Remittances and GDP Growth

Economists have recently turned their attention to estimating the impact of remittances on longer-term economic growth using modern growth theory.¹ Previous chapters introduced several possible mechanisms through which remittances may affect economic growth. The following discussion summarizes these mechanisms to motivate the empirical work that has examined the relationship between remittances and growth. The discussion groups the mechanisms into those through which remittances potentially have positive effects on growth and those through which they may have negative effects.

On the positive side, remittances may increase or enhance investment in physical capital. To the extent that there are frictions in domestic financial intermediation, imperfect capital mobility is present and remittances are not simply “disguised” capital flows. The receipt of temporary remittances in an economy may lead to an increase in the domestic investment rate, thus increasing economic growth. If financial constraints are significant—for example, a large group of households are rationed out of credit markets because of the lack of domestic financial development—then remittances may help to ease the constraints. This results in an increase in investment, provided that the rationed households also have access to productive investment opportunities and use the remittances to expand investment rather than consumption.² Alternatively, if remittances are primarily disguised capital flows—the recipients are investing on behalf of the remitter—then efficiency in investment is enhanced to the extent that the family member receiving the remittance flows possesses some informational advantage or expertise with respect to formal financial intermediaries.

Another channel through which remittances may have a positive impact on growth is via the facilitation of human capital formation. Even though, as discussed in Chapter 4, the majority of remitted funds are devoted to consumption and residential investment, significant productivity spillovers may result from the recipients’ improved nutrition and shelter, assuming that they participate in the labor market. In addition, the literature offers several examples in which a significant fraction of remittances are spent on education, assuming that those who receive the education do not in turn emigrate. As such, remittances may increase total factor productivity. These two channels through which remittances may have a positive impact on growth, the accumulation of physical capital and total factor productivity, are not mutually exclusive, but it may be possible to distinguish between them empirically.

In addition to enhancing investment and total factor productivity, remittances may also have a positive impact on growth through their effect on the recipient economy’s financial system. By increasing the recipient country’s demand for money, remittances are likely to expand the supply of funds to the banking system. This in turn may lead to enhanced financial development through a reduction in the external finance premium and thus to higher economic growth through one of two channels: (1) economies of scale, or (2) a political economy effect, whereby a larger constituency (depositors) is able to pressure the government into undertaking beneficial financial reform.

It is far from assured that remittances will have a positive impact on economic growth in any particular country, however. Each mechanism described in the foregoing discussion relies on a particular set of circumstances that may not be present in a given country; alternative conditions that reduce or eliminate the positive impact of remittances may be found there instead. In general, the greater the degree of capital mobility in a country, the less remittances will affect the domestic investment rate. Also, if remittances are perceived to be permanent, they will tend to be consumed in their entirety and therefore will not affect aggregate investment. A similar argument applies to credit-constrained households: even if the constraints are relaxed, additional funds from remittances may simply be consumed. Finally, the family member who receives a migrant’s remittances may actually be less skilled in investing than are financial intermediaries, which has important implications in this context if remittances are disguised capital flows.

Prevailing circumstances in a particular economy may likewise reduce the human capital and financial sector impacts of remittances. The consumption impact of remittances on labor productivity depends on recipient families’ standard of living. If a family’s standard of living is sufficiently high before the receipt of remittances that its basic needs are adequately met, then the labor productivity effect of remittances vanishes for that family. Also, any human capital accumulation impacts depend on the recipients’ participation in the labor force after accumulation of capital. In some remittance-receiving societies, education funded by remittances is intended to enable the recipients themselves to migrate. Finally, in terms of financial sector impacts, an increase in the size of the domestic banking system via an increase in the supply of funds does not necessarily reduce the external finance premium. The political economy mechanism arising from a larger banking system may have an adverse effect on financial development: depositors lobby the government for reforms favoring safety over intermediation, for example, causing banks to increase their holdings of safe assets rather than lending.

The question of whether remittances increase an economy’s growth is not simply a matter of whether conditions in the economy are favorable to the operation of the channels described previously. Remittances can also decrease economic growth through two means that operate differently than the positive channels. One that is increasingly mentioned in the literature is a Dutch disease effect, which requires that the traded goods sector of a remittance-receiving economy be the source of significant positive externalities that enhance other sectors’ productive capacity. If this condition is satisfied, a Dutch disease effect may arise from remittances to the extent that they cause the economy’s real exchange rate to appreciate. The third section of this chapter presents a separate discussion of the effect of remittances on the real exchange rate.

Acosta, Lartey, and Mandelman (2007) offer some empirical evidence on Dutch disease effects of remittances. These authors first develop a two-sector real business cycle model with remittances that produces Dutch disease effects in the remittance-receiving economy. The model is calibrated and simulated for El Salvador, and then impulse response functions from the model are compared to those calculated from a Bayesian vector autoregression (B-VAR) estimated on data for El Salvador from 1991–2006. The B-VAR response functions agree qualitatively with the model predictions, revealing evidence of Dutch disease effects of remittances (during this period, remittances increased 500 percent, and the country’s real exchange rate appreciated 30 percent).

A second means by which remittances may harm economic growth is through the moral hazard problem, an idea that was first formalized by Chami, Fullenkamp, and Jahjah (2003). Given that remittances are non-market income transfers that occur under asymmetric information and that monitoring and enforcement are extremely difficult because of the distance separating remitter and recipient, they may be plagued by severe moral hazard. The evidence discussed in Chapter 4 that remittances are compensatory transfers motivated by altruism supports such a view. The moral hazard problem manifests itself in two ways: recipients reduce their labor market effort and they make riskier investments. Anecdotal evidence of the labor effort effect is abundant, and academic studies have detected such an effect as well.³ The formal model developed in Chapter 6, which is designed to focus on households’ labor supply decisions, yields a similar conclusion. Reduced labor effort and increased investment risk lead to reduced economic growth.

An Instrument for Remittances

The set of variables used as an instrument for remittances in regressions is also an important potential source of the differences among estimation results in the studies discussed earlier in the chapter. Finding an appropriate instrument or set of instruments that corrects for the endogeneity of remittances has been a challenge for researchers. Two key features govern the selection of an instrument for remittances: the instrument must be correlated with remittances, and its effect on individual country growth must operate solely through its effect on remittances. Although two likely choices come to mind—GDP per capita, and growth in the developed countries where migrants from the remittance-receiving countries reside—both are also expected to have a direct impact on growth. Chapter 3 showed the first of these, GDP per capita, to be negatively correlated with remittance receipts, but it also affects economic growth directly through convergence. The second variable, GDP growth in developed countries where remitters reside, is likely to be correlated with trade flows,⁸ which in turn are expected to exert an independent impact on growth as well.

In general, the challenge in finding an appropriate instrument is that most variables that might explain remittances—domestic and foreign macroeconomic variables in particular—also tend to affect growth. For this reason, internal instruments (lagged right-hand-side variables) have been criticized (see especially World Bank, 2006), and migration and distance instruments have been suggested. These instruments may not be as great an improvement over internal instruments as they initially seem, however. Distance between migrants’ destination country and the remittance-receiving country is exogenous but time invariant, so it must be multiplied by host country GDP to obtain a time-varying instrument. Thus, distance instruments may be too strongly correlated with the growth rate in remittance-receiving countries. A similar argument can be made for migration instruments (migration shares are reported only periodically, so the migration shares are fixed and must be multiplied, again by host country GDP, to make the instrument time varying).

