Journal Issue
Share
Article

Impact of Remittances on Poverty and Financial Development in Sub-Saharan Africa

Author(s):
Sanjeev Gupta, Catherine Pattillo, and Smita Wagh
Published Date:
February 2007
Share
  • ShareShare
Show Summary Details

I. Introduction

The flow of remittances into developing countries is attracting increasing attention because of their rising volume and their impact on the receiving countries. In 2005, they totaled US$188 billion—twice the amount of official assistance developing countries received.2 Moreover, there is evidence that such flows are underreported. Remittances through informal channels could add at least 50 percent to the globally recorded flows (World Bank, 2006).3 Since 2000, remittances to developing countries have increased on average by 15 percent in annual terms. Though at least some part of the growth is attributable to better reporting by recipient countries, it appears that over the last decade remittances have outpaced private capital flows and official development assistance (World Bank, 2006).

Remittances are perceived as being more stable than other external flows. To the extent that migrants are motivated by altruism and send more money home in times of economic distress, remittances may actually be countercyclical. The stability of these inflows also opens up an opportunity for developing countries to lower borrowing costs in international capital markets by securitizing future flows of remittances.4 Because remittance receipts are widely dispersed, they may not cause the real exchange rate to appreciate; they may also obviate the deleterious effect on home country institutions observed in short-lived natural resource booms.

There are marked regional differences in remittance flows.5 Since the 1980s, remittances to countries in Latin America, the Caribbean, and the East Asia and Pacific regions have grown more rapidly than the average for developing countries generally. In 2005, the top three recipients—China, India and Mexico—accounted for more than one-third of the remittances to developing countries. Among the top 25 recipients of remittances, only one (Nigeria) is in sub-Saharan Africa (SSA) but three of the eight countries in South Asia (Bangladesh, India, and Pakistan) appear on the list.6

Studies using household-level data from individual countries in SSA have yielded some insights into how remittances are used at the micro level. In studying the impact of remittances at the aggregate level most analysts have concentrated on Latin America or South Asia, where the volumes swamp those going to SSA. But at their core remittances are private intrafamily/intracommunity income transfers that directly address the single most relevant challenge for SSA—poverty. Further, the long-term development potential of such transfers is determined by the use of the portion of remittances left over after basic consumption needs are met. The purpose of this paper is to study both these issues in a part of the world where the role of remittances has received comparatively little attention.

This paper analyzes the size and significance of remittance flows to SSA. Section II documents the volume and characteristics of remittances to the region, and discusses the dimensions and the related cost of brain drain from SSA countries. Section III estimates their impact; first the immediate consumption effect of remittances on poverty is investigated, using a cross-section dataset comprised in significant proportion of SSA countries. This is followed by the analysis of the indirect consequence of remittances. Because migrant transfers entail cross-border flows of relatively modest sums of money to low-income households, they present an opportunity for these households to access formal financial services. The paper therefore investigates how remittances affect financial development in SSA countries. Section IV concludes with a discussion of the market for money transfers in SSA and suggests how to enhance the effectiveness of remittances in the region.

II. Remittances TO Sub-Saharan Africa

A. Recent Trends

Sub-Saharan Africa has been part of the increasing global trend; remittances to SSA have increased by over 55 percent in U.S. dollar terms since 2000, while they increased for developing countries as a group by 81 percent.7 However, the recorded remittances are only a small fraction of total remittances to SSA. Freund and Spatafora (2005) estimate that informal remittances to SSA are relatively high at 45–65 percent of formal flows, compared to only about 5–20 percent in Latin America.

In 2005, remittances to the 34 SSA countries reporting are estimated to have been about US$6.5 billion. Remittance flows to SSA are relatively small, 4 percent of total remittances to developing countries and just 33 percent of those to India, which receives the most. In contrast, countries in Latin America and the Caribbean received 25 percent of all remittances, as did the countries of the East Asia and Pacific region.8

Relative to GDP, too, the volume of remittances to SSA is generally smaller than in other developing countries. On average remittances in the region are about 2.5 percent of GDP, compared to almost 5 percent for other developing countries. However, there are striking exceptions in SSA. In particular, remittances were almost 28 percent of GDP in Lesotho, and more than 5 percent in Cape Verde, Guinea-Bissau, and Senegal. In absolute terms, however, Kenya, Nigeria, and Senegal are the largest recipients of remittances in the region.

For some countries, remittances are also an important source of foreign exchange. For Lesotho, Cape Verde, Uganda, and Comoros, for instance, remittances have since 2000 amounted on average to more than 25 percent of export earnings (Figure 1).

Figure 1.Top Ten Recipients of Remittances in Sub-Saharan Africa

Source: IMF, Balance of Payments Yearbook, 2006; World Economic Outlook, 2006; World Bank staff estimates.

Note: Rankings are based on average remittance inflows for 2000–05.

In SSA, aid flows are considerably higher than remittance receipts (Figure 2). Since 2000 aid flows to the region increased on average by about 13 percent a year and reported remittances by almost 10 percent. However, during the 1990s, when aid flows to the region were more or less stagnant, remittances grew annually at more than 13 percent. And in 2005 when aid flows to the region (excluding Nigeria) fell, remittances were stable (OECD/DAC, 2006). While it is true that the region as a whole receives more aid than recorded remittances, for countries like Lesotho, Mauritius, Nigeria, Swaziland, and Togo, remittances are consistently greater than official assistance.

Figure 2.Inflows to SSA countries, 1975-2004

(Millions of U.S. dollars)

Source: IMF Balance of Payments Yearbook, 2006; IMF African Department database, 2006; OECD/DAC database 2006.

The balance of payments data used above probably underreports remittance flows between developing countries. Despite the paucity of records there is reason to believe that intraregional migration is common in SSA. Botswana and South Africa tend to attract migrants from neighboring countries (largely unskilled) in search of employment. The strong sociocultural ties in West Africa also encourage labor mobility. In East Africa, political turmoil seems to be the driving force in migration.

B. Characteristics of Remittances to SSA

One reason remittances have attracted attention is that they are seen as more stable than other foreign currency flows to developing countries. This is especially relevant to SSA, where official aid flows have fluctuated considerably from year to year. Remittances to SSA are not just consistently less volatile than official aid, they are also less volatile than FDI, which is usually seen as the most stable private flow (Figure 3a). In the 1990-2004 period however export earnings are more stable than remittances.

Remittances might also be expected to be countercyclical to the extent that they are motivated by the altruism of migrant workers and increase in times of economic distress in their home countries. Remittances to SSA are counter-cyclical only in the 1980s (Figure 3b). Since 1990 remittances have been procyclical, though less so than either official aid or export earnings. The low (though positive) correlation coefficient demonstrates the stability of remittances over time rather than any strong relationship to growth cycles.9 The countercyclicality of FDI flows in the latter time period must be viewed in the context of the very high volatility of such flows.

Figure 3.Volatility and Cyclicality of External Flows to Sub-Saharan Africa

Source: IMF Balance of Payments Yearbook, 2006; IMF African Department database, 2006; World Bank staff estimates; OECD/DAC, 2006.

Remittances can also contribute to stability by lowering the probability of current account reversals. Because they are a cheap and stable source of foreign currencies, remittances are likely to stem investor panic when international reserves are falling or external debt is rising. These beneficial effects are particularly strong for countries where remittances are above 3 percent of GDP (Bugamelli and Paterno, 2006). While the average SSA ratio is just below that threshold and currrent account reversals driven by investor panic are rare, for some countries this effect might be an additional benefit from remittances.

The impact of remittances on the real exchange rate and export competitiveness, their Dutch disease effect, is a matter of debate. As in the case of any other transfer (for instance, official aid) the effect depends on the proportion of such flows spent on domestic goods, in particular non-tradables (Gupta, Powell, and Yang, 2006). Since remittances are private transfers dispersed over a large number of poor households it has been argued that their impact on domestic demand differs from that of donor-funded infrastructure projects (World Bank, 2006). Remittances may in fact be self-correcting as an overvalued currency deters remittances, and hence Dutch disease effects are not sustained (Rajan and Subramanian, 2005). However, studies in Latin America (Amuedo-Dorantes and Pozo, 2004) and Cape Verde (Bourdet and Falck, 2006) have found evidence that remittances do have Dutch disease effects on the competitiveness of the tradable sector. In countries where remittances inflows are large compared to the size of the economy, where supply constraints are a significant hindrance to the expansion of the nontradables sector, and where a significant portion of remittances are spent on domestic goods policymakers will need to be alert to the possibility of a Dutch disease phenomenon.

C. Remittances and Brain Drain

Remittances are only one dimension of the phenomenon of migration from low-income countries. In particular, skilled migration has always been associated with concerns about brain drain, which might be especially costly for some SSA countries (Kapur and McHale, 2005; Carrington and Detragiache, 1998). Pond and McPake (2006) detail the human resource crisis in the health sector in SSA countries that arises as skilled health care professionals increasingly find employment in the high-demand OECD countries. They calculate that almost a quarter of the new overseas-trained physicians that registered with the U.K.’s National Health Service between 2002 and 2003 came from SSA. Similarly, Bach (2006) documents the high job vacancy rates in the public health systems of countries like Ghana due to large-scale migration. He estimates that in Zambia and Zimbabwe the annual rate of attrition in public health employment can range from 15 to 40 percent.

On average 20 percent of the SSA tertiary-educated population older than 15 work in OECD countries. Less than 10 percent of the comparator group from South Asia is found there. For some countries, such as Angola, Guinea-Bissau, and Mozambique, expatriation rates are in excess of 50 percent of the educated population.

