Financial liberalization and remittances: Recent panel evidence

Abstract

We investigate the impact of financial liberalization on remittances to 84 countries over the period 1986–2005. Explicitly accounting for the multidimensionality of financial reform, we find that the various dimensions impact remittances differently: Increased economic freedom in the financial sector, as captured by absence of direct government control over the allocation of credit, has a positive and immediate impact. However, the improved robustness of financial markets, as captured by the development of security markets, improvement in the quality of banking supervision, and removal of stringent restrictions on interest rates and international capital, has a negative and lagged effect. The net combined effect reveals that financial liberalization may have a modest negative impact on remittances in the long run.

Keywords:

• remittances
• financial liberalization
• economic freedom
• institutions

JEL Classifications:

• F22
• O15
• P48

Acknowledgements

We would like to thank the editor and two anonymous referees of this journal for their constructive suggestions and insights. An earlier version of this paper was presented at the 77th Annual Meeting of the Midwest Economic Association Conference in Columbus, Ohio and at the 88th Annual Meeting of the Western Economic Association at Seattle, Washington. We would like to thank Brad Humphreys, Scott Hegerty, Artatrana Ratha, and other participants in our sessions for comments and recommendations. We are also grateful to Terry Simpkins for his meticulous editorial assistance in improving this draft. A previous version of the paper has been released as IZA discussion paper #7497 (http://ftp.iza.org/dp7497.pdf) and as an EPRN paper (http://www.employmentpolicy.org/topic/21/research/financial-liberalization-and-remittances-recent-longitudinal-evidence). The usual caveats apply.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See La Porta et al. (1998), Guiso, Sapienza, and Zingales (2004), Rajan and Zingales (2003), and Chinn and Ito (2006) as influential contributions that respectively assert the salience of legal origin, the level of social capital, interest group activity, and the quality of institutions in determining the level of financial development of an economy.

2. Reinhart, Kirkegaard, and Sbrancia (2011, 22) define financial repression as occurring ‘when governments implement policies to channel to themselves funds that in a deregulated market environment would go elsewhere.’ They identify the specific policies as including all or a subset of the following: (a) directed lending to the government in the form of pension funds or domestic banks; (b) explicit or implicit ceilings on interest rates; (c) regulation of international capital movements; (d) a closer relationship between the state and the banking sector, either via direct state ownership or via coercion of privately owned banks; (e) relatively high reserve or liquidity requirements; (f) taxes on securities transactions; (g) prohibition of gold purchases; and (h) a high proportion of government debt being nonmarketable.

3. It should be stated that there is an emerging literature that asserts causality from growth to financial development, accounting, of course, for the eventuality that the causality could be bi-directional. On this, see Chow and Fung (2013) and references therein.

4. There is a debate in the literature on the impact of per capita GDP on aggregate remittance receipts: On one hand, the altruism motive for remittances predicts a negative relationship between the two variables since an increase in household income should reduce the reliance on remittance income for the purpose of consumption, insurance, or debt repayment. On the other hand, the investment motive for remittances predicts a positive relationship between the variables since a higher rate of growth will correlate with a higher rate of return on capital and hence, increase the incentive to remit for investment purposes. We refer the reader to Basu and Bang (forthcoming) for more on this debate.

5. The literature is again equivocal on the impact of government expenditure and we refer the reader to Basu and Bang (forthcoming) for more on the topic. To summarize, the demand-side perspective on remittances holds that such transfers substitute for needed social insurance programs such as public health care, unemployment insurance, crop insurance, and public pension schemes, which are typically absent in developing countries. The government expenditure variable, at least partially, captures the existence of such programs and should therefore be correlated negatively with aggregate remittance receipts, given the altruism motive for remitting. However, the relationship is more ambiguous if we consider the supply-side argument whereby remittances are motivated primarily by investment prospects in the home country. As noted by Basu and Bang (forthcoming), if public expenditure crowds out private investment, then higher state expenditure should reduce the incentive to remit. On the other hand, if public expenditure is directed at creating needed infrastructure and correcting market failures, this should invite higher volumes of remittances.