Thus, other determinants of remittances, such as their transaction costs, are likely candidates as instruments. In the absence of a direct observation of this cost variable, another (observable) variable might capture general trends in remittances throughout the world, including changes in transaction costs: the ratio of remittances to GDP of all other recipient countries (wrrow_i). Admittedly, this instrument does not eliminate all endogeneity, but it represents a significant improvement over internal, lag-driven instruments and over previous attempts at obtaining an external instrument. By excluding the remittances-to-GDP ratio of the country in question, wrrow_i is free of a direct causal link with other domestic macroeconomic variables. Furthermore, although one also expects wrrow_i to capture income growth in the developed world, the correlation with trade effects is diluted to the extent that, for a given country i, wrrow_i also incorporates the income movements in countries that have little trade with i. In other words, the diversification effect reduces any correlation between the instrument and the growth rate in the remittance-receiving country.

Using this variable as an instrument, the first-stage regression is given by

$w{r}_{it}={\alpha }_i+\beta wrro{w}_{it}+u_{it},$

View Expanded

where wr_it denotes the ratio of workers’ remittances to GDP in country i and year t, and wrrow_it denotes the ratio of workers’ remittances to GDP in the rest of the world—that is, in all countries except i—in year t. Thus, the first-stage regression includes the general world trend in remittances as an explanatory variable, along with a country-specific fixed effect to determine the average level of remittances for each given country. The second stage includes the fitted values from the first stage as a regressor.

Empirical Findings

Tables 7.1–7.4 present the main results of the growth regressions. For simplicity, the tables show only the coefficients for variables related to workers’ remittances, although three distinct combinations of a wide set of conditioning variables were included in the estimations. The basic conditioning set included initial per capita GDP; the ratios of trade and M2 to GDP, both in log terms; and the inflation rate. Through the addition of the ratio of domestic investment to GDP in the second conditioning set, a distinction could be made regarding whether remittances might have an impact on growth through higher investment or through higher TFP Finally, the full conditioning set included the following additional variables: foreign direct investment and the fiscal balance, both in relation to GDP; the rate of population growth; and the composite ICRG political risk indicator, as in Catrinescu and others (2006). Thus, each table includes results under each conditioning set, and both OLS and fixed-effects estimations are shown. Tables 7.1 and 7.2 report the results using wr as the remittance regressor, whereas Tables 7.3 and 7.4 report those using the second-stage fitted wr from the first-stage regression. In addition, results for the full country sample are reported in Tables 7.1 and 7.3, and those for the emerging economy subsample in Tables 7.2 and 7.4. Finally, each table shows the three alternative ways of including workers’ remittances identified in the previous subsection.

Table 7.1.

OLS and Fixed-Effects Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, All Countries

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable wr denotes the log of the ratio of workers’ remittances to GDP. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

*significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

Table 7.1.

OLS and Fixed-Effects Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, All Countries

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS	Fixed Effects	OLS	Fixed Effects	OLS	Fixed Effects
Sample: All Countries
Specification:
	1. wr only
		wr	0.200	0.172	0.171	0.202	0.056	−0.099
			(2.50)**	(0.90)	(2.21)**	(1.09)	(0.61)	(0.48)
		R 2	0.049	0.169	0.139	0.218	0.354	0.373
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66
	2. wr and wr-squared
		wr	0.125	−0.030	0.105	−0.017	−0.005	0.012
			(1.26)	(0.13)	(1.07)	(0.08)	(0.04)	(0.04)
		wr-squared	−0.033	−0.073	−0.028	−0.077	−0.028	0.032
			(1.27)	(1.48)	(1.09)	(1.62)	(0.74)	(0.59)
		R 2	0.051	0.175	0.139	0.226	0.352	0.375
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66
	3. wr and interaction with M2-GDP
		wr	0.798	0.550	0.570	0.423	0.039	−1.033
			(1.95)*	(0.83)	(1.45)	(0.65)	(0.08)	(1.18)
		wr × M2-GDP	−0.176	−0.111	−0.118	−0.065	0.005	0.296
			(1.49)	(0.60)	(1.03)	(0.35)	(0.03)	(1.09)
		R 2	0.052	0.170	0.176	0.218	0.350	0.379
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable wr denotes the log of the ratio of workers’ remittances to GDP. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

*significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

View Table

Table 7.1.

OLS and Fixed-Effects Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, All Countries

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS	Fixed Effects	OLS	Fixed Effects	OLS	Fixed Effects
Sample: All Countries
Specification:
	1. wr only
		wr	0.200	0.172	0.171	0.202	0.056	−0.099
			(2.50)**	(0.90)	(2.21)**	(1.09)	(0.61)	(0.48)
		R 2	0.049	0.169	0.139	0.218	0.354	0.373
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66
	2. wr and wr-squared
		wr	0.125	−0.030	0.105	−0.017	−0.005	0.012
			(1.26)	(0.13)	(1.07)	(0.08)	(0.04)	(0.04)
		wr-squared	−0.033	−0.073	−0.028	−0.077	−0.028	0.032
			(1.27)	(1.48)	(1.09)	(1.62)	(0.74)	(0.59)
		R 2	0.051	0.175	0.139	0.226	0.352	0.375
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66
	3. wr and interaction with M2-GDP
		wr	0.798	0.550	0.570	0.423	0.039	−1.033
			(1.95)*	(0.83)	(1.45)	(0.65)	(0.08)	(1.18)
		wr × M2-GDP	−0.176	−0.111	−0.118	−0.065	0.005	0.296
			(1.49)	(0.60)	(1.03)	(0.35)	(0.03)	(1.09)
		R 2	0.052	0.170	0.176	0.218	0.350	0.379
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable wr denotes the log of the ratio of workers’ remittances to GDP. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

*significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

Table 7.2.

OLS and Fixed-Effects Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, Emerging Economies

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable wr denotes the log of the ratio of workers’ remittances to GDP. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

Table 7.2.

OLS and Fixed-Effects Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, Emerging Economies

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS	Fixed Effects	OLS	Fixed Effects	OLS	Fixed Effects
Sample: Emerging Economies
Specification:
	1. wronly
		wr	0.207	0.192	0.182	0.223	0.053	−0.095
			(2.50)**	(0.97)	(2.28)**	(1.16)	(0.57)	(0.46)
		R 2	0.051	0.724	0.140	0.221	0.353	0.375
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	2. wr and wr-squared
		wr	0.119	−0.012	0.109	0.008	−0.005	0.008
			(1.16)	(0.05)	(1.08)	(0.03)	(0.04)	(0.03)
		wr-squared	−0.039	−0.073	−0.031	−0.076	−0.027	0.030
			(1.46)	(1.46)	(1.20)	(1.55)	(0.70)	(0.55)
		R 2	0.054	0.179	0.141	0.228	0.352	0.376
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	3. wr and interaction with M2-GDP
		wr	0.933	0.564	0.617	0.429	0.076	−1.068
			(2.14)**	(0.84)	(1.46)	(0.64)	(0.14)	(1.20)
		wr X M2-GDP	−0.216	−0.110	−0.129	−0.061	−0.007	0.308
			(1.69)*	(0.58)	(1.05)	(0.32)	(0.04)	(1.12)
		R 2	0.056	0.173	0.140	0.221	0.389	0.382
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable wr denotes the log of the ratio of workers’ remittances to GDP. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

View Table

Table 7.2.