We look into the issue of brain drain by calculating the difference between the expatriation rates of the educated over-15 population from country i and the rate at which the general over-15 population migrate to an OECD country.10 Because the emigrant population tends to be better educated, it is to be expected that in general the difference between the educated and the general expatriation rates will be positive. With a few exceptions, such as Mexico, Turkey, Bulgaria, and several OECD countries, this holds true. Moreover, the larger the difference between the educated and general expatriation rates, the higher the propensity of skilled workers to emigrate compared to the general propensity to emigrate.11

There are interesting regional differences in the extent to which the educated exceeds the general expatriation rate (Figure 4). Within the OECD countries there is almost no difference. Among developing countries the largest difference is observed for SSA countries, reflecting the strain on domestic economies from skilled emigration.12

Figure 4.Regional Expatriation to OECD Countries

(Percent)

Source: OECD, Trends in International Migration database, 2006.

Note: The data are from census and labor force surveys carried out in OECD countries in or about 2000.

For some countries in SSA the shortage of skilled personnel can be quite severe; more than a third of their educated workforce migrates (Table 1). Among the top 10 countries listed, six are from SSA. Among the top 20 countries, 75 percent are in SSA.

Table 1.Expatriation Rates: Top Ten Countries
Educated

Expatriation Rate
General

Expatriation Rate
Difference
Guinea-Bissau70.43.666.8
Haiti68.08.859.2
Mozambique52.30.851.5
Angola53.82.951.0
Trinidad and Tobago66.122.143.9
Jamaica72.630.642.0
Mauritius50.39.341.0
Guyana76.936.540.4
Gambia42.42.639.8
Burundi35.00.334.7
Source: OECD, Trends in International Migration database, 2006.Note: Countries are ranked by the difference between the educated and the general expatriation rates.
Source: OECD, Trends in International Migration database, 2006.Note: Countries are ranked by the difference between the educated and the general expatriation rates.

III. Impact OF Remittances

A. Direct Income and Consumption Effect of Remittances

In SSA, remittances are part of a private welfare system that transfers purchasing power from relatively richer to relatively poorer members of a family or community. They reduce poverty, smooth consumption, affect labor supply, provide working capital, and have multiplier effects through increased household spending. Anecdotal evidence suggests that most often women head the recipient households.

For the most part, remittances seem to be used to finance consumption or investment in human capital, such as education, health, and better nutrition.13 In Zimbabwe, for instance, households with migrants have less cultivated land but tend to be slightly better educated (de Haan, 2000). Quartey and Blankson (2004) find that migrant remittances to Ghana are in fact countercyclical and are effective in helping smooth household consumption and welfare over time, especially for food crop farmers, who are typically the most disadvantaged socioeconomic group. Similarly, using data from a large household survey Adams (2006) finds that international remittances significantly relieved poverty among the “poorest of poor households.” Ratha (2003) suggests that remittances that raise the consumption levels of rural households might have substantial multiplier effects because they are more likely to be spent on domestically produced goods. Some studies (Hanson and Woodruff, 2003; Cox Edwards and Ureta, 2003) have found evidence for “forward” linkages between remittances and human capital formation in Latin America.

The evidence on the direct impact of remittances on poverty and inequality seems to vary according to the sample (Adams, 1991; Barham and Boucher, 1998). Earlier studies posited that migration was likely to increase rural inequality because only relatively better-off households were able to finance a member’s search for better employment in urban areas or abroad (Stahl, 1982; Lipton, 1980). More recently, it has been found that migration patterns in East European and former Soviet Union countries are such that richer households receive greater remittances than do poorer households (World Bank, 2007). However, Koechlin and Leon (2006) find that as migrant communities form close networks in a foreign country, the cost of migration falls and remittances no longer reinforce inequalities in the recipient country. Other localized studies have concluded that remittances tend to improve the welfare of poorer rural households (Stark and Taylor, 1989; Adams, 1991). Studies covering a larger sample of countries have found evidence that remittances tend to lower poverty (Adams and Page, 2005; Spatafora, 2005).

In the rest of this section, we investigate the direct poverty-reducing impact of remittances using a sample that gives greater representation to SSA countries than other studies.14

Empirical model

We use a methodology similar to that of Adams and Page (2005), to examine the impact of incoming remittances on poverty. We build on their model by adopting the three-stage least squares estimation technique that allows for the simultaneous determination of poverty and remittances. Based on Ravallion (1997) and Ravallion and Chen (1997) we model poverty as a function of mean income, some measure of income distribution, and the variable of interest, remittances. The baseline specification is

(i = 1 ....N, t = 1 .....Ti),

where P is poverty in country i at time t; αi captures fixed effects; μ is per capita income, which functions as a measure of average consumption; g is income inequality as measured by the Gini index; and x is remittances. The model assumes that poverty is reduced as mean income rises; hence, β1 is expected to be negative. Based on previous studies we expect higher poverty to be associated with greater income inequality; hence, β2 is expected to be positive. Controlling for these two variables the model estimates the sign and magnitude of β3, which indicates the direct impact of remittances on poverty.

Data

Making use of poverty surveys beginning in 1980, the dataset consists of 76 countries and 233 observations.15 SSA countries are substantially represented: 23 percent of the observations come from the 24 SSA countries in the sample. To our knowledge giving this weight to SSA countries is atypical for cross-country studies on remittances.16

The poverty and inequality measures used here are from the World Bank’s PovcalNet database,17 which incorporates various measures of poverty: headcount poverty measures the percentage of the population living on less than one PPP dollar a day. The poverty gap, the mean distance below the poverty line as a proportion of the poverty line, tells us how poor the poor are—how far below the poverty line the average poor person’s income is. The squared poverty gap, which is the mean of the squared distance below the poverty line as a proportion of the poverty line, is more sensitive to the distribution of the poor below the poverty line. The income distribution measure, the Gini coefficient, is available from the same survey data.

Remittances are expressed as a ratio of the GDP of recipient countries. The income variable is per capita GDP in constant 2000 U.S. dollars. Other variables used in the three-stage estimation are educational attainment, proxied by average years of schooling for the over-25 population, and openness, measured by the ratio of imports plus exports to GDP. These variables are all measured as five-year averages corresponding to the survey year in the PovcalNet database. (Appendix Tables 3 and 4 provide detailed descriptions of the raw dataset.)

Results

The following estimation techniques were applied to equation 1. The log transformation of all the variables allows us to interpret the coefficients as elasticities. Regional dummies have been introduced to control for fixed effects.

Ordinary least squares (OLS) estimates from our sample conform to the predictions of the model (Appendix Table 5). Regardless of the measure of poverty used as the dependent variable, per capita income has a negative and significant coefficient. A positive and significant coefficient for the Gini index indicates that greater inequality is associated with higher poverty. We estimate a negative elasticity between poverty and incoming remittances; this result is quite consistent. Except in the case where the left side variable is the squared poverty gap, this result is always significant. Prima facie our findings indicate that a 10 percent rise in the inflow of remittances is associated with about a 1 percent fall in headcount poverty and the poverty gap.18 In keeping with the regional focus of the paper, we also introduce an interaction term between remittances and a dummy for SSA (Appendix Table 6). While the overall poverty reducing effect of remittances remains, the coefficient on the interaction term comes in with a positive sign. Although this effect is not always significant it raises the possibility that in SSA the severity of poverty might be motivating greater out-migration, so that poverty is positively associated with remittances.19 The issue of reverse causality is taken up next.

Ordinary least squares estimates are likely to be biased when any right side variable is endogenous. Moreover, we can argue that the relationship between poverty and remittances is unlikely to be unidirectional. To tackle this issue a system estimation technique that allows for both poverty and remittances to be determined simultaneously is adopted. Three-stage least squares is often described as the system equivalent of a two-stage least squares.20 The advantage is that estimating a system of equations where both poverty and remittances are endogenously determined allows us to observe not just the effect of remittances on poverty, but also the reverse effect of poverty of remittances. The price for this is that a misspecification error in one of the system equations is transmitted through the system.

The specification for the poverty equation is the same as in equation 1. We also estimate remittances (Rem) as a function of poverty (Pov), trade openness (Trade), schooling (Sch), distance (Dist) from the main remittance source country, a dummy for dual exchange markets (Dual), and lagged remittances (Remt-1).

(i = 1 .....N, t = 1 ....Ti),

Migration is the best determinant of remittances, but migration data are likely to suffer from the same problems as data on remittances. Thus we use other variables suggested by the literature on the motivation to migrate and remit. To the extent that remittances represent a private welfare transfer, we can expect them to be higher where there is widespread poverty; hence, we expect a positive sign for β1. If labor mobility and commodity trade are complementary in more open economies, we can also expect a positive sign on the openness variable. If, on the other hand, goods mobility substitutes for labor mobility, β2 would be less than zero.

The sign of β3 is subject to two countervailing influences. Because the general tendency is for the migrant population to be better educated than the general population, we can expect more schooling to be associated with greater migration and remittances. At the same time, educational attainment also serves as a proxy for development in the recipient country and hence more years of schooling may indicate less need to seek employment abroad. The distance variable here is the geographic distance between the recipient country and the OECD country with the largest migrant population from the recipient.21 The expectations for the sign of its coefficient are ambiguous. On the one hand, because distance captures the difficulty of migration, one can expect β4 to be less than zero. On the other hand, because of the implication that it takes a more educated migrant to overcome the higher cost of migration, one can expect higher remittances from the source country. Restrictions in the foreign exchange market can be deterrent to remittances (or at least the flows going through formal channels) and hence β5 is expected to have a negative sign. And finally, given the stability of out-migration and remittance flows, we can expect the previous period’s remittances to be a significant predictor of this period’s remittances, and hence β6 is expected to be greater than zero.

Table 1 reports the results from the three-stage least squares estimation. The hypothesis of reverse causality between poverty and remittances finds support in the positive coefficient on poverty as a right hand side variable when remittances are endogenously modeled. Trade openness is also a consistently positive and significant determinant of remittances in this two equation system. As expected lagged remittances are significant, positive predictor of current remittances. For our sample of countries, none of the other control variables are significant determinants of remittances.

