The number of skilled migrants from developing countries has increased dramatically over the past four decades. The United Nations estimated that the total number of highly-skilled South-North migrants between 1961 and 1972 was 300,000 (UNCTAD, 1975). By 1990, there were more than 2.5 million highly-educated immigrants from developing countries residing in the United States alone. Worldwide, average emigration rates amount to 5.5% for high-skill workers, compared to less than one percent for low-skill ones, (Docquier and Rapoport, 2004).
The United States is the major OECD destination country for skilled workers. (SOPEMI, 2004. OECD) The European Union is the second destination, followed by Canada and Australia. Among non-OECD countries, countries comprising the former USSR have the largest community of skilled expatriates (4.2 million), former Yugoslavia is second (2.2 million), followed by India (1.9 million).
Latin America and Africa are the two regions of the developing world that have the highest shares of skilled and highly skilled migrants residing in the developed countries; 14 of the 30 countries with the highest emigration rates of skilled workers are African. 7 Table 2.5 presents the numbers and percentages of the stock of expatriates of highly skilled migrants from Africa and Latin America in OECD countries. The five countries with the largest number of skilled immigrants from Latin America are Jamaica, Colombia, Brazil, Peru and Argentina. In the case of Africa, South Africa and Nigeria are comparable. Dumont and Lemaitre (2004) estimate emigration rates by educational attainment and country of origin, using the latest OECD database. In their sample Africa has nine of the 15 countries with the highest “emigration rates” of skilled people. Highly skilled migrants – doctors, nurses, lectures, engineers, scientists and technologists – have moved from Ghana, and recently, Nigeria and South Africa attracted by higher salaries and better living conditions abroad (Adepoyu, 2003). Latin America and the Caribbean account for the majority of the remainder, and Oceania the residual. Smaller countries in these regions can have more than 40% of their highly-skilled populations abroad.
The skilled labor market in the United States is perhaps the most interesting of the OECD countries from the standpoint of signaling international trends. United States immigration policy has consistently shown a bias toward highly skilled immigrants admitted under economic criteria. Table 2.6 shows the distribution of immigrants admitted under such preferences in 2003. Asia and Europe clearly dominate the share of workers admitted under these criteria. Africa accounts for less than four percent of immigrants admitted. The same situation is reflected in the admission of temporary workers. (Table 2.7) The number of skilled worker visas approved for immigrants from African countries is 24,249. Immigrants from Asia and Europe have received 284,087 and 469,545 visas, respectively.
3. Remittances: Trends and Determinants
Workers' remittances have emerged as a major source of external development finance in recent years. Given their large size, governments from developing and developed countries have focused attention on both the development impact of remittances and on regulatory issues in sending and receiving countries.8As in the case of migration, reliable data on remittances are hard to come by. While the International Monetary Fund publishes statistics on “worker’s remittances, compensation of employees and migrants transfers”, these data are neither comprehensively reported nor do they capture flows of monies that take place outside of formal financial channels.
Global transfers of remittances to developing countries have grown steadily in the last 10 years and exceed $100 billion worldwide (IMF 2005). For most countries, remittances exceed the volume of foreign aid and investment. Using the definition developed for its Global Development Finance, 2003 the World Bank estimates that global flows of migrant remittances were $204.5 billion in 2004, an increase of 43.5 percent from 2001. (Table 3.1) Developing countries received more than $144 billion in 2004, an increase of nearly 57 percent since 2001.
The growth of remittances has outpaced that of private capital flows and official development assistance during the last ten years (Figure 3.1 and 3.2). In 2004 remittance receipts were about 5 percent of developing countries’ imports and 8 percent of domestic investment and were larger than official flows and private non-FDI flows to developing countries. (World Bank, 2005). In Mexico they are larger than FDI. In many countries, remittances are larger than the earnings from their most important export. In Sri Lanka, they are larger than tea exports, and in Morocco they are larger than tourism receipts (World Bank, 2005).