OLS and Fixed-Effects Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, Emerging Economies

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS	Fixed Effects	OLS	Fixed Effects	OLS	Fixed Effects
Sample: Emerging Economies
Specification:
	1. wronly
		wr	0.207	0.192	0.182	0.223	0.053	−0.095
			(2.50)**	(0.97)	(2.28)**	(1.16)	(0.57)	(0.46)
		R 2	0.051	0.724	0.140	0.221	0.353	0.375
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	2. wr and wr-squared
		wr	0.119	−0.012	0.109	0.008	−0.005	0.008
			(1.16)	(0.05)	(1.08)	(0.03)	(0.04)	(0.03)
		wr-squared	−0.039	−0.073	−0.031	−0.076	−0.027	0.030
			(1.46)	(1.46)	(1.20)	(1.55)	(0.70)	(0.55)
		R 2	0.054	0.179	0.141	0.228	0.352	0.376
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	3. wr and interaction with M2-GDP
		wr	0.933	0.564	0.617	0.429	0.076	−1.068
			(2.14)**	(0.84)	(1.46)	(0.64)	(0.14)	(1.20)
		wr X M2-GDP	−0.216	−0.110	−0.129	−0.061	−0.007	0.308
			(1.69)*	(0.58)	(1.05)	(0.32)	(0.04)	(1.12)
		R 2	0.056	0.173	0.140	0.221	0.389	0.382
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable wr denotes the log of the ratio of workers’ remittances to GDP. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

Table 7.3.

OLS and Fixed-Effects Instrumental Variables Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, All Countries

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable fitted wr denotes the fitted value from a first-stage regression of the log of the ratio of workers’ remittances to GDP as a function of remittances in the rest of the world. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

Table 7.3.

OLS and Fixed-Effects Instrumental Variables Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, All Countries

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV
Sample: All Countries
Specification:
	1. wr only
		fitted wr	0.091	−5.667	0.044	−5.997	0.036	6.822
			(1.07)	(2.33)**	(0.53)	(2.56)**	(0.36)	(1.35)
		R 2	0.037	0.169	0.128	0.233	0.353	0.382
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66
	2. wr and wr-squared
		fitted wr	0.047	−5.525	0.011	−5.814	0.040	3.863
			(0.45)	(2.27)**	(0.10)	(2.49)**	(0.30)	(0.74)
		fitted wr-squared	−0.021	−0.951	−0.015	−1.245	0.002	−1.978
			(0.75)	(1.22)	(0.54)	(1.65)*	(0.04)	(1.88)*
	R 2		0.035	0.187	0.126	0.241	0.349	0.400
	Observations		383	383	374	374	189	189
	Countries		108	108	105	105	66	66
	3. wr and interaction with M2-GDP
		fitted wr	1.270	−3.005	0.854	−4.321	0.134	6.441
			(2.86)***	(1.10)	(1.98)**	(1.57)	(0.26)	(1.22)
		fitted wr X M2-GDP	−0.349	−0.652	−0.239	−0.389	−0.029	0.092
			(2.70)***	(2.07)**	(1.91)*	(1.16)	(0.19)	(0.25)
		R 2	0.052	0.195	0.134	0.237	0.350	0.382
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable fitted wr denotes the fitted value from a first-stage regression of the log of the ratio of workers’ remittances to GDP as a function of remittances in the rest of the world. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

View Table

Table 7.3.

OLS and Fixed-Effects Instrumental Variables Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, All Countries

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV
Sample: All Countries
Specification:
	1. wr only
		fitted wr	0.091	−5.667	0.044	−5.997	0.036	6.822
			(1.07)	(2.33)**	(0.53)	(2.56)**	(0.36)	(1.35)
		R 2	0.037	0.169	0.128	0.233	0.353	0.382
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66
	2. wr and wr-squared
		fitted wr	0.047	−5.525	0.011	−5.814	0.040	3.863
			(0.45)	(2.27)**	(0.10)	(2.49)**	(0.30)	(0.74)
		fitted wr-squared	−0.021	−0.951	−0.015	−1.245	0.002	−1.978
			(0.75)	(1.22)	(0.54)	(1.65)*	(0.04)	(1.88)*
	R 2		0.035	0.187	0.126	0.241	0.349	0.400
	Observations		383	383	374	374	189	189
	Countries		108	108	105	105	66	66
	3. wr and interaction with M2-GDP
		fitted wr	1.270	−3.005	0.854	−4.321	0.134	6.441
			(2.86)***	(1.10)	(1.98)**	(1.57)	(0.26)	(1.22)
		fitted wr X M2-GDP	−0.349	−0.652	−0.239	−0.389	−0.029	0.092
			(2.70)***	(2.07)**	(1.91)*	(1.16)	(0.19)	(0.25)
		R 2	0.052	0.195	0.134	0.237	0.350	0.382
		Observations	383	383	374	374	189	189
		Countries	108	108	105	105	66	66

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable fitted wr denotes the fitted value from a first-stage regression of the log of the ratio of workers’ remittances to GDP as a function of remittances in the rest of the world. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

Table 7.4.

OLS and Fixed-Effects Instrumental Variables Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, Emerging Economies

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable fitted wr denotes the fitted value from a first-stage regression of the log of the ratio of workers’ remittances to GDP as a function of remittances in the rest of the world. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

Table 7.4.

OLS and Fixed-Effects Instrumental Variables Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, Emerging Economies

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV
Sample: Emerging Economies
Specification:
	1. wr only
		fitted wr	0.094	−5.312	0.052	−5.732	0.035	6.877
			(1.06)	(2.08)**	(0.60)	(2.33)**	(0.34)	(1.34)
		R 2	0.038	0.183	0.128	0.233	0.388	0.384
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	2. wr and wr-squared
		fitted wr	0.039	−4.993	0.012	−5.348	0.039	3.926
			(0.36)	(1.95)*	(0.11)	(2.17)**	(0.30)	(0.74)
		fitted wr-squared	−0.027	−1.069	−0.019	−1.325	0.003	−2.145
			(0.91)	(1.29)	(0.64)	(1.66)*	(0.06)	(1.96)
		R 2	0.037	0.188	0.126	0.241	0.349	0.404
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	3. wr and interaction with M2-GDP
		fitted wr	1.536	−2.526	1.007	−3.998	0.187	6.488
			(3.22)***	(0.88)	(2.15)**	(1.38)	(0.34)	(1.20)
		fitted wr X M2-GDP	−0.431	−0.669	−0.284	−0.395	−0.046	0.092
			(3.08)***	(2.07)**	(2.07)**	(1.14)	(0.28)	(0.24)
		R 2	0.060	0.197	0.136	0.237	0.349	0.384
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable fitted wr denotes the fitted value from a first-stage regression of the log of the ratio of workers’ remittances to GDP as a function of remittances in the rest of the world. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

View Table

Table 7.4.

OLS and Fixed-Effects Instrumental Variables Regressions Explaining Per Capita GDP Growth as a Function of Workers’ Remittances and Different Conditioning Sets, Emerging Economies

			Conditioning Sets of Variables
			Basic Conditioning Set: Excludes Domestic Investment		Basic Conditioning Set Plus Domestic Investment		Full Conditioning Set, Including Institutional Variable
			OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV	OLS-IV	Fixed Effects-IV
Sample: Emerging Economies
Specification:
	1. wr only
		fitted wr	0.094	−5.312	0.052	−5.732	0.035	6.877
			(1.06)	(2.08)**	(0.60)	(2.33)**	(0.34)	(1.34)
		R 2	0.038	0.183	0.128	0.233	0.388	0.384
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	2. wr and wr-squared
		fitted wr	0.039	−4.993	0.012	−5.348	0.039	3.926
			(0.36)	(1.95)*	(0.11)	(2.17)**	(0.30)	(0.74)
		fitted wr-squared	−0.027	−1.069	−0.019	−1.325	0.003	−2.145
			(0.91)	(1.29)	(0.64)	(1.66)*	(0.06)	(1.96)
		R 2	0.037	0.188	0.126	0.241	0.349	0.404
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64
	3. wr and interaction with M2-GDP
		fitted wr	1.536	−2.526	1.007	−3.998	0.187	6.488
			(3.22)***	(0.88)	(2.15)**	(1.38)	(0.34)	(1.20)
		fitted wr X M2-GDP	−0.431	−0.669	−0.284	−0.395	−0.046	0.092
			(3.08)***	(2.07)**	(2.07)**	(1.14)	(0.28)	(0.24)
		R 2	0.060	0.197	0.136	0.237	0.349	0.384
		Observations	365	365	356	356	184	184
		Countries	104	104	101	101	64	64

