past month have you felt calm and peaceful?
How much of the time during the past month have you felt downhearted and blue?
How much of the time during the past month have you been a happy person?
How much of the time during the past month have you felt so down in the dumps that nothing could cheer you up?
Each question has six possible answers: (i) all of the time, (ii) most of the time, (iii) a good bit of the time, (iv) some of the time, (v) little of the time, and (vi) none of the time. Respondents who report an item as applying to them “all of the time” score 5 and respondents who report an item as applying to them “none of the time” score 0 (vice versa for positively phrased items). Summed scores provide an indication of psychological distress.
Baseline summary statistics (by amount)
Footnotes
Sarah Baird is an assistant professor of Global Health and Economics at George Washington University in Washington, D.C. Jacobus de Hoop is an impact evaluation specialist at the International Labor Office in Rome. Berk Özler is a senior economist at the Development Research Group, World Bank in Washington, D.C. The authors thank three anonymous referees, Jishnu Das, David McKenzie, participants at the National Bureau of Economic Research (NBER) Africa Project Research Conference, and at seminars at the Free University of Amsterdam, London School of Tropical Hygiene and Medicine, Paris School of Economics, University of Oxford, University of Otago, and the World Bank for valuable comments on earlier drafts of this paper. The authors gratefully acknowledge funding from the Global Development Network, Bill and Melinda Gates Foundation, NBER Africa Project, 3ie Open Window (Round 2), and World Bank. The data used in this article can be obtained beginning October 2013 through September 2016 from Berk Özler, World Bank, 1818 H St., NW, Washington, DC 20433, bozler{at}worldbank.org.
↵1. In a review of 11 epidemiological community studies from developed and developing countries, Patel et al. (2007) show that the prevalence rate of mental disorders among adolescents ranges from 8 percent in the Netherlands to 27 percent in Australia. Poor mental health also demands a heavy toll among young women in Sub-Saharan Africa: after HIV / AIDS and abortion, depression makes up the leading contribution to disability adjusted life-years (DALYs).
↵2. The age of onset for most mental disorders likely to persist into adult life, including depressive and anxiety disorders, falls within the 12–24 year age range (Patel et al. 2007).
↵3. Patel et al. (2007) state: “We were unable to identify a single intervention targeted at young people in low-income and middle-income countries that improved mental-health outcomes.” Ssewamala, Han, and Neilands (2009) finds positive effects of an asset-promotion program on the self-esteem of children, aged 11–17. Filmer and Schady (2009) show that a cash transfer intervention in Cambodia had a small effect on the mental health of its adolescent beneficiaries. Lund et al. (2011) review the impact of cash transfer interventions in Mexico and Ecuador on the mental health of young children.
↵4. Although the impacts were short-lived, they may still have substantial long-term consequences. For example, many women in developing countries, including Malawi, begin childbearing during adolescence. Maternal mental health problems are linked to preterm delivery, low birth weight, decreased infant growth, and higher mortality (Prince et al. 2007), while maternal stress during pregnancy can affect subsequent educational attainment of those infants exposed to elevated levels of stress (Aizer, Stroud, and Buka 2009). Thus, even transitory improvements in the psychological health of adolescents may translate into potentially long-lasting beneficial impacts on their children (Friedman and Sturdy 2011).
↵5. These common disorders are among the most important causes of morbidity in primary care settings, and make a significant contribution to the burden of disease and disability in low- and middle-income countries (Patel and Kleinman 2003; Lund et al. 2011).
↵6. There is also evidence on the relationship between income levels and subjective well-being or happiness. This literature was given impetus by Easterlin (1974). For references to this extensive literature we refer the reader to the introduction of Gardner and Oswald (2007). Friedman and Thomas (2008) and Das et al. (2008) both argue that measures of mental health appear to be distinct from subjective measures of welfare, such as happiness.
↵7. Schooling can have effects on mental health through other channels as well. Academic failure is a risk factor for poor mental health while involvement in school life and positive reinforcement from academic achievement are protective factors (Patel et al. 2007). Cash transfers tied to the girl also can provide the school-age girl with an opportunity for positive involvement in her family. Finally, schooling can increase an individual’s ability to solve problems, which is also considered a protective factor.
↵8. It should be noted, however, that the evidence of moderation of the effect of cash transfers on children’s stress through a reduction in mother’s depressive symptoms in Fernald and Gunnar (2009) is for young children (ages 2–6) and not school-age girls.
