Scholarly article on topic 'Poverty and perceived stress: Evidence from two unconditional cash transfer programs in Zambia'

Poverty and perceived stress: Evidence from two unconditional cash transfer programs in Zambia Academic research paper on "Health sciences"

CC BY-NC-ND
0
0
Share paper
Academic journal
Social Science & Medicine
OECD Field of science
Keywords
{"Perceived stress" / "Unconditional cash transfer" / "Food security"}

Abstract of research paper on Health sciences, author of scientific article — Lisa Hjelm, Sudhanshu Handa, Jacobus de Hoop, Tia Palermo

Abstract Introduction Poverty is a chronic stressor that can lead to poor physical and mental health. This study examines whether two similar government poverty alleviation programs reduced the levels of perceived stress and poverty among poor households in Zambia. Method Secondary data from two cluster randomized controlled trials were used to evaluate the impacts of two unconditional cash transfer programs in Zambia. Participants were interviewed at baseline and followed over 36 months. Perceived stress among female caregivers was assessed using the Cohen Perceived Stress Scale (PSS). Poverty indicators assessed included per capita expenditure, household food security, and (nonproductive) asset ownership. Fixed effects and ordinary least squares regressions were run, controlling for age, education, marital status, household demographics, location, and poverty status at baseline. Results Cash transfers did not reduce perceived stress but improved economic security (per capita consumption expenditure, food insecurity, and asset ownership). Among these poverty indicators, only food insecurity was associated with perceived stress. Age and education showed no consistent association with stress, whereas death of a household member was associated with higher stress levels. Conclusion In this setting, perceived stress was not reduced by a positive income shock but was correlated with food insecurity and household deaths, suggesting that food security is an important stressor in this context. Although the program did reduce food insecurity, the size of the reduction was not enough to generate a statistically significant change in stress levels. The measure used in this study appears not to be correlated with characteristics to which it has been linked in other settings, and thus, further research is needed to examine whether this widely used perceived stress measure appropriately captures the concept of perceived stress in this population.

Academic research paper on topic "Poverty and perceived stress: Evidence from two unconditional cash transfer programs in Zambia"

Contents lists available at ScienceDirect

Social Science & Medicine

journal homepage: www.elsevier.com/locate/socscimed

Poverty and perceived stress: Evidence from two unconditional cash transfer programs in Zambia

CrossMark

Lisa Hjelm a *, Sudhanshu Handa a b, Jacobus de Hoop a, Tia Palermo Zambia CGP and MCP Evaluation Teams

on behalf of the

a UNICEF Office of Research — Innocenti, Piazza SS, Annunziata, 12, 50122 Florence, Italy

b Carolina Population Center, University of North Carolina at Chapel Hill, 206 West Franklin St., Rm. 208, Chapel Hill, NC 27516, USA

ARTICLE INFO

Article history: Received 14 June 2016 Received in revised form 4 January 2017 Accepted 16 January 2017 Available online 19 January 2017

Keywords: Perceived stress Unconditional cash transfer Food security

ABSTRACT

Introduction: Poverty is a chronic stressor that can lead to poor physical and mental health. This study examines whether two similar government poverty alleviation programs reduced the levels of perceived stress and poverty among poor households in Zambia.

Method: Secondary data from two cluster randomized controlled trials were used to evaluate the impacts of two unconditional cash transfer programs in Zambia. Participants were interviewed at baseline and followed over 36 months. Perceived stress among female caregivers was assessed using the Cohen Perceived Stress Scale (PSS). Poverty indicators assessed included per capita expenditure, household food security, and (nonproductive) asset ownership. Fixed effects and ordinary least squares regressions were run, controlling for age, education, marital status, household demographics, location, and poverty status at baseline.

Results: Cash transfers did not reduce perceived stress but improved economic security (per capita consumption expenditure, food insecurity, and asset ownership). Among these poverty indicators, only food insecurity was associated with perceived stress. Age and education showed no consistent association with stress, whereas death of a household member was associated with higher stress levels. Conclusion: In this setting, perceived stress was not reduced by a positive income shock but was correlated with food insecurity and household deaths, suggesting that food security is an important stressor in this context. Although the program did reduce food insecurity, the size of the reduction was not enough to generate a statistically significant change in stress levels. The measure used in this study appears not to be correlated with characteristics to which it has been linked in other settings, and thus, further research is needed to examine whether this widely used perceived stress measure appropriately captures the concept of perceived stress in this population.

© 2017 UNICEF. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Stress is a determinant of poor mental health, a leading cause of disability in high-, middle- and low-income countries (Vos et al., 2015), and an important determinant of overall well-being. Therefore, it is important to measure stress as an outcome in its own right (Haushofer and Shapiro, 2016; Kling, 2007). Stress and mental health are both closely linked to poverty; studies from low-and middle-income countries have revealed a link between poor mental health and socioeconomic status (SES) indicators such as

* Corresponding author. UNICEF, Eastern and Southern Africa, PO Box 4414500100, Nairobi, Kenya.

E-mail addresses: lhjelm@unicef.org (L. Hjelm), shanda@email.unc.edu (S. Handa), jdehoop@unicef.org (J. de Hoop), tmpalermo@unicef.org (T. Palermo).

education, food insecurity, housing, social class, and financial stress (Lund et al., 2010). Given the adverse effects of poverty on mental health, this study hypothesized that a poverty-alleviation program (an unconditional cash transfer) would reduce poverty among poor households in Zambia and subsequently reduce stress in these households.

There are several hypothesized mechanisms through which poverty may influence mental health, including chronic stress, malnutrition, substance abuse, social exclusion, and exposure to trauma and violence. Known as the social causation hypothesis, it has been studied extensively (Johnson et al., 1999; Lund et al., 2011). In what is known as the social drift hypothesis, people with mental illness are at an increased risk of experiencing poverty through increased health expenditures, reduced productivity, and

http://dx.doi.org/10.1016/j.socscimed.2017.01.023

0277-9536/© 2017 UNICEF. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

stigma related to mental health (Lund et al., 2011). Thus, poverty and poor mental health mutually reinforce each other (Lorant et al., 2003; Lund et al., 2011). Poverty and low SES may also affect an individual's exposure to stress and stressful life events as well as his or her ability to cope with stress, as fewer social and psychological resources are usually available to overcome stressful events (Adler et al., 1994; Cohen, 1988; Cohen and Janicki-Deverts, 2012; Hamad et al., 2008).

