Does altruism matter for remittances?

Does altruism matter for remittances? Abstract We provide a direct test of the impact of altruism on remittances. From a sample of Indian migrant workers in Qatar, we elicit the propensity to share with others from their responses in a dictator game and use it as a proxy for altruism. For the entire sample, we find that altruism does not seem to matter. However, we document a strong positive relationship between altruism and remittances for a subset of migrants with a loan obligation, whereas indirect tests of altruism, typically used in the literature, would fail to establish this relationship. We explain the role of loan obligations with a standard remittance model extended with reference-dependent preferences. 1. Introduction Altruism is commonly, if not routinely, viewed as one of the most important motives for remittances in transnational families.1 However, because of its intangibility the actual impact of altruism on remitting behaviour is difficult to evaluate and, therefore, its assessment is typically attempted by indirect tests. A well-known example of such a test is found in Lucas and Stark (1985). They compare two alternative models of remittances, where one model is based on altruism and the other on self-interest. While both models predict that remittances increase with the migrant’s income, only in the altruistic model remittances decline with the recipient household’s income. Thus, a test for altruism is whether remittances decrease in the household’s income or, if aggregated, the receiving country’s agricultural GDP as in Bouhga-Hagbe (2006) or, if proxied, the number of other family members in migration as in Funkhouser (1995). Such indirect tests, however, can lack validity in the presence of confounding factors. For instance, other motives for remittances are also consistent with a negative relationship between remittances and household income. Under the strategic motive for remittances, migrants can increase their remittances in response to decreased levels of income in the home country in order to ease pressures for others to emigrate, which could negatively affect migrant earnings. Data limitations such as the absence of longitudinal data also tend to restrict the investigator from ruling out the moral hazard motive due to the recipient household reducing their labour effort in response to remittances. Furthermore, the worsening of the recipient’s financial conditions can trigger the control motive for remittances, the role of which, as recently demonstrated in Ashraf et al. (2015) or Batista et al. (2015), can be sizable.2 Lastly, recent empirical evidence on the role of asymmetric information on remittance flows further exacerbates the confounding problem. For instance, Ambler (2015) shows that migrants remit more if their families observe their earnings, implying that the role of altruism may be overestimated under symmetric information (also see Joseph et al., 2015; Seshan and Zubrickas, 2017).3 Given the increasing economic significance of remittances and the assumed importance of altruism, the existing approach to the study of the altruistic motive is hardly satisfactory. In an attempt to overcome limitations of indirect tests, we provide a direct test of the impact of altruism on remittances using a measure of altruism borrowed from behavioural economics. We administered a survey and conducted a behavioural experiment with a sample of 105 married male migrant workers from Kerala, India, working in Qatar whose spouses reside in India. The behavioural experiment consisted of tasks measuring social preferences, including a dictator game in which each participant received 100 Qatari Riyals (approximately, US$27) and decided how much of that amount to give to another, anonymous participant. From their responses, we elicit the propensity to share with others and use it to proxy for each participant’s degree of altruism. Our main interest lies with whether the measured variation in altruism across migrants helps to explain the observed variation in remittance behaviour. The proposed measure of altruism is widely employed in behavioural economics and psychology (see Forsythe et al., 1994; Camerer, 2003). Arguably, giving in a dictator game is initiated by sympathy or empathy in the mind of the giver and is not driven by ‘indirect selfishness,’ e.g. stemming from mutual cooperation, norm adherence through sanctions or other motives as can be relevant for remitting behaviour.4 However, an obvious question arises whether a measure constructed from migrants’ propensity to give to their peers is adequate for the purpose of assessing the propensity to give to their own family. In particular, according to the kin-selection theory of Hamilton (1964), altruism will be greater with stronger degree of kinship. While we do acknowledge limitations of our approach toward measuring intrafamilial altruism and that our findings need to be interpreted with corresponding reservations, we argue below that our constructed measure of altruism is at least positively correlated with the migrants’ intrafamilial sharing propensity. First of all, as argued by Kaplan and Gurven (2005), it is more social connectedness than kinship per se that motivates altruistic behaviour among humans, and the level of altruism varies inversely with social distance irrespective of the degree of relatedness.5 In this regard and as part of our design, there is a high degree of social connectedness among our sampled migrants. They form a close social group as not only they all originate from the same region in India and are of very similar socio-economic and cultural backgrounds but also reside in the same dormitory-styled accommodations in Qatar for a substantial period of time. The average transfer of 37% that we observe in a dictator game indicates their high, by any standard, propensity of sharing with peers.6 Second, following the dictator game we also conducted an ultimatum game with the same pool of migrants. The ultimatum game is the extension of a dictator game where the proposed division of the endowment takes place only if the receiver approves it. This extension gives rise to strategic motives for greater sharing lest the receiver find the division unfair. However, we observe very similar patterns of giving both in the dictator and ultimatum games (see Fig. A1).7 This suggests that, in our sample, giving does not seem to be motivated by strategic considerations. In particular, under the ‘indirect selfishness’ argument that a migrant who is altruistic toward his family could exhibit selfish behaviour toward his peers to save more money for his family, we would expect to observe changes in giving behaviour between the ultimatum and dictator games, but we do not. In our empirical analysis, we find no relationship between an individual-level measure of altruism and the sending of remittances when the entire sample is considered. This finding confirms our previous discussion about the complex interdependence of various factors and motives for remittances that, unless accounted for, may diminish the effect of the altruistic motive. In this study, we also uncover one such confounding factor—the possession of a loan obligation. We find that remittances rise with the degree of altruism only for the migrants with a loan obligation (who make about a half of the sample). Specifically, the estimated remittance schedule has a smaller intercept and a larger coefficient of altruism for the migrants with a loan obligation than for those without. It is important to note that the indirect test of altruism based on the recipient household’s income does not lead to the same conclusion. We argue that our empirical findings related to the possession of loan obligations are in line with the basic remittance model of altruism (Lucas and Stark, 1985; Stark, 1995, Ch.1) extended with reference-dependent preferences. In particular, we introduce a reference point for remittances, similar to benchmark remittances in Hoddinott (1994), against which a migrant’s utility from remittance is measured. The reference point can be thought of as a savings target or the amount of money that the migrant is expected to transfer home, as determined by existing obligations, family needs, or social comparisons. See Osili (2004); Amuedo-Dorantes et al. (2005); Yang (2011) for migrants’ target saving, which is particularly relevant for temporary, contract-based migration applicable to our sample. Lastly, we assume loss aversion (Kahneman and Tversky, 1979), where a migrant experiences a positive utility if his remittance is above the reference point and a negative utility if below, with losses looming larger than gains. In essence, our proposed model shows that altruism is subdued when a migrant is confronted with uncertainty over the amount of remittances to send. By contrast, a reduction of this uncertainty allows the altruistic motive to surface and more prominently influence the remitter’s behaviour akin to conventional models of remittances without uncertainty. Specifically, if migrants without loan obligations face more uncertainty about the reference point for remittances, an assumption that we justify later, we show that the theoretical predictions of the model closely match our empirical results. An increase in uncertainty about the reference point raises the risk and resultant disutility of falling short of the expectations back home if the amount remitted is relatively low, thus, prompting the loss-averse migrant to increase his remittance. But if the amount initially remitted is sufficiently large so that the risk of undershooting the reference point is negligible and is rather dominated by the risk of overshooting the reference point, then the relationship between remittances and uncertainty turns negative due to the diminishing utility of remittances past the reference point. Thus, given a positive relationship between remittance and altruism all else equal, the model predicts a flatter remittance schedule in altruism for migrants with less certainty about remittance expectations, i.e. without loan obligations, in accordance with our empirical results. All in all, the altruistic motive can be diminished by loss aversion coupled with uncertainty about remittance expectations. The remainder of the paper is organized as follows. In Section 2, we present the remittance model of altruism extended with reference-dependent preferences. In Section 3, we describe the data and compare the characteristics of migrants that report an explicit loan obligation to those that report none. In Section 4, we present our empirical findings, using which we discuss the role of reference dependence for modelling remittance behaviour in Section 5. The last section concludes the study. 2. Reference dependence and uncertainty Our theoretical approach builds on the remittance model of altruism (Lucas and Stark, 1985; Stark, 1995, Ch.1) extended with reference dependence. In the model, a migrant worker derives utility from own consumption and from his ability to meet a reference point for remittances, e.g. a savings target. The migrant experiences a psychological cost or gain if his remittance is below or, respectively, above the reference point, with losses looming larger than gains. We assume that the intensity of the psychological factor is in proportion to the migrant’s degree of altruism. As in Kőszegi and Rabin (2006), the reference point can be, however, uncertain, in which case the degree of uncertainty becomes an important factor for remittance decisions. Here, we take a reference point for remittances and its uncertainty as exogenously given. But see Seshan and Zubrickas (2017), where a threshold for remittances, equivalent to a reference point, is endogenously determined. In particular, the optimal (implicit) relationship contract between the migrant and the recipient household features a threshold for remittances if migrant earnings are private information. The threshold for remittances is determined by the degree of informational asymmetry, expressed by the cost of information acquisition, and distribution for earnings. Drawing on these findings, in this paper uncertainty about the reference point can be motivated by discrepancy in the recipient’s expectations about migrant earnings and the migrant’s beliefs about the recipient’s expectations. 2.1 Model Consider a migrant worker who has to divide his earned income Y > 0 between remittance R ≥ 0 sent to his family and private consumption, Y – R ≥ 0. The migrant’s utility from private consumption is given by function u(.) that satisfies u′(.)>0 and u′′(.)<0. The migrant’s utility from remittance R captures the family’s welfare and their expectations and is given by an increasing function μ(R−R¯), where R¯ is a reference point for remittances. We assume that μ(.) satisfies the properties of a ‘universal gain–loss function’ in Kőszegi and Rabin (2006). Specifically, the migrant experiences a negative utility from remittance if R<R¯ and a positive utility if R>R¯, where utility losses resonate more than gains, i.e. μ″(x)>0 for x < 0 but μ″(x)<0 for x ≥ 0. The reference point R¯ is uncertain and, for analytical convenience, assumed to be uniformly distributed over an interval [r−e,r+e]. The parameter e, 0<e<r, measures the migrant’s uncertainty about the reference point. Letting θ > 0 denote the migrant’s degree of altruism, measured as the weight the migrant puts on his utility from remittance, we write the migrant’s total utility as:   U(R;Y,e)=u(Y−R)+θ∫r−er+eμ(R−R¯)12edR¯. (1) The optimal level of remittance R∗=arg⁡max⁡RU(R;Y,e) is determined, assuming the interior solution exists, by the first-order condition:   −u′(Y−R∗)+θ2e∫r−er+eμ′(R∗−R¯)dR¯=0, (2) which, by the fundamental theorem of calculus, can be expressed as:   −u′(Y−R∗)+θ2e(μ(R∗−r+e)−μ(R∗−r−e))=0. (3) In other words, at the optimal level of remittance the marginal utility of private consumption is equal to the marginal utility from the remittance. Using that the second-order condition (SOC) has to hold, the internal derivative dR∗/dθ>0, taken from (3), implies that a higher degree of altruism results in larger remittances, and vice versa. Now suppose that the migrant becomes less certain about the reference point R¯, which we model by an increase in uncertainty parameter e. The effect of increased uncertainty on remittance is given by the internal derivative:   dR∗de=2u′(Y−R∗)−θ(μ′(R∗−r+e)+μ′(R∗−r−e))SOC. (4) Rewrite this derivative using (3) as:   dR∗de=θe(Δ(e)−Δ(−e))−SOC, (5) where:   Δ(e)=eμ′(R∗−r+e)−μ(R∗−r+e). (6) The denominator of (5) is positive, but the numerator can be both positive and negative. In particular, it depends on the size of R*, which is to say on the degree of altruism θ. Technically, at smaller values of R∗ (<r−e) so that μ(.) is convex, the numerator is positive because Δ′(e)>0. But at larger values of R∗ (>r+e) so that μ(.) is concave, the numerator turns negative because now Δ′(e)<0. Then, by continuity we obtain a flatter remittance schedule when the degree of uncertainty increases. Intuitively, consider a migrant who remits a relatively small amount due to his relatively low degree of altruism. Because of uncertainty there is a risk that the amount remitted falls short of the reference point. An increase in uncertainty also increases this risk, which prompts a loss-averse migrant to remit more as a hedging measure against the increased risk. Now consider a migrant who initially remits a large amount due to his high degree of altruism. The risk of falling short of the reference point is negligible even with an increased degree of uncertainty, but at the same time the risk of overshooting the reference point becomes significant. Because of the diminishing marginal utility from remittances above the reference point, an increase in uncertainty prompts a migrant to favour more private consumption and, as a result, to lower remittances. As an illustration, suppose that the reference point R¯ is certain and that the migrant’s initial remittance R* exceeds it by 100 riyals with the marginal utility from the remittance being θμ′(100). Let the reference point be no longer certain so that the initial remittance R* can exceed it by either 50 or 150 riyals with equal probability. Because of the diminishing marginal utility from remittances past the reference point, the expected marginal utility from the initial remittance becomes smaller, 0.5θ(μ′(50)+μ′(150))<θμ′(100). By condition (3), the migrant’s optimal response to the reduced marginal utility from remittances is to reduce his initial remittance in favour of more private consumption, which would put the marginal utilities of private consumption and from remittances back in balance. The opposite effect holds when the initial remittance is relatively low so that an increase in uncertainty increases the expected marginal utility from the remittance. 