Table 2.Three-Stage Least Squares Estimation
Headcount PovertyPoverty GapSquaredPoverty Gap
PovertyRemittancesPovertyRemittancesPovertyRemittances
Per capita GDP (constant 2000 dollars)-1.14***

(-10.06)
-1.33***

(-10.47)
-1.38***

(-9.43)
Gini coefficient1.97***

(4.16)
1.96***

(3.66)
2.44***

(4.04)
Inflow of remittances (ratio to GDP)-0.15***

(-2.86)
-0.11**

(-1.89)
-0.08

(-1.23)
Poverty0.21*

(1.86)
0.19**

(1.98)
0.21**

(2.08)
Schooling-0.05

(0.20)
-0.05

(0.19)
0.09

(0.34)
Trade openness0.65***

(2.71)
0.68***

(2.87)
0.65***

(2.65)
Distance0.01

(0.08)
0.08

(0.58)
0.10

(0.72)
Dual exchange market (dummy)-0.01

(-0.05)
-0.02

(-0.07)
-0.02

(-0.09)
Lagged remittances0.70***

(11.52)
0.69***

(11.52)
0.69***

(11.36)
Europe and Central Asia-1.94***

(-4.69)
-1.05**

(-2.28)
-0.38

(-0.72)
East Asia and Pacific-0.40

(-1.15)
-0.50

(-1.27)
-0.20

(-0.44)
Latin America and Caribbean-0.16

(-0.44)
0.60

(1.34)
0.71

(1.50)
Middle East and North Africa-1.86***

(-4.87)
-1.72***

(-4.03)
-1.57***

(-3.24)
Sub-Saharan Africa-0.70*

(-1.97)
-0.28

(-0.71)
-0.18

(-0.40)
Constant12.32***

(12.47)
-3.13*

(-1.96)
11.92***

(10.96)
-3.54**

(-2.24)
11.61***

(9.21)
-3.45**

(-2.10)
Observations156156155155152152
Adj R 2.720.530.700.540.640.55
F-Statistic51.4530.7145.9330.9534.6131.02
Note: ***,**,and *, indicate significance at the 1, 5 and 10 percent.
Note: ***,**,and *, indicate significance at the 1, 5 and 10 percent.

The effect of per capita income and income inequality is consistent with the OLS results. When endogenously determined in this manner, the poverty-reducing effect of remittances still remains, and the magnitude of this effect is very similar to the OLS estimates. However, the average remittance-inducing elasticity of poverty is consistently greater than the average poverty-reducing elasticity of remittances. This suggests that for SSA countries in the sample the impact of poverty on out-migration and remittances might be greater than the impact of remittances on poverty.

B. Impact on Financial Development

The immediate welfare-enhancing role of remittances is critical at both the household and the country level. However, it does not fully explain the usefulness of remittances as a source of development finance. To understand how remittances affect long-term growth potential we next turn our attention to an indirect consequence of cross-border money transfers: their impact on financial development. Because migrant transfers entail cross-border flows of relatively modest sums of money, they present an opportunity for low-income households to access formal financial services. This most likely begins with savings products but the growing interest that microfinance institutions have shown in this segment of the market raises the possibility of access to small business start-up capital for individuals previously excluded from the formal sector.

The impact of remittances on growth depends on how recipient households use them. Once again empirical studies yield an array of possibilities. One view is that remittances would mostly be used for consumption, sometimes even conspicuous consumption, and that the same community characteristics that led to migration also dampen the productive use of incoming remittances. Caceres and Saca (2006) find that in El Salvador remittances were accompanied by a sharp decline in savings, so that economic activity actually contracted. Yet Woodruff and Zenteno (2001) estimate that remittances accounted for about 20 percent of the capital invested in microenterprises in urban Mexico and that the impact is stronger for female-owned businesses. Lucas (1987) found that any effects on rural output of the loss of labor due to migration to South African mines from Botswana, Lesotho, and Malawi are offset in the long run by investments in farm technology. However, Rozelle, Taylor, and deBrauw (1999) estimate that farm investments only partially offset the decline in rural output due to migration.22

Given the decentralized decision-making process that characterizes the use of remittances, it is difficult to gauge their aggregate effect. The impact of remittances on growth in crosscountry studies is inconclusive. Studies that focus on the labor supply response of recipient households find that remittances lower growth (Chami, Fullenkamp, and Jahjah, 2003; Azam and Gubert, 2005). However studies that link remittances to investment, where remittances either substitute for or improve financial access, tend to conclude that remittances stimulate growth (Giuliano and Ruiz-Arranz, 2005; Toxopeus and Lensink, 2006). While the evidence on the contemporaneous impact of remittances on growth may be mixed, it is likely that remittances can affect long-term growth by fostering financial deepening.

The positive impact of financial development on growth has been extensively documented (Levine, 1997, 2004; Rajan and Zingales, 1998; Beck, Demirguc-Kunt, and Levine, 2004). For SSA countries in particular, lack of access to formal financial services is a significant impediment to financial deepening (Gulde and others, 2006). Migrant transfers can create an avenue for unbanked households to avail themselves of some of the products offered by formal financial providers.

Data and model

We investigate the impact of remittances on financial development in SSA countries using an unbalanced panel of 44 countries and six time periods, composed of five-year averages from 1975 through 2004. Our baseline specification closely follows Aggarwal, Demirguc-Kunt, and Peria (2006), but we restrict our sample to observations from SSA only. Financial development is alternatively proxied by the ratio of bank deposits to GDP and the ratio of M2 to GDP. Remittances are measured in relation to recipient country GDP, as defined elsewhere in the paper. The regressions also include the following control variables:23

  • The size of the economy is captured by the log of GDP in constant U.S. dollars.
  • Per capita GDP is a proxy for the degree of institutional development.
  • Inflation is measured as the annual change in the CPI.
  • A dummy variable signifies a dual exchange rate regime as a measure of capital account openness.
  • The ratio of import and exports to GDP proxies current account openness.
  • The sum of FDI and development assistance to GDP serves as an alternative measure of openness.

The core model can therefore be written as

where i identifies the cross-section and t the time period, Rem is the variable of interest, X is the vector of control variables, αi captures the country-specific effect, and uit is the error term.

Results

Table 3 reports the results from both the random and the fixed effects panel regressions. In all instances remittances are significant as a positive determinant of financial development. For SSA countries the size of the economy seems unrelated to financial development.24 Similarly, while per capita GDP seems to significantly affect financial development, the magnitude of the effect is surprisingly small. Capital and current account openness are both associated with greater financial development. 25

Table 3.Baseline Panel Estimation
Random EffectsFixed Effects
DepositsM2DepositsM2
Remittances to GDP0.65***

(2.66)
0.44***

(3.51)
0.56*

(1.87)
0.47***

(3.21)
Log(GDP)3.06***

(2.47)
1.68

(1.21)
2.32

(1.15)
0.06

(0.02)
Per capita GDP0.01***

(7.39)
0.01***

(5.48)
0.01***

(5.92)
0.01***

(4.57)
Inflation-0.003

(-0.52)
0.002

(0.43)
-0.001

(-0.11)
0.004

(0.71)
Dual-3.46**

(-1.98)
-3.04

(-1.61)
-3.92**

(-2.18)
-3.21*

(-1.66)
Trade openness0.05

(1.62)
0.09***

(2.51)
0.08**

(2.04)
0.15***

(3.18)
Other capital flows to GDP0.2

(1.87)
-0.02

(-0.18)
0.21*

(1.70)
-0.06

(0.46)
Constant-62.75**

(-2.30)
-23.32

(-0.76)
-50.69

(-1.18)
6.09

(0.12)
Observations150162150162
Adj R 20.460.350.360.29
Note:***,**, and * signify 1, 5, and 10 percent significance levels.
Note:***,**, and * signify 1, 5, and 10 percent significance levels.

Once again these estimates can be biased by endogeneity between financial development and remittances. It can be argued that better-developed financial institutions have a positive effect on remittances flowing through formal channels. To address this we adopt three instrumental variables from Aggarwal, Demirguc-Kunt, and Peria (2006) based on macroeconomic conditions in source countries. Unemployment, GDP growth, and per capita GDP in the source country, while related to remittances, are independent of financial development and other conditions in the recipient country. The results are reported in Table 4.

Table 4.Fixed Effects Panel Instrumental Variables Estimation
DepositsM2
(1)(2)(1)(2)
Instrumented Variable
Remittances to GDP3.47**

(2.61)
2.67***

(2.37)
0.39

(0.44)
4.75***

(2.99)
Exogenous Variables
Log(GDP)2.36

(0.68)
-3.80

(-0.61)
-2.07

(-0.43)
-10.75

(-1.25)
Per capita GDP0.02***

(6.40)
0.014***

(2.75)
0.02***

(5.64)
0.01**

(1.30)
Inflation0.01

(0.18)
-0.001

(-0.29)
0.004

(0.68)
0.003

(0.50)
Dual-3.85

(-1.52)
-4.10*

(-1.78)
-3.25

(-1.49)
-2.45

(0.71)
Trade openness0.05

(0.88)
-0.09

(-1.20)
0.18***

(3.48)
-0.05

(-0.42)
Other capital flows to GDP0.19

(1.09)
0.48***

(2.70)
-0.03

(-0.15)
0.14

(0.65)
Corruption-1.67

(-1.20)
-2.77

(-1.39
Internal conflict-1.97**

(2.46)
-2.80**

(-2.61)
Political risk0.69**

(2.66)
0.99***

(2.82)
Constant-64.15

(-0.86)
68.61

(0.52)
44.86

(0.44)
223.66

(1.25)
Observations1348914593
Cragg Donald F-statistic for weak instruments2.053.472.133.18
Adj R 20.080.120.400.41
Note:***,**, and * signify 1, 5, and 10 percent significance levels.
Note:***,**, and * signify 1, 5, and 10 percent significance levels.

