International Remittances, Domestic Sharing, and Income Inequality in Senegal

Abstract

This study examines the impact of international remittances on internal cash transfers in Senegal using instrumental variable analysis with a large sample of individuals aged 13 and above. The findings reveal a direct internal sharing effect of international remittances. Individuals receiving such remittances are 26 per cent more likely to make internal cash transfers, with wealthier recipients showing a stronger propensity. Notably, the poorest individuals benefit the most. When international remittances and internal cash transfers coexist, the Gini index is 3 percentage points lower than in scenarios without international remittances and 9.2 percentage points lower than in scenarios with international remittances but no internal sharing, emphasizing their redistributive nature. These results hold across recipient locations and various econometric approaches.

Keywords:

• Remittances
• redistribution
• economic development
• Africa
• Senegal

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Transfers received refers to the financial support or funds received by individuals aged 13 or older from sources outside of their household or from institutions within the past 12 months.

2 Transfers sent refer to the financial support or funds sent by individuals aged 13 years or older to individuals outside the household within the past 12 months.

3 It is important to note that 2 per cent of individuals receiving international remittances also engage in sending money abroad. However, for the purpose of this paper, which focuses on investigating internal sharing, we will consider these individuals, irrespective of whether they are also senders (See Figure S5 and Table S1 in the Supplementary Materials).

4 Descriptive statistics for internal transfer recipients are not reported; however, they constituted 2,468 individuals, or 14 per cent of the sample.

5 We focus on regular transfers in the main analysis for several reasons. Firstly, regular transfers are generally more stable and predictable, making them an appropriate subject for our main analysis. Irregular transfers, on the other hand, can vary considerably, both in terms of timing and amount, making it difficult to draw meaningful conclusions about their effects in the context of our study. Secondly, focusing on regular transfers could also mitigate memory bias and therefore measurement error. However, in the robustness section, we will test the robustness of our results by including both regular and irregular international remittances.

6 The magnitude of the great drought of 1983 is measured as a deviation from the historical average as follows: $${\mathit{\text{Shock}}}_{i1983}=\frac{\overline{R_i}-R_{i1983}}{R_i^{\mathit{\text{SD}}}}$$ where $\overline{R_i}$ is the historical annual average rainfall (for 22 years) at the district of residence i. $R_{i1983}$ indicates the 1983 annual rainfall in district i and $R_i^{\mathit{\text{SD}}}$ is the standard deviation of annual rainfall (calculated over the 22-year period).

7 All Figures and Tables referenced by S can be found in the Supplementary Materials.

8 The choice of linear probability is motivated by the fact that Angrist and Pischke (2008) has shown that the marginal effects of a dummy variable estimated by the logistic model and the linear probability model (LPM) are ‘indistinguishable.’

9 We included household characteristic variables as controls in our analysis in order to capture crucial dimensions of household context that may affect the dynamics of remittance behaviour and to isolate the specific impact of receiving international remittances on sending cash within the country. We recognize, however, that variables such as household composition and size could be influenced by the dynamics of receiving and sending remittances, raising potential concerns about their endogeneity. To address this, sensitivity analyses, with and without these variables, will be conducted to ensure the robustness of our results.

10 The Kleibergen-Paap Wald F rk statistic for weak identification significantly exceeds the critical values of Stock-Yogo. Moreover, the p-value associated with the Kleibergen-Paap LM rk statistic is below 5 per cent, allowing us to reject the null hypothesis of no correlation between the instruments and the endogenous variable. Additionally, the p-value of the Hansen test equals 0.149, indicating that we do not reject the null hypothesis of no correlation between the instruments and the error term. Thus, the exclusion restriction conditions are satisfied.

11 The Grimm vulnerability score is an indicator of objective and structural elements, which refers to the conception of poverty as a lack of capacity or vulnerability. The different components of the calculation of this score are: (1) an indicator of material living conditions (housing, water, electricity, toilets, type of fuel); (2) an indicator of the level of human capital calculated as the ratio between the number of years of education completed by household members and the maximum possible number of years of education, taking into account the age of each member; (3) a vulnerability indicator that takes into account the number of durable goods (bicycles, radios, televisions, etc.) available in each household. For each component, the maximum score corresponds to a high level of deprivation, while a score of zero means the absence of any impairment.

12 For a more comprehensive understanding of the methodology and detailed results, please refer to the Supplementary Material.