2.2 Loan obligations and uncertainty We assume that possessing a loan obligation reduces uncertainty about the reference point for remittances. This assumption can be motivated in two ways. First, in our sample of migrants we observe that about half possess an explicit loan obligation. Despite this difference, as detailed in Section 3, the migrants with an explicit loan obligation and those without earn and remit similar amounts and have otherwise very similar socioeconomic backgrounds and consumption preferences. In particular, both types report the same purpose of their loans or savings/remittances, which is housing-related. To the extent that achieving this purpose is a major influence on how much should be remitted, the presence of a loan obligation provides greater certainty about the reference point for remittances. Second, and more generally, even if not exposed to explicit loan obligations, migrants are frequently bound by implicit family loan contracts, typically made to finance migration costs (for the relevance of such implicit loan contracts, see Lucas and Stark, 1985; Hoddinott, 1994; Poirine, 1997). Remittances are then considered as dividends from the family investment, but explicit financial requirements—which can include a loan obligation—create greater certainty about the exact size of expected dividends for which a migrant is responsible. Assume that migrants without a loan obligation face more uncertainty about the reference point. Based on our theoretical analysis, we hypothesize: Hypothesis 1 The remittance schedule of migrants with loan obligations has a lower intercept but a larger coefficient of altruism than that of migrants without loan obligations.This hypothesis is illustrated in Fig. 1. Fig. 1. View largeDownload slide Predicted remittance schedule in altruism Fig. 1. View largeDownload slide Predicted remittance schedule in altruism 3. Data Our data come from two sources. First, data on migrants’ personal and household characteristics, income, remittances, assets, consumption, and loans were collected through a survey administered to a sample of migrant workers in Doha, Qatar. To limit heterogeneity, we screened for male, married workers from Kerala, India, whose spouses remained behind and who were of Hindu faith and had at most a high-school diploma. A working sample of 203 individuals from across seven dormitory-styled accommodations located in Doha’s Industrial Area completed a baseline survey in April and May 2012, conducted by a local survey firm staffed with migrants from Kerala.8 Furthermore, to better control for individual heterogeneity in subsequent empirical analysis, we constructed a measure of risk attitude from the surveyed migrants’ responses to exercises on their attitude toward risk, which involved real payouts.9 Our second source of data comes from the behavioural experiment we conducted with a subset of the migrant workers who had completed the survey. Specifically, a few weeks after the surveys were completed, we invited all migrants who took part in the survey to the campus of the Georgetown University in Doha on a weekend to participate in a series of behavioural games. Transportation was provided. A total of 105 migrants accepted the invitation. Table A1 in online Appendix A shows that migrants who visited the campus were on average slightly older, less likely to have post-secondary schooling, more likely to have loans and have wives participating in self-help groups (SHGs) and with lower household income in India. We control for these observed differences in the empirical analysis of remittance behaviour.10 On campus, the migrants played public good, trust, dictator, and ultimatum games with real payouts based on earnings in one randomly selected game. For the present study, which focuses on altruism, we mainly consider behaviour in the dictator game. In this game, the participants were randomly matched in pairs, but the identity of the paired participant was never revealed. Each participant was then asked to decide how much of a 100 Qatari Riyal endowment (approximately US$27, the equivalent of about two-days worth of the average migrant’s salary) he would share with his anonymous partner.11 The mean and median transfers made in the dictator game were 37% and 40% of the endowment, respectively.12 There is a discernible variation in the amount of endowment that was shared as seen in the non-parametric distribution plotted in Fig. A1 in Appendix A, and the distributions are similar by loan status (Fig. A2). Column 1 of Table 1 summarizes various individual and household characteristics of the participating migrants. By design, in our sample there is no variation in gender, place of origin, type of occupation, and marital status, and only little variation in religion and educational attainment. The average annual income earned in Qatar was the equivalent of US$6,456 of which 52% was remitted annually. The mean age was 40 years and the individuals were employed in Qatar for approximately 5.3 years. Only a quarter reported household members back in Kerala receiving an income, averaging US$327 in the past year. Table 1. Summary statistics   All  No loan  Loan  Difference  T-test        (1)  (2)  (3)  (2)–(3)  p-value  Migrant characteristics  Annual income (US$)  6,456  6,657  6,199  458  0.39  Annual remittances (US$)  3,387  3,414  3,353  61  0.81  Age in years  40.24  40.49  39.91  0.58  0.70  Post-high school (indicator)  0.10  0.12  0.07  0.05  0.36  Years in Qatar  5.29  5.69  4.76  0.93  0.41  Expected years in Qatar  5.47  5.29  5.70  −0.40  0.56  Altruism  0.37  0.37  0.37  0.00  0.92  Risk attitude  3.32  3.37  3.26  0.11  0.62  Household characteristics  Household income (indicator)  0.25  0.29  0.20  0.09  0.33  Household income (US$)  326.7  449.2  169.5  279.8  0.07  Wealth (US$)  31,029  28,161  34,708  −6,547  0.19  Landholdings (acres)  0.31  0.25  0.39  −0.15  0.11  Wife in SHG (indicator)  0.37  0.25  0.52  −0.27  0.00  Size (excluding the migrant)  3.70  3.81  3.54  0.27  0.16  Own home (indicator)  0.99  0.98  1.00  −0.02  0.38  Number of bedrooms  3.02  3.00  3.04  −0.04  0.72  Motor car (indicator)  0.05  0.05  0.04  0.01  0.86  Landline telephone (indicator)  0.62  0.71  0.50  0.21  0.03  Flat panel TV (indicator)  0.35  0.36  0.35  0.01  0.93  Refrigerator (indicator)  0.70  0.71  0.67  0.04  0.68  Computer (indicator)  0.06  0.05  0.07  −0.01  0.76  Number of Observations  105  59  46        All  No loan  Loan  Difference  T-test        (1)  (2)  (3)  (2)–(3)  p-value  Migrant characteristics  Annual income (US$)  6,456  6,657  6,199  458  0.39  Annual remittances (US$)  3,387  3,414  3,353  61  0.81  Age in years  40.24  40.49  39.91  0.58  0.70  Post-high school (indicator)  0.10  0.12  0.07  0.05  0.36  Years in Qatar  5.29  5.69  4.76  0.93  0.41  Expected years in Qatar  5.47  5.29  5.70  −0.40  0.56  Altruism  0.37  0.37  0.37  0.00  0.92  Risk attitude  3.32  3.37  3.26  0.11  0.62  Household characteristics  Household income (indicator)  0.25  0.29  0.20  0.09  0.33  Household income (US$)  326.7  449.2  169.5  279.8  0.07  Wealth (US$)  31,029  28,161  34,708  −6,547  0.19  Landholdings (acres)  0.31  0.25  0.39  −0.15  0.11  Wife in SHG (indicator)  0.37  0.25  0.52  −0.27  0.00  Size (excluding the migrant)  3.70  3.81  3.54  0.27  0.16  Own home (indicator)  0.99  0.98  1.00  −0.02  0.38  Number of bedrooms  3.02  3.00  3.04  −0.04  0.72  Motor car (indicator)  0.05  0.05  0.04  0.01  0.86  Landline telephone (indicator)  0.62  0.71  0.50  0.21  0.03  Flat panel TV (indicator)  0.35  0.36  0.35  0.01  0.93  Refrigerator (indicator)  0.70  0.71  0.67  0.04  0.68  Computer (indicator)  0.06  0.05  0.07  −0.01  0.76  Number of Observations  105  59  46      Notes: Household wealth refers to the total value of cash (in hand, banks, postal accounts), chitty funds, stocks, gold holdings and landholdings jointly held by the migrant and household members in Kerala. Only 74 observations were recorded for ‘Expected years in Qatar’ out of which 41 were for the No Loan group. 3.1 Loans One of the key findings of this study is that the possession of loan obligations can potentially affect remittance behaviour. Therefore, prior to presenting the empirical analysis we provide a short description of the loans and a comparison between the migrants with an explicit loan obligation and those without. In our sample, 46 migrants (44%) report having a loan obligation back home. Based on migrants’ and their households’ characteristics, we observe that these migrants are similar to the migrants who do not report any loan obligations (see columns 2 and 3 of Table 1). Both groups exhibit the same average levels of altruism (0.37 as a share of the endowment offered in the dictator game) and of risk aversion. Furthermore, we find that the two groups are similar in terms of not only the amounts remitted but also of how remittances are reportedly spent (see online Appendix B). In all but three variables (‘Household income’, ‘Wife in self-help group,’ and ‘Landline telephone’), the difference between the migrants with loans and those without is statistically insignificant at conventional levels. To further explore loan behaviour, we first observe that loans are primarily taken for housing investments. In fact, 76% of the migrants with a loan reported that it was taken to ‘buy a house, repair/build a house, or buy land’ (Fig. 2 shows migrants’ reasons for existing loans). Those that do not possess a loan share a similar preference for housing investments. When asked how the household plans to spend future savings, 73% of the migrants without a loan reported that they would like to ‘buy a house, repair/build a house, or buy land’ (see Fig. B1 in Appendix B), which is in line with studies on migrants’ target saving (e.g. Osili, 2004). If housing investments are a priority for both groups, a question arises as to why some migrant households obtained a loan while others did not. There could be differences in terms of personal and household characteristics or circumstances that, if not accounted for, would raise concerns about self-selection and omitted variable bias. A probit regression of loan take-up in Table 2 suggests that higher household income (weakly) and post-secondary schooling reduce the likelihood of taking a loan while having a wife who is a member of an SHG increases the probability of possessing a loan obligation.13 We also note that the migrant’s risk aversion does not matter for loan take-up. In the ensuing empirical analysis of remittance behaviour, we include the wife’s participation in a SHG and a host of personal and household characteristics to mitigate concerns about selectivity. Yet, we acknowledge that there may be unobserved characteristics of the migrant household that both explain loan behaviour and remittances and for which we cannot account. For instance, we possess no information on whether the recipient household is co-obligated with the migrant regarding the outstanding loan or even whether the family is aware of the loan. Table 2. Determinants of loan participation (probit)   (1)  (2)  (3)  Migrant’s income (ln)  −0.186  −0.195  −0.141    (0.121)  (0.125)  (0.169)  Household income (IHST)  −0.014  −0.013  −0.031*    (0.011)  (0.011)  (0.018)  Post-high school (indicator)  −0.203**  −0.166*  −0.080    (0.090)  (0.099)  (0.199)  Risk attitude measure  –  −0.025  −0.037    –  (0.029)  (0.044)  Altruism measure  –  –  0.059    –  –  (0.302)  Wife in SHG  0.160**  0.155**  0.261***    (0.079)  (0.082)  (0.101)  Landholdings (IHST)  0.093  0.080  0.062    (0.072)  (0.070)  (0.085)  Wealth (ln)  −0.0854  −0.071  −0.0217    (0.104)  (0.100)  (0.111)  Observations  203  191  105    (1)  (2)  (3)  Migrant’s income (ln)  −0.186  −0.195  −0.141    (0.121)  (0.125)  (0.169)  Household income (IHST)  −0.014  −0.013  −0.031*    (0.011)  (0.011)  (0.018)  Post-high school (indicator)  −0.203**  −0.166*  −0.080    (0.090)  (0.099)  (0.199)  Risk attitude measure  –  −0.025  −0.037    –  (0.029)  (0.044)  Altruism measure  –  –  0.059    –  –  (0.302)  Wife in SHG  0.160**  0.155**  0.261***    (0.079)  (0.082)  (0.101)  Landholdings (IHST)  0.093  0.080  0.062    (0.072)  (0.070)  (0.085)  Wealth (ln)  −0.0854  −0.071  −0.0217    (0.104)  (0.100)  (0.111)  Observations  203  191  105  Notes: The dependent variable is an indicator for having a loan repayment obligation. Coefficients show the marginal effects. Columns 1 and 2 use the full sample with the addition of the risk-attitude measure in column 2 (this measure could not be obtained from 12 individuals). Column 3 is limited to individuals who participated in campus visit and completed a set of behavioural