Stress as a mechanism that links poverty and health merits further investigation. Psychological stress, which occurs when the experience of environmental demands exceeds an individual's ability to cope with the situation (Lazarus and Folkman, 1984), is associated with a range of physical and mental health states. It has been linked to depressive disorder, depressive symptoms, cardiovascular disease, and the risk of progressing from HIV infection to AIDS (Cohen et al., 2007). Experimental studies have shown that acute and chronic stressors can produce biological stress reactions, including excessive inflammation (McEwen and Seeman, 1999). Over the course of a lifetime, these reactions may contribute to morbidity and mortality disparities and increased levels of cortisol, particularly for stressors of an uncontrollable nature (Miller et al., 2007). Poverty-induced chronic stress has also been hypothesized to accelerate the natural aging of the immune system (referred to as immunosenescence) (Aiello and Dowd, 2013). Studies have demonstrated that individuals of lower SES show an increased antibody response to persistent herpes viruses, which may be due to differential exposure to stress (Aiello and Dowd, 2013) and reduced resources to cope with it (Kristenson et al., 2004). Aiello and Dowd hypothesized that increased stress, caused bya range of poverty-associated factors such as continuously activated stress-related autonomic and neuroendocrine responses, impairs immunity and leads to poor health outcomes. Maternal perceived stress has been associated with low birth weight and poor childhood nutritional status (Dole et al., 2003; Lobel et al., 1992; Rondo et al., 2013; Torche, 2011).

The majority of studies that examine the relationship between stress, SES (Cohen and Janicki-Deverts, 2012; Matthews et al., 2010), and stressful life events are associated with higher levels of perceived stress (Dowd et al., 2014; van Eck et al., 1998). These variables have been studied less in sub-Saharan African countries, where food insecurity (Pike and Patil, 2006) and HIV infection (Garcia et al., 2013) are more widespread, which may have implications for variation in stress levels by SES. A South African study found that perceived stress was related to subjective social status but not to other socioeconomic indicators, such as education, employment, and income (Hamad et al., 2008). A Kenyan study among farmers demonstrated that elevated levels of cortisol and self-reported stress were induced by the absence of rain, which caused a negative income shock (Chemin et al., 2013). Another study found a reduction in self-reported stress due to unconditional cash transfers (a positive income shock), but no impact on cortisol levels (Haushofer and Shapiro, 2016). A key issue in all of these studies is the use of measures of stress that have not been validated in sub-Saharan Africa and that, therefore, may not be appropriate in low-income settings of sub-Saharan Africa.

This study posited that cash transfer programs to improve food security and smoothing consumption would lead to reduced stress levels in a poor- and food-insecure setting in sub-Saharan Africa. Cash transfer programs directly aim at alleviating poverty and not at improving outcomes in mental health and related areas. Thus, the impacts of the cash transfer must first work through household-level outcomes such as food security, economic security, time use and labor decisions, and general stress levels. Then, the impacts make their way to individual-level outcomes, such as physical and mental health, perceived stress, expectations, and

outlook.

To date, certain studies in Kenya and Malawi have demonstrated that social cash transfers have improved mental health by decreasing the rate of depressive symptoms (as measured respectively by the Center for Epidemiologic Studies Depression Scale [CES-D] and the General Health Questionnaire [GHQ-12]). Evidence from Malawi suggests that the effects of cash transfers on depressive symptoms depend on program design, specifically the combinations of conditions and transfer amounts. Other studies have reported mixed impacts on cortisol levels of cash transfer beneficiaries, including protective impacts among Mexican children and no impacts among adults in a Kenyan sample (Baird et al., 2013; L. Fernald and Gunnar, 2009; Haushofer and Shapiro, 2016; Kilburn et al., 2015).

Current evidence on the relationship between cash transfer programs and perceived stress is mixed. There are examples of studies that have examined this relationship in Latin America (Ozer et al., 2011; Schady and Paxson, 2007) and Africa (Haushofer and Shapiro, 2016). In Mexico, participation in the Oportunidades program was associated with lower depression, and reduced perceived stress [measured by the Perceived Stress Scale (PSS)] was found to be the mediating factor in women (Ozer et al., 2011). In contrast, in Ecuador, participation in an unconditional cash transfer program had no significant effect on perceived stress (measured using a four-item version of the PSS) or on symptoms of depression (Schady and Paxson, 2007). In Kenya, participation in a cash transfer program reduced perceived stress (measured by the Cohen PSS) but not cortisol levels in the overall sample. Nonetheless, reductions in cortisol were seen among some subsamples, such as female recipients and participants who received lump-sum transfers rather than monthly transfers (Haushofer and Shapiro, 2016). On a related note, two additional studies examined the impacts of loan access and the provision of health care on perceived stress. A Kenyan study found that health care receipts reduced perceived stress (Chemin et al., 2016). A South African study found that, among individuals who were initially not offered a small loan, a second chance to receive the loan increased the levels of perceived stress (Fernald et al., 2008).

As outlined above, the evidence to date on poverty alleviation and perceived stress is mixed, and therefore the present study aimed to investigate (1) whether participation in a cash transfer program reduced poverty-related outcomes and perceived stress and (2) which individual- and household-level characteristics are associated with higher levels of perceived stress. To investigate these questions, data from longitudinal impact evaluations of two government cash transfer programs in Zambia were used. It is important to note that neither program was designed to address stress, but rather to address food insecurity and extreme poverty. Nevertheless, given the theoretical link between poverty and stress, and the fact that food insecurity is a widespread problem in this population, it is of policy and public health interest to assess the link between the programs and perceived stress.

1. Method

1.1. Interventions

The Zambia Child Grant Program (CGP) is a government-run unconditional cash transfer program targeting households with a child under the age of five. The CGP's objectives include supplementation of household income, increased enrollment and attendance in primary school, reduced child morbidity, productive assets, food security, and improved mortality and nutrition. Districts for program implementation were targeted by the government because of their high rates of mortality, morbidity, stunting,

and wasting among children aged zero to three years. Households included in the program receive an amount equivalent to 11 USD per month, which is estimated to be sufficient to cover the cost of one meal per person per day in an average-sized household. Households "age-out" or graduate from the program when the index child turns five, although in practice this was not implemented until after the 36-month evaluation survey was conducted (American Institutes for Research, 2011).

