WPS6886 Policy Research Working Paper 6886 Cash Transfers and Temptation Goods A Review of Global Evidence David K. Evans Anna Popova The World Bank Africa Region Office of the Chief Economist May 2014 Policy Research Working Paper 6886 Abstract Cash transfers have been demonstrated to improve quantitative evidence on the impact of cash transfers on education and health outcomes and alleviate poverty temptation goods, as well as 11 studies that surveyed the in various contexts. However, policy makers and others number of respondents who reported they used transfers often express concern that poor households will use for temptation goods. Almost without exception, studies transfers to buy alcohol, tobacco, or other “temptation find either no significant impact or a significant negative goods.” The income effect of transfers will increase impact of transfers on temptation goods. In the only expenditures if alcohol and tobacco are normal goods, (two, non-experimental) studies with positive significant but this may be offset by other effects, including the impacts, the magnitude is small. This result is supported substitution effect, the effect of social messaging about by data from Latin America, Africa, and Asia. A growing the appropriate use of transfers, and the effect of shifting number of studies from a range of contexts therefore dynamics in intra-household bargaining. The net indicate that concerns about the use of cash transfers for effect is ambiguous. This paper reviews 19 studies with alcohol and tobacco consumption are unfounded. This paper is a product of the Office of the Chief Economist, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at devans2@worldbank.org and apopova@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Cash Transfers and Temptation Goods: A Review of Global Evidence David K. Evans Anna Popova World Bank World Bank Keywords: Cash Transfers, Temptation Goods, Impact Evaluation, Review, Consumption JEL Codes: E20, I38, O12 Acknowledgments: The authors are very grateful for comments and guidance from Javier Baez, Chris Blattman, Julio Elias, Francisco Ferreira, Mario Macis, and Norbert Schady. The findings, interpretations, and conclusions are entirely those of the authors. Corresponding author: David Evans (devans2@worldbank.org). Motivation Since the introduction of cash transfer programs, both conditional and unconditional, a major concern has been that households will misuse the cash. In Nicaragua, a senior government official expressed concern that “husbands were waiting for wives to return in order to take the money and spend it on alcohol” (Moore 2009). Interviews with stakeholders in Kenya revealed “widespread belief that cash transfers would either be abused or misdirected in alcohol consumption and other non-essential forms of consumption” (Ikiara 2009). A broader survey highlights that, “There is a widely held belief that cash given to poor people (especially to men) will be squandered on alcohol and other non-essentials” (Devereux 2002). Governments and aid agencies may worry that “men could control the cash provided and spend it on alcohol and cigarettes, rather than food for hungry children” (Harvey 2005). These concerns may explain why many countries prefer in-kind transfer programs, even though economic reasoning would suggest that cash transfers are more efficient (Case and Deaton 1998): Households can more easily meet their heterogeneous needs with liquid cash than with other, less liquid goods. Alcohol and tobacco have been referred to as “temptation goods” (Dasso and Fernandez 2013), a term used by Banerjee & Mullainathan (2010) to refer to “goods that generate positive utility for the self that consumes them, but not for any previous self that anticipates that they will be consumed in the future.” In an earlier literature, Musgrave (1959) used a related (but more normatively charged) term, “demerit goods,” to refer to goods that were so demeritorious (either to the consumer or to others) that the government may be correct in regulating their use. That term is sometimes used in reference to alcohol and tobacco in cash transfer studies. In this study, we use the term “temptation goods” principally to refer to alcohol and tobacco. 1 This study makes no normative assumption as to the value of alcohol and tobacco expenditures but merely seeks to systematically characterize the literature on the impact of cash transfers on these goods. Although alcohol and tobacco are the principal goods under consideration, some studies report other items as part of the same category, from doughnuts (Aker 2013) to soft drinks and Chinese food (Dasso and Fernandez 2013). The poor may wish to reduce spending on these items, as evidenced by a survey in Hyderabad, India, that asked households if they would like to eliminate any expenses in their budget: 28% of households identified at least one item, and the top items (44% of those) that households wanted to cut were alcohol and tobacco (Banerjee and Duflo 2007). Most cash transfer programs are not focused on either increasing or decreasing consumption of these goods specifically, and so most evaluations and the subsequent reviews have not been 1 Consumption of these goods may in some cases serve positive social purposes. For example, one study recounts the anecdote of demobilized soldiers returning home in Mozambique and using some of their demobilization grant on alcohol in the context of a village celebration to assist in their reintegration (Harvey 2005). 2 focused on these. Rather, reviews have focused on outcomes in schooling (Baird, et al. 2014, Saavedra and García 2013), health (Leroy, Ruel and Verhofstadt 2009, Ranganathan and Lagarde 2012), consumption (Fiszbein, et al. 2009), or a combination of these (IEG 2011). At the same time, many individual evaluations of cash transfer programs have included analysis of the impact on some set of temptation goods within their consumption analysis. Across 44 estimates from 19 studies, we find that almost without exception, studies find either no significant impact or a significant negative impact of transfers on expenditures on alcohol and tobacco. This finding is similar whether the analysis includes experimental and quasi- experimental designs or if it is restricted to randomized trials alone. Likewise, studies that have tried to quantify the proportion of beneficiaries who spend transfers on temptation goods find negligible effects. This result is consistent across the world, supported by data from Latin America, Africa, and Asia. It is also consistent across conditional and unconditional cash transfer programs. The evidence suggests that cash transfers are not used for alcohol and tobacco at any significant levels. Cash, Spending, and Consumption of Temptation Goods If alcohol and tobacco are normal goods, then as incomes rise, consumption of these goods will likewise rise (i.e., the income effect). Evidence from the United States suggests that alcohol is a normal good, whereas tobacco is an inferior good (Decker and Schwartz 2000); evidence from the United Kingdom suggests that alcohol expenditures rise with income, at least to a point (Banks, Blundell and Lewbel 1997). Banerjee and Duflo (2006; 2007) show expenditures for households living on under $1.08 per day and for households living on under $2.16 per day in 11 countries divided into urban and rural areas, resulting in 21 country-urban or country-rural combinations, due to missing data on urban Guatemala (Table 1). Of these 21 combinations, 14 increased or maintained the same percentage of spending on alcohol and tobacco combined when comparing $1.08 to $2.16 daily income, suggesting a likely increase in the total spending.2 That number rises to 20 out of 21 if one includes settings with minor decreases in the proportion, still consistent with increases in the total spending on these goods given the rise in income. These numbers suggest that alcohol and tobacco, when examining regular income, are normal goods. Beyond the income effect, there are at least three reasons that cash transfer income may affect spending on temptation goods differently from other income. First, conditional cash transfers in particular may induce a substitution effect, increasing the value of schooling and health 2 These percentages are best for distinguishing luxury goods (for which the proportion of spending increases with income) from necessity goods (for which the proportion falls), but they are suggestive of an increase in absolute spending (i.e., normal goods). 