games. Robust standard errors are displayed in parentheses. IHST stands for inverse hyperbolic sine transformation. *** p < 0.01, ** p < 0.05, * p < 0.1 Fig. 2. View largeDownload slide Reasons for existing loans Fig. 2. View largeDownload slide Reasons for existing loans 4. Empirical model We now turn to our study of how migrant characteristics, and particularly altruism, affect remittances. We first investigate how migrants’ background and household characteristics affect remittances. Consider a parsimonious remittance specification that excludes altruism:   ln⁡R=a+Xβ+Zγ+ɛ. (7) Above, ln⁡R is the log of the annual remittances; X is a vector of migrants’ background characteristics (income, age, risk attitude, and years employed in Qatar); Z is a vector of household characteristics (size, wife in SHG, income, and wealth), and ɛ is a normally distributed error term. A dummy variable is used to measure educational attainment, with a value of 1 if the migrant sought post high-school education, and 0, otherwise. The results for the entire sample of migrants participating in the behavioural games, the No Loan and Have Loan groups are reported in columns 1, 2, and 3 of Table 3, respectively. Table 3. Determinants of remittances (OLS)   All  No loan  Have loan        (1)  (2)  (3)  Migrant’s income (ln)  0.585***  0.571***  0.668***    (0.072)  (0.098)  (0.071)  Migrant’s age (ln)  0.132  0.275  0.071    (0.110)  (0.184)  (0.129)  Post-high school (indicator)  0.104  0.154  0.040    (0.092)  (0.128)  (0.150)  Years employed in Qatar (ln)  −0.027  0.009  −0.039    (0.029)  (0.036)  (0.034)  Household size in India  0.015  0.057  −0.016    (0.024)  (0.040)  (0.027)  Wife in SHG  0.043  0.044  0.021    (0.045)  (0.076)  (0.048)  Household income (IHST)  0.002  −0.007  0.020***    (0.011)  (0.013)  (0.007)  Wealth (ln)  0.034  −0.106  0.092    (0.057)  (0.086)  (0.073)  Landholdings (IHST)  −0.009  0.083  −0.063    (0.041)  (0.072)  (0.056)  Risk attitude measure  −0.010  −0.024  0.027    (0.025)  (0.033)  (0.018)  Constant  2.168**  2.694*  1.311    (0.919)  (1.515)  (0.831)  R-squared  0.54  0.46  0.78  Number of Observations  105  59  46    All  No loan  Have loan        (1)  (2)  (3)  Migrant’s income (ln)  0.585***  0.571***  0.668***    (0.072)  (0.098)  (0.071)  Migrant’s age (ln)  0.132  0.275  0.071    (0.110)  (0.184)  (0.129)  Post-high school (indicator)  0.104  0.154  0.040    (0.092)  (0.128)  (0.150)  Years employed in Qatar (ln)  −0.027  0.009  −0.039    (0.029)  (0.036)  (0.034)  Household size in India  0.015  0.057  −0.016    (0.024)  (0.040)  (0.027)  Wife in SHG  0.043  0.044  0.021    (0.045)  (0.076)  (0.048)  Household income (IHST)  0.002  −0.007  0.020***    (0.011)  (0.013)  (0.007)  Wealth (ln)  0.034  −0.106  0.092    (0.057)  (0.086)  (0.073)  Landholdings (IHST)  −0.009  0.083  −0.063    (0.041)  (0.072)  (0.056)  Risk attitude measure  −0.010  −0.024  0.027    (0.025)  (0.033)  (0.018)  Constant  2.168**  2.694*  1.311    (0.919)  (1.515)  (0.831)  R-squared  0.54  0.46  0.78  Number of Observations  105  59  46  Notes: The dependent variable is the log of remittances. Robust standard errors are displayed in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1 For the entire sample, the only statistically significant explanatory variable is the migrant’s income. A 10% increase in foreign income is associated with a 5.9% rise in remittances.14 Aside from the intercept, no other included variables have explanatory power. However, there appear to be differences for the two groups separately studied, as seen in columns 2 and 3. First, all the coefficients across the two groups are different.15 Second, for the group of migrants with loan obligations the household’s income has a positive effect on the amounts remitted by the migrant. Third, the R-squared coefficient for the group of migrants with loan obligations is much higher relative to their non-loan peers (0.78 vs 0.46), which suggests that the remittance behaviour is less predictable for the latter group. If we apply the indirect test of altruism based on household income, the previous analysis suggests that altruism does not seem to matter for the entire sample nor for each group.16 We proceed to investigate how robust this conclusion is by using our constructed measure of altruism. To further motivate our next specification that differentiates across the two groups of migrants, we provide scatter plots of altruism and remittances for each group in Fig. 3. We observe that for the migrants who report a monthly loan obligation (right diagram), remittances increase in altruism. However, for those without a loan obligation (left diagram), altruism is uncorrelated with remittances. Fig. 3. View largeDownload slide Scatter plot of altruism and remittances for migrants with and without loans Fig. 3. View largeDownload slide Scatter plot of altruism and remittances for migrants with and without loans Our basic specification that accounts for altruism is as follows:   ln⁡R=a0+a1ln⁡Y+a2LOAN+a3ALTR+a4ALTR*LOAN+ɛ, (8) where ln⁡Y is the log of the annual income of a migrant; LOAN is a dummy variable that takes a value of 1 if a migrant reports a monthly loan obligation and 0, otherwise; and ALTR, a proxy of altruism, is the share of the endowment (in decimals) that a migrant offers in the dictator game. Table 4 reports ordinary least squares (OLS) estimates for various models of specification (2). Columns 1–3 report the impact of altruism and of having a loan on remittance behaviour for the entire sample. Without an interaction between these two characteristics—i.e. when treating the entire pool of migrant workers as homogeneous—neither having a loan nor being altruistic matters. However, when we interact loan obligations with altruism, we find that both the loan dummy and the interaction term matter, as shown in column 4. To check robustness, we included the set of control variables that were originally part of specification (1) in Table 3. As shown in column 5, the magnitude and statistical significance of the loan variable and altruism–loan interaction variable do not change, and no other control variables are individually or jointly significant at conventional levels (F-test of joint significance yields a p-value of 0.68). Table 4. Remittances, loan obligations, and altruism   (1)  (2)  (3)  (4)  (5)  (6)  ln⁡Y  0.623***  0.617***  0.624***  0.614***  0.580***  0.633***    (0.082)  (0.077)  (0.075)  (0.069)  (0.061)  (0.181)  LOAN  0.036  –  0.036  −0.232**  −0.239**  −0.171    (0.038)    (0.039)  (0.094)  (0.112)  (0.108)  ALTR    −0.004  −0.009  −0.349  −0.401  1.641      (0.141)  (0.142)  (0.237)  (0.289)  (5.346)  ALTR*LOAN        0.727**  0.743***  0.484*          (0.283)  (0.317)  (0.288)  ALTR* ln⁡Y            −0.078              (0.512)  ALTR*Migrant’s age            0.252              (0.740)  ALTR*Post high school            0.332              (0.390)  ALTR*Years in Qatar            −0.217              (0.178)  ALTR*Hhold size            0.133              (0.169)  ALTR*Wife in SHG            0.464              (0.316)  ALTR*Hhold income            −0.043              (0.060)  ALTR*Wealth            −0.172              (0.342)  ALTR*Landholdings            −0.043              (0.258)  ALTR*Risk attitude            −0.189              (0.166)  Controls  No  No  No  No  Yes  Yes  R-squared  0.52  0.52  0.52  0.57  0.59  0.66  Observations  105  105  105  105  105  105                  (1)  (2)  (3)  (4)  (5)  (6)  ln⁡Y  0.623***  0.617***  0.624***  0.614***  0.580***  0.633***    (0.082)  (0.077)  (0.075)  (0.069)  (0.061)  (0.181)  LOAN  0.036  –  0.036  −0.232**  −0.239**  −0.171    (0.038)    (0.039)  (0.094)  (0.112)  (0.108)  ALTR    −0.004  −0.009  −0.349  −0.401  1.641      (0.141)  (0.142)  (0.237)  (0.289)  (5.346)  ALTR*LOAN        0.727**  0.743***  0.484*          (0.283)  (0.317)  (0.288)  ALTR* ln⁡Y            −0.078              (0.512)  ALTR*Migrant’s age            0.252              (0.740)  ALTR*Post high school            0.332              (0.390)  ALTR*Years in Qatar            −0.217              (0.178)  ALTR*Hhold size            0.133              (0.169)  ALTR*Wife in SHG            0.464              (0.316)  ALTR*Hhold income            −0.043              (0.060)  ALTR*Wealth            −0.172              (0.342)  ALTR*Landholdings            −0.043              (0.258)  ALTR*Risk attitude            −0.189              (0.166)  Controls  No  No  No  No  Yes  Yes  R-squared  0.52  0.52  0.52  0.57  0.59  0.66  Observations  105  105  105  105  105  105                Notes: The dependent variable is the log of remittances. The set of controls in columns 5 and 6 are those from Table 3. Robust standard errors are displayed in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1 Finally, in column 6 of Table 4 we include additional interactive terms between the set of control variables and the altruism measure. Besides mitigating concerns about omitted variables, the purpose is to see whether the impact of altruism on the group of migrants with loans stems from the possession of a loan per se rather than from other migrant characteristics correlated with having a loan. The coefficient on the altruism–loan interactive term remains stable, which suggests that explicit loan obligations matter for the effect of the altruistic motive.17 We also note that in column 6 the coefficient of the loan variable became somewhat smaller in the absolute value and is no longer significant, which can be explained by that the added control variables have some explanatory power on the decision to take a loan. Several observations can be made based on the empirical results. First of all, for the entire sample we find no relationship between altruism and remittances. But we find a positive and statistically significant relationship for the subsample of migrants with a loan obligation, unlike in the case when the indirect tests of altruism were applied. Using estimates from column 5 of Table 4, for the migrant with a loan obligation a 10 percentage point increase in the contribution made in the dictator game translates into a 3.4 percentage point higher remittance (p-value of 0.01). Furthermore, a negative coefficient on the dummy variable for loans indicates non-trivial differences in the amounts remitted across the two groups depending on the degree of altruism. If a migrant is selfish (0% contribution), then he remits by 20.9% less if he has a loan, but if he is of the average degree of altruism (a contribution of 37%), then he remits by 4.1% more if he has a loan.18 In other words, the remittance schedule in altruism estimated for the group of migrants with loans crosses the corresponding schedule of the other group from below, as shown in Fig. 1. As an additional robustness check, we also examine whether other variables, such as the ones in vectors X and Z in specification (1), have any explanatory power when interacted with the LOAN dummy. The results are reported in Table A2 in Appendix A. Each column represents a version of specification (2), in which the variable ALTR is replaced with a different variable. Aside from migrants’ own income, no other explanatory variable, including household income, in these regressions is statistically significant at conventional levels. 5. Discussion In this study, we find that altruism only seems to affect remittance behaviour for migrants with explicit loan obligations back home. We attribute the finding that altruism does not explain remittance behaviour for the entire sample to the fact that altruism can be offset by other motives for remittances or various confounding factors. In particular, we show that the observed linkage between altruism and loans can be consistent with the remittance model of altruism extended with reference-dependent preferences and loss-averse migrants. More generally, we suggest below that reference dependence can be a convenient way of modelling remittance behaviour. Utility theories of reference dependence are motivated by empirical evidence about the effects of contextual circumstances on the individual perception of utility. These effects are typically found to take the form of loss aversion with respect to some reference point, determined by the decision maker’s current position and expectations, as well as by social norms and comparisons. Regarding remittance behaviour, similarly to other realms of economic behaviour, there are strong reasons to believe that the subjective utility of remittances depends on contextual circumstances. Examples of such circumstances could be the history of remittances, individual or common beliefs about a ‘fair’ amount of remittances, recipients’ expectations, family or peer pressures, self-insurance and control motives for remittances. Each circumstance or motive can be characterized by whether they have an effect on one of three modelling blocks of reference dependence: (i) reference point, (ii) uncertainty; and (iii) loss aversion. For instance, social pressures and reciprocation can be attributed to the determinants of the level of reference point. According to Gardner (2012), a study on immigration to the Persian Gulf states, it is family pressures that are responsible for the decision to emigrate and for the amount of remittances to be sent home (also see Ilahi and Jafarey, 1999). The anthropological study of Osella and Osella (2000) describes a local status categorization of migrant workers from Kerala, based on their ability to earn money abroad, and privileges associated with high status. As another example, Chort et al. (2012) find that Senegalese migrants in France and Italy face ostracism and loss of access to services provided by a migrant network of their countrymen in the host country if remittances sent fall short of the expected norm. The degree of uncertainty about the reference point can be linked to the control motive for remittances. If remittances are sent toward a certain savings target, then migrants with a greater control over remittance uses will have more certainty about the reference point, which can have implications for their remitting behaviour as in Ashraf et al. (2015) or Batista et al. (2015).19 Lastly, the effects of the insurance motive on remittances as in Gubert (2002); Molina Millán (2015); Batista and Umblijs (2016) can be linked to the degree of loss aversion, i.e. the shape of ‘universal gain–loss function’. 