Note: This table shows results of panel data regressions of real per capita real GDP growth on remittances and different sets of conditioning variables, as explained below. t-statistics are reported in parentheses. The variable fitted wr denotes the fitted value from a first-stage regression of the log of the ratio of workers’ remittances to GDP as a function of remittances in the rest of the world. Basic conditioning set: five-year averages of initial GDP per capita, the trade-to-GDP ratio, the M2-to-GDP ratio (all in logs), and the inflation rate. A five-year average of the log of the investment-to-GDP ratio is included in the second conditioning set. The full conditioning set includes, in addition, five-year averages of logs of ratios to GDP of foreign direct investment, the fiscal balance, and population growth, and the log of the five-year average of the composite ICRG political risk indicator. The table reports adjusted R² values for OLS regressions and within-R² values for fixed-effects regressions.

* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

The first two tables, which show estimations using wr, provide little evidence of an impact of workers’ remittances on economic growth through the TFP channel; a positive significant impact of remittances on growth arises only in a few OLS regressions, and mainly in the first column, where the conditioning set does not include the investment ratio. Once the investment ratio or country fixed effects are included, many significant impacts on growth disappear. Furthermore, the square of wr is often negative but does not reach statistical significance in any of the regressions, thus ruling out a quadratic effect of remittances on growth. Regarding the credit constraint hypothesis, although the term for the interaction between wr and the financial deepening variable (the M2-to-GDP ratio) tends to be negative across most regressions, it is significant in only one case: the OLS estimation for emerging economies that excludes the investment ratio (first column of Table 7.2). Thus, there might be a small effect of remittances easing credit constraints in countries with small banking systems. However, it should be stressed that such an effect would operate primarily through investment volume, since the interaction term becomes nonsignificant once the investment ratio is included, and such an effect is difficult to separate from the country fixed effects, since it also disappears once these are included.

One additional feature of the estimations is that there are only minor differences between the results for the full country and emerging economy samples. Since most industrial countries historically have not tended to report remittance inflows, few observations are lost when the growth regressions exclude these countries. The main difference in results was highlighted in the preceding paragraph: that the negative interaction between remittances and financial deepening appears to be slightly stronger within the emerging economy sample.

Tables 7.3 and 7.4 show the results of the instrumental variables estimations, using fitted wr as the relevant explanatory variable. Two main results contrast with those discussed in the last paragraph. First, some of the fixed-effects estimations, primarily those obtained under the second conditioning set, reveal a significant negative impact of remittances on economic growth. That is, the portion of domestic remittance inflows that is related to global trends in remittances appears to have a negative impact on economic growth. Given that the second conditioning set includes the investment rate, this effect must be operating primarily through a reduction in TFP. In some cases, the squared term of the fitted remittances is also negative and significant.

Second, the significance of the (negative) term for the interaction between remittances and financial deepening increases in the instrumental variables estimation. However, it is not always clear that the direct impact of workers’ remittances is positive; thus, whereas it appears that higher remittances coupled with greater financial deepening may be related to lower rates of economic growth, it is not clear whether even at very low levels of financial deepening remittances can have a positive impact. For example, in the second column of Table 7.4, the interaction term is negative and significant, but the direct impact is negative and nonsignificant. Thus, even in the most underdeveloped financial system, remittances would not have a positive impact on economic growth.

The estimation results show that it is difficult to obtain a robust positive effect of workers’ remittances on economic growth. In many of the specifications, the remittances-to-GDP ratio has no significant correlation with economic growth. A positive and significant coefficient on remittances appears only when investment is excluded from the estimation and in the absence of country fixed effects. To the extent that country fixed effects proxy for differences in investment, this specification is a better indication of remittances’ true contribution to growth. On the other hand, the country fixed-effects results may be indicating that the contribution of remittances to growth is highly dependent on individual country circumstances. The results suggest, moreover, that remittances may be reducing economic growth in many countries. When endogeneity is controlled for, the effect of remittances becomes negative and significant regardless of whether investment is excluded. This negative effect appears to be operating through a reduction in TFP, in accordance with the theoretical descriptions of how remittances can reduce growth.

Remittances and Macroeconomic Volatility

This section examines the relationship between macroeconomic volatility and remittances and relates it to the findings of previous chapters. The evidence presented in Chapters 4 and 5 suggests that macro-economic fluctuations exert a strong influence on remittances. But this same evidence, as well as the evidence analyzed in the previous section, suggests that remittances also affect economic fluctuations. As in the case of the correlation between remittances and GDP growth, there exist multiple pathways through which remittances can influence economic volatility, and these pathways imply contradictory effects.

Much of the theoretical evidence examined in Chapter 4 suggests that remittances are motivated by altruism, which implies a desire among migrants to compensate their households for the negative impacts of economic fluctuations in the home country. In addition, the empirical evidence presented in that chapter shows that remittances tend to be compensatory rather than opportunistic. In other words, remittances enable recipient households to smooth their consumption over time. This implies that if they are large enough, remittances will reduce economic fluctuations in a remittance-receiving country.

Remittances can also reduce the volatility of investment through two distinct pathways. First, because firms rely mostly on internal financing to fund their investments, smoother consumption implies smoother business earnings and hence smoother investment. Second, to the extent that remittances flow through the financial system, they may make it easier for firms to borrow and hence can enable firms to smooth their investment expenditures over time.

On the other hand, remittances may change recipients’ behaviors in ways that tend to increase economic volatility. This is a further implication of the moral hazard argument discussed in the chapter’s first section. First, there is a moral hazard in terms of labor income. If remittance recipients reduce their labor effort, this will increase the likelihood of poor firm performance, effectively imposing more risk on firms. Risk-neutral firms will react by adjusting labor contracts in ways that shift this risk back onto the households: by increasing the dispersion of wages and employment levels over the business cycle. The increased dispersion of firm earnings and wage income will then lead to increased economic volatility. Furthermore, the theoretical model presented in Chapter 6 also indicates that remittances may generate increased economic volatility if the presence of remittances causes household labor supply to become more procyclical. There is also a moral hazard in terms of investment effort. Recipients will choose riskier projects, or expend less effort on their existing investment projects, leading to an increased dispersion of investment returns and hence an increase in output volatility.

Two recent studies, IMF (2005) and World Bank (2006), estimated the correlation between remittances and output volatility.⁹ The IMF (2005), in conjunction with the growth estimations described in the previous section, found negative and significant relationships between the total remittances–to–GDP ratio and several measures of volatility: GDP volatility, consumption volatility, and investment volatility (all defined as the standard deviation of the annual growth rate of the variable). The study also used the largest annual decline in GDP over the period as an alternate measure of volatility and obtained similar results. The World Bank (2006) performed a panel estimation of the determinants of output growth volatility in conjunction with the growth regressions discussed in the previous section. In addition to total remittances–to–GDP, control variables were inflation, monetary policy, and fiscal policy volatility; real exchange rate overvaluation; frequency of banking crises; trade openness; terms of trade and foreign growth rate volatilities; country fixed effects; and time period effects. The study found a negative and significant coefficient on the ratio of total remittances to GDP that was robust to the different instruments used.