3 investments relative to all other goods, which may shift households away from consumption of temptation goods (Fiszbein, et al. 2009). The relative strength of the income and substitution effects will vary across households depending on their baseline schooling and health investments. Those beneficiary households that make sufficient investments in their children’s education to satisfy the conditions of the program, before the program, will be less affected by the substitution effect. On the other hand, those who – before the program – are investing in education and health at levels below those required by program conditions will be more affected by the substitution effect. Second, while few cash transfer programs have explicit spending restrictions, they often come with strong social messaging. For example, Ecuador’s unconditional cash transfer program (Bono de Desarrollo Humano) was accompanied by an advertising campaign encouraging households to invest in their children’s human capital (Schady and Rosero 2008). In Zimbabwe, recipients of a cash transfer program were “instructed not to ‘waste’ the cash on drinks and other unproductive items” (Román 2010). In Nicaragua, a task of the community coordinators for the program was “promoting the use of cash transfers to buy goods and services which improve the nutritional, educational and health status of beneficiary families” (Adato and Roopnaraine 2004). Program officers often communicate to households that these resources are intended to improve education or health outcomes. As a result, households may be more likely to use the resources for expenditures related to education and health than on tempation goods, a manifestation of what has been termed the flypaper effect (Inman 2008). Finally, transfer income is often targeted at women, particularly in Latin America (Fiszbein, et al. 2009). This design choice is driven by the long-held idea that women are more likely to invest in children than men. The actual evidence on this is mixed. On one hand, researchers found that higher proportions of household income controlled by women led to greater food expenditures in Côte d’Ivoire (Hoddinott and Haddad 1995), greater expenditures on food and children’s goods in Mexico (Bobonis 2009), and to improved child health in Brazil (Thomas 1990). In Macedonia, randomly assigning cash transfers to mothers (versus the household head) significantly increased education expenditures as well as secondary school enrollment and achievement, but only when parents’ perceived returns to education were high (Armand 2013). On the other hand, two cash transfer programs that randomized whether the transfer was given to the woman or the man found no sigificant differences in outcomes for children. The outcome in the first study was health clinic visits for children in Burkina Faso (Akresh, de Walque and Kazianga 2012) and in the second study it was school participation in Morocco (Benhassine, et al. 2013). If men are indeed more likely to purchase temptation goods (as was explicitly documented in Côte d’Ivoire), then providing transfer income to women could reduce spending on those 4 goods, a household bargaining effect. The net effect – between the income effect, the substitution effect, the flypaper effect, and the household bargaining effect – is unclear theoretically: This paper seeks to characterize it empirically. A large literature has examined the impact of cash transfers on consumption, with a few studies explicitly contrasting transfer income with earned income. For example, Schady and Rosero (2008) show that food expenditures were much higher for transfer recipients than non- recipients in the Ecuador program, even when controlling for per capita expenditures (i.e., the income effect of cash transfers). This finding is contrary to Engel’s law, which states that “the proportion of income spent on food declines as income rises” (Houthakker 1957) and which has been empirically identified across many countries. In Nicaragua, Macours, Schady and Vakis (2012) use a similar strategy and find that cash transfer recipients shifted the composition of food expenditures to more expensive foods (i.e., more protein, fruits, and vegetables; fewer staples), even though total food expenditures were not different from other households with similar per capital expenditures. Case & Deaton (1998) demonstrate that pension income in South Africa increased food consumption and may have reduced alcohol and tobacco consumption, depending on the specification. These studies suggest that households may indeed treat transfer income differently from earned income. Methodology In this section we describe the criteria used to define the universe of literature relevant to this systematic review, as well as the search strategy employed to find papers conforming to these criteria. Scope of the review First, we describe the scope of the review in terms of the types of interventions, studies, and outcome variables of primary interest. We restrict our analysis to conditional cash transfers (CCTs) and unconditional cash transfers (UCTs) implemented in low and middle-income countries (as defined by the World Bank), with no other explicit population exclusion criteria. Since both CCTs and UCTs generally target poor and vulnerable households (often including school-aged children or pregnant women), the entire set of eligible interventions is largely targeted at disadvantaged populations. The review focuses on studies from 1997 to early 2014, which corresponds to the period following the onset of PROGRESA/Oportunidades, allowing for a relatively comparable group of cash transfer interventions, as in Baird et al. (2014). Eligible studies include both experimental and quasi-experimental designs. We limit the review to papers that compare cash transfer 5 recipients to a group that receives no transfers. Specifically, in the systematic review we consider the effects on consumption of all those goods which studies themselves identify as “temptation”, “demerit”, or “anti-social” goods, or those which reflect “misuse” or “waste”. In conducting the review we focus on the effects on alcohol and tobacco consumption, for comparability purposes. Consumption of temptation goods is measured in a number of ways across studies: notably, expenditure, share of expenditure, and share of individuals consuming the temptation good in the reference period. We include studies using all of these measures, although we focus on expenditure as our primary outcome of interest. We classify this universe of eligible studies into the following three categories: i. Impact Estimates: Randomized-control trials or quasi-experimental studies that estimate the impact of cash transfers on the consumption of temptation goods; ii. Level Estimates: Studies that use surveys or focus groups to characterize the number of beneficiaries or amount of transfers used to purchase temptation goods; and iii. Qualitative Reports: Studies that discuss reports of the use of transfers to purchase temptation goods, not necessarily by the interviewed household. Search methods for the identification of studies The remainder of this section describes how the literature was searched. The various phases of the search process are also summarized in chronological order in Table 2, together with the number of results they yielded. We restricted all searches to papers published since 1997. Our primary electronic search was conducted using Google Scholar. 3 The initial search was completed on January 20th, 2014; thus papers that were not yet available at that time are not included in this review. We searched for papers that included both the term "cash transfers", and any one of the terms “alcohol”, “tobacco”, “cigarettes”, "temptation goods", or "demerit goods". This search yielded a total of 4,290 articles. The titles and sources of these papers were reviewed and the majority of papers discarded due to irrelevant subject areas, leaving 434 papers. These were then checked for duplicates and 23 were removed, leaving 411 papers. These 411 papers were reviewed by reading their abstracts and conducting a word search within each article for appearances of the key search terms listed above, so as to identify the context within which they are referenced by the article. 179 papers were deemed irrelevant and removed on the basis of this review process – typically because the terms of interest 3 The search produced results drawn from databases including Science Direct, the Social Science Research Network, and the Wiley Online Library, as well as the databases of a number of international organizations – notably the Overseas Development Institute, United Nations Development Programme, the United Nations Children's Fund, the World Bank, the International Initiative for Impact Evaluation (3ie), the World Food Program – and universities. 6 appeared either in passing, or in a context other than that of cash transfers – leaving 232 papers. In addition to the Google Scholar search, we added 7 studies from our own knowledge as well as those recommended by other researchers. We investigated the bibliographies of relevant papers and systematic reviews of cash transfers uncovered through the previous steps and manually searched a number of databases to find relevant papers mentioned, yielding another 12 papers, bringing the total to 251 papers. 