6. Conclusion We study the relationship between altruism and remittances of migrants, using a measure of altruism elicited from a dictator game experiment. While we find little evidence of a universal relationship between altruism and remittances, we do observe a positive effect of altruism on remitting behaviour for migrants with loan obligations. Using the framework of reference-dependent preferences, we argue that the altruistic motive is subdued by uncertainty about remittance expectations and loss aversion. At the same time, the possession of loan obligations may reduce uncertainty about remittance expectations, subsequently making the altruistic motive more pronounced. Indirect tests may, however, fail to establish this relationship between altruism and remittances. More generally, we also suggest that our approach with reference dependence can be useful for modelling remittance behaviour. As a possible application, the remittance model with reference-dependent preferences could yield novel predictions on the effects of networks in explaining remittance behaviour. As a migrant worker spends a substantial amount of time interacting with his peers, the remittance behaviour of his peers, who most often come from the same community back home, may potentially influence expectations and reference levels and, through these, may also influence the migrant’s remittance behaviour. Supplementary material Supplementary material (the Appendix and data) is available online at the OUP website. Footnotes 1 Aside from its direct role on amounts remitted, altruism can also play a role in alleviating the problem of lack of enforcement in implicit familial contracts (Becker, 1991; Foster and Rosenzweig, 2001). 2 See Stark (1995), Rapoport and Docquier (2006) and Carling (2008) for reviews of altruism and other motives for remittances. 3 Further examples on altruism-confounding effects include de la Brière et al. (2002), Cox and Fafchamps (2007), Brown and Jimenez (2011), Yang (2011), McKenzie et al. (2013), De Arcangelis et al. (2015) and Batista and Narciso (2017). Section 5 provides further discussion. 4 For limitations of this approach toward measuring altruism, see Roth (1995), List (2007) and Andreoni et al. (2017). 5 For evidence from the psychology literature, see Jones and Rachlin (2006) and Rachlin and Jones (2008) and from economics—Roth (1995), Hoffman et al. (1996) and Bohnet and Frey (1999). 6 It is comparable with the average transfer of 39.6% in the dictator game of Batista et al. (2015), conducted in urban Mozambique, where the recipient is the closest person to the giver outside the giver’s household. 7 In the ultimatum game, the average transfer is 39% and not statistically different from the average in the dictator game. 8 Male interviewers would randomly select one migrant to be interviewed per room at each accommodation. We could not explicitly prohibit individuals of other faiths from participating but did so indirectly by stating that the campus visit would be on a Friday (which is a weekend day in Qatar), the day when church services were normally held in addition to the mid-day congregation prayer for Muslims. Some leeway was given to interviewers to enrol individuals with post-secondary education provided that this group formed a minority of the sample. 9 Subjects were asked to choose from one of six choices, each offering two possible rewards depending on a coin toss. The choices were: (i) 500 or 500; (ii) 450 or 950; (iii) 400 or 1200; (iv) 300 or 1500; (v) 100 or 1900; or (vi) 0 or 2000, where all the amounts are in Qatari Riyals. Lower numbered choices would reflect a greater degree of risk aversion. Workers were told that 40 of them (about 20%) would be selected at random to be paid for choices. 10 While our sample of 105 migrants is smaller than that in other studies of remittance behaviour (but see Osili, 2007 for a related study of the effects of altruism on remittances that uses a sample of similar size), our design employs behavioural tasks which take time to administer and for which subjects are paid. This necessarily limits our feasible sample size to one comparable to those used in laboratory experiments. 11 The English translation of the instructions of the dictator game from Malayalam, the local language of Kerala, is available upon request. Along with written instructions, the participants were also shown voiced PowerPoint presentations about the games played. Additionally, there were Malayalam-speaking assistants trained to help participants in better understanding the games. 12 These numbers fall within the range of mean and median offers observed in similar studies, which is between 10% and 50% (see Camerer, 2003, Table 2.4, pp. 57–8). 13 SHGs in Kerala typically consist of women who vouch for each other in fulfilling their loan obligations with a participating financial institution. Spouses involved in SHGs may have better access to loan opportunities relative to non-participating spouses. 14 Both remittances and income could be jointly determined. For example, the decision to migrate could be motivated by expectations about foreign income received and remittances that could be sent. 15 A Wald test of coefficient equality is rejected (p-value = 0.11). 16 A positive coefficient on the household’s income can instead be associated with exchange and investment motives (see Rapoport and Docquier, 2006, Table 2). 17 An F-test of the equality of the coefficient on ALTR*LOAN in columns 4 and 6, and also between columns 5 and 6 cannot be rejected (p-value of 0.41 and 0.37 respectively). 18 The results are qualitatively similar using coefficients from column 6 of Table 4. Evaluating at the mean of the control variables, a 10% increase in contribution made in the dictator game is associated with a 3.1 percentage point higher remittance for a migrant with a loan obligation. A migrant with a loan who is selfish sends 14.4% less, while those at the average degree of altruism remits by 2.4% more. 19 The possession of an explicit loan obligation could be related to the control motive and more certainty about remittance uses and, thus, the level of reference point. Acknowledgements We thank Henry Farber, Daniel Westbrook, two anonymous referees, and various seminar and conference audiences for valuable comments and suggestions. We also thank Vodafone Qatar for providing mobile Wi-Fi access that was used in the data collection process. References Ambler K. ( 2015) Don’t tell on me: experimental evidence of asymmetric information in transnational households, Journal of Development Economics , 113, 52– 69. Google Scholar CrossRef Search ADS PubMed  Amuedo-Dorantes C., Bansak C., Pozo S. ( 2005) On the remitting patterns of immigrants: evidence from Mexican survey data, Economic Review , 90, 37– 58. Andreoni J., Rao J.M., Trachtman H. ( 2017) Avoiding the ask: a field experiment on altruism, empathy, and charitable giving, Journal of Political Economy , 125, 625– 53. Google Scholar CrossRef Search ADS   Ashraf N., Aycinena D., Martínez A.C., Yang D. ( 2015) Savings in transnational households: a field experiment among migrants from El Salvador, Review of Economics and Statistics , 97, 332– 51. Google Scholar CrossRef Search ADS   Batista C., Narciso G. ( 2017) Migrant remittances and information flows: evidence from a field experiment, World Bank Economic Review , forthcoming. Batista C., Umblijs J. ( 2016) Do migrants send remittances as a way of self-insurance?, Oxford Economic Papers , 68, 108– 30. Google Scholar CrossRef Search ADS   Batista C., Silverman D., Yang D. ( 2015) Directed giving: evidence from an inter-household transfer experiment in Mozambique, Journal of Economic Behavior and Organization , 118, 2– 21. Google Scholar CrossRef Search ADS   Becker G.S. ( 1991) A Treatise on the Family , Harvard University Press, Cambridge, MA. Bohnet I., Frey B. ( 1999) The sound of silence in prisoner’s dilemma and dictator games, Journal of Economic Behavior and Organization , 38, 43– 57. Google Scholar CrossRef Search ADS   Bouhga-Hagbe J. ( 2006) Altruism and workers’ remittances: evidence from selected countries in the Middle East and Central Asia. IMF Working paper WP/06/130, Washington, DC. Google Scholar CrossRef Search ADS   Brown R.P.C., Jimenez E.V. ( 2011) Subjectively-assessed welfare and international remittances: evidence from Tonga, Journal of Development Studies , 47, 829– 45. Google Scholar CrossRef Search ADS   Camerer C. ( 2003) Behavioral game theory: experiments on strategic interaction , Princeton University Press, Princeton, NJ. Carling J. ( 2008) The determinants of migrant remittances, Oxford Review of Economic Policy , 24, 581– 98. Google Scholar CrossRef Search ADS   Chort I., Gubert F., Senne J.-N. ( 2012) Migrant networks as a basis for social control: remittance incentives among Senegalese in France and Italy, Regional Science and Urban Economics , 42, 858– 74. Google Scholar CrossRef Search ADS   Cox D., Fafchamps M. ( 2007) Extended family and kinship networks: economic insights and evolutionary directions, in Schultz T.P., Strauss J.A. (eds) Handbook of Development Economics , Volume 4, Elsevier-North Holland, Amsterdam, 3711– 84. Google Scholar CrossRef Search ADS   De Arcangelis G., Joxhe M., McKenzie D., Tiongson E., Yang D. ( 2015) Directing remittances to education with soft and hard commitments: evidence from a lab-in-the-field experiment and new product take-up among Filipino migrants in Rome, Journal of Economic Behavior and Organization , 111, 197– 208. Google Scholar CrossRef Search ADS   de la Brière B., Sadoulet E., de Janvry A., Lambert S. ( 2002) The roles of destination, gender, and household composition in explaining remittances: an analysis for the Dominican Sierra, Journal of Development Economics , 68, 309– 28. Google Scholar CrossRef Search ADS   Forsythe R., Horowitz J., Savin N.E., Sefton M. ( 1994) Fairness in simple bargaining experiments, Games and Economic Behavior , 6, 347– 69. Google Scholar CrossRef Search ADS   Foster A.D., Rosenzweig M.R. ( 2001) Imperfect commitment, altruism and the family: evidence from transfer behavior in low-income rural areas, Review of Economics and Statistics , 83, 389– 407. Google Scholar CrossRef Search ADS   Funkhouser E. ( 1995) Remittances from international migration: a comparison of El Salvador and Nicaragua, Review of Economics and Statistics , 77, 137– 46. Google Scholar CrossRef Search ADS   Gardner A. ( 2012) Why do they keep coming? Labor migrants in the Gulf States, in Kamrava M., Babar Z. (eds) Migrant Labor in the Persian Gulf , Hurst & Company, London, 41– 58. Gubert F. ( 2002) Do migrants insure those who stay behind? Evidence from the Kayes area (Western Mali), Oxford Development Studies , 30, 267– 87. Google Scholar CrossRef Search ADS   Hamilton W.D. ( 1964) The genetical evolution of social behaviour, I and II, Journal of Theoretical Biology , 7, 1– 52. Google Scholar CrossRef Search ADS PubMed  Hoddinott J. ( 1994) A model of migration and remittances applied to western Kenya, Oxford Economic Papers , 46, 459– 76. Google Scholar CrossRef Search ADS   Hoffman E., McCabe K., Smith V.L. ( 1996) Social distance and other regarding behavior in dictator games, American Economic Review , 86, 653– 60. Ilahi N., Jafarey S. ( 1999) Guest worker migration, remittances and the extended family: evidence from Pakistan, Journal of Development Economics , 58, 485– 512. Google Scholar CrossRef Search ADS PubMed  Kahneman D., Tversky A. ( 1979) Prospect theory: an analysis of decision under risk. Econometrica , 47, 263– 92. Google Scholar CrossRef Search ADS   Kőszegi B., Rabin M. ( 2006) A model of reference dependent preferences, Quarterly Journal of Economics , 121, 1133– 65. Google Scholar CrossRef Search ADS   Jones B., Rachlin H. ( 2006) Social discounting, Psychological Science , 17, 283– 6. Google Scholar CrossRef Search ADS PubMed  Joseph T., Nyarko Y., Wang S.-Y. ( 2015) Asymmetric information and remittances: evidence from matched administrative data., NBER Working Paper No. 20986, Cambridge, MA. Kaplan H., Gurven M. ( 2005) The natural history of human food sharing and cooperation: a review and a new multi-individual approach to the negotiation of norms, in Gintis H., Bowles S., Boyd R., Fehr E. (eds) Moral Sentiments and Material Interests: The Foundations of Cooperation in Economic Life , MIT Press, Cambridge, MA, 75– 113. List J. ( 2007) On the interpretation of giving in dictator games, Journal of Political Economy , 115, 482– 93. Google Scholar CrossRef Search ADS   Lucas R.E.B., Stark O. ( 1985) Motivations to remit: evidence from Botswana, Journal of Political Economy , 93, 901– 18. Google Scholar CrossRef Search ADS   McKenzie D., Gibson J., Stillman S. ( 2013) A land of milk and honey with streets paved with gold: do emigrants have over-optimistic expectations about incomes abroad?, Journal of Development Economics  102, 116– 27. Google Scholar CrossRef Search ADS   Molina Millán T. ( 2015) Regional migration, insurance and economic shocks: evidence from Nicaragua., IZA Discussion Paper No. 9494, Bonn. Osella F., Osella C. ( 2000) Migration, money and masculinity in Kerala, Journal of the Royal Anthropological Institute , 6, 117– 33. Google Scholar CrossRef Search ADS   Osili U.O. ( 2004) Migrants and housing investments: theory and evidence from Nigeria, Economic Development and Cultural Change , 52, 821– 49. Google Scholar CrossRef Search ADS   Osili U.O. ( 2007) Remittances and savings from international migration: theory and evidence using a matched sample, Journal of Development Economics , 83, 446– 65. Google Scholar CrossRef Search ADS   Poirine B. ( 1997) A theory of remittances as an implicit family loan arrangement, World Development , 25, 589– 611. Google Scholar CrossRef Search ADS   Rachlin H., Jones B. ( 2008) Altruism among relatives and non-relatives, Behavioural Processes , 79, 120– 3. Google Scholar CrossRef Search ADS PubMed  Rapoport H., Docquier F. ( 2006) The economics of migrants’ remittances, in Kolm S.-C., Ythier Mercier J. (eds) Handbook of the Economics of Giving, Altruism and Reciprocity , Volume 2, Elsevier-North Holland, Amsterdam, 1135– 98. Roth A. ( 1995) Bargaining experiments, in Kagel J.H., Roth A. (eds) Handbook of Experimental Economics , Princeton University Press, Princeton, NJ, 253– 348. Seshan G., Zubrickas R. ( 2017) Asymmetric information about migrant earnings and remittance flows, World Bank Economic Review , 31, 24– 43. Stark O. ( 1995) Altruism and Beyond: An Economic Analysis of Transfers and Exchanges Within Families and Groups , Cambridge University Press, Cambridge, MA. Google Scholar CrossRef Search ADS   Yang D. ( 2011) Migrant remittances, Journal of Economic Perspectives , 25, 129– 52. Google Scholar CrossRef Search ADS   © Oxford University Press 2017 All rights reserved http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Oxford Economic Papers Oxford University Press

Does altruism matter for remittances?

Loading next page...
 
/lp/ou_press/does-altruism-matter-for-remittances-woKVUAf0W1
Publisher
Oxford University Press
Copyright
© Oxford University Press 2017 All rights reserved
ISSN
0030-7653
eISSN
1464-3812
D.O.I.