The Zambia Multiple Category Cash Transfer Program (MCP) is another government-run unconditional cash transfer program in Zambia, also implemented by the Ministry of Community Development, Mother and Child Health. The objectives of the MCP are to assist the most vulnerable households in the society, allowing them to meet their basic needs related to health, education, food, and shelter. The program targets households that fall into any of the following categories: female headed and keeping orphans, having a disabled member, headed by an elderly person and keeping orphans, or special cases of critically vulnerable households that are not included in any of the aforementioned categories (American Institutes for Research, 2012).

1.2. Sample data

This study used secondary data that were collected as part of the impact evaluations of the CGP and the MCP. These impact evaluations were randomized controlled trials, designed and implemented by The American Institutes for Research and the University of North Carolina at Chapel Hill under contract to UNICEF-Zambia.

The CGP impact evaluation comprised 2515 households at baseline from 90 communities (randomized into treatment and control arms) in three districts - Kaputa, Kalabo, and Shang'ombo — for a total evaluation sample size of 14,565 individuals. Baseline data were collected in December 2010, and follow-up data were collected in September and October 2012 (24 months), June and July 2013 (30 months), and September and October 2013 (36 months). The MCP impact evaluation took place in 92 communities (randomized to treatment and control arms) within two districts, Luwingu and Serenje. MCP baseline data, including 3078 households and 15,630 individuals, were collected in November and December 2011, and follow-up data were collected in November and December 2013 (24 months) and November and December 2014 (36 months).

In the subsample under analysis in the present study, data were used from the observations of female caregivers of children (main household survey respondents). These caregivers were observed at baseline and 36 months for the MCP and at baseline, 30 months, and 36 months for the CGP. The official evaluation reports demonstrated balance at baseline (indicating successful randomization of the treatment and control arms) and no evidence of selective attrition between study arms (American Institutes for Research, 2011; 2012). Data collections and analyses plans went through ethical review at the American Institutes for Research in Washington, DC, and at the University of Zambia Ethical Review Committee. Informed consent was obtained from all study participants.

1.3. Measures

This study used the PSS to measure psychological stress. The PSS was developed based on the concept of stress as an interaction between environmental demands and the individual's capacity to cope (Cohen et al., 1983). Originally developed with 14 items, its creators later refined it to 10 items (the PSS10) of which 6 are negatively phrased and 4 are positively phrased (Cohen, 1988). These items consider the degree to which individuals experience

their lives as unpredictable, uncontrollable, and overloading (Cohen et al., 1983). This scale is one of the most frequently used measures of perceived stress and has been validated in many countries around the world; it is increasingly being used in sub-Saharan Africa (e.g., Garcia et al., 2013; Hamad et al., 2008; Lemma et al., 2012), but to current knowledge, it has never been validated there. The PSS has been associated with depressive symptomatology in many Western countries using the original English language questionnaire (Cohen et al., 1983; Eisenbarth, 2012; Hewitt et al., 1992) and in other regions and languages (e.g., Andreou et al., 2011; Chaaya et al., 2010; Reis et al., 2010; Wanget al., 2011).

Consistent with previous research, principal component analysis resulted in two factors in this study's samples, one consisting of the negatively worded items and the other consisting of the positively worded items (Cohen, 1988). However, the two subscales were not closely correlated. The "negative" subscale showed more variation in relation to happiness and optimism compared to the "positive" subscale, and the "positive" subscale did not show consistent associations. Thus, this study concluded that the positively worded items did not perform well in this setting, and a consolidated stress scale including both positively and negatively worded items would not be suitable as a single measure of perceived stress in these two samples. This study therefore included only the six negatively worded PSS items. The questions included in the final scale were:

In the last 4 weeks, how often have you been upset because of something that happened unexpectedly? In the last 4 weeks, how often have you felt that you were unable to control the important things in life? In the last 4 weeks, how often have you felt nervous and "stressed"? In the last 4 weeks, how often have you found that you could not cope with all things that you had to do? In the last 4 weeks, how often have you been angered because of things that were outside of your control? In the last 4 weeks, how often have you felt difficulties were piling up so high that you could not overcome them?

Likert-type responses ranged from zero (never) to four (very often/always). The scale was constructed by adding the total score from each question, resulting in a scale ranging from 0 to 24. Cronbach's a was 0.84 for the CGP and 0.83 for the MCP sample, indicating the high internal reliability in each sample. For MCP, 3% of the analysis had missing values for the PSS, while CGP had 6%. There was no systematic difference in treatment arm, age, education, or marital status between women who responded and those with missing values. Due to their relatively small number, these observations were dropped from the analysis.

Because this study hypothesized that cash transfers could alleviate stress through the poverty pathway, it also examined whether the program affected the following poverty-related outcomes: household consumption expenditures, food security, and the number of nonproductive assets owned. Food insecurity was measured using the previously validated (Knueppel et al., 2010; Maes et al., 2009) Household Food Insecurity Access Scale (HFIAS) (Coates et al., 2007). The scale consists of nine items capturing different severities of food security, from worrying about not having enough food to not eating at all because of lack of food. The questions referred to the previous four weeks. Severe food insecurity was calculated based on the occurrence of the more severe experiences of food insecurity included in the HFIAS questionnaire, for example, if there was no food to eat of any kind because of a lack of resources.

Monthly consumption expenditure per capita in Zambia kwacha (ZMW) was calculated using an expenditure module, which was

adopted from the Zambian Living Conditions Monitoring Survey, covering a broad range of expenditure categories and including more than 200 items. At follow-up, expenditures were deflated to baseline year ZMW values, 2010 for CGP and 2011 for MCP.

The asset ownership indicator is the sum of nonproductive assets owned from a list of 10 assets (clock, watch, mobile phone, DVD, television, radio, sofa, table, mattress, and bed) in the CGP and seven assets in the MCP (clock, watch, mobile phone, radio, sofa, table, and mattress).

1.4. Covariates

Individual-level control variables included age of the respondent, whether the respondent had ever attended school — and subsequently — highest grade attained in school; and whether the respondent was married, never married, divorced, or widowed. Control variables at household level included the total number of household members, number of household members of different age groups (0—5, 6—12, 13—18, 19—35, 36—55, 56—69, and 70 + years), and the district where the household was located. Control variables also included the poverty status at baseline in terms of food insecurity, log of per capita expenditures, and asset ownership (described above). To measure stressful life events, we examined any death in the household (recall period since last follow-up survey).