4 These papers were then examined more closely, and studies were removed that did not fall into any of the three categories described earlier (impact estimates, level estimates, and qualitative reports). This left 42 papers. The countries represented by these papers by impact estimates and level estimates, from all over the world, are illustrated in Figure 1. Several of these papers report multiple estimates: For example, estimates are reported using two different estimators (e.g., treatment-on-the-treated versus intent-to-treat, or matching versus instrumental variables) or in multiple time periods (e.g., the first tranche of transfers versus the second tranche). In order to capture the full range of possible impacts, we include all these estimates. We do not include estimates on sub-populations (e.g., female-headed households only) except when that is the only format in which results are reported in the original studies, because papers are very inconsistent in the sub-populations for which they report outcomes, and because relatively few do so. Results First, we discuss the evidence from estimates of program impacts on spending on alcohol and tobacco. Second, we discuss evidence from studies that surveyed the number of respondents who reported using transfers to purchase temptation goods. Impact Estimates Nineteen studies from 10 countries around the world (in Latin America, Africa, and Asia) report impacts of cash transfers on the level or proportion of expenditures on alcohol or tobacco, or the probability of consumption or abuse of these goods. These studies and the reported impacts are listed in Tables 3 and 4. The 19 studies include 44 impact estimates. To simplify, we group these estimates into four categories: 4 Of the 19 papers which were added manually, 4 contained relevant impact estimates and made it to our final sample of literature. Two of these (Dasso and Fernandez 2013 and Evans et al. 2014) were unpublished mimeos at the time of the original search and were therefore not picked up by our algorithm. Another study (Gilligan and Roy 2013) was not picked up because it does not include any of our search terms; specifically, it refers to estimates of consumption of “beer” only. The other study (The Kenya CT-OVC Evaluation Team 2010) was an in-depth internal evaluation report underlying a policy note identified using our algorithm. The remaining 15 papers in this group - a number of which were systematic reviews of cash transfers and social policies - contained relevant background material on cash transfers but no evidence on their effect on the consumption of temptation goods. 7 (1) negative and significant, (2) negative or zero and insignificant, (3) positive and insignificant, and (4) positive and significant. Across all 44 estimates (from the 19 studies), there are 12 estimates which are negative and significant, 24 which are negative and insignificant, 6 which are positive and insignificant, and 2 which are positive and significant (Table 5 Panel A). In other words, 82% of estimates are negative, and just 5% of estimates are significant and positive (Table 6 and Figure 2). One of those two positive results is an unconditional cash transfer program in Indonesia: in the first disbursement, the impact was slightly negative and highly significant, whereas in the second disbursement, the impact was slightly positive and mildly significant. The size of the coefficient is almost identical to that for expenditures on prepared food. The other positive result, from Peru’s Juntos program, is from a paper that uses two different methods, matching and instrumental variables, and finds opposite results from the two estimates on alcohol consumption: a moderately significant negative impact from the matching estimate and a weakly significant positive impact from the instrumental variables estimate. Estimates on other outcomes are mostly consistent across the two estimation methods. Thus, in both cases of positive significant results, the impacts are weakly significant and are not consistent across estimates within the same study. Furthermore, the effect sizes are very small: one is less than a penny, whereas the other is 21 cents. Even if those estimates accurately reflect changes in expenditures, the changes are trivial. If we instead consider only the 17 estimates from 8 randomized-control trials, we find 1 estimate which is negative and significant, 13 that are negative and insignificant, 3 that are positive and insignificant, and zero that are positive and significant (Table 5 Panel C). In other words, 82% are negative, and none are positive and significant (Table 6). Four studies explore the impacts on alternate measures of temptation goods. Two of them, rather than estimating the impact of transfers on expenditure levels or proportions, estimate the impact of the program on the share of individuals who consume any of the temptation good (i.e., who smoke or who drink). They both examine this impact for adolescents in Mexico, in the context of the Oportunidades transfer program. One paper examines the impact using 2004 data and an instrumental variables strategy. It separates the impact of program participation from the impact of total cumulative transfers, with community awareness of the program as an instrument for household participation in the program, and potential transfers (based on household demographics) as an instrument for actual transfers (Galárraga and Gertler 2009). The net effect is reported for smoking for men (a reduction from 30% of men smoking to 16%) and for alcohol consumption for women (a reduction from 22% of women 8 drinking to 13%). The other paper, using data from 1998 and 2000 and a propensity score matching approach, finds a reduction in adolescents that have had alcoholic drinks (12% reduction in rural areas, 2% in urban areas) and in those that have smoked (14% in rural areas, 4% in urban areas) (Gutiérrez, et al. 2004). A third study, from Uganda, uses a randomized design to estimate the impact of cash transfers provided through early childhood development centers on the number of days that children (aged 1 to 7 years) drank beer in the week preceding the survey (Gilligan and Roy 2013) 5. They find an insignificant reduction of 20% in the frequency of consumption. Finally, a randomized evaluation estimates the impact of Oportunidades on the probability of alcohol abuse, and finds that beneficiaries of the program are significantly less likely to have a habitual drinker present in the household than non-beneficiaries (Angelucci 2008). These studies add to the evidence that the net effect of cash is likely to be either insignificant or negative. The distribution of results is very similar if one includes only estimates on expenditure levels and not on the proportion of income spent or these other outcomes (Table 5 and Figure 2). There are several potential sources of heterogeneity in the impact of cash transfers on temptation goods, including program design (e.g., conditional versus unconditional cash transfers), geographic variation, or variation in how long households have been receiving the transfers. However, with so few significant effects, it is difficult to identify heterogeneity. We have 31 estimates from conditional cash transfer programs and 11 estimates from unconditional cash transfer programs.6 Table 7 shows the distribution of estimates (from negative and significant to positive and significant) for the two groups; we observe essentially the same pattern, with 84% and 73% of estimates being negative for conditional cash transfer programs and unconditional cash transfer programs, respectively. The proportions are almost identical when we consider estimates from randomized trials only. Likewise, if we separate the studies in Latin America from those in Asia and in Africa, there are no clear differences, albeit with slightly greater heterogeneity in Latin America (Table 8). We find that 79% of estimates in Latin America are negative, compared to 88% of estimates from other regions. There is also no evidence of differences related to the length of time people have been receiving transfers. The time during which beneficiaries have been receiving transfers is – on average – 1.9 years, and ranges from 6 months to 5.5 years, with no clear relationship 5 In the Karamoja sub‐region of Uganda, where this program was implemented, it is common to make a local homebrewed weak beer from sorghum and for both adults and young children to consume this, as well as the beer residue (Gilligan and Roy 2013). 6 For this comparison of conditional and unconditional transfer programs, we exclude the 2 estimates from the evaluation of The Kenya Cash Transfer Program for Orphans and Vulnerable Children, which imposed conditions in 3 of the 7 districts in which it was implemented. 