10.1093/oep/gpx035
Publisher site
See Article on Publisher Site

Abstract

Abstract We provide a direct test of the impact of altruism on remittances. From a sample of Indian migrant workers in Qatar, we elicit the propensity to share with others from their responses in a dictator game and use it as a proxy for altruism. For the entire sample, we find that altruism does not seem to matter. However, we document a strong positive relationship between altruism and remittances for a subset of migrants with a loan obligation, whereas indirect tests of altruism, typically used in the literature, would fail to establish this relationship. We explain the role of loan obligations with a standard remittance model extended with reference-dependent preferences. 1. Introduction Altruism is commonly, if not routinely, viewed as one of the most important motives for remittances in transnational families.1 However, because of its intangibility the actual impact of altruism on remitting behaviour is difficult to evaluate and, therefore, its assessment is typically attempted by indirect tests. A well-known example of such a test is found in Lucas and Stark (1985). They compare two alternative models of remittances, where one model is based on altruism and the other on self-interest. While both models predict that remittances increase with the migrant’s income, only in the altruistic model remittances decline with the recipient household’s income. Thus, a test for altruism is whether remittances decrease in the household’s income or, if aggregated, the receiving country’s agricultural GDP as in Bouhga-Hagbe (2006) or, if proxied, the number of other family members in migration as in Funkhouser (1995). Such indirect tests, however, can lack validity in the presence of confounding factors. For instance, other motives for remittances are also consistent with a negative relationship between remittances and household income. Under the strategic motive for remittances, migrants can increase their remittances in response to decreased levels of income in the home country in order to ease pressures for others to emigrate, which could negatively affect migrant earnings. Data limitations such as the absence of longitudinal data also tend to restrict the investigator from ruling out the moral hazard motive due to the recipient household reducing their labour effort in response to remittances. Furthermore, the worsening of the recipient’s financial conditions can trigger the control motive for remittances, the role of which, as recently demonstrated in Ashraf et al. (2015) or Batista et al. (2015), can be sizable.2 Lastly, recent empirical evidence on the role of asymmetric information on remittance flows further exacerbates the confounding problem. For instance, Ambler (2015) shows that migrants remit more if their families observe their earnings, implying that the role of altruism may be overestimated under symmetric information (also see Joseph et al., 2015; Seshan and Zubrickas, 2017).3 Given the increasing economic significance of remittances and the assumed importance of altruism, the existing approach to the study of the altruistic motive is hardly satisfactory. In an attempt to overcome limitations of indirect tests, we provide a direct test of the impact of altruism on remittances using a measure of altruism borrowed from behavioural economics. We administered a survey and conducted a behavioural experiment with a sample of 105 married male migrant workers from Kerala, India, working in Qatar whose spouses reside in India. The behavioural experiment consisted of tasks measuring social preferences, including a dictator game in which each participant received 100 Qatari Riyals (approximately, US$27) and decided how much of that amount to give to another, anonymous participant. From their responses, we elicit the propensity to share with others and use it to proxy for each participant’s degree of altruism. Our main interest lies with whether the measured variation in altruism across migrants helps to explain the observed variation in remittance behaviour. The proposed measure of altruism is widely employed in behavioural economics and psychology (see Forsythe et al., 1994; Camerer, 2003). Arguably, giving in a dictator game is initiated by sympathy or empathy in the mind of the giver and is not driven by ‘indirect selfishness,’ e.g. stemming from mutual cooperation, norm adherence through sanctions or other motives as can be relevant for remitting behaviour.4 However, an obvious question arises whether a measure constructed from migrants’ propensity to give to their peers is adequate for the purpose of assessing the propensity to give to their own family. In particular, according to the kin-selection theory of Hamilton (1964), altruism will be greater with stronger degree of kinship. While we do acknowledge limitations of our approach toward measuring intrafamilial altruism and that our findings need to be interpreted with corresponding reservations, we argue below that our constructed measure of altruism is at least positively correlated with the migrants’ intrafamilial sharing propensity. First of all, as argued by Kaplan and Gurven (2005), it is more social connectedness than kinship per se that motivates altruistic behaviour among humans, and the level of altruism varies inversely with social distance irrespective of the degree of relatedness.5 In this regard and as part of our design, there is a high degree of social connectedness among our sampled migrants. They form a close social group as not only they all originate from the same region in India and are of very similar socio-economic and cultural backgrounds but also reside in the same dormitory-styled accommodations in Qatar for a substantial period of time. The average transfer of 37% that we observe in a dictator game indicates their high, by any standard, propensity of sharing with peers.6 Second, following the dictator game we also conducted an ultimatum game with the same pool of migrants. The ultimatum game is the extension of a dictator game where the proposed division of the endowment takes place only if the receiver approves it. This extension gives rise to strategic motives for greater sharing lest the receiver find the division unfair. However, we observe very similar patterns of giving both in the dictator and ultimatum games (see Fig. A1).7 This suggests that, in our sample, giving does not seem to be motivated by strategic considerations. In particular, under the ‘indirect selfishness’ argument that a migrant who is altruistic toward his family could exhibit selfish behaviour toward his peers to save more money for his family, we would expect to observe changes in giving behaviour between the ultimatum and dictator games, but we do not. In our empirical analysis, we find no relationship between an individual-level measure of altruism and the sending of remittances when the entire sample is considered. This finding confirms our previous discussion about the complex interdependence of various factors and motives for remittances that, unless accounted for, may diminish the effect of the altruistic motive. In this study, we also uncover one such confounding factor—the possession of a loan obligation. We find that remittances rise with the degree of altruism only for the migrants with a loan obligation (who make about a half of the sample). Specifically, the estimated remittance schedule has a smaller intercept and a larger coefficient of altruism for the migrants with a loan obligation than for those without. It is important to note that the indirect test of altruism based on the recipient household’s income does not lead to the same conclusion. We argue that our empirical findings related to the possession of loan obligations are in line with the basic remittance model of altruism (Lucas and Stark, 1985; Stark, 1995, Ch.1) extended with reference-dependent preferences. In particular, we introduce a reference point for remittances, similar to benchmark remittances in Hoddinott (1994), against which a migrant’s utility from remittance is measured. The reference point can be thought of as a savings target or the amount of money that the migrant is expected to transfer home, as determined by existing obligations, family needs, or social comparisons. See Osili (2004); Amuedo-Dorantes et al. (2005); Yang (2011) for migrants’ target saving, which is particularly relevant for temporary, contract-based migration applicable to our sample. Lastly, we assume loss aversion (Kahneman and Tversky, 1979), where a migrant experiences a positive utility if his remittance is above the reference point and a negative utility if below, with losses looming larger than gains. In essence, our proposed model shows that altruism is subdued when a migrant is confronted with uncertainty over the amount of remittances to send. By contrast, a reduction of this uncertainty allows the altruistic motive to surface and more prominently influence the remitter’s behaviour akin to conventional models of remittances without uncertainty. Specifically, if migrants without loan obligations face more uncertainty about the reference point for remittances, an assumption that we justify later, we show that the theoretical predictions of the model closely match our empirical results. An increase in uncertainty about the reference point raises the risk and resultant disutility of falling short of the expectations back home if the amount remitted is relatively low, thus, prompting the loss-averse migrant to increase his remittance. But if the amount initially remitted is sufficiently large so that the risk of undershooting the reference point is negligible and is rather dominated by the risk of overshooting the reference point, then the relationship between remittances and uncertainty turns negative due to the diminishing utility of remittances past the reference point. Thus, given a positive relationship between remittance and altruism all else equal, the model predicts a flatter remittance schedule in altruism for migrants with less certainty about remittance expectations, i.e. without loan obligations, in accordance with our empirical results. All in all, the altruistic motive can be diminished by loss aversion coupled with uncertainty about remittance expectations. The remainder of the paper is organized as follows. In Section 2, we present the remittance model of altruism extended with reference-dependent preferences. In Section 3, we describe the data and compare the characteristics of migrants that report an explicit loan obligation to those that report none. In Section 4, we present our empirical findings, using which we discuss the role of reference dependence for modelling remittance behaviour in Section 5. The last section concludes the study. 2. Reference dependence and uncertainty Our theoretical approach builds on the remittance model of altruism (Lucas and Stark, 1985; Stark, 1995, Ch.1) extended with reference dependence. In the model, a migrant worker derives utility from own consumption and from his ability to meet a reference point for remittances, e.g. a savings target. The migrant experiences a psychological cost or gain if his remittance is below or, respectively, above the reference point, with losses looming larger than gains. We assume that the intensity of the psychological factor is in proportion to the migrant’s degree of altruism. As in Kőszegi and Rabin (2006), the reference point can be, however, uncertain, in which case the degree of uncertainty becomes an important factor for remittance decisions. Here, we take a reference point for remittances and its uncertainty as exogenously given. But see Seshan and Zubrickas (2017), where a threshold for remittances, equivalent to a reference point, is endogenously determined. In particular, the optimal (implicit) relationship contract between the migrant and the recipient household features a threshold for remittances if migrant earnings are private information. The threshold for remittances is determined by the degree of informational asymmetry, expressed by the cost of information acquisition, and distribution for earnings. Drawing on these findings, in this paper uncertainty about the reference point can be motivated by discrepancy in the recipient’s expectations about migrant earnings and the migrant’s beliefs about the recipient’s expectations. 2.1 Model Consider a migrant worker who has to divide his earned income Y > 0 between remittance R ≥ 0 sent to his family and private consumption, Y – R ≥ 0. The migrant’s utility from private consumption is given by function u(.) that satisfies u′(.)>0 and u′′(.)