1.5. Statistical analyses

This study first examined sample characteristics, including descriptive statistics and covariates for balance between treatment and control samples at baseline, and then performed two sets of multivariate analyses. In the first set of multivariate analyses, we investigated treatment effects on perceived stress and poverty-related indicators using ordinary least square (OLS) regressions. These analyses were conducted using cross-sections from the latest follow-up waves, specifically the 36-months follow-up, when perceived stress data were collected. To estimate program impacts, regressions with a treatment indicator (1 = treatment, 0 = otherwise) were run. The covariates listed above were controlled to improve the precision of the estimates. Additionally, regressions without controls were run, and it was concluded that adding controls did not significantly change the results. This study used baseline values of control variables because the program may have affected them, and thus, using contemporaneous values may have underestimated the treatment effect on the outcomes of interest. Control variables with missing values were replaced with —1, and then an indicator (+1) variable was added (0 = otherwise) to control for missing information (which is why control variables in Table 1 may have fewer observations than regressions in Tables 2 and 3). To estimate treatment impacts, these analyses relied on the successful randomization of the program, which created statistically equivalent treatment and control groups. As there was no pretreatment measure of the key outcome indicator, the estimated treatment effect assumes that this measure was balanced at baseline, which is consistent with the results of the balance tests reported in Table 1 over a range of outcomes.

In the second set of analyses, this study examined determinants of perceived stress by estimating associations between perceived stress and household- and individual-level characteristics using OLS regressions on the cross-sectional control samples at 36 months. Treatment individuals in this latter analysis were excluded as the program may have mitigated some of the risk factors for stress in the treatment arm. As a robustness check for the determinants of stress analysis, individual fixed-effects OLS regressions were run using observations from the CGP control sample

at 30 months and 36 months. As these regressions control for unobserved characteristics, which may simultaneously affect both risk factors and perceived stress, they may reduce any remaining bias in the estimates. All standard errors for clustering were adjusted at the community level, the level of program randomization. Analyses used Stata Version 14.

2. Results

Of the 2515 households included in the CGP at baseline, 2273 had a female caregiver who was observed at baseline and had completed the PSS questionnaire at 30 and 36 months. In the MCP, of the 3077 households included at baseline, 2490 households had a female caregiver who was observed at baseline and had completed the PSS questionnaire at 36 months.

Table 1 describes the covariates and the poverty-related outcome variables at baseline, and examines the balance between the treatment and control groups of the two samples at baseline. Treatment and control arms were balanced (i.e., there were no statistically significant differences between treatment and control arms) at baseline for both samples. Although the average age for women interviewed in the CGP households was approximately 30 years, women in MCP households were considerably older with an average age of 52 years. The poverty-related outcome variables illustrate the deprived background of the households benefitting from these programs. Per capita monthly expenditure at baseline was 40 ZMW for households in the CGP and 49 ZMW for households in the MCP, roughly equivalent to 8 USD (or 26 cents per person per day) for CGP households and 10 USD (or 33 cents per person per day) for MCP households. Approximately 90% of the CGP households and 81% of the MCP households were severely food insecure at baseline.

Within the individual items making up the PSS, among the women in the CGP and MCP control groups, respectively 6% and 22% had been upset because something happened unexpectedly; 8% and 19% had felt that they were unable to control the important things in life; 11% and 23% had felt nervous and stressed; 12% and 21% had found that they could not cope with all things that they had to do; 12% and 21% had been angered because things were outside their control; and 18% and 27% had felt that difficulties were piling up so high that they could not overcome them fairly often or very often/always in the four weeks preceding the survey.

Table 2 shows the program impacts on perceived stress and poverty-related indicators for the CGP (shown in Panel A) and MCP (Panel B). The results indicate that the program had no statistically significant impact on perceived stress (columns 1 and 2). Still, the program was successful in reducing poverty-related outcomes. It increased the monthly per capita expenditures by 10 ZMW (17 ZMW in the MCP; column 2), an increase of 20% (28% in MCP). It also reduced household food insecurity by three points (0.5 SD) on the HFIAS [three points (0.6 SD) in the MCP; column 3] and increased nonproductive assets by 0.7 items (0.4 in the MCP; column 4) on average.

Table 3 presents the results examining the determinants of perceived stress using only observations from the control groups. In the OLS regressions (columns 1 and 3), it was found that the only covariate associated with perceived stress was household food insecurity, which was associated with stress levels that were 0.15 points higher in the CGP and 0.27 points higher in the MCP on average. After controlling for unobserved factors in the fixed effects model in the CGP, this study found that death in the household was also associated with stress levels that were 1.63 points higher on average (column 2). Age, educational attainment, household consumption expenditures, and assets were not associated with stress levels in these samples.

Table 1

Household characteristics of the Zambia Child Grant Programme (CGP) and Multiple Category Cash Transfer Programme (MCP) samples at baseline.

Characteristic

All (proportion or mean) Control (proportion or mean) Treatment (proportion or mean) p

Panel A: CGP

Characteristics of women Age

Ever attended school Highest grade completed Married Never married Divorced Widowed Household demographics Household size Number of people ages 0-5 Number of people ages 6-12 Number of people ages 13 - 18 Number of people ages 19 - 55 Number of people ages 56 or older Material well-being Household food insecurity access scale (HFIAS) (0—24)a Severely food insecure households Total household expenditure per person in the household Assets owned (0—10) Minimum n Panel B: MCP Characteristics of women Age

Ever attended school Highest grade completed Married Never married Divorced Widowed Household demographics Household size Number of people ages 0 - 5 Number of people ages 6 - 12 Number of people ages 13 - 18 Number of people ages 19 - 55 Number of people ages 56 or older Material well-being Household food insecurity access scale (HFIAS) (0—24)a Severely food insecure households Total household expenditure per person in the household Assets owned (0—7) Minimum n

2272 29.79 2271 0.71 2261 3.93 2266 0.73 2266 0.10 2266 0.10 2266 0.06

2273 5.71 2273 1.91 2273 1.27 2273 0.56 2273 1.87 2273 0.09

2235 15.23

2243 0.90

2271 39.82

2273 0.82 2235

2490 51.62

2481 0.61

2458 3.03 2474 0.29 2474 0.05 2474 0.14 2474 0.51

2490 5.16

2490 0.75

2490 1.33

2490 0.98

2490 1.38

2490 0.72

2431 14.68

2459 0.81 2490 48.63 2490 0.54 2431

5.65 1.92 1.27 0.53 1.83 0.10

5.18 0.73 1.28 1.03 1.41 0.74

5.77 1.90 1.27 0.60 1.90 0.09

15.05 0.90 40.79 0.92 1106

5.14 0.77 1.38 0.94 1.34 0.71

0.65 0.40 0.09 0.73 0.78 0.10 0.87

0.51 0.68 1.00 0.15 0.18 0.95

0.55 0.90 0.46 0.12

0.50 0.42 0.64 0.83 0.01 0.96 0.37

0.84 0.51 0.19 0.11 0.46 0.55

0.76 0.09 0.91 0.35

Note. p values are reported from Wald tests on the equality of means of Treatment and Control for each variable. Standard errors are clustered at the community level. a One question was dropped from the Household Food Insecurity Access Scale (HFIAS) at baseline.