9 between the duration of treatment and the program effects on consumption of alcohol and tobacco. Level Estimates Other studies, while not estimating the impact of transfers on consumption of or expenditures on temptation goods, have sought to quantify how many beneficiaries use transfers for temptation goods, or how much of the transfers has been spent on temptation goods; these studies rely either on surveys or focus groups. We identified 11 such studies, representing programs in 8 countries: 6 in Africa, 1 in Asia, and 1 in the Middle East (Table 9). Four of the studies identified a proportion of beneficiaries or households that spent some or all of the transfer on temptation goods. The median proportion was 1.2%, a tiny fraction of households. Even in the one outlier, Lesotho’s Cash & Food Transfers Pilot Project, where about 6% of beneficiaries admitted to spending some of their transfer on alcohol and cigarettes, the study quotes a recipient as saying that it happens “only in rare and discreet cases.” Two more studies, from Malawi and Zimbabwe, identify the proportion of transfers spent on temptation goods: In both cases, the proportion is under 0.5%. The remaining studies simply report that they found no evidence that households were purchasing temptation goods, except one case that reports a “marginal increase.” This evidence is significantly less convincing than the impact estimates, which look at total expenditures rather than transfer expenditures alone, as transfer and other income are fungible. A household could, for example, use the transfer income entirely for education investments but at the same time decrease spending on education from regular income by ten percent. Then they could use that ten percent of regular income for temptation goods. In the respondent’s view, none of the transfer income would have been used for temptation goods, although clearly the transfer is what enabled the increased expenditures. Despite this caveat, these level estimates are consistent with the finding of insignificant quantities being spent on alcohol and tobacco that was already observed in the more reliable estimates on overall expenditures. Discussion In this section we discuss some of the implications and challenges related to this analysis. One principal concern when studying the consumption of goods such as alcohol and tobacco, especially in the context of a program where beneficiaries are encouraged not to use the resources on those goods, is that beneficiaries will report low expenditures on those goods 10 because they want to minimize the risk of expulsion from the program or other potential negative consequences. This is known as “social desirability bias.” There is some evidence from undergraduate students in the United States that self-reports of alcohol consumption can be biased downward (David, Thake and Vilhena 2010). In developing contexts, this is much less explored for alcohol and tobacco consumption. (For sexual behaviors, it has been explored extensively. 7) However, we do not expect this to be a major problem here, for the following reasons. First, the impact estimates presented here are usually based on detailed expenditure surveys that ask a household respondent how much the household spends on each of a long list of items. Alcohol and tobacco are not singled out. For the estimates of what proportion of households spent any resources on temptation goods, alcohol and tobacco may be singled out, which could explain why several studies found zero reports of any spending on temptation goods. However, those estimates merely provide supportive evidence to the more robust impact estimates. Second, transfer income is not asked about separately, so households would have to recall the amount of their overall income spent on temptation goods before the program and report a similar amount later. The simplest solution for households seeking to appease an interviewer would be to report zero or extremely low expenditures on alcohol and tobacco. This is especially true since household surveys are administered infrequently and so recalling previous reports may be difficult. In that case, we would expect to see a much starker pattern of significant negative impacts. On the contrary, we observe just 24% of all impacts on expenditures to be negative and significant, and 9% for randomized-control trials. The far more common result is an insignificant difference: the outcome in all eleven randomized trials (Figure 2 Panel D). This does not look like systematic social desirability bias. An additional concern could be that these studies were not sufficiently statistically powered to capture consumption impacts at all, whether on temptation goods or other categories of consumables. For this, we focus on the 6 positive and insignificant estimates in more detail. These 6 estimates come from 5 studies (each from different countries around the world), most of which report the estimated impact of cash transfers for total expenditure on temptation goods; Maluccio and Flores (2005) also present an estimate of the impact on the proportion of expenditures. For each of these studies, we examine whether the studies had sufficient statistical power to identify significant impacts on overall consumption using the same 7 This issue has been studied more extensively for sexual behavior in developing countries, and the evidence has been inconsistent: In Malawi and Kenya, for example, young women were more likely to report ever having had sex in a face-to-face interview, whereas they were likely to report more total partners in an audio computer- assisted self-interview (Mensch, et al. 2008). In Zimbabwe, respondents also reported fewer partners in face-to- face interviews (Gregson, et al. 2002). A study in Tanzania found female adolescents were more honest about sexual infection in face-to-face interviews, whereas males were less honest (Plummer, et al. 2004). A fuller list of relevant references is available in Handa et al. (2014). 11 estimation methodology (Table 10). We observe that in every case, the studies finding positive and insignificant estimates for temptation goods at the same time produce significant (positive) estimates for the impact on overall consumption. Because identifying impacts on individual consumption items or categories requires greater statistical power than identifying effects on total consumption, we also look at whether these studies find significant impacts on individual consumption items other than temptation goods (also in Table 10). We find that every study finding positive and insignificant estimates for temptation goods produces significant estimates for at least 20 percent of the disaggregated consumption items. This suggests that the insignificance of these temptation good estimates does not derive from a lack of statistical power. Rather, there is simply no quantitative evidence that beneficiaries use their transfers on alcohol and tobacco. An alternative comparison to the one central to this study is the relative impact of cash transfers versus in-kind food transfers. In the course of our search, we identified three studies – all randomized trials – that quantitatively estimate this relative impact: one in Mexico (Cunha 2012), one in Uganda (Gilligan and Roy 2013), and one in Yemen (Schwab, Margolies and Hoddinott 2013). The Mexico study finds (insignificantly) increased expenditures on alcohol among cash recipient households relative to food recipient households. The Uganda study, focusing on children age 1-7, finds that children of cash transfer recipients are (insignificantly) less likely to consume sorghum beer. The Yemen study estimates (insignificantly) higher expenditures on tobacco and khat for cash transfer recipients. In none of the cases is there a significant difference between the impact of cash and in-kind transfers on expenditures on temptation goods. The first two studies also include a pure comparison group and are included in the analysis above (and in Tables 3 and 4). Qualitative Results While the impact estimates suggest zero average effect, and the level estimates suggest only tiny fractions of beneficiaries using transfer resources to purchase temptation goods, qualitative reports sometimes tell a different story. Consider the following examples: • In Malawi, researchers reported from focus groups that “In our village, there were certain men who wasted their money even though they had families and children” and “We heard of four men who received their rations on a Thursday. They all went to a nearby popular drinking bar” (Devereux, Mvula and Solomon 2006). • In Bolivia, “Of the 35 subjects interviewed, 20 admitted they knew people who misspent the cash transfers.” However, “Many mentioned the media as their main source of information regarding any misspending” (Vaughan 2010). 12 • In Kenya, “Cases of misuse of funds were reported in the two sites: according to key informants, in some cases, male recipients have used some of the cash to buy alcohol, although this is relatively rare (only three cases reported, with the majority of the cash being used for consumption and investment)” (Onyango-Ouma and Samuel 2012). • In Swaziland, a focus group participant reported that “Men don’t return home on pay- days; some have found other women to spend the money with” (Devereux and Jere, 2008). • In Uganda, participants and informants observed that “Some beneficiaries – especially men – have used the cash transfer in over-drinking alcohol” and “Some older men especially drink all the money” (Bukuluki and Watson 2012). How do we reconcile these anecdotes with the extremely insignificant or even negative effects we observed earlier? First, the results previously discussed do not indicate that no single beneficiary uses his or her transfer on alcohol. For example, the Malawi anecdote above comes from a study that measured the proportion of transfers that were spent on alcohol; the proportion was 0.1%. So although interviewees had “heard of four men” or knew “certain men”, these numbers seem very small. What the quantitative results earlier claim is that, on average, there is no positive impact of transfers on alcohol expenditures. Second, most of these reports are not with reference to one’s own household, but rather to other individuals who respondents may know who spend the money on alcohol and tobacco. However, multiple respondents may well know the same person in the community who has a reputation for high levels of alcohol or tobacco consumption. These anecdotes can be subject to “saliency bias”, in which individuals pay attention to highly noticeable factors and dramatic events: A village drunkard stands out and is likely to come up disproportionately in discussions. An alternative possibility is that the respondents in household surveys are unaware of how their household resources are spent. For example, if a husband takes household resources and spends them on alcohol without the wife’s knowledge and the wife is the survey respondent, then such spending might show up in qualitative reports from other households but be missing in the impact estimates. However, it seems unlikely both that (1) the surveys consistently interview the non-drinking member of the household, and (2) this member is consistently ignorant of these expenditures, particularly in low-income households with limited liquid income. These results underline the importance of complementing qualitative reports with quantitative data and are reconcilable with the earlier quantitative finding that, on average, there is no increase in the consumption of temptation goods. 