The instrumented remittances variable comes in with a positive coefficient of a magnitude greater than previously estimated. While the impact of per capita GDP on financial development is consistent with the panel estimation, in this specification current and capital account openness are less significant. Source country variables do not perform very strongly as instruments for remittances in our sample, although the Cragg Donald statistic is above the critical value. 26In general, however, the estimated effect of remittances on financial development in SSA compares well with the effect estimated by Aggarwal and others (2006) using a larger sample.

IV. Improving the Effectiveness of Remittance Flows

A. Channeling Remittance Flows to Formal Providers

While remittances can facilitate the entry of households into formal financial markets, only a fraction of the sums remitted by migrant workers from SSA finds its way into the formal system. The high fees formal providers charge is a deterrent for poor migrants who want to send small sums of money home, and even if a migrant has access to banks the recipient may not. So migrants rely more on import-export operators, retail shops, and currency dealerships—but there are no records of the transactions these conduct (Sander and Maimbo, 2005). Informal money transfer systems modeled closely on the hawala system in the Middle East dominate the remittance market in several African countries (El Qorchi, Maimbo,and Wilson, 2003). Informal providers offer numerous client-friendly features, such as anonymity, minimal paperwork, and speed.

The cost of transferring funds, especially small sums, is indeed high. A survey of money transfer operators (MTO) in the U.K. found that the fee on money transfers was lower in high-volume corridors like U.K.-India and higher for UK-Africa (Table 5).27

Table 5.Fees for Remittances Sent Through Money Transfer Operators in the U.K.(Percent of amount)
GhanaKenyaNigeriaIndiaSpeed of transfer
Transfer amount£100£500£100£500£100£500£100£500
Cheque point3364.255n.a.n.a.Up to 24 hours
First Remit54.2n.a.n.a.54.254.2Up to 24 hours
Money Gram127.2127.27512410 minutes/instant
Travelex Money Transfer7.54.8n.a.n.a.7.54.8n.a.4.810 minutes/instant
Western Union126.4147.4126.440.810 minutes/instant
Source: DFID, 2006.Note: Since the fees can change due to exchange rate changes, the number should be interpreted as indicative rather than precise.
Source: DFID, 2006.Note: Since the fees can change due to exchange rate changes, the number should be interpreted as indicative rather than precise.

The market in money transfers between developing countries in SSA is underserved by formal institutions, and the prohibitive fees they charge severely depress their use. A study in South Africa (Genesis Analytics, 2003) found that the comparative cost of an R250 international transfer was the lowest when a “friend” or taxi driver was used to effect the transfer and highest when banks were used. Though cross-border Post Office transfers are competitively priced, they are not as fast or as secure. Table 6 compares the cost of remitting R300 from South Africa by provider and method of transfer.

Table 6.Fees for Remittances from South Africa(Percent of 300 Rand Transfer)
Amount Received in Local Currency
Method of

Transfer
ProviderTransfer

Fee
BotswanaLesothoMalawiMozambique
Bank draftFNB52.6120.8142.12688.5142.1
Nedbank68.276.495.52005.595.5
Standard Bank35.0195.0195.0195.0195.0
ElectronicABSA33.3178.3200.04370.0815660.0
FNB52.6120.8142.12688.5142.1
Nedbank62.590.0112.52362.5112.5
Postbank19.2242.5254.0
Standard Bank61.7115.0115.0115.0115.0
Mail transferPostbank8.2275.5278.5275.5
MoneygramStandard Bank25.3224.0n.a.
OnlineiKobo6.2247.7281.51239469.9
Source: DFID and FinMark Trust, 2006.Note: Shaded cells indicate the rand value of the money transfer since the transfer is converted to local currency by the receiving organization.
Source: DFID and FinMark Trust, 2006.Note: Shaded cells indicate the rand value of the money transfer since the transfer is converted to local currency by the receiving organization.

The absence in South Africa of a major MTO like Western Union further limits competition among the players in the formal market and increases the likelihood that migrant workers will use informal channels to send money home. The terrorist attacks of September 11, 2001, have increased the scrutiny of international money transfers and many banks are imposing more identification requirements on both individuals and small MTOs (Sander and Maimbo, 2003). In South Africa only authorized dealers, who must have a banking license and have invested in an expensive exchange control reporting system, can remit funds. By further increasing the effective cost the rules discourage remittances through formal channels (Genesis Analytics 2005).

Banks are not always interested in the small remittances market. Most analysts see significant opportunities for banks to reduce the transaction costs on remittances, especially small remittances sent by poor migrants. Freund and Spatafora (2005) find that concentration in the banking sector, financial risk, and exchange rate variability typically increase transaction costs. Financial sector reforms that address any or all of these structural problems in the receiving and sending countries are also likely to lower the cost of remittances.28 Cross-border uniformity in the regulations related to remittances and regulatory interventions where fees are prohibitive have been proposed as other cost-reducing measures (Ratha and Riedberg, 2005; Sanders and Maimbo, 2005).

Among formal providers many smaller banks and microfinance institutions have already gauged the untapped potential of this market. Where there is a long history of migration some small banks have adapted to the needs of the migrant community. For instance, Theba Bank, a miners’ bank, offers low-cost transfers from South Africa to families that have bank accounts in Mozambique and Swaziland (Orozco, 2003). International Remittance Network (IRnet) consists of about 200 credit unions that offer low-cost services in 40 countries in Asia, Africa, Europe, and Latin America (Samuels, 2003). The network does not require that the receiving family have an account with a credit union.

Lately, in well-developed financial markets like the United States the growing demand for remittance services has caught the attention of major commercial banks like Citizen’s Bank and Wells Fargo. These banks see remittance services as an effective way to draw the attention of a significant unbanked population to their more mainstream financial products. In an arrangement with two banks in Cape Verde Citizen’s Bank offers Cape Verdean migrants a remittance facility that is low cost compared to Western Union. In three years of operation this program has made over 1,000 formerly unbanked migrants Citizen’s customers.29

There are already signs that the window of opportunity for financial institutions to tap into this highly profitable and rapidly innovating market might be narrow. Recent strides in cell phone encryption technology have facilitated fast, low-cost money transfers between OECD countries and recipient countries as diverse as the Philippines and Zambia, allowing customers to avoid the higher fees and longer waiting periods associated with MTOs and banks (Jordan, 2006).

B. Using Remittances Effectively

Bringing recipient households into the formal financial sector is only the first step in using remittances more effectively. Country-specific surveys indicate that while typically a large proportion of remittances are spent the propensity to save from remittances among some households can be as high as 40 percent (UNDP, 2005). For policy-makers the challenge is to channel these savings into productive uses.

Most studies indicate that remittances not used to pay for the immediate consumption needs of the recipient household are used for human capital development or conspicuous consumption. While the long-term benefits of the former are apparent, not all conspicuous consumption is wasteful. The construction of very large houses for migrant workers in West Africa has spurred local economic activity through multiplier effects. In Mexico, the Sociedad Hipotecaria Federal, a government financial institution established to build primary and secondary mortgage markets, provides long-term financing and partial mortgage insurance to Mexican sofols (mortgage providers) that extend loans to immigrants for housing construction (Serrano, 2006). The loans are denominated in Mexican pesos. Migrant workers are given some flexibility about the method of income verification and there is no credit history requirement. Mortgage payments are made in the workers place of residence. Inadequate financial infrastructure makes launching of similar schemes in Africa more challenging, but they can spur a sustained housing boom with positive spillovers on both real and financial sectors of the economy.

By bundling financial services like savings products and entrepreneurial loans for remittance-receiving households, financial institutions, especially banks, can activate the investment channels through which remittances can promote growth.30 Given the paucity of assets that can serve as collateral in SSA a steady future flow of migrant remittances could be used to secure small business loans—though small retail businesses started entirely with remittance savings face expansion limits unless they can access additional long-term funding. 31

The surge in remittances to India over the last few years is attributed in some part to incentive schemes launched by the government such as the Resurgent India Bond to encourage the inflow of diaspora savings. While such flows are more likely to be subject to speculative reversals than intrafamily transfers they can significantly supplement domestic investible resources (World Bank, 2006).32

Remittances are not a panacea for all that ails low-income countries. They cannot be a substitute for a sustained, domestically engineered development effort. Moreover, large-scale migration can have a deleterious effect on domestic labor markets in specific sectors, particularly where those leaving are largely skilled workers. Nevertheless, migrant transfers can help ease the immediate budget constraints of recipient households. For developing countries as a whole they are a larger transfer of resources than all development assistance and have a more direct impact on poverty. And the vast untapped market in money transfers is an opportunity for small savers to gain a foothold in the formal financial sector.