3. Discussion

This study examined whether participation in an unconditional cash transfer program reduced perceived stress. Although the program was successful in reducing poverty as measured by three different household-level indicators, this study found no program impacts on perceived stress levels. These findings are similar to those found in Ecuador, where participation in an unconditional cash transfer program had no effect on perceived stress (Schady and Paxson, 2007). However, the findings of this current study contrast with quantitative findings from Kenya and Mexico, which found that cash transfers reduced perceived stress (as measured by the PSS) (Haushofer and Shapiro, 2016; Ozer et al., 2011).

When this study examined correlates of perceived stress, only food insecurity and household deaths were related to higher levels of stress, which suggests that the experience of not having enough food in the household is a key source of stress compared to other aspects of poverty that were examined. High levels of food insecurity are a persistent challenge in sub-Saharan Africa (FAO et al., 2015). In sub-Saharan African rural settings, food insecurity may be a more prominent source of insecurity than in other locations,

particularly in subsistence farming settings with distinct seasonal variations in the accessibility of food (Hadley and Patil, 2008). The insecurity of not knowing where the next meal will come from, having to deprioritize other necessities, and the shame of not having enough food for the children, causes stress among those that are living with food insecurity (Hadley etal., 2012). Hunger and food insecurity have been identified as important stressors in sub-Saharan Africa (Pike and Patil, 2006), and the experience of food insecurity as a stressful event raises the levels of perceived stress (Addo et al., 2011). The stress of food insecurity potentially explains the link between household food insufficiency and increased risk for depression that has been found in several studies in sub-Saharan Africa (Hadley et al., 2008; Maes et al., 2010; Palermo et al., 2013; Tsai et al., 2016), particularly in Zambia (Cole and Tembo, 2011).

There are several possible explanations regarding why this study did not find an association between stress and consumption expenditures, asset ownership, or cash transfer receipt. One is that relative poverty or the perceived SES could be more important in determining stress levels than absolute levels of poverty (measured by expenditures and asset ownership), and this is supported by

Table 2

Treatment effect on stress and indicators of poverty in Zambia's Child Grant Programme (CGP) and Multiple Category Cash Transfer Programme (MCP) at the 36-months follow-up.

Statistic Perceived Stress Expenditure Household Food Insecurity Number of non-productive

Scale (0-24)a per capitab Access Scale (0-27)b assets ownedb c

(1) (2) (3) (4)

Panel A: CGP

Treatment effect 0.07 10.43*** -2.86*** 0.72***

(t statistic) (0.21) (4.32) (-7.63) (6.29)

Number of observations 2273 2273 2269 2272

Control mean (SD) 7.60 (4.20) 50.98 (36.97) 13.50 (5.20) 0.86(1.44)

Treatment mean (SD) 7.70 (4.03) 62.52 (37.56) 10.54 (4.84) 1.71 (1.80)

Panel B: MCP

Treatment effect -0.42 16.68*** -3.02*** 0.37***

(t statistic) (-1.17) (4.76) (-6.94) (5.73)

Number of observations 2490 2490 2490 2490

Control mean (SD) 9.92 (4.73) 60.54 (40.53) 14.50 (5.54) 0.49 (0.87)

Treatment mean (SD) 9.58 (4.64) 76.87 (53.62) 11.52 (5.11) 0.84 (0.97)

Note. Impact estimated based on OLS regressions with and without controls. t statistics are based on standard errors clustered at the community level. *p < 0.05. **p < 0.01. ***p < 0.001.

a Control variables include age, education and marital status of the woman as well as poverty status at baseline (expenditure per capita, food insecurity and asset ownership), district, household size and demographic composition of the household.

b Control variables include poverty status at baseline (expenditure per capita, food insecurity and asset ownership), district, household size and demographic composition of the household.

c Number of assets owned range from 0 to 10 in the CGP and 0—7 in the MCP.

Table 3

Fixed-effect and cross-sectional OLS regressions of individual and household characteristics associated with the perceived stress scale, control groups only.

Characteristic or statistic CGP MCP

Cross-sectional OLS regression (1) Fixed-effect regression (2) Cross-sectional OLS regression (3)

Age 0.02 - 0.01

(1.58) - (0.64)

Education (attended school) 0.48 - -0.16

(1.74) - (-0.52)

Any death in the household 0.53 1.63** -0.61

(0.58) (2.91) (-0.84)

Household Food Insecurity Access Scale (0—27) 0.15** 0.12** 0.27***

(3.21) (2.77) (6.36)

Expenditure per capita 0.01 0.00 -0.00

(1.78) (1.04) (-0.78)

Number of non-productive assets owneda -0.01 -0.23 -0.08

(-0.05) (-1.37) (-0.42)

Constant 4.82*** 6.42*** 6.63***

(4.28) (9.18) (5.40)

Number of women 1139 1145 1227

R2 0.09 0.04 0.18

Observations 1139 2285 1227

Note. Robust t-statistics in parentheses based on standard errors clustered at the community level. Cross-sectional models control for marital status, district, household size and demographic composition of the household. *** *p < 0.05. **p < 0.01. ***p < 0.001.

research done in South Africa (Hamad et al., 2008). Several studies from high-income countries have shown that relative experience of deprivation can be a source of stress and cause ill health (Adler et al., 1994; Jennifer Beam Dowd et al., 2008; Schulz et al., 2012). The possibility that the program did affect perceived social status remains; however, our study did not measure this outcome. The measures of poverty included in this study are in more absolute terms, and thus do not reflect a person's subjective perception of their rank in society, which may be a larger comparative driver of stress.