Given that the IMF and World Bank studies did not use the preferred definition of remittances, it is important to conduct new volatility estimations using this variable. Therefore, a cross-sectional regression was estimated for a sample of 70 countries, comprising 16 advanced economies and 54 developing countries. The dependent variable in the regression is defined as the standard deviation of real per capita GDP growth over the 1970–2004 period.¹⁰ The explanatory variables are similar to those that have been used in other studies examining output volatility (e.g., Easterly, Islam, and Stiglitz, 2001; Kose, Prasad, and Terrones, 2003)—relative income, relative income squared, terms of trade volatility, trade openness, financial openness, government consumption, institutional quality, an indicator of financial sector development, a trade concentration ratio, and an indicator of the commodity composition of exports—plus the ratio of workers’ remittances to GDP. Data sources and definitions of the variables are discussed in the appendix to this chapter. The explanatory variables are constructed as averages over the 1970–2004 period, except for the relative income variable, which is measured using its value in 1970, with at least 15 years of available data for a particular country required for inclusion of that country in the sample for the variable. Also, the averages for the variables are calculated including only data for those years for which data are present for all the explanatory variables included in the regression.

An OLS regression was estimated including all the possible explanatory variables in the regression. A preferred-specification regression was also conducted, with insignificant variables dropped from the regression, using a country sample identical to that employed in the regression that included all of the explanatory variables. The results of the cross-country regression, presented in Table 7.5, indicate that there is a negative relationship between workers’ remittances and the volatility of output and that this relationship is of marginal statistical significance. In practical terms, an increase in the workers’ remittances–to–GDP ratio of one percentage point leads to a reduction of 0.164 percent in the standard deviation of GDP growth, according to the regression results. This implies that countries with high workers’ remittances–to–GDP ratios experience significantly lower economic volatility than they would in the absence of remittances. Interestingly, the estimated coefficient on remittances has the same sign and nearly the same magnitude as the estimated coefficients in the studies mentioned previously. Figure 7.1 plots output volatility against the ratio of workers’ remittances to GDP for the countries in the regression sample and suggests that the negative relationship found in the regression would have been stronger if not for the presence of one outlier (Jordan).

Table 7.5.

Cross-Sectional Regression Explaining GDP Volatility as a Function of Workers’ Remittances and Conditioning Variables

Note: This table shows results of a cross-sectional regression explaining volatility of per capita output growth, defined as the standard deviation of the real GDP per capita growth rate over 1970–2004. The explanatory variables include remittances plus a set of conditioning variables, as explained below. t-statistics are reported in parentheses. All variables are calculated as averages over the 1970–2004 period. The conditioning set includes Terms of trade volatility, measured as the standard deviation of the annual change in the terms of trade; Financial openness, the ratio of the stock of foreign liabilities and foreign assets to GDP; Commodity export composition, measured as the share of primary commodities in total exports; Government consumption to GDP, measured as the ratio of government consumption to GDP in constant 2000 prices, both on its own and interacted with a dummy variable, Industrial, denoting industrial countries.

* significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

Table 7.5.

Cross-Sectional Regression Explaining GDP Volatility as a Function of Workers’ Remittances and Conditioning Variables

Conditioning Set of Variables
Workers’ remittances to GDP		−0.164
		(−1.68)*
Terms of trade volatility		0.090
		(1.95)*
Financial openness		0.012
		(2.04)**
Commodity export composition		0.016
		(1.79)*
Government consumption to GDP		0.066
		(1.75)*
Government consumption to GDP		−0.064
* Industrial		(−2.28)**
	R 2	0.374
	Observations	70

Note: This table shows results of a cross-sectional regression explaining volatility of per capita output growth, defined as the standard deviation of the real GDP per capita growth rate over 1970–2004. The explanatory variables include remittances plus a set of conditioning variables, as explained below. t-statistics are reported in parentheses. All variables are calculated as averages over the 1970–2004 period. The conditioning set includes Terms of trade volatility, measured as the standard deviation of the annual change in the terms of trade; Financial openness, the ratio of the stock of foreign liabilities and foreign assets to GDP; Commodity export composition, measured as the share of primary commodities in total exports; Government consumption to GDP, measured as the ratio of government consumption to GDP in constant 2000 prices, both on its own and interacted with a dummy variable, Industrial, denoting industrial countries.

* significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

View Table

Table 7.5.

Cross-Sectional Regression Explaining GDP Volatility as a Function of Workers’ Remittances and Conditioning Variables

Conditioning Set of Variables
Workers’ remittances to GDP		−0.164
		(−1.68)*
Terms of trade volatility		0.090
		(1.95)*
Financial openness		0.012
		(2.04)**
Commodity export composition		0.016
		(1.79)*
Government consumption to GDP		0.066
		(1.75)*
Government consumption to GDP		−0.064
* Industrial		(−2.28)**
	R 2	0.374
	Observations	70

Note: This table shows results of a cross-sectional regression explaining volatility of per capita output growth, defined as the standard deviation of the real GDP per capita growth rate over 1970–2004. The explanatory variables include remittances plus a set of conditioning variables, as explained below. t-statistics are reported in parentheses. All variables are calculated as averages over the 1970–2004 period. The conditioning set includes Terms of trade volatility, measured as the standard deviation of the annual change in the terms of trade; Financial openness, the ratio of the stock of foreign liabilities and foreign assets to GDP; Commodity export composition, measured as the share of primary commodities in total exports; Government consumption to GDP, measured as the ratio of government consumption to GDP in constant 2000 prices, both on its own and interacted with a dummy variable, Industrial, denoting industrial countries.

* significant at 10 percent; ** significant at 5 percent; ***significant at 1 percent.

View Full Size

Output Volatility and the Ratio of Workers’ Remittances to GDP

(In percent)

Sources: World Bank (2006) and authors’ calculations.

• Download Figure
• Download figure as PowerPoint slide

The results of the foregoing estimation seem to imply that the volatility-dampening effects of remittances out-weigh the volatility-increasing effects described earlier. But there are several reasons why the empirical results do not necessarily support this conclusion. First, there is a data measurement issue. The theoretical model underlying the moral hazard effect is a business cycle model, similar to that used in Chapter 6, so its predictions are most relevant to variables measured at business cycle frequencies, say, quarterly or annually. The empirical exercises use long-run estimates of volatility out of necessity, so their results are not directly applicable to the theoretical model. Second, the theoretical model assumes that resources, most notably labor, are fully employed. In reality, most countries that receive remittances have high rates of under- or unemployment. Thus, the average household in remittance-receiving countries probably does not exhibit the strong labor-leisure trade-off present in the theoretical model, like those calibrated in the cash and credit economies of Chapter 6. Finally, given the evidence on remittances and investment presented in Chapter 4 as well as in the preceding section, any impact remittances may have on the riskiness of investment is probably too small to detect in aggregate data.

Nevertheless, the empirical results support the idea that remittances reduce macroeconomic volatility over long horizons. Yet the exercise does not shed light on the exact mechanism by which remittances reduce such volatility. Given the analysis on motives, intended uses, and end uses of remittances from Chapter 4 and the simulation results from Chapter 6, it appears likely that remittances reduce output volatility at the aggregate level because they dampen consumption volatility at the household level. The increased smoothness of consumption has a direct impact on measured GDP volatility, since consumption accounts for a large share of GDP. This will remain a conjecture, however, until detailed longitudinal studies of household consumption and investment, including both households that do and those that do not receive remittances, are conducted.