APPENDIX
VariableSource
Remittances (sum of receipts of worker remittances, employee compensation, migrant transfers)Balance of Payments (supplemented by World Bank staff estimates for 2005)
Poverty Regressions
Poverty indicatorsPovcalNet database (available at http://iresearch.worldbank.org/PovcalNet/jsp/index.jsp.)
Gini indexPovcalNet database (available at http://iresearch.worldbank.org/PovcalNet/jsp/index.jsp.)
Per capita GDP (constant 2000 US dollar)World Development Indicators
Schooling (average schooling years among over 25 population)Barro-Lee database
Trade openness ((imports + exports)/GDP)World Development Indicators
Dual exchange market dummyAnnual Report on Exchange Arrangements and Exchange Restrictions, IMF
Financial Development Regressions
M2/GDPInternational Financial Statistics
Bank deposits/GDPInternational Financial Statistics
GDP (constant 2000 U.S.$)World Development Indicators
Per capita GDP (constant 2000 U.S $)World Development Indicators
Inflation (annual percentage change in CPI)World Development Indicators
Trade openness ((imports + exports)/GDP)World Development Indicators
Foreign direct investmentWorld Economic Outlook
Official development assistanceOECD/DAC database
Dual exchange market dummyAnnual Report on Exchange Arrangements and Exchange Restrictions, IMF
General and Educated Expatriation RateOECD Trends in International Migration database
CorruptionICRG database
Internal conflictICRG database
Political RiskICRG database
List of Countries
AngolaCote d’IvoireMadagascarSierra Leone
BeninEquatorial GuineaMalawiSouth Africa
BotswanaEritreaMaliSwaziland
Burkina FasoEthiopiaMauritiusTanzania
BurundiGabonMozambiqueTogo
CameroonGambia, TheNamibiaUganda
Cape VerdeGhanaNigerZambia
Central African RepublicGuineaNigeriaZimbabwe
ChadGuinea-BissauRwanda
ComorosKenyaSão Tomé & Príncipe
Congo, Rep. ofLesothoSenegal
Congo, Dem. Rep. ofLiberiaSeychelles
Appendix Table 1.Workers’ remittances, compensation of employees, and migrant transfers(Millions of U.S. Dollars)
20052006
1995199619971998199920002001200220032004estimateestimate
Angola..5....................
Benin1008671907787847655555555
Botswana595048433426262739393939
Burkina Faso808080808067505050505050
Burundi..................000
Cameroon111111111111111111111111
Cape Verde10610076747987818592929292
Central African Republic........................
Chad........................
Comoros121212121212121212121212
Congo, Dem. Rep.........................
Congo, Rep.485212101113151111
Cote d’Ivoire151147136143138119116120142148148148
Equatorial Guinea00....................
Eritrea......343............
Ethiopia27169273453183347134134134
Gabon466646536666
Gambia, The192066714778888
Ghana1728263031324644658299120
Guinea111561915111424242
Guinea-Bissau..22222101823232323
Kenya298288352348432538517395494494494494
Lesotho411388379295276252209194288355355355
Liberia........................
Madagascar141112111211111716161616
Malawi111111111111
Mali1121119284867388137154155155155
Mauritius132160168180178177215215215215215215
Mozambique596164463837425370585757
Namibia161413111099712161616
Niger845192714221926262626
Nigeria8049471,9201,5441,3011,3921,1671,2091,0632,2732,2732,273
Rwanda21555578791099
Sao Tome and Principe......110111111
Senegal146150150147186233305344511511511511
Seychelles11000322571111
Sierra Leone2425620227722262522
South Africa105102206283327344297288435523658658
Swaziland837684787074746288898989
Tanzania1192127816129111616
Togo15292619233469104148148148148
Uganda........233238338416285347642642
Zambia........................
Zimbabwe........................
Source: World Bank (2006)
Source: World Bank (2006)
Appendix Table 2:Poverty Dataset Details
CountrySurvey Year
Albania1996
Albania2002
Algeria1988
Algeria1995
Benin2003
Bolivia1990
Bolivia1997
Bolivia2002
Botswana1985
Botswana1993
Brazil1981
Brazil1987
Brazil1992
Brazil1997
Brazil2002
Burkina Faso1994
Burkina Faso1998
Burkina Faso2003
Cambodia1997
Cameroon1996
Cameroon2001
Central African Rep.1993
Chile1987
Chile1992
Chile1998
Chile2000
China1984
China1987
China1992
China1997
China2001
Colombia1980
Colombia1988
Colombia1991
Colombia1996
Colombia2003
Costa Rica1981
Costa Rica1986
Costa Rica1993
Costa Rica1997
Costa Rica2001
Côte d’Ivoire1987
Côte d’Ivoire1993
Côte d’Ivoire1998
Côte d’Ivoire2002
Croatia1988
Croatia1998
Croatia2001
Czech Rep.1988
Czech Rep.1993
Czech Rep.1996
Dominican Rep.1986
Dominican Rep.1992
Dominican Rep.1996
Dominican Rep.2000
Ecuador1987
Ecuador1994
Ecuador1998
Egypt1990
Egypt1995
El Salvador1989
El Salvador1997
El Salvador2002
Estonia1988
Estonia1993
Estonia1998
Estonia2002
Ethiopia1981
Ethiopia1995
Ethiopia2000
Gambia, The1992
Gambia, The1998
Ghana1988
Ghana1991
Ghana1998
Guatemala1987
Guatemala1998
Guatemala2002
Guyana1992
Guyana1998
Haiti2001
Honduras1986
Honduras1992
Honduras1998
Honduras2003
India1977
India1983
India1987
India1992
India1997
Indonesia1987
Indonesia1993
Indonesia1998
Indonesia2002
Iran1986
Iran1994
Iran1998
Jamaica1988
Jamaica1992
Jamaica1996
Jamaica2000
Jordan1986
Jordan1992
Jordan1997
Jordan2002
Kenya1992
Kenya1997
Kyrgyz Rep.1988
Kyrgyz Rep.1993
Kyrgyz Rep.1997
Kyrgyz Rep.2002
Laos1992
Laos1997
Laos2002
Lesotho1986
Lesotho1993
Lesotho1995
Lithuania1988
Lithuania1993
Lithuania1998
Lithuania2002
Madagascar1980
Madagascar1993
Madagascar1997
Madagascar2001
Malawi1997
Malawi2004
Malaysia1984
Malaysia1987
Malaysia1992
Malaysia1997
Mali1989
Mali1994
Mali2001
Mauritania1987
Mauritania1993
Mauritania1995
Mauritania2000
Mexico1984
Mexico1989
Mexico1992
Mexico1996
Mexico2002
Morocco1984
Morocco1990
Morocco1998
Mozambique1996
Namibia1993
Nepal1995
Nepal2003
Nicaragua1993
Nicaragua1998
Nicaragua2001
Niger1992
Niger1995
Nigeria1985
Nigeria1992
Nigeria1996
Nigeria2003
Pakistan1987
Pakistan1992
Pakistan1996
Panama1979
Panama1989
Panama1991
Panama1997
Panama2002
Paraguay1990
Paraguay1997
Paraguay2002
Peru1985
Peru1990
Peru1996
Peru2002
Philippines1988
Philippines1994
Philippines1997
Philippines2000
Poland1987
Poland1992
Poland1998
Poland2002
Romania1989
Romania1992
Romania1998
Romania2002
Russia1988
Russia1993
Russia1998
Russia2002
Rwanda1984
Rwanda1999
Senegal1991
Senegal2001
Sierra Leone1989
Slovak Rep.1988
Slovak Rep.1992
Slovak Rep.1996
Slovenia1987
Slovenia1993
Slovenia1998
South Africa1993
South Africa1995
South Africa2000
Sri Lanka1985
Sri Lanka1990
Sri Lanka1995
Sri Lanka2002
St Lucia1995
Swaziland1994
Thailand1981
Thailand1988
Thailand1992
Thailand1996
Thailand2002
Trinidad & Tobago1988
Trinidad & Tobago1992
Tunisia1985
Tunisia1990
Tunisia1995
Tunisia2000
Turkey1987
Turkey1994
Turkey2002
Venezuela1981
Venezuela1987
Venezuela1993
Venezuela1997
Venezuela2000
Yemen1992
Yemen1998
Zimbabwe1990
Zimbabwe1995
Appendix Table 3.Descriptive Statistics of Regression Variables
ObservationsMeanMedianStandard

deviation
Range
Headcount poverty23317.79.419.479.3
Poverty gap2336.42.78.851.4
Squared poverty gap2333.30.95.637.9
Gini index2330.40.40.10.5
Per capita income2281,770.31,352.71,581.18,361.2
Remittances to GDP2163.51.17.572.9
Trade openess22470.060.837.1213.3
Schooling1874.94.72.310.0
Note: These are raw data series, before the log transformation.
Note: These are raw data series, before the log transformation.
Appendix Table 4.Bivariate Correlations of Regression Variables
Headcount

Poverty
Poverty

Gap
Squared

Poverty

Gap
Gini IndexPer capita

income
Remittances

to GDP
Trade

Openness
Schooling
Headcount poverty1.00
Poverty gap0.94*1.00
Squared poverty gap0.84*0.97*1.00
Gini index0.20*0.31*0.35*1.00
Per capita income-0.58*-0.49*-0.41*0.021.00
Remittances to GDP0.010.060.070.12*-0.14*1.00
Trade openess-0.25*-0.16*-0.11*0.050.21*0.26*1.00
Schooling-0.61*-0.55*-0.48*-0.22*0.56*-0.070.30*1.00
Note: * indicates significant at 10 percent.
Note: * indicates significant at 10 percent.
Appendix Table 5.Ordinary Least Squares Estimation (With and Without Regional Dummies)
Headcount PovertyPoverty GapSquared Poverty Gap
(1)(2)(1)(2)(1)(2)
Per capita GDP (constant 2000 dollars)-1.21***

(-10.56)
-1.07***

(-6.62)
-1.26***

(-10.58)
-1.20***

(-5.93)
-1.22***

(-10.29)
-1.19***

(-5.07)
Gini coefficient3.30***

(6.76)
1.95***

(3.74)
3.66***

(7.56)
2.03***

(3.39)
3.80***

(7.00)
2.36***

(3.60)
Inflow of remittances (ratio to GDP)-0.14***

(-2.53)
-0.11**

(-2.38)
-0.13**

(-2.07)
-0.08

(-1.48)
-0.10

(-1.55)
-0.05

(-0.81)
Europe and Central Asia-2.01***

(-3.84)
-1.46**

(-2.01)
-0.10

(-1.10)
East Asia and Pacific-0.48

(-1.00)
-0.65

(-0.98)
-0.45

(-0.60)
Latin America and Caribbean-0.26

(-0.51)
0.27

(0.40)
0.39

(0.49)
Middle East and North Africa-1.88***

(-3.21)
-1.78***

(-2.43)
-1.64**

(-1.94)
Sub-Saharan Africa-0.62

(-1.46)
-0.28

(-0.49)
-0.11

(-0.16)
Constant13.22***

(16.06)
11.86***

(9.53)
12.59***

(16.59)
11.22***

(7.64)
11.59***

(14.24)
10.45***

(5.89)
Observations212212211211208208
Adj R 20.600.720.580.680.530.61
F-Statistic58.9344.1087.2153.3071.3533.58
Note: ***,**,and * indicate significant at 1, 5,and 10 percent. T-Statistics are reported in parentheses. Standard errors are clustered by country to eliminate any downward bias.
Note: ***,**,and * indicate significant at 1, 5,and 10 percent. T-Statistics are reported in parentheses. Standard errors are clustered by country to eliminate any downward bias.
Appendix Table 6.Ordinary Least Squares Estimation (With Interaction Term)
Headcount PovertyPoverty GapSquared Poverty Gap
(1)(2)(1)(2)(1)(2)
Per capita GDP (constant 2000 dollars)-1.24***