6. To be specific, we consider the total emigration rates to six major OECD destinations, namely, Canada, Australia, United States, United Kingdom, France, and Germany (Defoort 2008). Focusing on these six destinations is less restrictive than may appear to be the case: These countries accounted for 77% of the OECD skilled immigration stock in the year 2000 (Beine et al. 2011). This is significant considering that 90% of all high skilled international migrants were found to be living in the OECD in that year (Docquier et al., 2007. Further, the United States, Germany, France, Canada, and the United Kingdom were, in descending order, the five largest remittance-sending countries in 2005, together accounting for approximately half of the global remittance flow (Ratha and Shaw 2007). Australia was the ninth largest, being further superseded by Saudi Arabia, Spain, and Hong Kong in descending order. See Beine et al. (2011) and Mitra et al. (forthcoming) for other studies based on this convention.

7. There may be a concern that the Polity IV Index may not be the appropriate choice of control for institutional quality. We address this issue at the end of Section 5, with the corroborating results appearing in Table A2 in the appendix.

8. It should be clarified that what the Abiad, Detragiache, and Tressel (2010) subindices measure is the existence (or absence) of policies designed to further these eight objectives and not the outcomes of these policies. As a case in point, the subindex that captures participatory constraints in the banking sector is based on the absence of barriers to entry by international banks, restrictions on the establishment of branches by both domestic and international banks, and limitations on the range of financial activities performed by both types of banks. It does not, for example, measure the actual number of foreign banks in the economy, the average number of branches per bank, or the average range of permitted financial activities.

9. We acknowledge that this leaves the problem of reverse causality unresolved. One problem with addressing this issue comes with finding suitable instruments for financial liberalization that will pass the tests for strength and validity. While the GMM-style instruments suggested by the Arellano–Bond estimator satisfy the conditions for strong instruments, they are also correlated with remittances, and therefore do not pass the Sargan test for instrument validity. We refer the reader to Bazzi and Clemens (2013) for a more complete discussion of these pitfalls to the Arellano–Bond and related GMM estimators.

10. The unique part of the decomposed variance can be seen as a residual, consisting of a random component and measurement error. Specifically, the uniqueness factor reported in Table 3 consists of the total variability of each variable minus the sum of its squared factor loadings.

11. As can be seen from Table 3, we lose nothing from ignoring the fourth factor retained by the EFA since the first three factors together account for all of the common variance in the data.

12. The magnitudes underlying Figures 2–4 are calculated in the same way as described in footnote 5 of Section 4.

13. Including a lagged value for financial reform does not require us to sacrifice observations because the time series for the financial reform database begins in 1981, whereas the lagged differences for our dependent variable needed to estimate the difference GMM model are only available beginning for the five-year period ending in 1990. Ideally, we would like to be able to test for the optimal lag length in the impact of financial reform. However, given the shortness of our sample (t = 4), including additional lagged values of financial reform would require sacrificing an entire time period of observations, which would call into question the consistency of our estimated coefficients.

14. The estimated combined impact of a standard deviation change in both factors is about 0.4012; the estimated standard error of this linear combination is about 0.2500, resulting in a z-statistic of about 1.60 and a p-value of about 0.109.

15. The point estimate for the linear combination is 0.5850 and the standard error is 0.2850, yielding a test statistic of 2.04 and a p-value of 0.040.

16. The point estimate for the second period impact is 0.894(β_{freedom(t − 1)}) + 0.831(β_{robustness(t − 1)}) + 0894(β_{remit(t − 1)}) (β_freedom(t)) + 0.831(β_{remit(t − 1)}) (β_freedom(t)) = −0.6916. Since the model contains nonlinear combinations of the coefficients, a simple z-test is invalid. However, we can perform a Wald test. The value of the Wald statistic for this test is 4.13, and the Bonferroni-adjusted p-value for the test is 0.042. Hence, the nonlinear effect is statistically significant.

17. The cumulative effect on remittances after two periods is −0.1071, but this effect is not statistically significant.

18. To conserve space, we have not reported the full results of these robustness checks in the paper, but will make any or all of them available on request.

19. It is important to note that balancing the panel leads reduces the number of countries considerably (from 84 to 56) and the countries that drop out are not selected at random. In fact, not only do we lose a considerable number of non-OECD countries, but many of the countries that drop out are transition economies. Since these are precisely the societies which have experienced some of the most significant episodes of financial liberalization during this time, omitting these countries may increase the possibility of selection bias.

20. We are grateful to a referee for drawing our attention to this literature.

21. This is, in fact, an ongoing debate in the literature. Giuliano and Ruiz-Arranz (2009), by contrast, maintain that the growth impact of remittances is greater in less financially developed economies, since remittances provide an alternative source of finance for productive investment that would otherwise be inhibited by the credit and liquidity constraints.