13 Conclusion We have investigated evidence from around the developing world, including Latin America, Africa, and Asia. There is clear evidence that transfers are not consistently used for alcohol or tobacco in any of these environments. This is particularly true when relying on the randomized trials. For all studies, the only evidence for a positive, significant effect is inconsistent across estimates within the studies themselves; and in those cases, the size of the impact is trivial. Thus, it seems that the flypaper effect and the effect of women controlling more resources (the household bargaining effect) likely compensate for the income effect, leading to no significant net change in alcohol and tobacco consumption. We see no difference between conditional and unconditional cash transfer programs, so this does not seem to be a function of conditions. We also observe no difference depending on the region of the transfer program. These results provide strong evidence that concerns that transfers will be used on alcohol and tobacco are unfounded. We do have estimates from Peru that beneficiaries are more likely to purchase a roasted chicken at a restaurant or some chocolates soon after receiving their transfer (Dasso and Fernandez 2013), but hopefully even the most puritanical policymaker would not begrudge the poor a piece of chocolate. 14 Works Cited Adato, Michelle, and Terry Roopnaraine. 2004. “A Social Analysis of Red de Protección Social in Nicaragua”. Washington, DC. Aheeyar, M.M.M. 2006. “Cash Delivery Mechanisms in Tsunami-Affected Districts of Sri Lanka”. London. Aker, Jenny. 2013. “Cash or Coupons? Testing the Impacts of Cash versus Vouchers in the Democratic Republic of Congo.” Akresh, Richard, Damien de Walque, and Harounan Kazianga. 2012. “Alternative Cash Transfer Delivery Mechanisms: Impacts on Routine Preventative Health Clinic Visits in Burkina Faso”. Washington DC. Angelucci, Manuela. 2008. “Love on the Rocks: Domestic Violence and Alcohol Abuse in Rural Mexico.” The B.E. Journal of Economic Analysis & Policy 8 (1). Armand, Alex. 2013. “Who Wears the Trousers in the Family? Intra-Household Resource Control, Subjective Expectations and Human Capital Investment”. University College London. Attanasio, Orazio, Erich Battistin, Emla Fitzsimons, Alice Mesnard, and Marcos Vera-Hernández. 2005. “How Effective Are Conditional Cash Transfers? Evidence from Colombia.” Attanasio, Orazio, and Luis Carlos Gómez. 2004. “Evaluación Del Impacto Del Programa Familias En Acción - Subsidios Condicionados de La Red de Apoyo Social. Informe Del Primer Seguimiento (Ajustado)”. Bogotá. Attanasio, Orazio, and Alice Mesnard. 2005. “The Impact of a Conditional Cash Transfer Programme on Consumption in Colombia”. The Institute of Fiscal Studies. Bailey, Sarah. 2009. “An Independent Evaluation of Concern Worldwide’s Emergency Response in North Kivu, Democratic Republic of the Congo”. London. Baird, Sarah, Francisco H.G. Ferreira, Berk Özler, and Michael Woolcock. 2014. “Relative Effectiveness of Conditional and Unconditional Cash Transfers for Schooling Outcomes in Developing Countries: A Systematic Review.” Journal of Development Effectiveness (September). doi:10.4073/csr.2013.8. Banerjee, Abhijit, and Esther Duflo. 2006. “The Economic Lives of the Poor.” ———. 2007. “The Economic Lives of the Poor.” Journal of Economic Perspectives 21 (1) (January): 141– 167. doi:10.1257/jep.21.1.141. Banerjee, Abhijit, and Sendhil Mullainathan. 2010. “The Shape of Temptation: Implications for the Economic Lives of the Poor”. Cambridge, MA. Banks, James, Richard Blundell, and Arthur Lewbel. 1997. “Quadratic Engel Curves and Consumer Demand.” Review of Economics and Statistics LXXIX (4): 527–539. 15 Bazzi, Samuel, Sudarno Sumarto, and Asep Suryahadi. 2012. “Evaluating Indonesia’s Unconditional Cash Transfer Programme, 2005‒2006.” Benhassine, Najy, Florencia Devoto, Esther Duflo, Pascaline Dupas, and Victor Pouliquen. 2013. “Turning a Shove into a Nudge? A "Labeled Cash Transfer”. Cambridge, MA. Bhowmik, Sharit K., Indira Gartenberg, and Kanchan Sarker. 2013. “Food or Cash: Assisting the Urban Poor in India.” In The Food Crisis: Implications for Labor, edited by Cristoph Scherrer and Debdulal Saha. ICDD. Biron, Aurélie. 2012. “Adapting to Urban Displacement: The Use of Cash Transfers in Uraban Areas”. Sciences Po Grenoble. Bobonis, Gustavo J. 2009. “Is the Allocation of Resources within the Household Efficient? New Evidence from a Randomized Experiment.” Journal of Political Economy (November): 453–503. Braido, Luis H. B., Pedro Olinto, and Helena Perrone. 2012. “Gender Bias in Intrahousehold Allocation: Evidence from an Unintentional Experiment.” Review of Economics and Statistics. Brewin, Mike. 2008. “Evaluation of Concern Kenya’s Kerio Valley Cash Transfer Pilot (KVCTP), April - June 2008.” Bukuluki, Paul, and Carol Watson. 2012. “Transforming Cash Transfers: Beneficiary and Community Perspectives on the Senior Citizen Grant (SCG) in Uganda”. London, UK, and Kampala, Uganda. Case, Anne, and Angus Deaton. 1998. “Large Cash Transfers to the Elderly in South Africa.” The Economic Journal: 1330–1361. Creti, Pantaleo, and Susanne Jaspars. 2006. Cash-Transfer Programming in Emergencies. Oxford: Oxfam GB. Cunha, Jesse M. 2012. “Testing Paternalism: Cash versus In-Kind Transfers.” Unpublished Mimeo. Naval Postgraduate School. Dasso, Rosamaría, and Fernandez Fernando. 2013. “Temptation Goods and Conditional Cash Transfers in Peru”. Washington, DC. Davis, Christopher G., Jennifer Thake, and Natalie Vilhena. 2010. “Social Desirability Biases in Self- Reported Alcohol Consumption and Harms.” Addictive Behaviors: 302–311. Decker, Sandra L., and Amy Ellen Schwartz. 2000. “Cigarettes and Alcohol: Substitutes or Complements?” Devereux, S, and M Mhlanga. 2008. “Cash Transfers in Lesotho: An Evaluation of World Vision’s Cash and Food Transfers Pilot Project”. Brighton, UK, and Maseru, Lesotho. 16 Devereux, Stephen. 2002. “Social Protection for the Poor: Lessons from Recent International Experience”. Brighton, UK. Devereux, Stephen, and Paul Jere. 2008a. “‘Choice, Dignity and Empowerment?’Cash and Food Transfers in Swaziland: An Evaluation of Save the Children’s Emergency Drought Response, 2007/08”. Brighton, UK, and Lilongwe, Malawi. ———. 2008b. “Cash and Food Transfers in Swaziland: An Evaluation of Save the Children’s Emergency Drought Response, 2007/08”. Brighton, UK, and Lilongwe, Malawi. Devereux, Stephen, Catherine Mthinda, Fergus Power, Patrick Sakala, and Abigail Suka. 2007. “An Evaluation of Concern Worldwide’s Dowa Emergency Cash Transfer Project (DECT) in Malawi, 2006/07”. Brighton, UK, and Lilongwe, Malawi. Devereux, Stephen, Peter Mvula, and Colette Solomon. 2006. “After the FACT : An Evaluation of Concern Worldwide ’ S Food and Cash Transfers Project in Three Districts of Malawi , 2006”. Brighton, UK, and Zomba, Malawi. Edmonds, Eric. 2002. “Reconsidering the Labeling Effect for Child Benefits: Evidence from a Transition Economy.” Economic Letters: 303–309. European Commission. 2013. “The Use of Cash and Vouchers in Humanitarian Crises.” Evans, David K., Stephanie Hausladen, Katrina Kosec, and Natasha Reese. 2014. Community-Based Conditional Cash Transfers in Tanzania: Results from a Randomized Trial. Washington, DC: World Bank: © World Bank. Farrington, John, and Rachel Slater. 2009. “Lump Sum Cash Transfers in Developmental and Post- Emergency Contexts: How Well Have They Performed?” London. Fiszbein, Ariel, and Norbert R. Schady. 2009. Conditional Cash Transfers: Reducing Present and Future Poverty. Washington, DC: The World Bank. doi:10.1596/978-0-8213-7352-1. Galárraga, Omar, and Paul J. Gertler. 2009. “Conditional Cash & Adolescent Risk Behaviors: Evidence from Urban Mexico.” Gangopadhyay, Shubhashis, Robert Lensink, and Bhupesh Yadav. 2013. “Cash or Food Security Through the Public Distribution System? Evidence from a Randomized Controlled Trial in Delhi, India.” Evidence from a …. Gilligan, Daniel O., and Shalini Roy. 2013. “Resources, Stimulation ,and Cognition: How Transfer Programs and Preschool Shape Cognitive Development in Uganda”. Washington, DC. Gitter, SR. 2006. “Women and Targeted Cash Transfers in Nicaragua.” Aae.wisc.edu. 17 Gregson, Simon, Tom Zhuwau, Joshua Ndlovu, and Constance A Nyamukapa. 