Patterns of remittances to developing countries
Rich countries are the leading originators of global remittances, with the United States dominant. South-South remittance flows are believed to be large, even in relation to North-South flows, but data are severely limited. Estimates are that in East Asia, South Asia, and Sub-Saharan Africa, more than two-thirds of emigrants from poor countries migrate to a country in the same region, and in South Asia and Sub-Saharan Africa, most of them migrate to another developing country. Upper-middle-income developing countries are an important source of remittance flows. Saudi Arabia, Malaysia, Russia and China are among the top 20 source countries of remittances (World Bank, 2005). Remittances from South Africa and India are also believed to be large (CGAP, 2005).
Table 3.1 provides some insight into the dynamics of remittance growth across income categories and regions of the developing world. Lower middle income countries have historically dominated the share of remittances received by developing countries, followed by low income countries. The rate of growth of remittances since 1990, however, has favored low income countries with more than a five fold increase.
Among regions, the Middle East and North Africa dominated the remittance picture in 1990 with remittance income more than twice that of the next region, Latin America and the Caribbean. By 2004 Latin America and East Asia and the Pacific had become the largest regions in terms of remittance receipts. South Asia has also experienced a dramatic increase in the volume of remittances. African remittances receipts began at a low base in 1990 and continue to lag other developing regions in terms of the absolute volume of remittances received. Its growth rate of remittance income similarly lags those of the more dynamic receiving regions.
Table 3.2 shows the top 20 recipients of remittances among developing countries in 2004, according to the IMF9. India, Mexico, the Philippines and Egypt were the top recipients among developing countries. World Bank estimates move China into the number one position, with more than $20 billion of remittance receipts. There were no sub-Saharan African countries included among the twenty largest receivers of remittances.
It is not surprising that large countries and more populous regions are among the top recipients of remittances in dollar terms. However, when remittances are expressed in per capita terms or as a share of GDP, the global picture changes. (Figures 3.3 and 3.4) In 2003 upper middle income countries received the equivalent of US$73.55 in remittances per capita, compared to US$ 15.87 per capita in low income countries. Latin America and The Middle East and North Africa received the largest remittances per capita; Sub-Saharan Africa received the smallest amount of remittances in per capita terms (US$ 8.52 in 2003), and had a slower growth rate in remittances per capita. (Table 3.3)
Figure 3.4 shows the share of remittances in gross domestic product by income group and region. For South Asia and the Middle East and North Africa, remittances are an important share of GDP. In contrast, remittances accounted for 1.37 percent of the GDP in Sub-Saharan Africa in 2003. Small countries such as Haiti, Tonga, Lebanon and Jordan dominate the top recipients in terms of contribution to national income (Table 3.2). Regionally, small countries from the Middle East, Central America and the Caribbean, and the former Soviet Union are strongly represented, reflecting their close proximity to labor importing countries. Two African countries, Lesotho and Cape Verde are among the top twenty recipients in terms of remittances as a share of GDP.
Official data on remittances are believed to be underestimated, perhaps severely, but there is little agreement as to their magnitude. A recent International Monetary Fund study (El-Qorchi, Maimbo and Wilson, 2003) estimated that unofficial transfers of remittances to the developing world currently amount to $10 billion per annum. Another study estimates that global remittances are about 2.5 times the size of recorded remittances reported in the IMF Balance of Payments data (AITE 2005). These estimates differ by a factor of 25!