Workers’ Remittances and the Equilibrium Real Exchange Rate

As indicated in Chapter 5, one of the most important potential macroeconomic effects of remittance inflows is on the recipient country’s equilibrium real exchange rate. Changes in the equilibrium real exchange rate not only affect the distributional impacts of remittance inflows, both by altering the returns to factors employed in the traded and nontraded goods sectors and by affecting the relative price of traded and non-traded consumption goods, but may also be one of the mechanisms through which remittance flows exert their main impact on long-run growth, via Dutch disease effects. Deriving the theoretical implications of remittance inflows for the long-run equilibrium real exchange rates in recipient countries requires the use of a macroeconomic model. Surprisingly, there has been relatively little analytical work on t his issue. The discussion in this section is based on Montiel (2006), which explored the effects of remittance receipts on the longrun equilibrium real exchange rate in the context of a fairly standard two-sector open economy model. The equilibrium real exchange rate in the model is defined in Nurksian terms as the value of the real exchange rate that is simultaneously consistent with internal and external balance, conditioned on sustainable values of the economy’s underlying real fundamentals.

External balance refers to a situation in which the ongoing current account deficit is financed by sustainable capital inflows. In the model of Montiel (2006), this condition generates a positive trade-off between the real exchange rate and real domestic consumption (measured in units of traded goods), because an increase in domestic consumption creates an excess demand for traded goods, requiring a real depreciation to generate the offsetting excess supply required to maintain the trade balance at the level that can be financed by sustainable capital flows. The resulting locus is depicted as curve EB in Figure 7.2. On the other hand, internal balance refers to a situation in which the market for nontraded goods is in equilibrium at full employment. This condition generates a negative trade-off between the real exchange rate and real domestic consumption, since the excess demand for nontraded goods caused by an increase in domestic consumption requires a real appreciation to sustain equilibrium in the market for nontraded goods. The implied internal balance locus is depicted as curve IB in Figure 7.2. The intersection of these loci at point A in Figure 7.2, where external and internal balance hold simultaneously, determines the long-run equilibrium real exchange rate.

View Full Size

Remittances and the Equilibrium Real Exchange Rate

• Download Figure
• Download figure as PowerPoint slide

Interpreted as an exogenous transfer from the rest of the world, workers’ remittances are a component of the current account and thus affect the position of the EB curve. However, since remittances have no direct effect on the market for nontraded goods, changes in remittance flows leave the IB curve undisturbed.¹¹ For a given level of capital inflows, a larger inflow of remittances permits the economy to sustain a larger trade deficit, and thus a more-appreciated real exchange rate, without violating the external balance condition. Thus, an increase in remittance receipts shifts the EB curve downward, resulting in a more-appreciated long-run equilibrium real exchange rate and a higher level of real domestic consumption, as shown at point B. Note that the quantitative effect of the change in remittance flows on the equilibrium real exchange rate depends on the elasticities of the external and internal balance curves, such that the more elastic these curves are, the smaller the change in the equilibrium real exchange rate resulting from a change in the amount of remittances. Thus, alterations in remittance flows have a smaller impact on the equilibrium real exchange rate the greater the degree of substitutability between traded and nontraded goods in both production and consumption.

Elements of the preceding analysis underlie the standard presumption that higher remittance flows are likely to result in an appreciation of the equilibrium real exchange rate. However, the analysis relies on several special assumptions. Montiel (2006) shows the following:

• If remittance inflows include a component that is inversely related to real income in the recipient country, a change in the exogenous component of remittance receipts has weaker effects on the equilibrium than in the absence of this endogenous component.

• Similarly, if remittance receipts are disproportionately devoted to spending on traded goods (say, because remittances are transferred in kind, or because they are used to import consumer durables), their effects on the long-run equilibrium real exchange rate tend to be weakened, and in the limit can be eliminated altogether. For example, if all remittance receipts are devoted to spending on traded goods, then the EB and IB curves in Figure 7.2 both shift to the right by exactly equal amounts in response to an increase in remittances. In that case, long-run domestic real consumption increases by the amount of increased remittance income, but the long-run equilibrium real exchange rate remains unchanged.

• Somewhat surprisingly, if the receiving country’s external creditors treat remittance receipts as part of that country’s wealth, and if an increase in remittance receipts consequently lowers the risk premium the country faces in international capital markets, then a permanent increase in such receipts gives rise to a transitory consumption boom in the recipient country, but its long-run equilibrium real exchange rate remains unaffected.

Based on these considerations, the presumption from theory is that a permanent increase in remittance inflows is associated with an appreciation of the recipient economy’s long-run equilibrium real exchange rate. However, under certain empirically plausible conditions, this effect may be weak, or even absent altogether. Thus the effect of remittance receipts on the equilibrium real exchange rate is an empirical question.

The next step for understanding the relationship between remittances and the real exchange rate, beyond casual data analysis, is to include remittances in the standard exchange rate estimation. Despite their empirical importance for many countries and the strong theoretical presumption that remittance inflows affect the equilibrium real exchange rate, the literature on equilibrium real exchange rate estimation has not typically incorporated remittance flows into the set of real exchange rate fundamentals. The studies that do incorporate remittances into this set of fundamentals have focused on the experience both of individual countries and of various country groupings as well. Such studies typically include remittance flows in the set of fundamentals that enter a cointegrating equation for the real exchange rate, thus controlling for other potential real exchange rate determinants in a single-country or panel context. An early single-country study by Bourdet and Falck (2003) examined the effect of workers’ remittances on the equilibrium real exchange rate in Cape Verde over the period 1980–2000, confirming the conventional view that an increase in remittance receipts is associated with an appreciation of the equilibrium real exchange rate. Similar results were obtained by Hyder and Mahboob (2005), who found that higher remittance inflows tended to appreciate the equilibrium real exchange rate in Pakistan during 1978–2005, as well as by Saadi-Sedik and Petri (2006), who derived the same result for Jordan over 1964–2005.

For a sample of six Central American and Caribbean countries, Izquierdo and Montiel (2006) obtained mixed results over the period 1960–2004. They followed a procedure aimed at identifying cointegrating relationships between the real effective exchange rate and a set of nonstationary fundamental variables. Starting with a full set of possible fundamental variables—which included, in addition to the ratio of workers’ remittances to GDP, a measure of average labor productivity; the ratios to GDP of government consumption and of the international investor position; trade openness; and terms of trade—their procedure eliminated those variables that did not form part of a cointegrating vector or those whose sign was not theoretically appropriate. Table 7.6 shows the final cointegrating vectors Izquierdo and Montiel obtained. In Honduras, Jamaica, and Nicaragua, the authors found no influence of workers’ remittances on the equilibrium real exchange rate—that is, workers’ remittances were not part of a cointegrating vector with the real effective exchange rate—despite the fact that these countries received very large remittance inflows over the last half of their sample. For the Dominican Republic, El Salvador, and Guatemala, however, remittance inflows turned out to be important determinants of the equilibrium real exchange rate, with an increase in remittance inflows having a much more powerful effect on equilibrium real exchange rate appreciation in El Salvador and Guatemala than in the Dominican Republic.

Table 7.6.

Cointegrating Relations for the Real Exchange Rate, 1960–2004

Source: Izquierdo and Montiel (2006), Table 1. Note: The log of the real effective exchange rate is included in the cointegrating equation. An increase (decrease) in the real effective exchange rate corresponds to a depreciation (appreciation). Standard errors are shown in parentheses.

Table 7.6.