2002. “Methods to Reduce Social Desirability Bias in Sex Surveys in Low-Development Settings: Experience in Zimbabwe.” Sexually Transmitted Diseases: 568–575. Gutiérrez, Juan Pablo, Sergio Bautista, Paul Gertler, Mauricio Hernández, and Stefano M. Bertozzi. 2004. “External Evaluation of the Impact of the Human Development Program Oportunidades:” Mexico City. Handa, Sudhanshu, Carolyn Tucker Halpern, Audrey Pettifor, and Harsha Thirumurthy. 2014. “The Government of Kenya’s Cash Transfer Program Reduces the Risk of Sexual Debut among Young People Age 15-25.” PloS One 9 (1) (January). doi:10.1371/journal.pone.0085473. Harvey, Paul. 2007. “Cash-Based Responses in Emergencies”. London. Harvey, Paul, and Kevin Savage. 2006. “No Small Change. Oxfam GB Malawi and Zambia Emergency Cash Transfer Projects: A Synthesis of Key Learning”. London. Haushofer, Johannes, and Jeremy Shapiro. 2013. “Household Response to Income Changes: Evidence from an Unconditional Cash Transfer Program in Kenya”. Massachusetts Institute of Technology. Hoddinott, John, and Lawrence Haddad. 1995. “Does Female Income Share Influence Household Expenditures - Evidence from Cote-Divoire.” Oxford Bulletin of Economics and Statistics 57: 77–96. doi:10.1111/j.1468-0084.1995.tb00028.x. Houthakker, HS. 1957. “An International Comparison of Household Expenditure Patterns, Commemorating the Centenary of Engel’s Law.” Econometrica 25 (4): 532–551. Humphreys, Rowena. 2008. “Evaluation of the Cash Transfers for Development Project in Vietnam”. London. IEG. Evidence and Lessons Learned from Impact Evaluations on Social Safety Nets. Washington, DC: World Bank. Ikiara, Gerrishon K. 2009. “The Political Economy of Cash Transfers in Kenya”. London. Inman, Robert P. 2008. “The Flypaper Effect”. Cambridge, MA. Jones, Nicola, Rosana Vargas, and Eliana Villar. 2007. “Conditional Cash Transfers in Peru: Tackling the Multi-Dimensionality of Childhood Poverty and Vulnerability. Social Protection Initiatives for Families, Women and.” In Forthcoming Chapter in Alberto Minujin et Al. (ed.) 2007. Social Protection Initiatives for Families, Women and Children: An Analysis of Recent Experiences. New York: New School and UNICEF. Khera, Reetika. 2013. “Cash vs In-Kind Transfers: Indian Data Meets Theory.” 18 Leroy, Jef L., Marie Ruel, and Ellen Verhofstadt. 2009. “The Impact of Conditional Cash Transfer Programmes on Child Nutrition: A Review of Evidence Using a Programme Theory Framework.” Journal of Development Effectiveness 1 (2) (June 11): 103–129. doi:10.1080/19439340902924043. Macours, Karen, Norbert Schady, and Renos Vakis. 2012. “Cash Transfers, Behavioral Changes, and the Cognitive Development of Young Children: Evidence from a Randomized Experiment.” American Economic Journal: Applied Economics: 247–273. Maluccio, John A., and Rafael Flores. 2005. “Impact Evaluation of a Conditional Cash Transfer Program”. Washington, DC. Mensch, Barbara S., Paul C. Hewett, Richard Gregory, and Stephane Helleringer. 2008. “Sexual Behavior and STI/HIV Status among Adolescents in Rural Malawi: An Evaluation of the Effect of Interview Mode on Reporting.” Studies in Family Planning 39 (4) (December): 321–34. Miller, Candace, Maxton Tsoka, and Kathryn Reichert. 2008. “Impact Evaluation Report External Evaluation of the Mchinji Social Cash Transfer Pilot”. Boston, USA, and Zomba, Malawi. Mohiddin, Lili, Manohar Sharma, and Anette Haller. 2007. “Comparing Cash and Food Transfers: Findings from a Pilot Project in Sri Lanka”. London, UK, Washington DC, USA, and Rome, Italy. Moore, Charity. 2009. “Nicaragua’s Red de Protección Social: An Exemplary but Short-Lived Conditional Cash Transfer Programme.” Musgrave, Richard. 1959. The Theory of Public Finance: A Study in Public Economy. McGraw-Hill. Onyango-Ouma, W., and Fiona Samuels. 2012. “Transforming Cash Transfers: Beneficiary and Community Perspectives on the Cash Transfer for Orphans and Vulnerable Children Programme in Kenya”. London, UK, and Nairobi, Kenya. Perova, Elizaveta. 2011. “Three Essays on Intended and Not Intended Impacts of Conditional Cash Transfers”. University of California, Berkeley. Phiri, Swedi. 2012. “A Comparative Assessment of the Impact of Unconditional Cash Transfers for Urban Vulnerable Households Headed by Elderly and Non-Elderly Women in Mucheke Ward”. University of Zimbabwe. Plummer, M. L., D. A Ross, D. Wight, J. Changalucha, G. Mshana, J. Wamoyi, J. Todd, et al. 2004. “‘A Bit More Truthful’: The Validity of Adolescent Sexual Behaviour Data Collected in Rural Northern Tanzania Using Five Methods.” Sexually Transmitted Infections 80 (II) (December): ii49–56. doi:10.1136/sti.2004.011924. Ranganathan, Meghna, and Mylene Lagarde. 2012. “Promoting Healthy Behaviours and Improving Health Outcomes in Low and Middle Income Countries: A Review of the Impact of Conditional Cash Transfer Programmes.” Preventive Medicine (November): S95–S105. doi:10.1016/j.ypmed.2011.11.015. 19 Román, Elena Ruiz. 2010. “Zimbabwe Emergency Cash Transfer (ZECT) Pilot Programme: Monitoring Consolidated Report, November 2009 to March 2010.” Rubalcava, Luis, Graciela Teruel, and Duncan Thomas. 2002. “Welfare Design, Women’s Empowerment and Income Pooling”. Mexico City. Schluter, Christian, and Jackline Wahba. 2004. “Are Poor Parents Altruistic? Evidence from Mexico.” Slater, Rachel, and Matselio Mphale. 2008. “Cash Transfers, Gender and Generational Relations: Evidence from a Pilot Project in Lesotho”. London. The Kenya CT-OVC Evaluation Team. 2010. “Cash Transfer Program for Orphans and Vulnerable Children (CT-OVC), Kenya: Operational and Impact Evaluation, 2007 – 2009 Final Report”. Oxford Policy Management. Vaughan, Andres. 2010. “Unconditional Cash Transfers: Will Redistribution in Bolivia Work?” Oregon State University. Willibald, Sigrid. 2006. “Does Money Work? Cash Transfers to Ex-Combatants in Disarmament, Demobilisation and Reintegration Processes.” Disasters 30 (3) (September): 316–39. doi:10.1111/j.0361-3666.2005.00323.x. 20 Tables Table 1: Consumption of Alcohol and Tobacco as a Share of Total Consumption Household living on less than… $1 per $2 per Ratio day day ($2/$1) Rural Côte d'Ivoire 2.7% 2.2% 81.5% Guatemala 0.4% 0.5% 125.0% India - UP/Bihar 3.1% 3.0% 96.8% Indonesia 6.0% 6.8% 113.3% Mexico 8.1% 6.5% 80.2% Nicaragua 0.1% 0.6% 600.0% Pakistan 3.1% 2.9% 93.5% Papua New Guinea 4.1% 5.1% 124.4% Peru 1.0% 1.3% 130.0% South Africa 2.5% 3.4% 136.0% Timor-Leste 0.0% 0.0% 100.0% Urban Côte d'Ivoire 3.5% 3.3% 94.3% India - Hyderabad 2.5% 2.7% 108.0% Indonesia 5.5% 6.3% 114.5% Mexico 3.6% 4.2% 116.7% Nicaragua 1.0% 0.7% 70.0% Pakistan 3.0% 2.9% 96.7% Papua New Guinea 0.6% 4.4% 733.3% Peru 0.2% 0.8% 400.0% South Africa 5.0% 5.1% 102.0% Timor-Leste 0.0% 0.0% 100.0% Note: Adapted from Banerjee & Duflo (2006). 21 Table 2: Steps used to select papers in the systematic review Review Number of Phase Procedures Used Papers 1 Google Scholar search 4,290 2 Review titles and eliminate irrelevant papers 434 3 Eliminate 23 duplicate papers 411 Review abstracts and conduct word search and remove 4 232 papers that do not seem to look at the impact of income on consumption of temptation goods Add 7 papers recommended by colleagues and 12 5 251 papers referenced in bibliographies of papers identified in previous phases Read papers and remove those without impact 6 estimates, level estimates, or qualitative reports of the 42 impact of cash transfers on consumption of temptation goods 7 Categorize papers into 3 groups: i) Papers with impact estimates 19 ii) Papers with level estimates 11 12 iii) Qualitative reports 22 Table 3: Studies with estimated impact of transfer on alcohol or tobacco expenditures Country Program name Temptation good Impact Detail on impact Methodology CCT/UCT Reference Report total expenditures (presented in 2012 PPP) Brazil Bolsa Alimentaria & Bolsa Alcohol, tobacco, and -1.961 (1.86) Difference in differences CCT Braido, Olinto, & Perrone 2012 Escola gambling Colombia Familias en Acción Alcohol and tobacco 2.822 (4.64) Urban Difference in differences CCT Attanasio & Mesnard 2005 -1.536 (3.31) Rural India Unconditional Cash Alcohol -0.001 Before-after UCT Bhowmik, Gartenberg, & Transfer Pilot Tobacco -0.001ǂ Sarker 2009 India Unconditional Cash Alcohol 0.080 (0.53) Impact of transfer Difference in differences UCT Gangopadhyay, Lensink, & Transfer Pilot -0.455 (0.53) Impact of transfer & Yadav 2013 bank account Indonesia Unconditional cash Alcohol and tobacco -0.0001*** (0.00) First disbursement Difference in differences UCT Bazzi, Sumarto, & Suryahadi transfer 0.0001* (0.00) Second disbursement 2012 0.0000 (0.00) Average across both Kenya The GiveDirectly Alcohol -0.017 (0.02) RCT UCT Haushofer & Shapiro 2013 Unconditional Cash Tobacco -0.003 (0.00) Transfer Program Kenya The Kenya Cash Transfer Alcohol -0.024ǂ Difference in differences Conditionalities Kenya CT-OVC Evaluation Program for Orphans and Tobacco 0.000 in 3/7 districts Team 2012 Vulnerable Children Mexico Programa de Apoyo Alcohol 0.336 (0.40) RCT UCT Cunha 2012 Alimentario Tobacco -0.218 (0.14) Mexico PROGRESA Tobacco -0.029 (0.26) Benefit (dummy) RCT CCT Schluter & Wahba 2004 -0.001 (0.00) Benefit (level) Nicaragua Red de Protección Social Alcohol and tobacco -0.010* (0.01) First year Difference in differences CCT Gitter 2006 -0.001 (0.01) Second year Nicaragua Red de Protección Social Alcohol and tobacco 4.251 