Undercounting arises from two sources. First, most remittance source countries do not require reporting of “small” transactions.10 Remittances through post offices, exchange bureaus and other agents of money transfer companies are often not reflected in the official statistics (World Bank, 2005).11 Second, official data do not capture remittance flows through informal channels. Remittances transferred through agents such as informal operators or hand carried by travelers may be nearly as large as remittances through official channels. Many household surveys (Bangladesh, Pakistan, Moldova and Uganda) show widespread use of informal channels of remittances. The fact that in several Asian countries (China, Pakistan, and India) recorded remittances quadrupled, tripled or doubled between 2001 and 2003 may be in part due to a shift in flows from informal to formal channels in response to tightened regulatory scrutiny since September 11, 2001 (World Bank, 2005).12 A recent World Bank study by Sander and Maimbo (2003) reports that unrecorded flows appear to be high in Africa. In Sudan, for example, informal remittances are estimated to account for 85 percent of total remittance receipts. Preliminary findings from Mazzucato, van den Boom and Nsowah-Nuamah (2004) of the Ghana Transnational Networks research program in Amsterdam find that as much as 65 percent of total remittances to Ghana may be sent informally and the Bank of Ghana estimates that informal flows are at least as high as recorded flows. In South Africa an informal money remittance system exists side-by-side with the formal system, and the bulk of remittances to neighboring countries flows through informal, rather than formal channels.13 In Comoros informal transfers account for approximately 80 percent of remittances (da Cruz, Fegler, Schwartzman, 2004). One explanation for the generalized use of informal channels is the weakness of the Comoros banking sector. Comoros has only one commercial bank.
One example of an informal remittance transfer system is the Somali xawilaad. The xawilaad is an informal system of value transfer that operates in almost every part of the world (Horst and Van Hear 2002). The system is operated by Somalis and mainly used by Somalis. Interviews conducted in Virginia (one of the areas with the largest Somalian migrant population) report that there are two large companies providing transfer of remittances to the Somalian community: Dahbbshil and Amal.14 The system relies heavily on telecommunications. For that reason xawilaad companies have invested in telephones, mobile radio systems, computer networks, and satellite telecommunications facilities (Motclos and Kagwanja 2000, Gundel 2003). Transfers by xawiilaad are fast and made with great efficiency (Montclos 2002). However, it is very difficult to estimate the amount of remittances sent through this system to Kenya (the largest refugee site of Somalis) and to Somalia.
An estimate of total remittances.
In the absence of systematic studies of the magnitude of informal, unrecorded remittances it is difficult to assess their impact or the policy significance of efforts to move them into formal financial channels. In this section we use a simple econometric technique to estimate the extent to which unrecorded remittances may exceed official estimates, using data on 143 observations on migration and remittances for developing countries taken from Adams and Page (forthcoming).
The Adams and Page data reveals two types of situations in which it is likely that international remittances are underreported: first, observations where there is international migration but no recorded official remittances (N=35); and second, observations where international migration as a share of country population is much larger than official remittances as share of country GDP (N=41). In each of these situations it is likely that there is a large volume of informal, unofficial international remittances flowing back to the labor-exporting countries.
To predict total remittances, and hence derive an estimate of unofficial, unreported remittances, we assume that recorded remittances are less than or equal to total remittances. We also assume that remittances per migrant in the labor-importing country are proportional to per capita income in the labor-exporting country and are influenced by other factors such as the educational level and macroeconomic stability of the labor-exporting country. If these assumptions hold, country observations that have high levels of official remittances as a share of GDP relative to the share of migrants in the population, controlling for other migrant and macroeconomic characteristics, are likely to define the "true"relationship between total remittances and these variables. These observations define an "outer-bound" relationship between total remittances (official and unofficial) and their determinants.
To predict total remittances (official and unofficial) we specify the following equation for the 67 observations in the Adams-Page data set which have positive values for both migrants as a share of population and official remittances as a share of GDP:
(1) REMit= a0+ a1MIGit+ a2BMit+a3EDSit +bjDj+ e
where for labor-exporting country i at time tREM is the share of official recorded remittances in country GDP; MIG is migrants asshare of country population, BM is the black market exchange rate premium ((black market rate/official exchange rate-1) x 100) in the country, EDS is the share of country population over 25 years that has completed secondary education, and e is an asymmetric error term that constrains most observations to lie below the regression plane. Five regional dummy variables, Dj are also included in the model to allow for fixed effects.