Cointegrating Relations for the Real Exchange Rate, 1960–2004

		Dominican Republic	El Salvador	Guatemala	Honduras	Jamaica	Nicaragua
Workers’ remittances/GDP		−1.085	−35.255	−44.611
		(0.720)	(0.444)	(2.937)
Other nonstattonary fundamentals
	Average productivity of labor	−0.174	6.626		−5.340	−7.544	22.091
		(0.409)	(0.662)		(0.814)	(0.725)	(5.088)
	Government consumption/GDP		−0.410	−0.010	−0.028	−0.025	−0.008
			(0.012)	(0.006)	(0.008)	(0.002)	(0.019)
	Trade openness				0.008	0.003
				(0.008)	(0.000)
	Terms of trade			−0.006	−0.003	−0.007	−0.065
				(0.001)	(0.001)	(0.000)	(0.010)
	International investor position/GDP				−0.004	−0.003
					(0.000)	(0.000)
Time trend			0.241	0.069	−0.032	−0.061	0.222
			(0.004)	(0.005)	(0.005)	(0.003)	(0.065)

Source: Izquierdo and Montiel (2006), Table 1. Note: The log of the real effective exchange rate is included in the cointegrating equation. An increase (decrease) in the real effective exchange rate corresponds to a depreciation (appreciation). Standard errors are shown in parentheses.

View Table

Table 7.6.

Cointegrating Relations for the Real Exchange Rate, 1960–2004

		Dominican Republic	El Salvador	Guatemala	Honduras	Jamaica	Nicaragua
Workers’ remittances/GDP		−1.085	−35.255	−44.611
		(0.720)	(0.444)	(2.937)
Other nonstattonary fundamentals
	Average productivity of labor	−0.174	6.626		−5.340	−7.544	22.091
		(0.409)	(0.662)		(0.814)	(0.725)	(5.088)
	Government consumption/GDP		−0.410	−0.010	−0.028	−0.025	−0.008
			(0.012)	(0.006)	(0.008)	(0.002)	(0.019)
	Trade openness				0.008	0.003
				(0.008)	(0.000)
	Terms of trade			−0.006	−0.003	−0.007	−0.065
				(0.001)	(0.001)	(0.000)	(0.010)
	International investor position/GDP				−0.004	−0.003
					(0.000)	(0.000)
Time trend			0.241	0.069	−0.032	−0.061	0.222
			(0.004)	(0.005)	(0.005)	(0.003)	(0.065)

Source: Izquierdo and Montiel (2006), Table 1. Note: The log of the real effective exchange rate is included in the cointegrating equation. An increase (decrease) in the real effective exchange rate corresponds to a depreciation (appreciation). Standard errors are shown in parentheses.

Given the small set of countries examined in single-country studies to date, it is difficult to generalize from these results. However, other researchers have used panel methods to examine the effects of remittance inflows on the real exchange rate for larger samples of countries. Amuedo-Dorantes and Pozo (2004), for example, used a panel with 13 Latin American and Caribbean countries over the period 1978–98 and found that an increase in workers’ remittances appreciated the real exchange rate. Holzner (2006) derived the same result using a worldwide sample. In contrast with these results, Rajan and Subramanian (2005) found, for a sample of 15 countries during the 1990s, that higher remittance receipts were not associated with slower growth in manufacturing industries with higher labor intensity or greater export orientation, as one might expect if remittance receipts are associated with Dutch disease effects operating through an appreciated real exchange rate.

Thus, although neither the single-country nor panel evidence speaks with a single voice, most of the research to date is consistent with the conventional presumption that higher remittance receipts tend to appreciate the equilibrium real exchange rate. The implication is that if Dutch disease effects are indeed present, the beneficial short-run effects of remittance inflows on economic welfare in the recipient countries through higher and more stable levels of consumption may come at the expense of reduced long-run growth. Chapter 8 considers the policy challenge posed by this trade-off.

Remittances, Fiscal Policy, and Debt Sustainability

The notion that remittances have a significant effect on fiscal policy and debt sustainability may at first be surprising, since governments have no direct claims on these person-to-person transfers. The fact that remittances enter the recipient economy through family transfers means that remittances affect fiscal policy and debt sustainability indirectly through the activities of remittance-receiving households, primarily through their consumption decisions and saving patterns. In this respect remittances are quite different from natural resources, which governments may own and from which they derive revenue, and public aid transfers, which enter the government budget constraint directly. Since remittances contribute to higher consumption of domestic and imported goods, they may affect government revenues through consumption- and trade-based taxation.¹² Furthermore, remittances may lead to increased deposits in the banking system and, to the extent that the marginal propensity to consume is less than unity, they may increase the level of private saving. Both of these channels may affect fiscal policy through credit market activity. As a result, remittances can play an important role in the assessment of a country’s debt sustainability, since they alter the fiscal balance and the evolution of the stocks of public and private sector liabilities over time.

To illustrate this concept, this section examines a simplified economy in which the government issues only domestic-currency-denominated debt. Furthermore, it is assumed that the household in this simplified economy receives remittances only in terms of domestic currency.¹³ Sustainability conditions are derived from the household and government budget constraints to illustrate the channel through which remittances alter the accumulation of liabilities and affect debt sustainability. After debt sustainability conditions are derived in this simplified setting, the more complex case with foreign-currency-denominated debt and the need to transfer remittances across the exchange rate is considered.

The government’s intertemporal budget constraint in the presence of remittances, derived explicitly in Box 7.1, is useful for understanding the impact of remittances on fiscal policy choices. Since the stock of debt issued during the previous period, B_t, must be taken as a given state variable, increases in remittances that do not result in a one-to-one increase in household consumption will support new sequences of taxes, money growth, and bond issuance. For example, given a stream of future tax revenues chosen by the fiscal authority and a future stream of money growth chosen by the monetary authority, an increase in remittances will be met with an increase in future debt issuance, B_t+i+1, for some future period. The increase in government bond issuance can be viewed as a mechanism to absorb additional levels of household saving, S_t+i, and support the household’s desire to smooth consumption across time. If these flows were instead channeled through the financial system, an increase in remittances that led to additional household saving could also lead to increases in financial system liabilities, which in turn could lead to additional private and public sector credit provision.

Box 7.1.

Remittances and Fiscal Sustainability

Remittances are unrequited, nonmarket personal transfers between households across countries, and as such they enter the household budget constraint as an addition to income separate from the domestic production process. Previously accumulated stocks of money balances (M) and real government bonds (B), income from production (Y) net of taxes (T), and real remittance transfers (Rem) are all used to finance household expenditures (C). In this simplified setting the aggregate household budget constraint is

$C_t+\frac{M_{t+1}}{P_t}+B_{t+1}=Y_t+Re{m}_t-T_t+\frac{M_t}{P_t}+B_t{R}_t,$

View Expanded

where R is the gross domestic interest rate, or R = (1 + r), where r is the net domestic interest rate, and P is the price level. The government uses taxes, money creation, and real bond issuance to finance its expenditures (G) according to

$G_t+B_t{R}_t\frac{M_t}{P_t}=T_t+\frac{M_{t+1}}{P_t}+B_{t+1}.$

View Expanded

Under the assumption that household and government consumption includes public and private investment, the economy-wide resource constraint is

$C_t+G_t=Y_t+Re{m}_t.$

View Expanded

A clearer picture of the effect of remittances on fiscal policy choices, including debt creation, can be obtained through examination of the intertemporal government budget constraint. Substituting for successive bond terms in the government budget constraint yields

$B_t={\displaystyle \sum _{i=0}^I{q}_i}\left[T_{t+i}-G_{t+i}+\frac{M_{t+1}}{P_{t+i}}\left({\mu }_{t+i+1}-1\right)\right]+q_I{B}_{t+I+1},$

View Expanded

where μ is the growth rate of nominal money balances and

$q_i={\displaystyle \prod _{n=1}^{i+1}\frac{1}{R_{t+n-1}}.}$

View Expanded

The usual interpretation of this exercise is that a positive stock of debt in the present period must eventually be paid for by generating fiscal surpluses or money creation. Using the economy-wide resource constraint to substitute for the future sequence of government spending results in

$B_t={\displaystyle \sum _{i=0}^I{q}_i}\left[T_{t+i}-\left(\underset{s_{t+i}}{\underbrace{Y_{t+i}+Rem-C_{t+i}}}\right)+\frac{M_{t+1}}{P_{t+i}}\left({\mu }_{t+i+1}-1\right)\right]+q_I{B}_{t+I+1,}$

View Expanded

where S_t+i is the level of household saving.