A second possible explanation could be that the grant amount was not sufficient to affect stress levels. The cash may work its way along the pathway from poverty, which it did affect, but may not

have been enough to continue along the pathway to affect stress. Considering the effect sizes, it can be concluded that the treatment effect on food insecurity is large, with a 2.9-point reduction on the food insecurity scale for CGP beneficiaries. However, although the association between food security and perceived stress was statistically significant, the coefficient was small. In the CGP, a one-point decrease on the food insecurity scale resulted in a 0.12-point reduction on the PSS. Consequently, the hypothesized treatment effect on perceived stress through reduced food insecurity would result in a 0.35-point (2.9 x 0.12) or 0.08-SD reduction in perceived stress for CGP beneficiaries. The corresponding figure for the MCP would be a 0.8-point or 0.17 SD reduction in perceived stress, which is slightly higher but still not large enough to be

detected in this study sample. Beyond the size of the grant, other aspects of program implementation such as predictability of payment could also affect stress. Yet, in both the programs, payments were made regularly and on-time; therefore, this is unlikely to explain the lack of impacts on PSS.

A third theory is that the PSS10, which was first developed in a Western setting, does not adequately capture stress in a poor, rural setting in sub-Saharan Africa. It has been argued that the concept of "stress" may not be used and understood the same way in all cultural contexts (Pike and Patil, 2006) and that a context-specific stress scale based on local conditions and expectations may better capture the experience of stress (Ice et al., 2012). The possible inadequacy of the PSS to capture stress in this context could also explain why age and education, which are factors that are typically associated with levels of perceived stress in other settings (Cohen and Janicki-Deverts, 2012; Dowd et al., 2014; Remor, 2006), showed no consistent findings in this study's samples. Although there was no association between age and perceived stress in these samples, the relatively older women in the MCP households report more stress. This finding may be a function of stress increasing with age or with the targeting criteria for the MCP, which targets households in potentially demanding situations, such as caring for orphans and disabled household members. In contrast, CGP beneficiaries are targeted based on the sole criteria of having a child under the age of five in the household (in very poor geographic areas).

Nonetheless, our study did find that one measured stressful life event—a death in the family—was associated with higher perceived stress, which is confirmed by findings elsewhere (Cohen et al., 1983). Thus, the PSS10 may be more sensitive to stressful life events in this setting but less sensitive to chronic daily stress. Chronic daily stress is a key hypothesized mediator for poverty alleviation programs, and it is therefore important to measure in impact evaluations of cash transfer programs (Haushofer and Shapiro, 2016; Kling, 2007).

Finally, the divergence between findings on food insecurity and PSS may be explained by the fact that HFIAS includes items that capture both material outcomes (having enough to eat) and mental outcomes (worrying about food), and the material outcomes may drive the impact on the overall scale. To explore this hypothesis, this study estimated program impacts on each individual item in the scale separately and found significant impacts on each item (results not displayed). The key difference, of course, is that the mental items in the HFIAS relate specifically to food, whereas the mental items in the PSS are more general. This underscores the idea that food insecurity is the critical dimension of well-being in this population.

3.1. Limitations

There are some limitations to this study. Although the PSS is widely used and validated globally, it has never, to our knowledge, been validated in a sub-Saharan African setting. Further, psychometric problems mean that this study was not able to use all 10 items on the PSS10 scale; instead, this study used a subset of six questions, excluding the positively worded question items because they performed poorly in this setting.

The key outcome used in this paper was only collected at follow-up. Thus, causal inference relies on the assumption of baseline equivalence in the mean value of the stress indicator across study arms. This limitation is mitigated by the fact that in both studies used random assignment to each study arm, and baseline equivalence was established among a range of other indicators, including a key determinant of stress—household food security.

A final limitation concerns the potential generalizability of these

findings to other rural populations in sub-Saharan Africa. The two cash transfer programs targeted ultra-poor households with unique demographic profiles, including households with a pre-school child in the CGP, and households with a disabled member or an elderly or female head caring for orphans in the MCP. The combination of absolute poverty and demographic vulnerability may result in unique stressors which may not be relevant for other types of poor households in rural sub-Saharan Africa.

3.2. Future studies

The findings of this study can explain how different aspects of poverty may differentially affect mental health in low-income countries, with stress as a potential pathway between poverty and mental health. Future studies, particularly in sub-Saharan Africa and specifically Zambia, should investigate the concept of stress and how poverty relates to different measures of stress. The inconsistency in findings of the present study compared to those of previous studies indicates that the stress scale is inadequately measuring stress in this context. The PSS should be validated in a sub-Saharan African setting, and a more regionally appropriate scale should be developed.

Acknowledgements

The CGP and MCP impact evaluations were commissioned by the Government of Zambia (GRZ) through the Ministry of Community Development, Mother and Child Health to the American Institutes of Research (AIR) and the University of North Carolina at Chapel Hill (UNC) and funded by a consortium of donors including DFID, UNICEF, Irish Aid, and the Government of Finland. Palermo, Handa, and Hjelm received additional funding from the Swedish International Development Cooperation Agency (G41102) to the UNICEF Office of Research - Innocenti for analysis of the data and drafting of the manuscript.

The members of the CGP evaluation team, listed by affiliation and then alphabetically within affiliation are:

Principal Investigators: David Seidenfeld (AIR) and Sudhanshu Handa (UNC); AIR: Juan Bonilla, Rosa Castro Zarzur, Leah Prencipe, Dan Sherman, David Seidenfeld; UNICEF-Zambia: Charlotte Har-land Scott, Paul Quarles van Ufford; Government of Zambia: Vandras Luywa, Stanfield Michelo, Manzunzo Zulu; DFID-Zambia: Kelley Toole; Palm Associates: Alefa Banda, Chiluba Goma, Liseteli Ndiyoi, Gelson Tembo, NathanTembo; UNC: Sudhanshu Handa; UNICEF Office of Research — Innocenti: Sudhanshu Handa, Tia Palermo, Amber Peterman, Leah Prencipe.

The members of the MCP evaluation team, listed by affiliation and then alphabetically within affiliation are:

Principal Investigators: David Seidenfeld (AIR) and Sudhanshu Handa (UNC); AIR: Juan Bonilla, Alvaro Ballarin Cabrera, Thomas De Hoop, Gilbert Kiggundu, Nisha Rai, Hannah Reeves, Joshua Sennett, Dan Sherman, Jonathan Sokoll, Amy Todd, Rosa Castro Zarzur; Palm Associates: Alefa Banda, Liseteli Ndiyoi, Nathan Tembo; UNC: Sudhanshu Handa; UNICEF Office of Research - Innocenti: Tia Palermo, Amber Peterman, Leah Prencipe.