<0. The migrant’s utility from remittance R captures the family’s welfare and their expectations and is given by an increasing function μ(R−R¯), where R¯ is a reference point for remittances. We assume that μ(.) satisfies the properties of a ‘universal gain–loss function’ in Kőszegi and Rabin (2006). Specifically, the migrant experiences a negative utility from remittance if R<R¯ and a positive utility if R>R¯, where utility losses resonate more than gains, i.e. μ″(x)>0 for x < 0 but μ″(x)<0 for x ≥ 0. The reference point R¯ is uncertain and, for analytical convenience, assumed to be uniformly distributed over an interval [r−e,r+e]. The parameter e, 0<e<r, measures the migrant’s uncertainty about the reference point. Letting θ > 0 denote the migrant’s degree of altruism, measured as the weight the migrant puts on his utility from remittance, we write the migrant’s total utility as:   U(R;Y,e)=u(Y−R)+θ∫r−er+eμ(R−R¯)12edR¯. (1) The optimal level of remittance R∗=arg⁡max⁡RU(R;Y,e) is determined, assuming the interior solution exists, by the first-order condition:   −u′(Y−R∗)+θ2e∫r−er+eμ′(R∗−R¯)dR¯=0, (2) which, by the fundamental theorem of calculus, can be expressed as:   −u′(Y−R∗)+θ2e(μ(R∗−r+e)−μ(R∗−r−e))=0. (3) In other words, at the optimal level of remittance the marginal utility of private consumption is equal to the marginal utility from the remittance. Using that the second-order condition (SOC) has to hold, the internal derivative dR∗/dθ>0, taken from (3), implies that a higher degree of altruism results in larger remittances, and vice versa. Now suppose that the migrant becomes less certain about the reference point R¯, which we model by an increase in uncertainty parameter e. The effect of increased uncertainty on remittance is given by the internal derivative:   dR∗de=2u′(Y−R∗)−θ(μ′(R∗−r+e)+μ′(R∗−r−e))SOC. (4) Rewrite this derivative using (3) as:   dR∗de=θe(Δ(e)−Δ(−e))−SOC, (5) where:   Δ(e)=eμ′(R∗−r+e)−μ(R∗−r+e). (6) The denominator of (5) is positive, but the numerator can be both positive and negative. In particular, it depends on the size of R*, which is to say on the degree of altruism θ. Technically, at smaller values of R∗ (<r−e) so that μ(.) is convex, the numerator is positive because Δ′(e)>0. But at larger values of R∗ (>r+e) so that μ(.) is concave, the numerator turns negative because now Δ′(e)<0. Then, by continuity we obtain a flatter remittance schedule when the degree of uncertainty increases. Intuitively, consider a migrant who remits a relatively small amount due to his relatively low degree of altruism. Because of uncertainty there is a risk that the amount remitted falls short of the reference point. An increase in uncertainty also increases this risk, which prompts a loss-averse migrant to remit more as a hedging measure against the increased risk. Now consider a migrant who initially remits a large amount due to his high degree of altruism. The risk of falling short of the reference point is negligible even with an increased degree of uncertainty, but at the same time the risk of overshooting the reference point becomes significant. Because of the diminishing marginal utility from remittances above the reference point, an increase in uncertainty prompts a migrant to favour more private consumption and, as a result, to lower remittances. As an illustration, suppose that the reference point R¯ is certain and that the migrant’s initial remittance R* exceeds it by 100 riyals with the marginal utility from the remittance being θμ′(100). Let the reference point be no longer certain so that the initial remittance R* can exceed it by either 50 or 150 riyals with equal probability. Because of the diminishing marginal utility from remittances past the reference point, the expected marginal utility from the initial remittance becomes smaller, 0.5θ(μ′(50)+μ′(150))<θμ′(100). By condition (3), the migrant’s optimal response to the reduced marginal utility from remittances is to reduce his initial remittance in favour of more private consumption, which would put the marginal utilities of private consumption and from remittances back in balance. The opposite effect holds when the initial remittance is relatively low so that an increase in uncertainty increases the expected marginal utility from the remittance. 2.2 Loan obligations and uncertainty We assume that possessing a loan obligation reduces uncertainty about the reference point for remittances. This assumption can be motivated in two ways. First, in our sample of migrants we observe that about half possess an explicit loan obligation. Despite this difference, as detailed in Section 3, the migrants with an explicit loan obligation and those without earn and remit similar amounts and have otherwise very similar socioeconomic backgrounds and consumption preferences. In particular, both types report the same purpose of their loans or savings/remittances, which is housing-related. To the extent that achieving this purpose is a major influence on how much should be remitted, the presence of a loan obligation provides greater certainty about the reference point for remittances. Second, and more generally, even if not exposed to explicit loan obligations, migrants are frequently bound by implicit family loan contracts, typically made to finance migration costs (for the relevance of such implicit loan contracts, see Lucas and Stark, 1985; Hoddinott, 1994; Poirine, 1997). Remittances are then considered as dividends from the family investment, but explicit financial requirements—which can include a loan obligation—create greater certainty about the exact size of expected dividends for which a migrant is responsible. Assume that migrants without a loan obligation face more uncertainty about the reference point. Based on our theoretical analysis, we hypothesize: Hypothesis 1 The remittance schedule of migrants with loan obligations has a lower intercept but a larger coefficient of altruism than that of migrants without loan obligations.This hypothesis is illustrated in Fig. 1. Fig. 1. View largeDownload slide Predicted remittance schedule in altruism Fig. 1. View largeDownload slide Predicted remittance schedule in altruism 3. Data Our data come from two sources. First, data on migrants’ personal and household characteristics, income, remittances, assets, consumption, and loans were collected through a survey administered to a sample of migrant workers in Doha, Qatar. To limit heterogeneity, we screened for male, married workers from Kerala, India, whose spouses remained behind and who were of Hindu faith and had at most a high-school diploma. A working sample of 203 individuals from across seven dormitory-styled accommodations located in Doha’s Industrial Area completed a baseline survey in April and May 2012, conducted by a local survey firm staffed with migrants from Kerala.8 Furthermore, to better control for individual heterogeneity in subsequent empirical analysis, we constructed a measure of risk attitude from the surveyed migrants’ responses to exercises on their attitude toward risk, which involved real payouts.9 Our second source of data comes from the behavioural experiment we conducted with a subset of the migrant workers who had completed the survey. Specifically, a few weeks after the surveys were completed, we invited all migrants who took part in the survey to the campus of the Georgetown University in Doha on a weekend to participate in a series of behavioural games. Transportation was provided. A total of 105 migrants accepted the invitation. Table A1 in online Appendix A shows that migrants who visited the campus were on average slightly older, less likely to have post-secondary schooling, more likely to have loans and have wives participating in self-help groups (SHGs) and with lower household income in India. We control for these observed differences in the empirical analysis of remittance behaviour.10 On campus, the migrants played public good, trust, dictator, and ultimatum games with real payouts based on earnings in one randomly selected game. For the present study, which focuses on altruism, we mainly consider behaviour in the dictator game. In this game, the participants were randomly matched in pairs, but the identity of the paired participant was never revealed. Each participant was then asked to decide how much of a 100 Qatari Riyal endowment (approximately US$27, the equivalent of about two-days worth of the average migrant’s salary) he would share with his anonymous partner.11 The mean and median transfers made in the dictator game were 37% and 40% of the endowment, respectively.12 There is a discernible variation in the amount of endowment that was shared as seen in the non-parametric distribution plotted in Fig. A1 in Appendix A, and the distributions are similar by loan status (Fig. A2). Column 1 of Table 1 summarizes various individual and household characteristics of the participating migrants. By design, in our sample there is no variation in gender, place of origin, type of occupation, and marital status, and only little variation in religion and educational attainment. The average annual income earned in Qatar was the equivalent of US$6,456 of which 52% was remitted annually. The mean age was 40 years and the individuals were employed in Qatar for approximately 5.3 years. Only a quarter reported household members back in Kerala receiving an income, averaging US$327 in the past year. Table 1. Summary statistics   All  No loan  Loan  Difference  T-test        (1)  (2)  (3)  (2)–(3)  p-value  Migrant characteristics  Annual income (US$)  6,456  6,657  6,199  458  0.39  Annual remittances (US$)  3,387  3,414  3,353  61  0.81  Age in years  40.24  40.49  39.91  0.58  0.70  Post-high school (indicator)  0.10  0.12  0.07  0.05  0.36  Years in Qatar  5.29  5.69  4.76  0.93  0.41  Expected years in Qatar  5.47  5.29  5.70  −0.40  0.56  Altruism  0.37  0.37  0.37  0.00  0.92  Risk attitude  3.32  3.37  3.26  0.11  0.62  Household characteristics  Household income (indicator)  0.25  0.29  0.20  0.09  0.33  Household income (US$)  326.7  449.2  169.5  279.8  0.07  Wealth (US$)  31,029  28,161  34,708  −6,547  0.19  Landholdings (acres)  0.31  0.25  0.39  −0.15  0.11  Wife in SHG (indicator)  0.37  0.25  0.52  −0.27  0.00  Size (excluding the migrant)  3.70  3.81  3.54  0.27  0.16  Own home (indicator)  0.99  0.98  1.00  −0.02  0.38  Number of bedrooms  3.02  3.00  3.04  −0.04  0.72  Motor car (indicator)  0.05  0.05  0.04  0.01  0.86  Landline telephone (indicator)  0.62  0.71  0.50  0.21  0.03  Flat panel TV (indicator)  0.35  0.36  0.35  0.01  0.93  Refrigerator (indicator)  0.70  0.71  0.67  0.04  0.68  Computer (indicator)  0.06  0.05  0.07  −0.01  0.76  Number of Observations  105  59  46        All  No loan  Loan  Difference  T-test        (1)  (2)  (3)  (2)–(3)  p-value  Migrant characteristics  Annual income (US$)  6,456  6,657  6,199  458  0.39  Annual remittances (US$)  3,387  3,414  3,353  61  0.81  Age in years  40.24  40.49  39.91  0.58  0.70  Post-high school (indicator)  0.10  0.12  0.07  0.05  0.36  Years in Qatar  5.29  5.69  4.76  0.93  0.41  Expected years in Qatar  5.47  5.29  5.70  −0.40  0.56  Altruism  0.37  0.37  0.37  0.00  0.92  Risk attitude  3.32  3.37  3.26  0.11  0.62  Household characteristics  Household income (indicator)  0.25  0.29  0.20  0.09  0.33  Household income (US$)  326.7  449.2  169.5  279.8  0.07  Wealth (US$)  31,029  28,161  34,708  −6,547  0.19  Landholdings (acres)  0.31  0.25  0.39  −0.15  0.11  Wife in SHG (indicator)  0.37  0.25  0.52  −0.27  0.00  Size (excluding the migrant)  3.70  3.81  3.54  0.27  0.16  Own home (indicator)  0.99  0.98  1.00  −0.02  0.38  Number of bedrooms  3.02  3.00  3.04  −0.04  0.72  Motor car (indicator)  0.05  0.05  0.04  0.01  0.86  Landline telephone (indicator)  0.62  0.71  0.50  0.21  0.03  Flat panel TV (indicator)  0.35  0.36  0.35  0.01  0.93  Refrigerator (indicator)  0.70  0.71  0.67  0.04  0.68  Computer (indicator)  0.06  0.05  0.07  −0.01  0.76  Number of Observations  105  59  46      Notes: Household wealth refers to the total value of cash (in hand, banks, postal accounts), chitty funds, stocks, gold holdings and landholdings jointly held by the migrant and household members in Kerala. Only 74 observations were recorded for ‘Expected years in Qatar’ out of which 41 were for the No Loan group. 3.1 Loans One of the key findings of this study is that the possession of loan obligations can potentially affect remittance behaviour. Therefore, prior to presenting the empirical analysis we provide a short description of the loans and a comparison between the migrants with an explicit loan obligation and those without. In our sample, 46 migrants (44%) report having a loan obligation back home. Based on migrants’ and their households’ characteristics, we observe that these migrants are similar to the migrants who do not report any loan obligations (see columns 2 and 3 of Table 1). Both groups exhibit the same average levels of altruism (0.37 as a share of the endowment offered in the dictator game) and of risk aversion. Furthermore, we find that the two groups are similar in terms of not only the amounts remitted but also of how remittances are reportedly spent (see online Appendix B). In all but three variables (‘Household income’, ‘Wife in self-help group,’ and ‘Landline telephone’), the difference between the migrants with loans and those without is statistically insignificant at conventional levels. To further explore loan behaviour, we first observe that loans are primarily taken for housing investments. In fact, 76% of the migrants with a loan reported that it was taken to ‘buy a house, repair/build a house, or buy land’ (Fig. 2 shows migrants’ reasons for existing loans). Those that do not possess a loan share a similar preference for housing investments. When asked how the household plans to spend future savings, 73% of the migrants without a loan reported that they would like to ‘buy a house, repair/build a house, or buy land’ (see Fig. B1 in Appendix B), which is in line with studies on migrants’ target saving (e.g. Osili, 2004). If housing investments are a priority for both groups, a question arises as to why some migrant households obtained a loan while others did not. There could be differences in terms of personal and household characteristics or circumstances that, if not accounted for, would raise concerns about self-selection and omitted variable bias. A probit regression of loan take-up in Table 2 suggests that higher household income (weakly) and post-secondary schooling reduce the likelihood of taking a loan while having a wife who is a member of an SHG increases the probability of possessing a loan obligation.13 We also note that the migrant’s risk aversion does not matter for loan take-up. In the ensuing empirical analysis of remittance behaviour, we include the wife’s participation in a SHG and a host of personal and household characteristics to mitigate concerns about selectivity. Yet, we acknowledge that there may be unobserved characteristics of the migrant household that both explain loan behaviour and remittances and for which we cannot account. For instance, we possess no information on whether the recipient household is co-obligated with the migrant regarding the outstanding loan or even whether the family is aware of the loan. Table 2. Determinants of loan participation (probit)   (1)  (2)  (3)  Migrant’s income (ln)  −0.186  −0.195  −0.141    (0.121)  (0.125)  (0.169)  Household income (IHST)  −0.014  −0.013  −0.031*    (0.011)  (0.011)  (0.018)  Post-high school (indicator)  −0.203**  −0.166*  −0.080    (0.090)  (0.099)  (0.199)  Risk attitude measure  –  −0.025  −0.037    –  (0.029)  (0.044)  Altruism measure  –  –  0.059    –  –  (0.302)  Wife in SHG  0.160**  0.155**  0.261***    (0.079)  (0.082)  (0.101)  Landholdings (IHST)  0.093  0.080  0.062    (0.072)  (0.070)  (0.085)  Wealth (ln)  −0.0854  −0.071  −0.0217    (0.104)  (0.100)  (0.111)  Observations  203  191  105    (1)  (2)  (3)  Migrant’s income (ln)  −0.186  −0.195  −0.141    (0.121)  (0.125)  (0.169)  Household income (IHST)  −0.014  −0.013  −0.031*    (0.011)  (0.011)  (0.018)  Post-high school (indicator)  −0.203**  −0.166*  −0.080    (0.090)  (0.099)  (0.199)  Risk attitude measure  –  −0.025  −0.037    –  (0.029)  (0.044)  Altruism measure  –  –  0.059    –  –  (0.302)  Wife in SHG  0.160**  0.155**  0.261***    (0.079)  (0.082)  (0.101)  Landholdings (IHST)  0.093  0.080  0.062    (0.072)  (0.070)  (0.085)  Wealth (ln)  −0.0854  −0.071  −0.0217    (0.104)  (0.100)  (0.111)  Observations  203  191  105  Notes: The dependent variable is an indicator for having a loan repayment obligation. Coefficients show the marginal effects. Columns 1 and 2 use the full sample with the addition of the risk-attitude measure in column 2 (this measure could not be obtained from 12 