However, should the fiscal authority not want to increase future debt levels, the increase in remittances given a future stream of monetary policy will be met by an increase in future tax revenue. Though remittances do not enter directly into the government budget constraint and empirical evidence suggests that governments do not tax remittances directly, governments receive tax revenue from remittances indirectly. As shown in Chapter 6, additional tax revenue may be collected with the least amount of distortion through a consumption-based tax system.¹⁴ Consequently, remittances can be viewed as part of the potential tax base in addition to labor income from production, depending on the tax structure in place. Finally, for a given stream of future tax revenue and bond issuance by the fiscal authority, the monetary authority may also meet an increase in remittances with an increase in money creation. In this regard, inflows of remittances have macroeconomic policy implications similar to those of inflows of market-based capital, and countries that operate within inflation-target regimes need to conduct liquidity operations to deal with any changes in liquidity resulting from remittances.

To examine some of the implications of remittances for fiscal policy and debt levels, data were assembled from the IMF World Economic Outlook database and the World Bank’s World Development Indicators for the top 20 remittance-receiving economies, based on the average ratio of remittances to GDP between 1990 and 2005. Averages were computed for the ratio of workers’ remittances to GDP, credit to the government as a percentage of total credit, the net general government debt-to-GDP ratio, and government consumption as a percentage of GDP. The minimum and maximum average remittances-to-GDP ratios in this sample were 3.1 and 18.3 percent, respectively. As the countries in the data set are a subset of remittance-receiving economies, any results drawn from this data set should be interpreted as relating to economies that can be characterized as significantly remittance dependent.¹⁵

As shown in the first panel of Figure 7.3, the ratio of workers’ remittances to GDP has a positive relationship with the ratio of credit to the government to total credit, indicating that the banking system tends to channel additional household saving from remittances into credit provision to the government as opposed to the private sector. The figure’s second panel plots the relationship between the ratio of average workers’ remittances to GDP and the average net general government debt-to-GDP ratio. The positive relationship suggests that remittances also tend to result in higher levels of public sector debt. Finally, the third panel in Figure 7.3 plots the relationship between the ratios of average workers’ remittances to GDP and average government consumption to GDP in remittance-dependent economies. The data indicate that the government’s additional financing, whether channeled through the banking system or through an increase in government debt issuance, results in higher government consumption.

View Full Size

Remittance-Dependent Economies and Fiscal Policy, 1990–2005

Sources: IMF (2006b) and World Bank (2006).Note: Variables are averages for each country between 1990 and 2005. Remittance data include workers’ remittances while excluding employee compensation and migrants’ transfers.

• Download Figure
• Download figure as PowerPoint slide

Given the discussion in earlier chapters regarding the stability of remittance flows and the focus in this chapter on total household resources (i.e., income from production plus remittances) when the intertemporal budget constraint is examined, a more accurate representation of debt sustainability for a country that receives significant remittance flows should employ as a base a more-aggregated measure of income than GDP. For example, gross national disposable income (GNDI) could be used instead of GDP when the evolution of liabilities in an open economy setting is computed. However, GNDI includes net factor income from nonresidents and public transfers, including grants, and may therefore be inappropriate, depending on the composition of flows that a particular country receives. Net factor income may not be suitable for inclusion as part of the potential revenue base of the fiscal authority, and public sector transfers may be lumpy and inconsistent over time. An alternative to using GNDI would be to construct a measure of GDP plus net current private transfers.

Normalizing the intertemporal government budget constraint in Box 7.1 by the sum of income from production and remittances yields the traditional debt sustainability relationship in terms of the debt-to-GNDI ratio, less net factor income and public transfers. Setting i = 0 and ignoring the use of money creation in debt financing results in the following equation describing the approximate evolution of debt in terms of growth rates:

View Expanded

where b is the stock of domestic-currency-denominated government debt, r is the net interest rate, π is the growth rate of the GDP deflator, γ is the growth rate of real GDP, η is the growth rate of remittances in domestic currency terms, and (t_t − g_t) is taxes less noninterest government spending, or the primary fiscal balance. In this case, for a given set of remaining variables, increases in the growth rate of remittances in domestic currency units improve debt sustainability.

A more complete derivation of (7.1) to assess debt sustainability would also include exchange rate effects on remittances and foreign currency debt issued by the public sector. In addition to the sustainability of public sector debt, remittances also affect external sustainability through their inclusion in net transfers as part of the noninterest current account balance. Box 7.2 discusses the role of remittances and external sustainability. Including the government’s foreign currency debt and the need to translate remittance flows across currencies results in a debt evolution equation of

View Expanded

Box 7.2.

Remittances and External Sustainability

Assessments of external sustainability are a key element in IMF surveillance of member countries and involve forming a view of how outstanding stocks of liabilities are likely to evolve over time. External debt evolves according to

$D_{t+1}=\left(1+r\right)D_t-CA{B}_t,$

View Expanded

where D represents the stock of public and private external debt and CAB the noninterest current account balance in U.S. dollars. Separating the components of the noninterest current account means the equation can also be written as

$D_{t+1}=\left(1+r\right)D_t-\left[T{B}_t+In{c}_t+T{r}_t\right],$

View Expanded

where TB is the balance on goods and services, Inc represents the balance on income less interest, and Tr is net current transfers, which includes workers’ remittances.

Normalizing the preceding equation by nominal GDP results in an external debt-to-GDP ratio of

$d_{t+1}=\frac{\left(1+r\right)}{\left(1+\gamma \right)\left(1+\rho \right)}{d}_t-t{b}_t-in{c}_t-t{r}_t,$

View Expanded

where γ is the growth rate of real GDP and ρ is the growth rate of the U.S. dollar value of the GDP deflator.¹ Here, lowercase variables (d, tb, inc, tr) are used to denote ratios to GDP (of D, TB, Inc, and Tr, respectively). The change in the debt-to-GDP ratio is then

$d_{t+1}-d_t=\frac{\left(r-\gamma -\rho -\gamma \rho \right)}{\left(1+\gamma +\rho +\gamma \rho \right)}{d}_t-t{b}_t-in{c}_t-t{r}_t.$

View Expanded

According to this debt dynamics equation, an increase in the level of the remittances-to-GDP ratio, all else equal, improves external sustainability. Remittances also have indirect beneficial effects on debt dynamics to the extent that their presence reduces external borrowing costs and causes the domestic currency to appreciate.

However, as mentioned in the chapter text, the improvement in debt dynamics should be qualified when the empirical results regarding the cyclicality of remittance flows and any potential adverse effects of real exchange rate appreciation are considered. An increase in domestic GDP relative to GDP abroad will improve debt dynamics through the coefficient on d_t but will be offset somewhat by a decline in remittances (e.g., transfers) and any increased demand for imports. Conversely, a relative increase in GDP abroad will lead to higher remittance inflows and possibly increase the demand for exports, both of which should lead to improvements in sustainability. Finally, external sustainability will improve if remittance transfers result in a real appreciation of the domestic currency, but such an appreciation will also adversely affect exports.

1 External sustainability analysis uses variables in U.S. dollars following traditional balance of payments accounting, whereas fiscal sustainability is analyzed in terms of domestic currency. The variable ρ captures the growth rate of the U.S. dollar value of the GDP deflator, which is similar to the use of changes in the GDP deflator (π) and the exchange rate (ε) in equations (7.1) and (7.2).