References

Addo, A.A., Marquis, G.S., Lartey, A.A., Perez-Escamilla, R., Mazur, R.E., Harding, K.B., 2011. Food insecurity and perceived stress but not HIV infection are independently associated with lower energy intakes among lactating Ghanaian women. Maternal Child Nutr. 7, 80—91. Adler, N.E., Boyce, T., Chesney, M.A., Cohen, S., Folkman, S., Kahn, R.L., et al., 1994.

SES and health: the challenge of the gradient. Am. Psychol. 49,15—23. Aiello, A.E., Dowd, J.B., 2013. Socio-economic status and immunosenescence. In: Jos A.Bosch, J.A., Anna C. Phillips, A.C., Lord, J.M. (Eds.), Immunosenescence. Springer, New York, pp. 145—157.

American Institutes for Research, 2011. Zambia's Child Grant Program: Baseline Report. American Institutes for Research, Washington, DC.

American Institutes for Research, 2012. Zambia's Multiple Category Cash Transfer Program: Baseline Report. American Institutes for Research, Washington, DC.

Andreou, E., Alexopoulos, E.C., Lionis, C., Varvogli, L., Gnardellis, C., Chrousos, G.P., et al., 2011. PSS: reliability and validity study in Greece. Int. J. Environ. Res. Public Health 8, 3,287—3,298.

Baird, S., De Hoop, J., Ozler, B., 2013. Income shocks and adolescent mental health. J. Hum. Resour. 48, 370—403.

Chaaya, M., Osman, H., Naassan, G., Mahfoud, Z., 2010. Validation of the Arabic version of the Cohen PSS (PSS-10) among pregnant and postpartum women. BMC Psychiatry 10, 111.

Chemin, M., De Laat, J., Haushofer, J., 2013. Negative Rainfall Shocks Increase Levels of the Stress Hormone Cortisol Among Poor Farmers in Kenya. Available at: SSRN 2294171. http://ssrn.com/abstract=2294171. http://dx.doi.org/10.2139/ ssrn.2294171.

Chemin, M., Haushofer, J., Jang, C., 2016. Health Insurance Reduces Stress: Evidence from a Randomized Experiment in Kenya. Available at: http://econ.as.nyu.edu/ docs/IO/42071/Chemin_Haushofer_Jang_Insurance_2016.03.10.pdf.

Coates, J., Swindale, A., Bilinsky, P., 2007. Household Food Insecurity Access Scale (HFIAS) for Measurement of Food Access: Indicator Guide. Food and Nutrition Technical Assistance Project, Academy for Educational Development, Washington, DC.

Cohen, S., 1988. Perceived stress in a probability sample of the United States. In: Spacapan, S., Oskamp, S. (Eds.), The Social Psychology of Health. Sage, Newbury Park, CA, pp. 31 —6 .

Cohen, S., Janicki-Deverts, D., Miller, G.E., 2007. Psychological stress and disease. JAMA 298,1,685—1,687.

Cohen, S., Janicki-Deverts, D., 2012. Who's stressed? Distributions of psychological stress in the United States in probability samples from 1983, 2006, and 2009. J. Appl. Soc. Psychol. 42, 1,320—1,334.

Cohen, S., Kamarck, T., Mermelstein, R., 1983. A global measure of perceived stress. J. Health Soc. Behav. 24, 385—396.

Cole, S.M., Tembo, G., 2011. The effect of food insecurity on mental health: panel evidence from rural Zambia. Soc. Sci. Med. 73,1,071—1,079.

Dole, N., Savitz, D.A., Hertz-Picciotto, I., Siega-Riz, A.M., McMahon, M.J., Buekens, P., 2003. Maternal stress and preterm birth. Am. J. Epidemiol. 157,14—24.

Dowd, J.B., Haan, M.N., Blythe, L., Moore, K., Aiello, A.E., 2008. Socioeconomic gradients in immune response to latent infection. Am. J. Epidemiol. 167,112—120.

Dowd, J.B., Palermo, T., Chyu, L., Adam, E., McDade, T.W., 2014. Race/ethnic and socioeconomic differences in stress and immune function in the National Longitudinal Study of Adolescent Health. Soc. Sci. Med. 115, 49—55.

Eisenbarth, C., 2012. Does self-esteem moderate the relations among perceived stress, coping, and depression? Coll. Stud. J. 46,149—157.

FAO, IFAD, WFP, 2015. The State of Food Insecurity in the World 2015. Meeting the 2015 International Hunger Targets: Taking Stock of Uneven Progress. FAO, Rome.

Fernald, L., Gunnar, M.R., 2009. Effects of a poverty-alleviation intervention on salivary cortisol in very low-income children. Soc. Sci. Med. 68, 2,180—2,189.

Fernald, L.C., Hamad, R., Karlan, D., Ozer, E.J., Zinman, J., 2008. Small individual loans and mental health: a randomized controlled trial among South African adults. BMC Public Health 8, 409.

Garcia, J., Hromi-Fiedler, A., Mazur, R.E., Marquis, G., Sellen, D., Lartey, A., et al., 2013. Persistent household food insecurity, HIV, and maternal stress in peri-urban Ghana. BMC Public Health 13, 215.

Hadley, C., Patil, C.L., 2008. Seasonal changes in household food insecurity and symptoms of anxiety and depression. Am. J. Phys. Anthropol. 135, 225—232.

Hadley, C., Stevenson, E.G.J., Tadesse, Y., Belachew, T., 2012. Rapidly rising food prices and the experience of food insecurity in urban Ethiopia: impacts on health and well-being. Soc. Sci. Med. 75, 2,412—2,419.

Hadley, C., Tegegn, A., Tessema, F., Cowan, J., Asefa, M., Galea, S., 2008. Food insecurity, stressful life events and symptoms of anxiety and depression in east Africa: evidence from the Gilgel Gibe growth and development study. J. Epidemiol. Community Health 62, 980—986.

Hamad, R., Fernald, L.C., Karlan, D.S., Zinman, J., 2008. Social and economic correlates of depressive symptoms and perceived stress in South African adults. J. Epidemiol. Community Health 62, 538—544.

Haushofer, J., Shapiro, J., 2016. The short-term impact of unconditional cash transfers to the poor: Experimental Evidence from Kenya. Q. J. Econ. 131, 1973—2042.

Hewitt, P.L., Flett, G.L., Mosher, S.W., 1992. The PSS: factor structure and relation to depression symptoms in a psychiatric sample. J. Psychopathol. Behav. Assess. 14, 247—25 .