RCT CCT Maluccio & Flores 2005 Peru Juntos Alcohol -0.113** (0.05) Propensity score matching CCT Perova 2011 0.210* (0.12) Instrumental variables Peru Juntos Alcohol: beer, -0.002 (0.00) 2009 estimate Compare recently paid to CCT Dasso & Fernandez 2013 whisky, rum, pisco 0.005 (0.00) 2010 estimate less recently paid Tanzania TASAF CCT Pilot Program Cigarettes, tobacco -3.322 (3.16) ETT midline RCT CCT Evans, Hausladen, Kosec, & & snuff -3.098 (4.24) ITT midline Reese 2014 -2.386 (3.16) ETT endline -2.312 (4.31) ITT endline Notes: * denotes statistical significance at the 10% level, ** denotes statistical significance at the 5% level, *** denotes statistical significance at the 1 % level, ǂ denotes that statistical significance is reported but not a standard error, and ~ denotes that statistical significance is not reported. CCT is Conditional cash transfer; UCT is unconditional cash transfer. ETT is estimate of treatment on the treated. ITT is the intent-to-treat estimator. RCT is randomized control trial. TASAF is the Tanzania Social Action Fund. The reported impacts on total expenditures (and corresponding standard errors) presented in 2012 PPP are calculated by inflating the impact in local currency in the various base years (the year the data were collected, or as close to that as could be inferred) to their 2012 values using the inflation GDP deflator (annual %), before dividing by the 2012 PPP conversion factors for private consumption (LCU per international $). Both indicators used in the PPP conversion come from the World Development Indicators database available at http://data.worldbank.org/data- catalog/world-development-indicators 23 Table 4: Studies with alternative estimated impact of transfer on alcohol or tobacco consumption Country Program name Temptation good Impact Detail on impact Methodology CCT/UCT Reference Report proportion of expenditures Brazil Bolsa Alimentaria & Bolsa Alcohol, tobacco, and −0.003 (0.002) Difference in differences CCT Braido, Olinto, & Perrone 2012 Escola gambling Mexico Oportunidades Alcohol and tobacco -0.0025 (0.0018) RCT CCT Rubalcava, Teruel, & Thomas 2002 Mexico PROGRESA Tobacco -0.02 (0.09) Benefit (dummy) RCT CCT Schluter & Wahba 2004 -0.51 (0.398) Benefit (level) Nicaragua Red de Protección Social Alcohol and tobacco 0.1 RCT CCT Maluccio & Flores 2005 Report probability of alcohol abuse in household Mexico Oportunidades Alcohol abuse -0.042*** (0.016) RCT CCT Angelucci 2008 Report probability of consumption (for adolescents only) Mexico Oportunidades Alcohol -11%*** (0.026) Propensity score matching CCT Gutiérrez, Bautista, Gertler, Rural, 10 - 21 year olds Hernández, & Bertozzi 2004 incorporated in 1998 -13%***(0.029) Rural, 10 - 21 year olds incorporated in 2000 -4%***(0.015) Urban, 15 - 21 year olds Tobacco -15%***(0.029) Rural, 10 - 21 year olds incorporated in 1998 -13%***(0.024) Rural, 10 - 21 year olds incorporated in 2000 -2%**(0.007) Urban, 15 - 21 year olds Mexico Oportunidades Alcohol -40%~ Females Instrumental variables CCT Galárraga & Gertler 2009 Tobacco -46%~ Males Report number of days consumed in past week (for children 1-7 only) Uganda WFP Cash Transfers to Beer -0.198 (0.198) ITT RCT CCT Gilligan & Roy 2012 UNICEF‐supported ECD centers Notes: * denotes statistical significance at the 10% level, ** denotes statistical significance at the 5% level, *** denotes statistical significance at the 1 % level, ǂ denotes that statistical significance is reported but not a standard error, and ~ denotes that statistical significance is not reported. CCT is Conditional cash transfer; UCT is unconditional cash transfer. ETT is estimate of treatment on the treated. ITT is the intent-to-treat estimator. RCT is randomized control trial. TASAF is the Tanzania Social Action Fund. The reported impacts on total expenditures (and corresponding standard errors) presented in 2012 PPP are calculated by inflating the impact in local currency in the various base years (the year the data were collected, or as close to that as could be inferred) to their 2012 values using the inflation GDP deflator (annual %), before dividing by the 2012 PPP conversion factors for private consumption (LCU per international $). Both indicators used in the PPP conversion come from the World Development Indicators database available at http://data.worldbank.org/data- catalog/world-development-indicators 24 Table 5: Distribution of Estimates of the Impact of Cash Transfers on Temptation Goods Negative & Negative (or 0) Positive & Positive & significant & insignificant insignificant significant Total Panel A: All estimates Estimates 12 24 6 2 44 From [--] studies 6 14 5 2 19 From [--] interventions 6 12 5 2 13 Panel B: Only expenditure levels Estimates 5 17 5 2 29 From [--] studies 5 11 5 2 14 From [--] interventions 5 10 5 2 11 Panel C: All estimates - RCTs only Estimates 1 13 3 0 17 From [--] studies 1 6 2 0 8 From [--] interventions 1 5 2 0 7 Panel D: Only expenditure levels - RCTs only Estimates 0 9 2 0 11 From [--] studies 0 4 2 0 5 From [--] interventions 0 4 2 0 5 Table 6: Percentage distribution of Estimates of the Impact of Cash Transfers on Temptation Goods Negative & Negative (or 0) Positive & Positive & significant & insignificant insignificant significant Total All estimates 27% 55% 14% 5% 100% Only expenditure levels 17% 59% 17% 7% 100% All estimates - RCTs only 6% 76% 18% 0% 100% Only expenditure levels - RCTs only 0% 82% 18% 0% 100% Table 7: Percentage distribution of Estimates of the Impact of Cash Transfers on Temptation Goods – Conditional Cash Transfers (CCTs) versus Unconditional Cash Transfers (UCTs) Negative & Negative (or 0) Positive & Positive & significant & insignificant insignificant significant Total All estimates, CCTs 29% 55% 13% 3% 100% All estimates, UCTs 18% 55% 18% 9% 100% All estimates, CCTs - RCTs only 8% 77% 15% 0% 100% All estimates, UCTs - RCTs only 0% 75% 25% 0% 100% Notes: This table excludes the evaluation of The Kenya Cash Transfer Program for Orphans and Vulnerable Children, which imposed conditions in 3 of the 7 districts in which it was implemented. 25 Table 8: Percentage distribution of Estimates of the Impact of Cash Transfers on Temptation Goods by Region Negative & Negative (or 0) Positive & Positive & significant & insignificant insignificant significant Total All estimates, Latin America 32% 46% 18% 4% 100% All estimates, other regions 19% 69% 6% 6% 100% All estimates, Latin America - RCTs only 10% 60% 30% 0% 100% All estimates, other regions - RCTs only 0% 100% 0% 0% 100% Table 9: Studies with estimated survey or focus group levels of transfer on alcohol or tobacco expenditure Temptation Country Program name Impact Reference good Democratic Republic Doughnuts and <1% of households Aker 2013 of Congo beer Jordan UNHCR cash grants Alcohol, 1.3% of households Biron 2012 tobacco, and medicines Kenya Kerio Valley Cash Transfer General No reports of use on Brewin 2008 Pilot temptation goods Lesotho World Vision Cash and Food Alcohol & No significant Slater & Mphale Transfers Pilot Project tobacco increase 2008 Lesotho The Cash and Food Alcohol and 6.4% of recipients Devereux & Transfers Pilot Project cigarettes Mhlanga 2008 Malawi Mchinji Social Cash Transfer Alcohol 1.1% of recipients Miller, Tsoka & Pilot Scheme Reichert 2008 Malawi Food & Cash Transfers Alcohol, 0.1% of transfer Devereux, Mvula cigarettes, & Solomon 2006 entertainment Malawi & Zambia Oxfam's cash transfers Alcohol No reports of use on Harvey & Savage temptation goods. 2006 Vietnam Non-emergency cash grants Alcohol and No reports of use on Humphreys 2008 in An Loc commune gambling temptation goods. Zimbabwe Government of Zimbabwe Alcohol Marginal increase in Phiri 2012 Harmonised Social Cash consumption Transfer Zimbabwe Zimbabwe Emergency Cash Alcohol & <0.5% of transfer Román 2010 Transfer (ZECT) Pilot tobacco used on temptation Program goods 26 Table 10: Overall Consumption Impacts for Studies with Positive and Insignificant Estimates on Temptation Goods Number of significant Percentage of disaggregated Significant impact on disaggregated consumption Country Program name consumption estimates that Reference total consumption estimates / total disaggregated are significant consumption estimates Report total expenditures Colombia Familias en Acción x 17/34 50% Attanasio & Mesnard 2005 India Unconditional Cash Transfer Pilot x 4/6 67% Gangopadhyay 2013 Mexico Programa de Apoyo Alimentario x 7/32 22% Cunha 2012 Nicaragua Red de Protección Social x 9/16 56% Maluccio and Flores 2005 Peru Juntos x 4/14 29% Dasso & Fernandez 2013 Report proportion of expenditures Nicaragua Red de Protección Social x 9/16 56% Maluccio and Flores 2005 Notes: This table presents an analysis of the statistical power of evaluations to identify significant impacts on consumption, for those studies which find positive insignificant impact estimates on the consumption of temptation goods. To do this, we present both whether or not these studies find significant impacts on total consumption, as well as the number and percentage of significant estimates they find for disaggregated consumption items. We are conservative in our calculations of the latter, counting only the most disaggregated estimates in a given study (for example, we exclude the estimates for grains in studies which further disagreggate this into estimates for rice, pasta, and cereal). When considering disaggregated consumption estimates, we exclude estimates on alcohol and tobacco in these calculations so as to compare the statistical power of the evaluations to identify non-temptation good consumption estimates with that for identifying temptation good estimates. 