Remittance Concentration and Volatility: Evidence from 72 Developing Countries

ABSTRACT

This paper contributes to the literature by introducing the role of geographic concentration of the source of remittances. Specifically, using data over 2010–2015 for 72 developing countries, we study the impact of (i) large remittances and (ii) the geographic concentration of the source of remittances on economic volatilities. Results suggest that while (i) large remittances can be stabilizing on average, (ii) high remittance concentration from source countries can aggravate economic volatilities in recipient countries. Results are robust to global shocks affecting both source and recipient countries, and volatility in the remittance-sending country.

KEYWORDS:

• Remittances
• volatility
• remittance concentration
• developing countries

Disclosure statement

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

Notes

1 The determinants of remittances are a separate strand of the literature. See Rapoport and Docquier (2006) for a comprehensive survey.

2 For example, if remittances increase under a flexible exchange rate regime, the resulting currency appreciation would offset the increase in aggregate demand from increased expenditure.

3 Adeniyi et al. (2019) and Ahamada and Coulibaly (2011) present empirical support to the importance of financial development in how remittances can affect macroeconomic volatility.

4 Chami et al. (2008) define remittance-dependent countries as those with remittances-to-GDP ratios of 5 percent or more.

5 Both samples for real GDP growth volatility have similar standard deviations, although that with large remittances has a more positively skewed distribution compared to that without large remittances (skewness of 2.19 compared to 1.88). The sample with large remittances also has a fatter tail than that without large remittances, with more observations around the median (kurtosis of 9.35 compared to 7.53).

6 We use NPV instead of nominal debt as most of our sample is Low Income-Countries, where their debt stock is predominantly external concessional loans. Each dot in Figure 7 reprsents country years.

7 The sample contains 72 developing countries, as defined according to IMF (2013, 2015).

8 Given the very short time dimension, which is constrained by data availability on remittance concentration, dynamic panel methods such as GMM or pooled mean group estimators are not applicable.

9 While having instrumental variables would have been ideal, it is difficult to find an instrument in a cross-sectional dimension that is correlated with remittances but not economic volatility. Bugamelli and Paterno (2009b) use distance and GDP growth in the source country as instruments. Physical distance may not be a good instrument for a number of reasons. These include the role of technology where the costs of sending remittances may vary greatly, even between close neighbors (see also the example of footnote 20). Moreover, some studies have shown that neighbors can have highly synchronized business cycles (see for example Sethapramote, 2015) and thus distance would also be correlated with the dependent variables. Alternatively, internal instruments if one uses a dynamic panel method are also not feasible in our context given the very small T dimension. 

10 About 23 percent of our sample, of 385 country-year observations, lies below the large remittance-to-GDP definition of 1 percent.

11 Data, which starts in 2010, is available online on the World Bank website: https://www.worldbank.org/en/topic/migrationremittancesdiasporaissues/brief/migration-remittances-data

12 The CPIA data is available starting 2005, here: http://data.worldbank.org/data-catalog/CPIA. The CPIA, since 1980, has been used to allocate World Bank International Development Association (IDA) resources to eligible client countries. It is produced annually for more than 70 IDA eligible countries.

13 The nominal exchange rate is defined in LCU/USD such that higher values indicate a depreciation against the U.S. dollar.

14 Pooled OLS is also estimated, in all regressions, assuming that the intercept and the slopes do not change across units or over time. Results of the F-test, comparing the unconstrained FE model with the constrained OLS model, are highly significant, indicating that pooled OLS would be biased and inconsistent (Baltagi, 2008). We also apply the Breusch and Pagan (1980) Lagrangian multiplier test, and we reject the null hypothesis of no individual effects, rendering the OLS inconsistent. Regarding the static panel estimation methodologies, a FE model considers the individual effect to be fixed (hence the name fixed), as opposed to the RE model which assumes the specific individual effect is randomly distributed (hence the name random). The RE model, which is a feasible Generalized Least Squares (GLS) developed by Swamy and Arora (1972), is generally assumed to be a more efficient estimator than the FE model. The Hausman test (1978) basically checks a more efficient model against a less efficient but consistent model to make sure that the more efficient model also gives consistent results. See Hosny (2011) for a similar application in the context of a trade gravity model. In what follows, we report either the FE or the RE panel model, depending on the result of diagnostic tests.

15 Results are the same, with or without the exchange rate dummy, where flexible de facto regimes take the value of 1 for (flexible) countries with yearly nominal changes of ± 2 percent, and 0 (fixed) otherwise, following the definition used in Ben Naceur et al. (2019), Hosny et al. (2015) and Shambaugh (2004).

16 We also measure real GDP volatility as 5-year rolling windows for robustness, and results are largely similar.

17 Focusing exclusively on the remittances-institutions relationship, in a sample of 111 countries, Abdih et al. (2012b) present evidence that higher ratio of remittances to GDP may lead to lower indices of control of corruption, government effectiveness, and rule of law, even after controlling for potential reverse causality.

18 Although concentration may sometimes be beneficial, as shown by Vaaler (2013) who finds that geographically-concentrated diasporas can increase their home-country venture funding access.

19 See https://datacatalog.worldbank.org/dataset/remittance-prices-worldwide

20 This issue is also observed in higher-income emerging economies, where for example it is more expensive to send remittances from Indonesia's first remittance source country (Malaysia, at around 6 percent, despite the geographic proximity) than from its second source country (Saudi Arabia, at around 4 percent). 

Additional information

Notes on contributors

Amr Hosny

Amr Hosny is a national of the Arab Republic of Egypt, and is currently an Economist at the International Monetary Fund. He holds a PhD in Economics from the University of Wisconsin-Milwaukee, and his research areas are in the fields of open economy macroeconomics and economic development.