Ice, G.H., Sadruddin, A.F., Vagedes, A., Yogo, J., Juma, E., 2012. Stress associated with caregiving: an examination of the stress process model among Kenyan Luo elders. Soc. Sci. Med. 74, 2,020—2,02 .

Johnson, J.G., Cohen, P., Dohrenwend, B.P., Link, B.G., Brook, J.S., 1999. A longitudinal investigation of social causation and social selection processes involved in the association between SES and psychiatric disorders. J. Abnorm. Psychol. 108,490.

Kilburn, K., Thirumurthy, H., Halpern, C.T., Pettifor, A., Handa, S., 2015. Effects of a

large-scale unconditional cash transfer program on mental health outcomes of young people in Kenya. J. Adolesc. Health 58, 223—229.

Kling, J.R., 2007. Methodological Frontiers of Public Finance Field Experiments. National Bureau of Economic Research, Cambridge, MA.

Knueppel, D., Demment, M., Kaiser, L., 2010. Validation of the household food insecurity access scale in rural Tanzania. Public Health Nutr. 13, 360—367.

Kristenson, M., Eriksen, H.R., Sluiter, J.K., Starke, D., Ursin, H., 2004. Psychobiological mechanisms of socioeconomic differences in health. Soc. Sci. Med. 58, 1,511—1,522.

Lazarus, R.S., Folkman, S., 1984. Stress, Appraisal, and Coping. Springer publishing company.

Lemma, S., Gelaye, B., Berhane, Y., Worku, A., Williams, M.A., 2012. Sleep quality and its psychological correlates among university students in Ethiopia: a cross-sectional study. BMC Psychiatry 12, 23 .

Lobel, M., Dunkel-Schetter, C., Scrimshaw, S.C., 1992. Prenatal maternal stress and prematurity: a prospective study of socioeconomically disadvantaged women. Health Psychol. 11, 32—40.

Lorant, V., Deliege, D., Eaton, W., Robert, A., Philippot, P., Ansseau, M., 2003. Socioeconomic inequalities in depression: a meta-analysis. Am. J. Epidemiol. 157, 98—112.

Lund, C., Breen, A., Flisher, A.J., Kakuma, R., Corrigall, J., Joska, J.A., et al., 2010. Poverty and common mental disorders in low and middle income countries: a systematic review. Soc. Sci. Med. 71, 517—528.

Lund, C., De Silva, M., Plagerson, S., Cooper, S., Chisholm, D., Das, J., et al., 2011. Poverty and mental disorders: breaking the cycle in low-income and middle-income countries. Lancet 378,1,502—1,514.

Maes, K.C., Hadley, C., Tesfaye, F., Shifferaw, S., 2010. Food insecurity and mental health: surprising trends among community health volunteers in Addis Ababa, Ethiopia during the 2008 food crisis. Soc. Sci. Med. 70, 1,450—1,457.

Maes, K.C., Hadley, C., Tesfaye, F., Shifferaw, S., Tesfaye, Y.A., 2009. Food insecurity among volunteer AIDS caregivers in Addis Ababa, Ethiopia was highly prevalent but buffered from the 2008 food crisis. J. Nutr. 139, 1,758—1,764.

Matthews, K.A., Gallo, L.C., Taylor, S.E., 2010. Are psychosocial factors mediators of SES and health connections? Ann. N. Y. Acad. Sci. 1186, 146—173.

McEwen, B.S., Seeman, T., 1999. Protective and damaging effects of mediators of stress: elaborating and testing the concepts of allostasis and allostatic load. Ann. N. Y. Acad. Sci. 896, 30—47.

Miller, G.E., Chen, E., Zhou, E.S., 2007. If it goes up, must it come down? Chronic stress and the hypothalamic-pituitary-adrenocortical axis in humans. Psychol. Bull. 133, 25—45.

Ozer, E.J., Fernald, L.C., Weber, A., Flynn, E.P., VanderWeele, T.J., 2011. Does alleviating poverty affect mothers' depressive symptoms? A quasi-experimental investigation of Mexico's Oportunidades program. Int. J. Epidemiol. 40, 1,565—1,576.

Palermo, T., Rawat, R., Weiser, S.D., Kadiyala, S., 2013. Food access and diet quality are associated with quality of life outcomes among HIV-infected individuals in Uganda. PLoS One 8, e62353.

Pike, I.L., Patil, C.L., 2006. Understanding women's burdens: preliminary findings on psychosocial health among Datoga and Iraqw women of northern Tanzania. Cult. Med. Psychiatry 30, 299—330.

Reis, R.S., Hino, A.A.F., Anez, C.R.R., 2010. PSS reliability and validity study in Brazil. J. health Psychol. 15,107—114.

Remor, E., 2006. Psychometric properties of a European Spanish version of the PSS (PSS). Span. J. Psychol. 9, 86—93.

Rondo, P., Rezende, G., Lemos, J., Pereira, J., 2013. Maternal stress and distress and child nutritional status. Eur. J. Clin. Nutr. 67, 348—352.

Schady, N., Paxson, C.H., 2007. Does Money Matter? the Effects of Cash Transfers on Child Health and Development in Rural Ecuador. World Bank Policy Research Working Paper No 4226.

Schulz, A.J., Mentz, G., Lachance, L., Johnson, J., Gaines, C., Israel, B.A., 2012. Associations between SES and allostatic load: effects of neighborhood poverty and tests of mediating pathways. Am. J. Public Health 102,1706—1714.

Torche, F., 2011. The effect of maternal stress on birth outcomes: exploiting a natural experiment. Demography 48,1,473—1,491.

Tsai, A.C., Tomlinson, M., Comulada, W.S., Rotheram-Borus, M.J., 2016. Food insufficiency, depression, and the modifying role of social support: evidence from a population-based, prospective cohort of pregnant women in peri-urban South Africa. Soc. Sci. Med. 151, 69—77.

van Eck, M., Nicolson, N.A., Berkhof, J., 1998. Effects of stressful daily events on mood states: relationship to global perceived stress. J. Personal. Soc. Psychol. 75, 1,572—1,585.

Vos, T., Barber, R.M., Bell, B., Bertozzi-Villa, A., Biryukov, S., Bolliger, I., et al., 2015. Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990—2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 386, 743—800.

Wang, Z., Chen, J., Boyd, J.E., Zhang, H., Jia, X., Qiu, J., et al., 2011. Psychometric properties of the Chinese version of the PSS in policewomen. PLoS One 6, e28610.