↵9. DSM-IV refers to the fourth (and the latest) edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders. ICD-10 refers to the World Health Organization’s International Classification of Diseases.
↵10. However, further psychiatric assessment is necessary to determine whether those who are classified as cases using these screening tools suffer from actual pathology.
↵11. GHQ-12 has been translated to, and validated in, all major languages and many smaller languages, such as Malay (Yusoff, Rahim, and Yaacob 2009), Persian (Montazeri et al. 2003), and Yoruba (Gureje and Obikoya 1990). For this study, the GHQ-12 was carefully translated (including back-translation) into Chichewa, the main language spoken in the study area.
↵12. Nor did we find any validation studies of the GHQ-12 for Malawi.
↵13. This section draws heavily from Section II in Baird, McIntosh, and Özler (2011).
↵14. The “conditionality” experiment was not conducted among baseline dropouts. As the sample size for this cohort is relatively small, dividing the treatment group into a conditional and an unconditional treatment arm would yield an experiment with low statistical power.
↵15. Measures of psychological distress were not collected at baseline, so we cannot provide direct evidence of the balance of this outcome across treatment groups. Similarly, data on household support for education, wearing shoes, and consumption of proteins during the past seven days were only collected at followup; hence we are not able to provide evidence of baseline balance for these variables.
↵16. 12 girls were excluded from the panel due to missing mental health data in Round 2 or 3. We were able to construct our main binary indicator of psychological distress for another 14 observations, for which there is a missing value for at least one GHQ-12 item, as each of these cases could be categorized as a 0 or 1 regardless of the missing item(s). The summed GHQ-12 indicators are missing for these observations.
↵17. 97 percent of the control group among baseline schoolgirls has a score of 9 or lower.
↵18. Detailed results are available from the authors upon request.
↵19. When the GHQ-12 is scored as a Likert scale, responses of “much less than usual” and “less than usual” are assigned a score of 0, “same as usual” 1, “more than usual” 2, and “much more than usual” 3 (and vice versa for positively phrased items). The scores are then summed to create a total score between 0 and 36.
↵20. The GHQ-12 scored bimodally has a large number of 0 or small values with a long tail. Hence, with its distribution skewed to the left, it can be analyzed using a count model rather than OLS. Analyzing this variable using a negative binomial regression does not alter the findings.
↵21. We have not collected mental health data for the parents of the school-age study participants. However, we asked her whether “conflict / fighting in her household was higher or lower compared to 12 months ago.” We find no effects on this outcome in either treatment arm, so we exclude it from the analysis of channels of impact.
↵22. As we found no impacts among baseline dropouts, the channels of impact analysis in Tables 5 and 6 is limited to the cohort of baseline schoolgirls in Round 2.
↵23. In a multivariate regression analysis conducted in the control group among baseline schoolgirls, all of these variables are significantly associated with psychological distress.
↵24. It should be noted that wearing shoes also can be a health channel by providing protection against hookworm infections, which can result in lack of energy and anemia.
↵25. Including the followup values of indicators to account for the channels of impact in Equation 2 leaves the size and statistical significance of the parental transfer size gradients unchanged; indicating that the detrimental marginal effect of increased transfers offered conditionally to the parents is independent of the hypothesized mechanisms.
↵26. We are not aware of any studies of cash transfer programs that investigate effects on the mental health or subjective well-being of untreated households.
↵27. We refer to these 69 eligible girls in the study sample as siblings even though we do not know their exact relationship to the core respondent.
↵28. Households containing a single adolescent girl eligible for the program may differ from those with more than one eligible girl. To estimate unbiased impacts on girls in these two types of households, we control for the number of eligible girls in Columns 3–6.
↵29. Program effects on these outcomes among this group are available from the authors upon request.
↵30. We cannot rule out the possibility that the increase in psychological distress among untreated girls in treatment EAs is due to the fact that they viewed being randomly excluded from the program as being “unfair” rather than their utility being dependent on the income (or status) of others in their reference group.
- Received July 2011.
- Accepted April 2012.
References
Jump to section
Related Articles
Cited By...
- The Economics of Child Mental Health: Introducing the Causes and Consequences of Child Mental Health Special Issue
- Improving Mental Health of Adolescent Girls in Low- and Middle-Income Countries: Causal Evidence from Life Skills Programming
- Child Control in Education Decisions: An Evaluation of Targeted Incentives to Learn in India