individuals). Column 3 is limited to individuals who participated in campus visit and completed a set of behavioural games. Robust standard errors are displayed in parentheses. IHST stands for inverse hyperbolic sine transformation. *** p < 0.01, ** p < 0.05, * p < 0.1 Fig. 2. View largeDownload slide Reasons for existing loans Fig. 2. View largeDownload slide Reasons for existing loans 4. Empirical model We now turn to our study of how migrant characteristics, and particularly altruism, affect remittances. We first investigate how migrants’ background and household characteristics affect remittances. Consider a parsimonious remittance specification that excludes altruism:   ln⁡R=a+Xβ+Zγ+ɛ. (7) Above, ln⁡R is the log of the annual remittances; X is a vector of migrants’ background characteristics (income, age, risk attitude, and years employed in Qatar); Z is a vector of household characteristics (size, wife in SHG, income, and wealth), and ɛ is a normally distributed error term. A dummy variable is used to measure educational attainment, with a value of 1 if the migrant sought post high-school education, and 0, otherwise. The results for the entire sample of migrants participating in the behavioural games, the No Loan and Have Loan groups are reported in columns 1, 2, and 3 of Table 3, respectively. Table 3. Determinants of remittances (OLS)   All  No loan  Have loan        (1)  (2)  (3)  Migrant’s income (ln)  0.585***  0.571***  0.668***    (0.072)  (0.098)  (0.071)  Migrant’s age (ln)  0.132  0.275  0.071    (0.110)  (0.184)  (0.129)  Post-high school (indicator)  0.104  0.154  0.040    (0.092)  (0.128)  (0.150)  Years employed in Qatar (ln)  −0.027  0.009  −0.039    (0.029)  (0.036)  (0.034)  Household size in India  0.015  0.057  −0.016    (0.024)  (0.040)  (0.027)  Wife in SHG  0.043  0.044  0.021    (0.045)  (0.076)  (0.048)  Household income (IHST)  0.002  −0.007  0.020***    (0.011)  (0.013)  (0.007)  Wealth (ln)  0.034  −0.106  0.092    (0.057)  (0.086)  (0.073)  Landholdings (IHST)  −0.009  0.083  −0.063    (0.041)  (0.072)  (0.056)  Risk attitude measure  −0.010  −0.024  0.027    (0.025)  (0.033)  (0.018)  Constant  2.168**  2.694*  1.311    (0.919)  (1.515)  (0.831)  R-squared  0.54  0.46  0.78  Number of Observations  105  59  46    All  No loan  Have loan        (1)  (2)  (3)  Migrant’s income (ln)  0.585***  0.571***  0.668***    (0.072)  (0.098)  (0.071)  Migrant’s age (ln)  0.132  0.275  0.071    (0.110)  (0.184)  (0.129)  Post-high school (indicator)  0.104  0.154  0.040    (0.092)  (0.128)  (0.150)  Years employed in Qatar (ln)  −0.027  0.009  −0.039    (0.029)  (0.036)  (0.034)  Household size in India  0.015  0.057  −0.016    (0.024)  (0.040)  (0.027)  Wife in SHG  0.043  0.044  0.021    (0.045)  (0.076)  (0.048)  Household income (IHST)  0.002  −0.007  0.020***    (0.011)  (0.013)  (0.007)  Wealth (ln)  0.034  −0.106  0.092    (0.057)  (0.086)  (0.073)  Landholdings (IHST)  −0.009  0.083  −0.063    (0.041)  (0.072)  (0.056)  Risk attitude measure  −0.010  −0.024  0.027    (0.025)  (0.033)  (0.018)  Constant  2.168**  2.694*  1.311    (0.919)  (1.515)  (0.831)  R-squared  0.54  0.46  0.78  Number of Observations  105  59  46  Notes: The dependent variable is the log of remittances. Robust standard errors are displayed in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1 For the entire sample, the only statistically significant explanatory variable is the migrant’s income. A 10% increase in foreign income is associated with a 5.9% rise in remittances.14 Aside from the intercept, no other included variables have explanatory power. However, there appear to be differences for the two groups separately studied, as seen in columns 2 and 3. First, all the coefficients across the two groups are different.15 Second, for the group of migrants with loan obligations the household’s income has a positive effect on the amounts remitted by the migrant. Third, the R-squared coefficient for the group of migrants with loan obligations is much higher relative to their non-loan peers (0.78 vs 0.46), which suggests that the remittance behaviour is less predictable for the latter group. If we apply the indirect test of altruism based on household income, the previous analysis suggests that altruism does not seem to matter for the entire sample nor for each group.16 We proceed to investigate how robust this conclusion is by using our constructed measure of altruism. To further motivate our next specification that differentiates across the two groups of migrants, we provide scatter plots of altruism and remittances for each group in Fig. 3. We observe that for the migrants who report a monthly loan obligation (right diagram), remittances increase in altruism. However, for those without a loan obligation (left diagram), altruism is uncorrelated with remittances. Fig. 3. View largeDownload slide Scatter plot of altruism and remittances for migrants with and without loans Fig. 3. View largeDownload slide Scatter plot of altruism and remittances for migrants with and without loans Our basic specification that accounts for altruism is as follows:   ln⁡R=a0+a1ln⁡Y+a2LOAN+a3ALTR+a4ALTR*LOAN+ɛ, (8) where ln⁡Y is the log of the annual income of a migrant; LOAN is a dummy variable that takes a value of 1 if a migrant reports a monthly loan obligation and 0, otherwise; and ALTR, a proxy of altruism, is the share of the endowment (in decimals) that a migrant offers in the dictator game. Table 4 reports ordinary least squares (OLS) estimates for various models of specification (2). Columns 1–3 report the impact of altruism and of having a loan on remittance behaviour for the entire sample. Without an interaction between these two characteristics—i.e. when treating the entire pool of migrant workers as homogeneous—neither having a loan nor being altruistic matters. However, when we interact loan obligations with altruism, we find that both the loan dummy and the interaction term matter, as shown in column 4. To check robustness, we included the set of control variables that were originally part of specification (1) in Table 3. As shown in column 5, the magnitude and statistical significance of the loan variable and altruism–loan interaction variable do not change, and no other control variables are individually or jointly significant at conventional levels (F-test of joint significance yields a p-value of 0.68). Table 4. Remittances, loan obligations, and altruism   (1)  (2)  (3)  (4)  (5)  (6)  ln⁡Y  0.623***  0.617***  0.624***  0.614***  0.580***  0.633***    (0.082)  (0.077)  (0.075)  (0.069)  (0.061)  (0.181)  LOAN  0.036  –  0.036  −0.232**  −0.239**  −0.171    (0.038)    (0.039)  (0.094)  (0.112)  (0.108)  ALTR    −0.004  −0.009  −0.349  −0.401  1.641      (0.141)  (0.142)  (0.237)  (0.289)  (5.346)  ALTR*LOAN        0.727**  0.743***  0.484*          (0.283)  (0.317)  (0.288)  ALTR* ln⁡Y            −0.078              (0.512)  ALTR*Migrant’s age            0.252              (0.740)  ALTR*Post high school            0.332              (0.390)  ALTR*Years in Qatar            −0.217              (0.178)  ALTR*Hhold size            0.133              (0.169)  ALTR*Wife in SHG            0.464              (0.316)  ALTR*Hhold income            −0.043              (0.060)  ALTR*Wealth            −0.172              (0.342)  ALTR*Landholdings            −0.043              (0.258)  ALTR*Risk attitude            −0.189              (0.166)  Controls  No  No  No  No  Yes  Yes  R-squared  0.52  0.52  0.52  0.57  0.59  0.66  Observations  105  105  105  105  105  105                  (1)  (2)  (3)  (4)  (5)  (6)  ln⁡Y  0.623***  0.617***  0.624***  0.614***  0.580***  0.633***    (0.082)  (0.077)  (0.075)  (0.069)  (0.061)  (0.181)  LOAN  0.036  –  0.036  −0.232**  −0.239**  −0.171    (0.038)    (0.039)  (0.094)  (0.112)  (0.108)  ALTR    −0.004  −0.009  −0.349  −0.401  1.641      (0.141)  (0.142)  (0.237)  (0.289)  (5.346)  ALTR*LOAN        0.727**  0.743***  0.484*          (0.283)  (0.317)  (0.288)  ALTR* ln⁡Y            −0.078              (0.512)  ALTR*Migrant’s age            0.252              (0.740)  ALTR*Post high school            0.332              (0.390)  ALTR*Years in Qatar            −0.217              (0.178)  ALTR*Hhold size            0.133              (0.169)  ALTR*Wife in SHG            0.464              (0.316)  ALTR*Hhold income            −0.043              (0.060)  ALTR*Wealth            −0.172              (0.342)  ALTR*Landholdings            −0.043              (0.258)  ALTR*Risk attitude            −0.189              (0.166)  Controls  No  No  No  No  Yes  Yes  R-squared  0.52  0.52  0.52  0.57  0.59  0.66  Observations  105  105  105  105  105  105                Notes: The dependent variable is the log of remittances. The set of controls in columns 5 and 6 are those from Table 3. Robust standard errors are displayed in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1 Finally, in column 6 of Table 4 we include additional interactive terms between the set of control variables and the altruism measure. Besides mitigating concerns about omitted variables, the purpose is to see whether the impact of altruism on the group of migrants with loans stems from the possession of a loan per se rather than from other migrant characteristics correlated with having a loan. The coefficient on the altruism–loan interactive term remains stable, which suggests that explicit loan obligations matter for the effect of the altruistic motive.17 We also note that in column 6 the coefficient of the loan variable became somewhat smaller in the absolute value and is no longer significant, which can be explained by that the added control variables have some explanatory power on the decision to take a loan. Several observations can be made based on the empirical results. First of all, for the entire sample we find no relationship between altruism and remittances. But we find a positive and statistically significant relationship for the subsample of migrants with a loan obligation, unlike in the case when the indirect tests of altruism were applied. Using estimates from column 5 of Table 4, for the migrant with a loan obligation a 10 percentage point increase in the contribution made in the dictator game translates into a 3.4 percentage point higher remittance (p-value of 0.01). Furthermore, a negative coefficient on the dummy variable for loans indicates non-trivial differences in the amounts remitted across the two groups depending on the degree of altruism. If a migrant is selfish (0% contribution), then he remits by 20.9% less if he has a loan, but if he is of the average degree of altruism (a contribution of 37%), then he remits by 4.1% more if he has a loan.18 In other words, the remittance schedule in altruism estimated for the group of migrants with loans crosses the corresponding schedule of the other group from below, as shown in Fig. 1. As an additional robustness check, we also examine whether other variables, such as the ones in vectors X and Z in specification (1), have any explanatory power when interacted with the LOAN dummy. The results are reported in Table A2 in Appendix A. Each column represents a version of specification (2), in which the variable ALTR is replaced with a different variable. Aside from migrants’ own income, no other explanatory variable, including household income, in these regressions is statistically significant at conventional levels. 5. Discussion In this study, we find that altruism only seems to affect remittance behaviour for migrants with explicit loan obligations back home. We attribute the finding that altruism does not explain remittance behaviour for the entire sample to the fact that altruism can be offset by other motives for remittances or various confounding factors. In particular, we show that the observed linkage between altruism and loans can be consistent with the remittance model of altruism extended with reference-dependent preferences and loss-averse migrants. More generally, we suggest below that reference dependence can be a convenient way of modelling remittance behaviour. Utility theories of reference dependence are motivated by empirical evidence about the effects of contextual circumstances on the individual perception of utility. These effects are typically found to take the form of loss aversion with respect to some reference point, determined by the decision maker’s current position and expectations, as well as by social norms and comparisons. Regarding remittance behaviour, similarly to other realms of economic behaviour, there are strong reasons to believe that the subjective utility of remittances depends on contextual circumstances. Examples of such circumstances could be the history of remittances, individual or common beliefs about a ‘fair’ amount of remittances, recipients’ expectations, family or peer pressures, self-insurance and control motives for remittances. Each circumstance or motive can be characterized by whether they have an effect on one of three modelling blocks of reference dependence: (i) reference point, (ii) uncertainty; and (iii) loss aversion. For instance, social pressures and reciprocation can be attributed to the determinants of the level of reference point. According to Gardner (2012), a study on immigration to the Persian Gulf states, it is family pressures that are responsible for the decision to emigrate and for the amount of remittances to be sent home (also see Ilahi and Jafarey, 1999). The anthropological study of Osella and Osella (2000) describes a local status categorization of migrant workers from Kerala, based on their ability to earn money abroad, and privileges associated with high status. As another example, Chort et al. (2012) find that Senegalese migrants in France and Italy face ostracism and loss of access to services provided by a migrant network of their countrymen in the host country if remittances sent fall short of the expected norm. The degree of uncertainty about the reference point can be linked to the control motive for remittances. If remittances are sent toward a certain savings target, then migrants with a greater control over remittance uses will have more certainty about the reference point, which can have implications for their remitting behaviour as in Ashraf et al. (2015) or Batista et al. (2015).19 Lastly, the effects of the insurance motive on remittances as in Gubert (2002); Molina Millán (2015); Batista and Umblijs (2016) can be linked to the degree of loss aversion, i.e. the shape of ‘universal gain–loss function’. 