27 Figures Figure 1: Countries with Estimates of the Impact of Cash Transfers on Temptation Goods or with Estimates of the Level of Consumption of Temptation Goods from Transfer Income Note: Areas in red are countries covered (in part or entirely) in our impact and level estimates. 28 Figure 2: Distribution of Estimates of the Impact of Cash Transfers on Temptation Goods Panel A: All estimates Panel B: Only expenditure levels Number of estimates Negative & Negative & significant significant Negative or zero Only Negative or zero All estimates 12 24 6 2 & insignificant* expenditure 5 17 5 2 & insignificant* levels Positive & Positive & insignificant insignificant Positive & Positive & significant significant 0% 100% 0% 50% 100% Panel C: All estimates - RCTs only Panel D: Only expenditure levels - RCTs only Negative & Negative & significant significant Negative or zero Only Negative or zero All estimates - & insignificant* expenditure & insignificant* 1 13 30 0 9 20 RCTs only levels - RCTs Positive & only Positive & insignificant insignificant Positive & Positive & significant significant 0% 50% 100% 0% 50% 100% 29 Figure 3: Distribution of Estimates of the Impact of Cash Transfers on Temptation Goods by Program Type Panel A: Conditional cash transfers – All estimates Panel B: Unconditional cash transfers – All estimates Negative & Negative & significant significant Negative or zero Negative or zero CCTs: All & insignificant* UCTs: All 9 17 41 2 6 2 1 & insignificant* estimates estimates Positive & Positive & insignificant insignificant Positive & Positive & significant significant 0% 50% 100% 0% 50% 100% Figure 4: Distribution of Estimates of the Impact of Cash Transfers on Temptation Goods by Region Panel A: Latin America - All estimates Panel B: Other regions – All estimates Negative & significant Negative & significant Negative or zero Latin America: & insignificant* Negative or zero 9 13 5 1 Other regions: All estimates 3 11 11 & insignificant* Positive & All estimates insignificant Positive & insignificant Positive & significant Positive & significant 0% 50% 100% 0% 50% 100% 30 Annex 1: Studies Included in the Systematic Review Impact Estimates Angelucci, Manuela. 2008. “Love on the Rocks: Domestic Violence and Alcohol Abuse in Rural Mexico.” The B.E. Journal of Economic Analysis & Policy 8 (1). Attanasio, Orazio, Erich Battistin, Emla Fitzsimons, Alice Mesnard, and Marcos Vera-Hernández. 2005. “How Effective Are Conditional Cash Transfers? Evidence from Colombia.” Bazzi, Samuel, Sudarno Sumarto, and Asep Suryahadi. 2012. “Evaluating Indonesia’s Unconditional Cash Transfer Programme, 2005‒2006.” Bhowmik, Sharit K., Indira Gartenberg, and Kanchan Sarker. 2013. “Food or Cash: Assisting the Urban Poor in India.” In The Food Crisis: Implications for Labor, edited by Cristoph Scherrer and Debdulal Saha. ICDD. Braido, Luis H. B., Pedro Olinto, and Helena Perrone. 2012. “Gender Bias in Intrahousehold Allocation: Evidence from an Unintentional Experiment.” Review of Economics and Statistics. Cunha, Jesse M. 2012. “Testing Paternalism: Cash versus In-Kind Transfers.” Unpublished Mimeo. Naval Postgraduate School. Dasso, Rosamaría, and Fernandez Fernando. 2013. “Temptation Goods and Conditional Cash Transfers in Peru”. Washington, DC. Evans, David K., Stephanie Hausladen, Katrina Kosec, and Natasha Reese. 2014. Community-Based Conditional Cash Transfers in Tanzania: Results from a Randomized Trial. Washington, DC: World Bank: © World Bank. Galárraga, Omar, and Paul J. Gertler. 2009. “Conditional Cash & Adolescent Risk Behaviors: Evidence from Urban Mexico.” Gangopadhyay, Shubhashis, Robert Lensink, and Bhupesh Yadav. 2013. “Cash or Food Security Through the Public Distribution System? Evidence from a Randomized Controlled Trial in Delhi, India.” Evidence from a …. Gilligan, Daniel O., and Shalini Roy. 2013. “Resources, Stimulation ,and Cognition: How Transfer Programs and Preschool Shape Cognitive Development in Uganda”. Washington, DC. Gitter, SR. 2006. “Women and Targeted Cash Transfers in Nicaragua.” Aae.wisc.edu. Gutiérrez, Juan Pablo, Sergio Bautista, Paul Gertler, Mauricio Hernández, and Stefano M. Bertozzi. 2004. “External Evaluation of the Impact of the Human Development Program Oportunidades:” Mexico City. 31 Haushofer, Johannes, and Jeremy Shapiro. 2013. “Household Response to Income Changes: Evidence from an Unconditional Cash Transfer Program in Kenya”. Massachusetts Institute of Technology. Maluccio, John A., and Rafael Flores. 2005. “Impact Evaluation of a Conditional Cash Transfer Program”. Washington, DC. Perova, Elizaveta. 2011. “Three Essays on Intended and Not Intended Impacts of Conditional Cash Transfers”. University of California, Berkeley. Rubalcava, Luis, Graciela Teruel, and Duncan Thomas. 2002. “Welfare Design, Women’s Empowerment and Income Pooling”. Mexico City. Schluter, Christian, and Jackline Wahba. 2004. “Are Poor Parents Altruistic? Evidence from Mexico.” The Kenya CT-OVC Evaluation Team. 2010. “Cash Transfer Program for Orphans and Vulnerable Children (CT-OVC), Kenya: Operational and Impact Evaluation, 2007 – 2009 Final Report”. Oxford Policy Management. Level Estimates Aker, Jenny. 2013. “Cash or Coupons? Testing the Impacts of Cash versus Vouchers in the Democratic Republic of Congo.” Biron, Aurélie. 2012. “Adapting to Urban Displacement: The Use of Cash Transfers in Uraban Areas”. Sciences Po Grenoble. Brewin, Mike. 2008. “Evaluation of Concern Kenya’s Kerio Valley Cash Transfer Pilot (KVCTP), April - June 2008.” Devereux, S, and M Mhlanga. 2008. “Cash Transfers in Lesotho: An Evaluation of World Vision’s Cash and Food Transfers Pilot Project”. Brighton, UK, and Maseru, Lesotho. Devereux, Stephen, Peter Mvula, and Colette Solomon. 2006. “After the FACT : An Evaluation of Concern Worldwide ’ S Food and Cash Transfers Project in Three Districts of Malawi , 2006”. Brighton, UK, and Zomba, Malawi. Harvey, Paul, and Kevin Savage. 2006. “No Small Change. Oxfam GB Malawi and Zambia Emergency Cash Transfer Projects: A Synthesis of Key Learning”. London. Humphreys, Rowena. 2008. “Evaluation of the Cash Transfers for Development Project in Vietnam”. London. Miller, Candace, Maxton Tsoka, and Kathryn Reichert. 2008. “Impact Evaluation Report External Evaluation of the Mchinji Social Cash Transfer Pilot”. Boston, USA, and Zomba, Malawi. 32 Phiri, Swedi. 2012. “A Comparative Assessment of the Impact of Unconditional Cash Transfers for Urban Vulnerable Households Headed by Elderly and Non-Elderly Women in Mucheke Ward”. University of Zimbabwe. Román, Elena Ruiz. 2010. “Zimbabwe Emergency Cash Transfer (ZECT) Pilot Programme: Monitoring Consolidated Report, November 2009 to March 2010.” Slater, Rachel, and Matselio Mphale. 2008. “Cash Transfers, Gender and Generational Relations: Evidence from a Pilot Project in Lesotho”. London. Qualitative Reports Aheeyar, M.M.M. 2006. “Cash Delivery Mechanisms in Tsunami-Affected Districts of Sri Lanka”. London. Bukuluki, Paul, and Carol Watson. 2012. “Transforming Cash Transfers: Beneficiary and Community Perspectives on the Senior Citizen Grant (SCG) in Uganda”. London, UK, and Kampala, Uganda. Devereux, Stephen, and Paul Jere. 2008. “Cash and Food Transfers in Swaziland: An Evaluation of Save the Children’s Emergency Drought Response, 2007/08”. Brighton, UK, and Lilongwe, Malawi. Devereux, Stephen, Catherine Mthinda, Fergus Power, Patrick Sakala, and Abigail Suka. 2007. “An Evaluation of Concern Worldwide’s Dowa Emergency Cash Transfer Project (DECT) in Malawi, 2006/07”. Brighton, UK, and Lilongwe, Malawi. Farrington, John, and Rachel Slater. 2009. “Lump Sum Cash Transfers in Developmental and Post- Emergency Contexts: How Well Have They Performed?” London. Harvey, Paul. 2007. “Cash-Based Responses in Emergencies”. London. Jones, Nicola, Rosana Vargas, and Eliana Villar. 2007. “Conditional Cash Transfers in Peru: Tackling the Multi-Dimensionality of Childhood Poverty and Vulnerability. Social Protection Initiatives for Families, Women and.” In Forthcoming Chapter in Alberto Minujin et Al. (ed.) 2007. Social Protection Initiatives for Families, Women and Children: An Analysis of Recent Experiences. New York: New School and UNICEF. Khera, Reetika. 2013. “Cash vs In-Kind Transfers: Indian Data Meets Theory.” Mohiddin, Lili, Manohar Sharma, and Anette Haller. 2007. “Comparing Cash and Food Transfers: Findings from a Pilot Project in Sri Lanka”. London, UK, Washington DC, USA, and Rome, Italy. Onyango-Ouma, W., and Fiona Samuels. 2012. “Transforming Cash Transfers: Beneficiary and Community Perspectives on the Cash Transfer for Orphans and Vulnerable Children Programme in Kenya”. London, UK, and Nairobi, Kenya. 33 Vaughan, Andres. 2010. “Unconditional Cash Transfers: Will Redistribution in Bolivia Work?” Oregon State University. Willibald, Sigrid. 2006. “Does Money Work? Cash Transfers to Ex-Combatants in Disarmament, Demobilisation and Reintegration Processes.” Disasters 30 (3) (September): 316–39. doi:10.1111/j.0361-3666.2005.00323.x. 34