(-10.69)
-1.08***

(-6.70)
-1.29***

(-10.61)
-1.21***

(-6.04)
-1.25***

(-10.26)
-1.21***

(-5.17)
Gini coefficient3.29***

(6.86)
1.93***

(3.77)
3.66***

(7.76)
2.01***

(3.39)
3.80***

(7.16)
2.35***

(3.59)
Inflow of remittances (ratio to GDP)-0.18***

(-2.65)
-0.16***

(-2.45)
-0.17**

(-2.15)
-0.12

(-1.61)
-0.15*

(-1.68)
-0.09

(-1.01)
Remittances*Sub-Saharan Africa

(interaction term)
0.16*

(1.70)
0.14*

(1.91)
0.16*

(1.71)
0.14

(1.54)
0.16

(1.42)
0.12

(1.11)
Europe and Central Asia-2.06***

(-3.99)
-1.50**

(-2.10)
-1.03

(-1.16)
East Asia and Pacific-0.53

(-1.11)
-0.69

(-1.06)
-0.49

(-0.67)
Latin America and Caribbean-0.28

(-0.56)
0.25

(0.37)
0.38

(0.48)
Middle East and North Africa-1.86***

(-3.28)
-1.76***

(-2.47)
-1.62**

(-1.95)
Sub-Saharan Africa-0.62

(-1.51)
-0.28

(-0.51)
-0.11

(-0.17)
Constant13.38***

(16.40)
11.96***

(9.70)
12.76***

(16.83)
11.32***

(7.70)
11.76***

(14.44)
10.55***

(5.92)
Observations212212211211208208
Adj R 20.610.720.580.680.530.61
F-Statistic46.5243.9171.2955.2958.8633.97
Note: ***,**,and * indicate significant at 1, 5,and 10 percent. T-Statistics are reported in parentheses. Standard errors are clustered by country to eliminate any downward bias.
Note: ***,**,and * indicate significant at 1, 5,and 10 percent. T-Statistics are reported in parentheses. Standard errors are clustered by country to eliminate any downward bias.
Appendix Table 7.Ordinary Least Squares Estimation for Rem>1 Sample
Headcount

Poverty
Poverty

Gap
Squared

Poverty

Gap
Per capita GDP (constant 2000 dollars)-1.28***

(-6.52)
-1.29***

(-5.55)
-1.14***

(-4.57)
Gini coefficient3.03***

(-4.20)
3.28***

(4.42)
3.74***

(4.817
Inflow of remittances (ratio to GDP)-0.26**

(-2.06)
-0.19*

(-1.35)
-0.22

(-1.40)
Europe and Central Asia-1.80***

(-2.94)
-1.50*

(-1.84)
-1.36

(-1.42)
East Asia and Pacific-0.74

(-1.26)
-1.08

(-1.44)
-0.87

(-1.18)
Latin America and Caribbean-0.27

(-0.45)
0.13

(0.17)
0.08

(0.09)
Middle East and North Africa-1.86***

(-3.30)
-1.96***

(-2.81)
-1.96**

(-2.40)
Sub-Saharan Africa-0.75*

(-1.45)
-0.37

(-0.59)
-0.16

(-0.21)
Constant14 44***

(9.31)
13.20***

(7.72)
11.69***

(6.34)
Observations112111109
Adj R 20.750.740.71
F-Statistic36.3941.5433.10
Note: ***,**,and * indicate significant at 1, 5,and 10 percent. T-Statistics are reported in parentheses. Standard errors are clustered by country to eliminate any downward bias
Note: ***,**,and * indicate significant at 1, 5,and 10 percent. T-Statistics are reported in parentheses. Standard errors are clustered by country to eliminate any downward bias
Appendix Table 8.Descriptive Statistics for Regression Variables
ObservationsMeanStandard

Deviation
MinimumMaximum
Bank deposits to GDP18818.0614.271.1693.21
M2 to GDP23326.8118.040.81165.25
Remittances to GDP1983.629.94075.33
Log(GDP)24521.381.4217.4225.68
Per capita GDP245807.161,210.5984.767,164.45
Inflation20759.85459.36-5.616424.99
Trade openness24471.5337.0712.88224.21
Other capital flows to GDP24814.9114.66-2.09104.61
Appendix Table 9.Bivariate Correlations of Regression Variables
Bank Deposits

to GDP
M2 to

GDP
Remittances

to GDP
Log(GDP)Per Capita

GDP
InflationDual

Exchange

Rate
Trade

Openness
Other

Capital

Flows to

GDP
Bank deposits to GDP1.00
M2 to GDP0.971.00
Remittances to GDP0.220.151.00
Log(GDP)0.140.01-0.251.00
Per capita GDP0.620.38-0.080.141.00
Inflation-0.1-0.06-0.040.08-0.051.00
Dual exchange rate0.070.050.080.230.03-0.011.00
Trade openness0.430.340.33-0.240.45-0.030.041.00
Other capital flows to GDP-0.240.010.10-0.600.230.040.020.151.00
Appendix Table 10.Baseline Panel Estimation
Random EffectsFixed Effects
DepositsM2DepositsM2
Remittances to GDP0.74**

(2.10)
1.66***

(3.30)
0.76*

(1.90)
1.72***

(2.90)
Log(GDP)3.69***

(2.99)
2.79**

(1.94)
0.97

(0.21)
-5.54

(-0.81)
Per capita GDP0.00***

(2.63)
0.01

(0.76)
0.01***

(2.75)
0.01

(1.01)
Inflation-0.02

(-0.59)
0.002

(0.52)
-0.01

(-0.29)
0.003

(0.56)
Dual-2.16

(-1.29)
-0.98

(-0.42)
-3.57**

(-1.87)
-0.82

(-0.30)
Trade openness-0.29

(-0.80)
0.01

(0.22)
-0.01

(-0.14)
0.08

(0.98)
Other capital flows to GDP0.24**

(2.31)
0.03

(0.25)
0.32**

(2.46)
-0.04

(-0.27)
Corruption-0.28

(-0.30)
0.21

(0.20)
0.43

(-0.42)
-1.20

(-0.78)
Internal conflict-1.14***

(-2.49)
-1.50**

(-2.26)
-1.06**

(-1.99)
-1.69**

(-2.17)
Political risk0.41***

(3.37)
0.48***

(2.77)
0.39**

(2.29)
0.61***

(2.43)
Constant-62.75**

(-2.30)
-57.28*

(-1.81)
-29.74

(-0.30)
117.82

(0.83)
Observations89938993
Adj R 20.450.360.380.26
Note:***,**, and * signify 1, 5, and 10 percent significance levels.The ICRG database measures political risk on a scale of 1 to 100 with higher values implying less risk. So a positive coefficient on political risk indicates that lower political risk is associated with greater financial development.
Note:***,**, and * signify 1, 5, and 10 percent significance levels.The ICRG database measures political risk on a scale of 1 to 100 with higher values implying less risk. So a positive coefficient on political risk indicates that lower political risk is associated with greater financial development.
Appendix Table 11.Fixed Effects Panel Instrumental Variables Estimation
DepositsM2
(1)(2)(1)(2)
Instrumented Variable
Remittances to GDP4.04***

(2.39)
1.81*

(1.67)
3.71**

(1.91)
7.99***

(2.74)
Exogenous Variables Log(GDP)2.7

(0.70)
-1.64

(-0.29)
12.34

(1.12)
-16.30

(-1.29)
Per capita GDP0.02***

(5.84)
0.01***

(2.79)
0.01**

(2.04)
.014

(1.25)
Inflation0.002

(0.20)
-0.001

(0.30)
0.01

(0.66)
0.004

(0.38)
Dual-3.65

(-1.30)
-3.86*

(-1.89)
-6.07

(-1.17)
-4.17

(-0.83)
Trade openness0.04

(0.66)
-0.05

(-0.76)
0.13

(1.08)
-0.18

(1.02)
Other capital flows to GDP0.18

(0.90)
0.40***

(2.53)
-0.64

(-1.36)
0.35

(1.03)
Corruption-1.11

(-0.88)
-4.45

(-1.49)
Internal conflict-1.56**

(-2.11)
-3.98

(-2.38)
Political risk0.56**

(2.32)
1.39

(2.53)
Constant-72.4024.16-266.39336.65


(-0.88)


(0.20)


(-1.13)


(1.28)
Cragg Donald F-statistic for weak instruments1.703.152.192.93
Observations1348914593
Adj R 20.500.300.290.02
Note:***,**, and * signify 1, 5, and 10 percent significance levels. Instruments weighted by expatriation rate
Note:***,**, and * signify 1, 5, and 10 percent significance levels. Instruments weighted by expatriation rate

Appendix Figure 1.Remittances to Developing Countries by Region, 1975-2005

(Millions of US dollars)

Source: IMF Balance of Payments Yearbook, 2006; World Bank staff estimates

Appendix Figure 2.Regional Shares of Remittances to Developing Countries, 2000-05

(Millions of U.S. dollars)

Source: IMF, Balance of Payments Yearbook, 2006; World Bank staff estimates.

References

    AdamsJr., Richard H., 1991,”The Effects of International Remittances on Poverty, Inequality, and Development in Rural Egypt,” IFPRI Research Report 86 (Washington: IFPRI).

    • Search Google Scholar
    • Export Citation

    AdamsJr., Richard H., 2006, “Remittances and Poverty in Ghana,” World Bank Policy Research Paper 3838, (Washington: World Bank).

    AdamsJr., Richard H., and JohnPage, 2005, “Do International Migration and Remittance Reduce Poverty in Developing Countries?” World Development, Vol. 33 (10), pp. 1,645–69.

    • Search Google Scholar
    • Export Citation

    Aggarwal, Reena, AsliDemirguc-Kunt, and MariaPeria, 2006, “Do Workers’ Remittances Promote Financial Development?” World Bank Policy Research Working Paper No. 3957 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    Amuedo-Dorantes, Catalina, and SusanPozo, 2004, “Workers’ Remittances and the Real Exchange Rate: A Paradox of Gifts,” World Development, Vol. 32 (8), pp. 1,407–17.

    • Search Google Scholar
    • Export Citation

    Azam, Jean-Paul, and FloreGubert, 2005, “Migrant Remittances and Economic Development in Africa: A Review of Evidence,” IDEI Working Paper 354 (Toulouse: Institut d’Économie Industrielle).

    • Search Google Scholar
    • Export Citation

    Bach, Stephen, 2006, “International Mobility of Health Professionals: Brain Drain or Brain Exchange?” UNU-WIDER Research Paper No. 2006.82 (Helsinki: World Institute for Development Economics Research).

    • Search Google Scholar
    • Export Citation

    Barham, Bradford, and StephenBoucher, 1998, “Migration, Remittances, and Inequality: Estimating the Net Effects of Migration on Income Distribution,” Journal of Development Economics, Vol. 55, pp. 307–31.

    • Search Google Scholar
    • Export Citation

    Beck, Thorsten, AsliDemirguc-Kunt, and RossLevine, 2004, “Finance, Inequality and Poverty: Cross-Country Evidence,” NBER Working Paper 10979 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    Bourdet, Yves, and HansFalck, 2006, “Emigrants’ Remittances and Dutch Disease in Cape Verde,” International Economic Journal, forthcoming.

    • Search Google Scholar
    • Export Citation

    Bugamelli, Matteo, and FrancescoPaterno, 2006, “Workers’ Remittances and Current Account Reversals,” Bank of Italy Economic Research Paper No. 573 (Rome: Bank of Italy).

    • Search Google Scholar
    • Export Citation

    Caceres, Luis Rene, and Nolvia N.Saca, 2006, “What Do Remittances Do? Analyzing the Private Remittance Transmission Mechanism in El Salvador,” IMF Working Paper 06/250 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Carrington, William J., and EnricaDetragiache, 1998, “How Big Is the Brain Drain?” IMF Working Paper 98/102 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Chami, Ralph, ConnelFullenkamp, and SamirJahjah, 2003, “Are Immigrant Remittance Flows a Source of Capital for Development?” IMF Working Paper 03/189 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Cox Edwards, Alejandra, and ManuelitaUreta, 2003, “International Migration, Remittances, and Schooling: Evidence from El Salvador,” Journal of Development Economics, Vol. 72 (December), pp. 429–61.

    • Search Google Scholar
    • Export Citation

    De Haan, Harjaan2000, “Migrants, Livelihoods, and Rights: The Relevance of Migration in Development Policies,” DFID Social Development Working Paper 4 (London: Department for International Development).

    • Search Google Scholar
    • Export Citation

    Detragiache, Enrica, PoonamGupta, and ThierryTressel, 2005, “Finance in Lower-Income Countries: An Empirical Exploration, IMF Working Paper 05/167 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    DFID, 2006, “Sending Money Home: A Survey of Remittance Products and Services in the United Kingdom,” available on the internet at: http://www.dfid.gov.uk/pubs/files/sendmoneyhome.pdf.

    • Search Google Scholar
    • Export Citation

    El Qorchi, Mohammed, Samuel M.Maimbo, and John F.Wilson, 2003, “Informal Funds Transfer Systems: An Analysis of the Informal Hawala System,” IMF Occasional Paper No. 222, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Freund, Caroline, and NikolaSpatafora, 2005, “Remittances: Transaction Costs, Determinants, and Informal Flows,” World Bank Policy Research Working Paper No. 3704 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    Genesis Analytics, 2003, “African Families, African Money: Bridging the Money Transfer Divide,” study prepared for FinMark Trust, South Africa; available on the internet at: http://www.finmarktrust.org.za/documents/2003/APRIL/Genesis_AFAM.pdf.

    • Search Google Scholar
    • Export Citation

    Genesis Analytics, 2005, “Supporting Remittances in Southern Africa: Estimating Market Potential and Assessing Regulatory Obstacles,” study prepared for Consultative Group to Assist the Poor (CGAP) and FinMark Trust; available on the internet at: http://www.finmarktrust.org.za/documents/2005/JULY/remittances.pdf.

    • Search Google Scholar
    • Export Citation

    Giuliano, Paola, and MartaRuiz-Arranz, 2005, “Remittances, Financial Development, and Growth,” IMF Working Paper 05/234 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Gubert, F., 2002, “Do Migrants Insure Those Who Stay Behind? Evidence from the Kayes Area,” Oxford Development Studies, Vol. 30 (3), pp. 267–87.

    • Search Google Scholar
    • Export Citation

    Gulde, Anne-Marie, CatherinePattillo, JakobChristensen, KevinCarey, and SmitaWagh, 2006, Sub-Saharan Africa: Financial Sector Challenges (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Gupta, Sanjeev, RobertPowell, and YongzhengYang, 2006, The Macroeconomic Challenges of Scaling Up Aid to Africa: A Checklist for Practitioners (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Hanson, Gordon H., and ChristopherWoodruff, 2003, “Emigration and Educational Attainment in Mexico,” University of California photocopy.

    • Search Google Scholar
    • Export Citation

    Isern, Jennifer, RaniDeshpande, and Judith vanDoorn, 2005, “Crafting a Money Transfers Strategy: Guidance for Pro-Poor Financial Service Providers,” CGAP Occasional Paper, No. 10 (Washington: Consultative Group to Assist the Poor).

    • Search Google Scholar
    • Export Citation

    Jordan, Mary, 2006, “New Conductors Speed Global Flows of Money,” Washington Post, Oct3, 2006.

    Kapur, Devesh, and JohnMcHale, 2005, “The Global Migration of Talent: What Does it Mean for Developing Countries?” CGD Brief (Washington: Center for Global Development); available on the internet at http://www.cgdev.org/content/publications/detail/4473/.

    • Search Google Scholar
    • Export Citation

    Kasekende, Louis A, 2000, “Capital Account Liberalization: The Ugandan Experience,” paper presented at the Overseas Development Institute, June 21, London.

    • Search Google Scholar
    • Export Citation

    Koechlin, Valerie, and GianmarcoLeon, 2006, “International Remittances and Income Inequality: An Empirical Investigation,” IADB Working Paper No. 571 (Washington: Inter-American Development Bank).

    • Search Google Scholar
    • Export Citation

    Levine, Ross, 1997, “Financial Development and Economic Growth: Views and Agenda,” Journal of Economic Literature, Vol. 35 (June), pp. 688–726.

    • Search Google Scholar
    • Export Citation

    Levine, Ross, 2004, “Finance and Growth: Theory and Evidence,” NBER Working Paper 10766 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    Lipton, M., 1980, “Migration from Rural Areas of Poor Countries: The Impact of Rural Productivity and Income Distribution,” World Development, Vol. 8. pp. 1–24.

    • Search Google Scholar
    • Export Citation

    Lucas, Robert E.B., 1987, “Emigration to South Africa’s Mines,” American Economic Review, Vol. 77 (3), pp. 313–30.

    Lucas, Robert E.B., and OdedStark, 1985, “Motivations to Remit: Evidence from Botswana,” Journal of Political Economy, Vol. 93 (Oct), pp. 901–18.

    • Search Google Scholar
    • Export Citation

    Lueth, Erik, and MartaRuiz-Arranz, 2006, “A Gravity Model of Workers’ Remittances,” IMF Working Paper 06/290 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    OECD/DAC, 2006, “Final Data on ODA for 2005,” available on the internet at; http://www.oecd.org/dataoecd/52/18/37790990.pdf

    Orozco, Manuel, 2003, “Worker Remittances: An International Comparison,” Working Paper commissioned by the Multilateral Investment Fund, Inter-American Development Bank, February.

    • Search Google Scholar
    • Export Citation

    Pond, Bob, and BarbaraMcPake, 2006, “The Health Migration Crisis: The Role of Four Organisation for Economic Cooperation and Development Countries,” The Lancet, Vol. 367 (April), pp. 1,448–55.

    • Search Google Scholar
    • Export Citation

    Quartey, Peter, and TheresaBlankson, 2004, “Do Migrant Remittances Minimize the Impact of Macro-Volatility on the Poor in Ghana?” Final report submitted to the Global Development Network; available on the internet at: http://www.gdnet.org/pdf2/gdn_library/global_research_projects/macro_low_income/Quartey.pdf.

    • Search Google Scholar
    • Export Citation

    Rajan, Raghuram, and ArvindSubramanian, 2005, “What Undermines Aid’s Impact on Growth?” IMF Working Paper 05/126 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Rajan, Raghuram, and LuigiZingales, 1998, “Financial Dependence and Growth,” American Economic Review, Vol. 88 (June), pp. 559-86.

    Ratha, Dilip, 2003, “Worker’s Remittances: An Important and Stable Source of External Development Finance,” in Global Development Finance, 2003 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    Ratha, Dilip, and JanRiedberg. 2005. “On Reducing Remittance Costs,” unpublished paper of the Development Research Group (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    Ravallion, M., 1997, “Can High-Inequality Developing Countries Escape Absolute Poverty?” Economics Letters, Vol. 56, pp. 51–57.

    Ravallion, M., and Chen, S., 1997, “What Can new Survey Data Tell us About Recent Changes in Distribution and Poverty?” World Bank Economic Review, Vol. 11, pp. 357–82.

    • Search Google Scholar
    • Export Citation

    Rozelle, Scott, J. EdwardTaylor, and AlandeBrauw, 1999, “Migration, Remittances, and Agricultural Productivity in China,” American Economic Review, Vol. 89 (May), pp. 287–91.

    • Search Google Scholar
    • Export Citation

    Samuels, George, 2003, “Banking Unbanked Immigrants through Remittances,” Communities and Banking, Fall2003; available on the internet at: http://www.bos.frb.org/commdev/c&b/2003/fall/unbanked.pdf.

    • Search Google Scholar
    • Export Citation

    Sander, Cerstin, 2004, “Capturing a Market Share? Migrant Remittances and Money Transfers as a Microfinance Service in Sub-Saharan Africa,” Small Enterprise Development, Vol. 15 (March), pp. 20-34.

    • Search Google Scholar
    • Export Citation

    Sander, Cerstin, and S.M.Maimbo, 2005, “Migrant Remittances in Africa–A Regional Perspective,” in Remittances–Development Impact and Future Prospects, ed. by Samuel MunzeleMaimbo and DilipRatha(Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    Serrano, Carlos, 2006, “Leveraging Remittances for Housing Finance: Mortgage Program for Mexican Migrant Workers,” prepared for the Second International Conference on Migrant Remittances, London, UK, November2006; available on the internet at: http://siteresources.worldbank.org/INTTOPCONF3/Resources/1588024-1163704792404/S6_P2_SERRANO.pdf

    • Search Google Scholar
    • Export Citation

    Spatafora, Nicola, 2005, “Two Current Issues Facing Developing Countries,” in World Economic Outlook, 2005 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Stahl, C., 1982, “Labor Emigration and Economic Development,” International Migration Review, Vol. 16, pp. 868–99.

    Stark, O., and J. E.Taylor, 1989, “Relative Deprivation and International Migration,” Demography, Vol. 26, pp. 1–4.

    Toxopeus, Helen S., and RobertLensink, 2006, “Remittances and Financial Inclusion in Development,” draft (The Netherlands: University of Groningen).

    • Search Google Scholar
    • Export Citation

    Woodruff, Christopher, and ReneZenteno, 2001, “Remittances and Microenterprises in Mexico,” UCSD Graduate School of International Relations and Pacific Studies Working Paper, (University of California San Diego)

    • Search Google Scholar
    • Export Citation

    World Bank, 2006, Global Economic Prospects: Economic Implications of Remittances and Migration (Washington: World Bank).

    World Bank, 2007, Migration and Remittances: Eastern Europe and the Former Soviet Union, (Washongton: World Bank).

    UNDP, 2005, “The Potential Role of Remittances in Achieving the Millennium Development Goals-An Exploration,” Background Note for the Roundtable on Remittances and MDGs, October2005, available on the internet at: http://tcdc1.undp.org/Remittances_Oct102005B.pdf

    • Search Google Scholar
    • Export Citation
1

The authors thank, without implicating, Benedicte Vibe Christensen, Anne-Marie Gulde, Theirry Tressel, Dilip Ratha, and Sanket Mohapatra for their helpful comments.

2

As in other studies on the topic, the remittance data referred to here are aggregate worker remittances, compensation to employees, and migrant transfers series from the IMF Balance of Payments database, supplemented by the data from World Bank (2006). All 2005 remittance data are estimates provided by Dilip Ratha of the World Bank. Appendix Table 1 has details on remittance flows to SSA countries over the last ten years.

3

Even where migrants use formal channels, the reporting of “small” remittances is not mandatory in most countries.

4

However, since remittances are private transfers foreign borrowing against such flows would only be possible with additional stipulations like surrender requirements, prohibition of foreign currency accounts and/or taxes on remittances.

6

Sub-Saharan Africa refers to the 44 countries listed in the Data Appendix.

7

This growth at least in some part reflects better reporting. Also, since the underlying data are in U.S. dollars, changes in the value of the dollar are captured in measuring the growth of nondollar remittances.

9

Lueth and Ruiz-Arranz (2006) also find that remittances do not increase after a natural disaster, and are in fact aligned with business cycles for 11 recipient countries in Asia and Central Europe.

10

“Educated” refers to the segment of the population that has received a tertiary education. Expatriation rates are calculated as the ratio of emigrant population to the total population (emigrant plus resident) within a group.

11

This makes it possible to distinguish countries like Barbados where both general and educated migration rates are high from countries like Burundi where educated migration rates are far greater than the general propensity to migrate. The two phenomena are likely to impact the local labor markets quite differently, but the latter is closer to what we understand as brain drain.

12

Since the data refer to migration to OECD countries, they may overstate the difference between general and skilled migration from SSA. General expatriation rates for SSA are likely to be underestimated given the high volumes of intraregional, undocumented migration by low-skilled workers.Low-skilled workers in SSA do not have the same geographic proximity to OECD countries as those in North Africa or East and Central Asia or Latin America, so in SSA intraregional migration is a more likely option for low-skilled workers. At the same time, the high expatriation rates of skilled workers reflect at least to some extent the small base of such workers in SSA populations.

13

Altruism may not completely explain the intrafamily transfer of resources. Often migrant workers remit money to maintain their stake in family property, perhaps with a view to returning in the future. Lucas and Stark (1985) found that in Botswana not only do remittances rise with the size of the migrant’s income but there is also a positive relationship between the level of remittances and the receiving household’s preremittance income. The insurance motive for remittances was supported by a study using survey data from Western Mali (Gubert, 2002).

14

Using poverty surveys restricts the number of data points so that estimation results from any single regional group are not significant.

15

Appendix Table 2 lists the countries and survey years of the dataset.

16

Almost 32 percent of the observations in the dataset come from Latin American and Caribbean countries, 11 percent from the East Asia and Pacific region, almost 18 percent from East Europe and Central Asia, 10 percent from the Middle East and North Africa, and 6 percent from South Asia.

17

For details on this and other data sources see the Data Appendix.

18

Appendix Table 7 reports the OLS results when the sample is restricted to countries where remittances amount to more than 1 percent of GDP. This is a macro replication of the micro idea that the poverty-reducing effect of remittances is likely to be enhanced when the sample includes only households that have migrant workers—those that actually receive income transfers. The higher elasticities with this restricted sample support the idea that a more general sample dilutes the poverty-reducing impact of remittances.

19

A postestimation test of the OLS coefficients suggests that the sum of the average effect of remittances for all countries and the coefficient on the interaction term is not different from zero. While this does suggest that the relationship between poverty and remittances in SSA might be different there are not enough observations from the region to pursue this issue. Instead we explore the issue of reverse causality using a three-stage least squares estimation for the full sample.

20

The three-stage least squares technique involves simultaneously generating two-stage least squares estimates of all the equations in the system. The technique allows for nonzero contemporaneous correlations between the disturbances in different equations. If the disturbances are uncorrelated, the three-stage least squares technique reduces to a two-stage least squares.

21

Since only OECD members keep detailed records of their immigrant population we are restricted to using only these countries as the source countries for remittances. This assumption may be questioned in SSA where intraregional migration is common, and where, for instance, South Africa might be a more significant source country.

22

Asymmetric information does raise the possibility of moral hazard on the recipient’s side. Since migrant workers are typically unable to monitor the use of their transfers, there is an incentive for household members to curtail their own labor effort, using the supplemental income from remittances to maintain their standard of living. Azam and Gubert (2005) found that in the Kayes region of western Mali widespread migration lowered recipient productivity.

23

The dataset is described in detail in Appendix Tables 8 and 9.

24

This result holds even when South Africa is excluded from the sample.

25

Recent studies have emphasized the role of non-economic factors in financial development among low-income countries (Detragiache, Gupta, and Tressel, 2005). In Appendix Tables 10 we include corruption, internal conflict and political risks as additional control variables, though the limited time series availability of these variables restricts our observations to less than 60 percent of those reported in Table 2. The results indicate that even when the signifcant effect of internal conflict and political risk on financial development in SSA is taken into account, remittances are still positively and signicantly associated with financial development.

26

We also weight the source country variables by the general expatriation rate to improve the fit of the instruments. The results are reported in Appendix Table 11 and are not materially different from those reported above.

27

DfID, 2006. This pattern also holds for the high-volume U.S.-Mexico corridor, where since 2000 the cost of remitting money has almost halved (Serrano, 2006).

28

For instance, eliminating the discrepancies between the official and parallel market exchange rates in either the sending or the receiving country can make formal channels more attractive. In Uganda measures permitting residents to open foreign currency accounts led to a dramatic surge in private transfers in the early 1990s (Kasekende, 2000, cited in Ratha, 2003).

29

However, since most such programs require that the migrant open a checking or savings account, they are unlikely to appeal to undocumented workers.

30

At present the market is dominated by specialized MTOs like Western Union that are less likely to offer ancillary financial products to their clients.

31

This micro-level replication of recipient countries gaining favorable access to capital markets by securitizing future remittance flows is likely to be perceived as less risky by local financial institutions if accompanied by entrepreneurial training for receiving households.

32

Funds invested directly at attractive rates in deposit schemes or bonds are not strictly speaking remittances; because they are not intrahousehold transfers and there is a monetary quid pro quo. However, since such funds are typically converted to local currency and stay in the recipient country they can be an important source of savings.

Other Resources Citing This Publication