6. Conclusion We study the relationship between altruism and remittances of migrants, using a measure of altruism elicited from a dictator game experiment. While we find little evidence of a universal relationship between altruism and remittances, we do observe a positive effect of altruism on remitting behaviour for migrants with loan obligations. Using the framework of reference-dependent preferences, we argue that the altruistic motive is subdued by uncertainty about remittance expectations and loss aversion. At the same time, the possession of loan obligations may reduce uncertainty about remittance expectations, subsequently making the altruistic motive more pronounced. Indirect tests may, however, fail to establish this relationship between altruism and remittances. More generally, we also suggest that our approach with reference dependence can be useful for modelling remittance behaviour. As a possible application, the remittance model with reference-dependent preferences could yield novel predictions on the effects of networks in explaining remittance behaviour. As a migrant worker spends a substantial amount of time interacting with his peers, the remittance behaviour of his peers, who most often come from the same community back home, may potentially influence expectations and reference levels and, through these, may also influence the migrant’s remittance behaviour. Supplementary material Supplementary material (the Appendix and data) is available online at the OUP website. Footnotes 1 Aside from its direct role on amounts remitted, altruism can also play a role in alleviating the problem of lack of enforcement in implicit familial contracts (Becker, 1991; Foster and Rosenzweig, 2001). 2 See Stark (1995), Rapoport and Docquier (2006) and Carling (2008) for reviews of altruism and other motives for remittances. 3 Further examples on altruism-confounding effects include de la Brière et al. (2002), Cox and Fafchamps (2007), Brown and Jimenez (2011), Yang (2011), McKenzie et al. (2013), De Arcangelis et al. (2015) and Batista and Narciso (2017). Section 5 provides further discussion. 4 For limitations of this approach toward measuring altruism, see Roth (1995), List (2007) and Andreoni et al. (2017). 5 For evidence from the psychology literature, see Jones and Rachlin (2006) and Rachlin and Jones (2008) and from economics—Roth (1995), Hoffman et al. (1996) and Bohnet and Frey (1999). 6 It is comparable with the average transfer of 39.6% in the dictator game of Batista et al. (2015), conducted in urban Mozambique, where the recipient is the closest person to the giver outside the giver’s household. 7 In the ultimatum game, the average transfer is 39% and not statistically different from the average in the dictator game. 8 Male interviewers would randomly select one migrant to be interviewed per room at each accommodation. We could not explicitly prohibit individuals of other faiths from participating but did so indirectly by stating that the campus visit would be on a Friday (which is a weekend day in Qatar), the day when church services were normally held in addition to the mid-day congregation prayer for Muslims. Some leeway was given to interviewers to enrol individuals with post-secondary education provided that this group formed a minority of the sample. 9 Subjects were asked to choose from one of six choices, each offering two possible rewards depending on a coin toss. The choices were: (i) 500 or 500; (ii) 450 or 950; (iii) 400 or 1200; (iv) 300 or 1500; (v) 100 or 1900; or (vi) 0 or 2000, where all the amounts are in Qatari Riyals. Lower numbered choices would reflect a greater degree of risk aversion. Workers were told that 40 of them (about 20%) would be selected at random to be paid for choices. 10 While our sample of 105 migrants is smaller than that in other studies of remittance behaviour (but see Osili, 2007 for a related study of the effects of altruism on remittances that uses a sample of similar size), our design employs behavioural tasks which take time to administer and for which subjects are paid. This necessarily limits our feasible sample size to one comparable to those used in laboratory experiments. 11 The English translation of the instructions of the dictator game from Malayalam, the local language of Kerala, is available upon request. Along with written instructions, the participants were also shown voiced PowerPoint presentations about the games played. Additionally, there were Malayalam-speaking assistants trained to help participants in better understanding the games. 12 These numbers fall within the range of mean and median offers observed in similar studies, which is between 10% and 50% (see Camerer, 2003, Table 2.4, pp. 57–8). 13 SHGs in Kerala typically consist of women who vouch for each other in fulfilling their loan obligations with a participating financial institution. Spouses involved in SHGs may have better access to loan opportunities relative to non-participating spouses. 14 Both remittances and income could be jointly determined. For example, the decision to migrate could be motivated by expectations about foreign income received and remittances that could be sent. 15 A Wald test of coefficient equality is rejected (p-value = 0.11). 16 A positive coefficient on the household’s income can instead be associated with exchange and investment motives (see Rapoport and Docquier, 2006, Table 2). 17 An F-test of the equality of the coefficient on ALTR*LOAN in columns 4 and 6, and also between columns 5 and 6 cannot be rejected (p-value of 0.41 and 0.37 respectively). 18 The results are qualitatively similar using coefficients from column 6 of Table 4. Evaluating at the mean of the control variables, a 10% increase in contribution made in the dictator game is associated with a 3.1 percentage point higher remittance for a migrant with a loan obligation. A migrant with a loan who is selfish sends 14.4% less, while those at the average degree of altruism remits by 2.4% more. 19 The possession of an explicit loan obligation could be related to the control motive and more certainty about remittance uses and, thus, the level of reference point. Acknowledgements We thank Henry Farber, Daniel Westbrook, two anonymous referees, and various seminar and conference audiences for valuable comments and suggestions. We also thank Vodafone Qatar for providing mobile Wi-Fi access that was used in the data collection process. References Ambler K. ( 2015) Don’t tell on me: experimental evidence of asymmetric information in transnational households, Journal of Development Economics , 113, 52– 69. Google Scholar CrossRef Search ADS PubMed  Amuedo-Dorantes C., Bansak C., Pozo S. ( 2005) On the remitting patterns of immigrants: evidence from Mexican survey data, Economic Review , 90, 37– 58. Andreoni J., Rao J.M., Trachtman H. ( 2017) Avoiding the ask: a field experiment on altruism, empathy, and charitable giving, Journal of Political Economy , 125, 625– 53. Google Scholar CrossRef Search ADS   Ashraf N., Aycinena D., Martínez A.C., Yang D. ( 2015) Savings in transnational households: a field experiment among migrants from El Salvador, Review of Economics and Statistics , 97, 332– 51. Google Scholar CrossRef Search ADS   Batista C., Narciso G. ( 2017) Migrant remittances and information flows: evidence from a field experiment, World Bank Economic Review , forthcoming. Batista C., Umblijs J. ( 2016) Do migrants send remittances as a way of self-insurance?, Oxford Economic Papers , 68, 108– 30. Google Scholar CrossRef Search ADS   Batista C., Silverman D., Yang D. ( 2015) Directed giving: evidence from an inter-household transfer experiment in Mozambique, Journal of Economic Behavior and Organization , 118, 2– 21. Google Scholar CrossRef Search ADS   Becker G.S. ( 1991) A Treatise on the Family , Harvard University Press, Cambridge, MA. Bohnet I., Frey B. ( 1999) The sound of silence in prisoner’s dilemma and dictator games, Journal of Economic Behavior and Organization , 38, 43– 57. Google Scholar CrossRef Search ADS   Bouhga-Hagbe J. ( 2006) Altruism and workers’ remittances: evidence from selected countries in the Middle East and Central Asia. IMF Working paper WP/06/130, Washington, DC. Google Scholar CrossRef Search ADS   Brown R.P.C., Jimenez E.V. ( 2011) Subjectively-assessed welfare and international remittances: evidence from Tonga, Journal of Development Studies , 47, 829– 45. Google Scholar CrossRef Search ADS   Camerer C. ( 2003) Behavioral game theory: experiments on strategic interaction , Princeton University Press, Princeton, NJ. Carling J. ( 2008) The determinants of migrant remittances, Oxford Review of Economic Policy , 24, 581– 98. Google Scholar CrossRef Search ADS   Chort I., Gubert F., Senne J.-N. ( 2012) Migrant networks as a basis for social control: remittance incentives among Senegalese in France and Italy, Regional Science and Urban Economics , 42, 858– 74. Google Scholar CrossRef Search ADS   Cox D., Fafchamps M. ( 2007) Extended family and kinship networks: economic insights and evolutionary directions, in Schultz T.P., Strauss J.A. (eds) Handbook of Development Economics , Volume 4, Elsevier-North Holland, Amsterdam, 3711– 84. Google Scholar CrossRef Search ADS   De Arcangelis G., Joxhe M., McKenzie D., Tiongson E., Yang D. ( 2015) Directing remittances to education with soft and hard commitments: evidence from a lab-in-the-field experiment and new product take-up among Filipino migrants in Rome, Journal of Economic Behavior and Organization , 111, 197– 208. Google Scholar CrossRef Search ADS   de la Brière B., Sadoulet E., de Janvry A., Lambert S. ( 2002) The roles of destination, gender, and household composition in explaining remittances: an analysis for the Dominican Sierra, Journal of Development Economics , 68, 309– 28. Google Scholar CrossRef Search ADS   Forsythe R., Horowitz J., Savin N.E., Sefton M. ( 1994) Fairness in simple bargaining experiments, Games and Economic Behavior , 6, 347– 69. Google Scholar CrossRef Search ADS   Foster A.D., Rosenzweig M.R. ( 2001) Imperfect commitment, altruism and the family: evidence from transfer behavior in low-income rural areas, Review of Economics and Statistics , 83, 389– 407. Google Scholar CrossRef Search ADS   Funkhouser E. ( 1995) Remittances from international migration: a comparison of El Salvador and Nicaragua, Review of Economics and Statistics , 77, 137– 46. Google Scholar CrossRef Search ADS   Gardner A. ( 2012) Why do they keep coming? Labor migrants in the Gulf States, in Kamrava M., Babar Z. (eds) Migrant Labor in the Persian Gulf , Hurst & Company, London, 41– 58. Gubert F. ( 2002) Do migrants insure those who stay behind? Evidence from the Kayes area (Western Mali), Oxford Development Studies , 30, 267– 87. Google Scholar CrossRef Search ADS   Hamilton W.D. ( 1964) The genetical evolution of social behaviour, I and II, Journal of Theoretical Biology , 7, 1– 52. Google Scholar CrossRef Search ADS PubMed  Hoddinott J. ( 1994) A model of migration and remittances applied to western Kenya, Oxford Economic Papers , 46, 459– 76. Google Scholar CrossRef Search ADS   Hoffman E., McCabe K., Smith V.L. ( 1996) Social distance and other regarding behavior in dictator games, American Economic Review , 86, 653– 60. Ilahi N., Jafarey S. ( 1999) Guest worker migration, remittances and the extended family: evidence from Pakistan, Journal of Development Economics , 58, 485– 512. Google Scholar CrossRef Search ADS PubMed  Kahneman D., Tversky A. ( 1979) Prospect theory: an analysis of decision under risk. Econometrica , 47, 263– 92. Google Scholar CrossRef Search ADS   Kőszegi B., Rabin M. ( 2006) A model of reference dependent preferences, Quarterly Journal of Economics , 121, 1133– 65. Google Scholar CrossRef Search ADS   Jones B., Rachlin H. ( 2006) Social discounting, Psychological Science , 17, 283– 6. Google Scholar CrossRef Search ADS PubMed  Joseph T., Nyarko Y., Wang S.-Y. ( 2015) Asymmetric information and remittances: evidence from matched administrative data., NBER Working Paper No. 20986, Cambridge, MA. Kaplan H., Gurven M. ( 2005) The natural history of human food sharing and cooperation: a review and a new multi-individual approach to the negotiation of norms, in Gintis H., Bowles S., Boyd R., Fehr E. (eds) Moral Sentiments and Material Interests: The Foundations of Cooperation in Economic Life , MIT Press, Cambridge, MA, 75– 113. List J. ( 2007) On the interpretation of giving in dictator games, Journal of Political Economy , 115, 482– 93. Google Scholar CrossRef Search ADS   Lucas R.E.B., Stark O. ( 1985) Motivations to remit: evidence from Botswana, Journal of Political Economy , 93, 901– 18. Google Scholar CrossRef Search ADS   McKenzie D., Gibson J., Stillman S. ( 2013) A land of milk and honey with streets paved with gold: do emigrants have over-optimistic expectations about incomes abroad?, Journal of Development Economics  102, 116– 27. Google Scholar CrossRef Search ADS   Molina Millán T. ( 2015) Regional migration, insurance and economic shocks: evidence from Nicaragua., IZA Discussion Paper No. 9494, Bonn. Osella F., Osella C. ( 2000) Migration, money and masculinity in Kerala, Journal of the Royal Anthropological Institute , 6, 117– 33. Google Scholar CrossRef Search ADS   Osili U.O. ( 2004) Migrants and housing investments: theory and evidence from Nigeria, Economic Development and Cultural Change , 52, 821– 49. Google Scholar CrossRef Search ADS   Osili U.O. ( 2007) Remittances and savings from international migration: theory and evidence using a matched sample, Journal of Development Economics , 83, 446– 65. Google Scholar CrossRef Search ADS   Poirine B. ( 1997) A theory of remittances as an implicit family loan arrangement, World Development , 25, 589– 611. Google Scholar CrossRef Search ADS   Rachlin H., Jones B. ( 2008) Altruism among relatives and non-relatives, Behavioural Processes , 79, 120– 3. Google Scholar CrossRef Search ADS PubMed  Rapoport H., Docquier F. ( 2006) The economics of migrants’ remittances, in Kolm S.-C., Ythier Mercier J. (eds) Handbook of the Economics of Giving, Altruism and Reciprocity , Volume 2, Elsevier-North Holland, Amsterdam, 1135– 98. Roth A. ( 1995) Bargaining experiments, in Kagel J.H., Roth A. (eds) Handbook of Experimental Economics , Princeton University Press, Princeton, NJ, 253– 348. Seshan G., Zubrickas R. ( 2017) Asymmetric information about migrant earnings and remittance flows, World Bank Economic Review , 31, 24– 43. Stark O. ( 1995) Altruism and Beyond: An Economic Analysis of Transfers and Exchanges Within Families and Groups , Cambridge University Press, Cambridge, MA. Google Scholar CrossRef Search ADS   Yang D. ( 2011) Migrant remittances, Journal of Economic Perspectives , 25, 129– 52. Google Scholar CrossRef Search ADS   © Oxford University Press 2017 All rights reserved

Journal

Oxford Economic PapersOxford University Press

Published: Jan 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial