Is Informal Redistribution Costly? Evidence from a Lab-in-the-Field Experiment in Senegal

Abstract

In Sub-Saharan Africa, individuals frequently transfer a substantial share of their resources to members of their social networks. Social pressure to redistribute, however, can induce disincentive effects on resource allocation decisions. This paper measures and characterizes the costs of redistributive pressure by estimating individuals’ willingness to pay (WTP) to hide their income. The study estimates a social tax due to informal redistribution of 10 percent. Moreover, it shows that individuals are willing to escape from the redistributive pressure exerted mainly by extended family members.

1. Introduction

In Sub-Saharan Africa, individuals frequently transfer a substantial share of their resources to members of their social networks (Baland, Guirkinger, and Mali 2011; di Falco and Bulte 2011; Baland et al. 2016). These transfers can fulfill a risk-sharing role (Cox and Fafchamps 2007) and can also be driven by social motives such as altruism, charity, social prestige, and strong social obligations for redistribution (Platteau 2014). While transfers motivated by a risk-sharing purpose are potentially reciprocal in nature, transfers motivated by social obligations for redistribution are akin to a tax and may create some inefficiencies. Recent studies have found that social pressure to redistribute can induce direct disincentive effects on resource accumulation decisions, such as labor supply or investment (Grimm, Hartwig, and Lay 2013; Hadness, Vollan, and Kosfeld 2013; Squires 2018), and indirect distortions in resource allocation choices (Baland, Guirkinger, and Mali 2011; di Falco and Bulte 2011; Boltz 2015; Baland et al. 2016; Goldberg 2017).

This paper measures and characterizes the costs of redistributive pressure by estimating individuals’ willingness to pay (WTP) to hide their income. To this end, the study conducted a lab-in-the-field in May and June 2014 in seven different poor communities of the Dakar region in Senegal. In addition to providing a direct monetary measure of the inefficiency induced by the social pressure to share, this experiment answers three questions: (i) To what extent are individuals willing to hide their income? (ii) Who are the individuals who are willing to pay to hide their resources? (iii) From whom are they hiding?¹ Few papers have been able to properly identify the costs of redistributive pressure. Among those who have relied on lab-in-field experiments to assess the WTP to hide income, Jakiela and Ozier (2016) and Beekman, Gatto, and Nillesen (2015) are the closest papers to this article. This study's main contribution to this literature is to correlate, in a single framework, the WTP to three experimental variations in the pool of observers that have been shown to be relevant one by one: whether a kin member outside the household is present (see, for instance, Jakiela and Ozier 2016), whether a household member is present (see, for instance, Castilla and Walker 2013), and whether an acquaintance is present (see, for instance, Goldberg 2017).²

2. Experimental Design

The baseline sample consisted of 920 individuals, randomly selected within the surveyed communities. Moreover, in half of the households sampled, one participant was selected, while two participants were selected in the other half. The lab phase took place within a few days following the baseline interviews. It included four sessions covered on a single day per community. On average 30 individuals from the same or nearby dwelling blocks were invited to the same session.

To measure the cost of redistributive pressure, the study elicited the WTP to hide income for all participants in the lab phase, through a multiple-price list method. During a session, participants were invited to a private interview followed by a private lottery. In this lottery, participants chose cards at random from a box; the cards determined their income from the experiment and whether some of their income would be announced publicly to other attendees. Before the lottery took place, to reveal their preference for income unobservability, participants were asked to state their preferences between two options of payoffs for different cards: option A corresponds to receiving 9,000 CFA francs (FCFA) in public,³ that is, in front of the other session participants, while option B means receiving 1,000 FCFA in public and 8,000 FCFA minus some varying amount p in private. Here, p is the price of the income-hiding option and equals 0, 200, 500, 700, or 1,000 FCFA. Each choice, that is, for each value of p, was presented one after the other, in ascending order. The enumerator made clear to the participant that the choices made would be implemented after the lottery draw.⁴ The WTP for unobservable income can be directly recovered from these responses. The final experimental sample is of 788 participants, two-thirds of whom were women, 19 percent of household heads, and 41 percent of the sample contribute to the household food expenses.⁵

3. Results

What is the Willingness to Pay to Hide?

Table 1 provides some descriptive statistics of the WTP. More than two-thirds of subjects preferred to receive their gains in private rather than in public, and among them, they were ready to forgo on average 10 percent of their unobserved income (the mean WTP being equal to 805 FCFA for individuals with a positive WTP). This share is smaller for women (9.8 percent) than for men (10.5 percent).⁶

Going one step further, the study provides an additional measure of the redistributive pressure by reporting the relative WTP computed as the ratio of the WTP to the household daily food expenditure. This supplemental measure shows that individuals living in poorer households reveal levels of WTP that are twice as large as those living in richer households, in both men and women samples.

In sum, the study finds a social tax of 10 percent, more than twice as high as the one found by Jakiela and Ozier (2016) in the context of rural Kenya. Redistribution pressure seems to arise mainly from extended family members, especially for women, and not from within the household. In addition, it appears strongly regressive, poorer individuals facing a relatively higher redistributive burden.

Who is Willing to Hide Her Income?

To estimate the determinants to the WTP to hide, the study runs an interval-censored-data regression model.⁷Tables 2 and 3 present the results for all the covariates of the interval-censored data model respectively on the whole sample and on the subsamples below and above the median household food consumption.⁸ Coefficients represent the additional price people with the considered characteristics are willing to pay.

Looking at columns (1w) and (1m) in table 2, it appears clearly that men and women do not share the same determinants of their WTP to hide income. This is not so surprising, since in the context of West Africa, men and women have generally different social realms; hence the pressure to redistribute may vary across gender. Moreover, men and women do not have the same role as contributors to household expenses. Therefore, the study disentangles its analysis hereafter along the gender dimension.

For women, the characteristics correlated with a higher WTP to hide income are closely linked to the position they hold in their household and in their community. A woman who is a household head is willing to pay 755 FCFA more relative to other women. A woman who has always been living in the community is willing to pay almost 500 FCFA more. Similarly, the increase in the WTP to hide income for single women relative to married women is about 500 FCFA. These two last effects are actually driven by the sample of women from poorer households (table 3). On this sample, the study also observes that women's labor income is positively correlated with the WTP to hide. A possible interpretation is that those women, whether in the middle of the labor income distribution and/or single and/or community natives, are more likely to be taxed by others.

Determinants of the WTP to hide income for men fall into two broad dimensions: on the one hand, characteristics related to the economic position—a better economic position is correlated with a higher WTP—and on the other hand, having a good social position in the community correlates with a lower WTP. Regarding the social dimension, holding a responsible position in the community induces a WTP 2683 FCFA lower. Men choosing a responsible position within the community may have a higher internalized redistributive duty. Alternatively, they may fear less being observed because they may have access to alternative strategies allowing them to avoid taxation. Looking at proxies for economic status, it can be seen that being single, that is, being financially responsible for fewer people, encourages men to increase their WTP to hide income by 1089 FCFA. Proxies for household wealth are positively correlated with the WTP: the level of daily food expenditure (positively), or whether the household is renting a house (negatively).

From Whom are People Hiding?

To answer this question, the study relies on experimental variations of the composition of the pool of participants per session (first three rows of tables 2 and 3). For men, the study finds no significant effect from the experimental variations of the group composition of a lab session on the maximum price they were willing to pay. For women, each additional kin attending the same session (other than the player's potential paired household member) significantly increased the WTP to hide income by 955 FCFA. This effect holds whether women come from poorer or from richer households. Being randomly selected along with another household member appears to have an opposite effect on players’ WTP, depending on whether they come from a relatively poorer or a richer household: it is positive for the former, negative for the latter.⁹

Therefore, redistributive pressure, from which individuals try to escape, appears to come more from people outside the household for women from richer households, and from both household members and from extended family members for individuals from poorer households.

4. Conclusion

The study estimates a social tax due to informal redistribution of 10 percent. Moreover, it shows that people are willing to escape from redistributive pressure exerted mainly by the extended family. More specifically, women are found to hide from nonhousehold family members and also from household members for the poorest ones.¹⁰

These results thus indicate that a well-designed savings scheme would allow two-thirds of the population to save around 10 percent of their earnings. Such programs can have a large leverage in terms of poverty reduction, as the population at stake is the poorest. Further research is necessary to capture the general equilibrium effects, including the long-term benefits of social redistribution in terms of risk-sharing, and the distortionary effects on resource allocation and income generating activities.

Author Biographical

Marie Boltz is an assistant professor at the Centre d’Économie de la Sorbonne, Université Paris 1; her email address is [email protected]. Karine Marazyan is an assistant professor IEDES-University Paris 1, UMR Développement et Sociétés; her email address is [email protected]. Paola Villar is a postdoctoral fellow at the University of Namur; her email address is [email protected]. The authors are thankful to the CEPREMAP, the PSE Research Fund, the Sarah Andrieux Fund, the Chair G-Mond, the IRD, and the UMR Développement et Sociétés for their financial support for this research project. They particularly would like to thank Jean-Marie Baland, Xavier Gine, Laurent Gobillon, Isabel Günther, Marc Gurgand, Sergei Guriev, Elise Huillery, Kenneth Houngbedji, Pamela Jakiela, Sylvie Lambert, Dan Levin, Karen Macours, Alice Mesnard, Owen Ozier, and numerous conference and seminar participants for their valuable comments. This project received the IRB certification of J-Pal Europe (CE/2014-005). All remaining errors are ours.

Footnotes

References