From an economic standpoint, the level of international remittances received in a country will depend heavily on the number of migrants produced by that country. The relationship between remittances share and share of migrants in equation (1) should therefore be positive and significant. Various studies have suggested that the larger the black market premium (that is, the difference between black market and officialexchange rates), the more remittances will be remitted through unofficial, rather than official, channels. The relationship between remittances share and black market premium in equation (1) is thus expected to be negative and significant. With respect to the educational variable, human capital theory generally argues that more educated people are more likely to migrate, and some micro-level studies have found that since more educated people earn more, they are also tend to remit more of their earnings. It is therefore expected that the relationship between remittances share and education will be positive and significant.
If the error term in equation (1) is assumed to the distributed normally with zero mean, the predicted values of total remittances (official and unofficial) derived from estimating the model using official remittance data as the dependent variable will under predict the "true" value of total remittances. One approach to this problem would be to estimate equation (1) using an asymmetric error term or a composite error consisting of both a symmetric and asymmetric component. This "outer-bound" function would conform to our assumption that official remittances must be equal to or less than total remittances, and would allow us to predict total remittances for observations in which total remittances were un- or under-reported on the basis of the outer-bound parameters.
We know nothing, however, about the likely distribution of the asymmetric component of the error, making application of the outer-bound method problematic. For this reason, we allow the regression plane to lie above the "average" estimate in migration-remittance space by defining a dummy variable (MIGI) to be 0 in cases where the ratio of migrants as a share of population to remittances as a percentage of GDP is greater than 2, and 1 otherwise. This decision rule is equivalent to assuming that severe underreporting of remittances occurs in 60 percent of the observations for which we have data on both migration and remittances. The MIGI dummy variable, thus represents a simple approach to deal with the asymmetry of the error term implied by the assumptions above.
Table 3.4 summarizes the results obtained from estimating equation (1) by OLS with a normally distributed error. All of the variables have the expected signs and, except for the MIGI dummy variable, they are all statistically significant. The magnitude and precision of the parameter estimates of the explanatory variables are quite robust to changes in the percentage of observations constrained to lie below the regression plane in migration-remittance space. The range of the estimated parameter of the MIG1 dummy variable is also reasonably compact (1.98 - 1.44), when .the percentage of cases of underreporting is allowed to vary from 30 - 75 percent of the observations. The significance of the parameter estimate on the MIG1 dummy variable increases, not surprisingly, as we constrain a smaller proportion of the observations to lie below the regression plane.
To predict total remittances (official and unofficial), we apply the parameters from equation (1) to the 76 observations in the data set where either: (a) migrants as share of country population divided by official remittances as share of country GDP is greater than 2 (that is, migration as a share of population is much larger than remittances as share of GDP); or (b) there are migrants, but no recorded official remittances. In this step the MIGI dummy variable is set to 1 for all predicted values. For cases that do not meet the above two criteria (those which define the "outer-bound" sub-sample) we accept the reported level of official remittances as the estimate of total remittances (official and unofficial).15
The results of our estimates are reported in Table 3.5 and Table 3.6. The share of unrecorded (unofficial) remittances in total remittances is reported for each developing region, together with an estimate of the volumes of these remittances. Our results support the widely held belief that unrecorded remittances are large. The share of unrecorded remittances in total remittances in our estimates averages 48 percent worldwide, ranging from 73 per cent in Sub-Saharan Africa to a negligible amount in South Asia.16 Sub-Saharan Africa has the highest share of unrecorded remittances, which may reflect the fact that informal channels are common in many African countries because the formal financial infrastructure is limited. (Sander and Maimbo 2003). In terms of absolute amounts our estimates suggest that total unrecorded remittances worldwide to developing countries may have been on the order of $57.53 billions in 2004, with East Asia and Pacific leading in terms of the regional distribution.
The results are very similar to estimates presented in other recent studies using very different methods. For example, Adams argues that, while no one knows the level of unofficial or informal remittances, some estimate that unofficial remittances may amount to 40 to 50% of official remittances.17 The ILO-ARTEP (1989) and Puri and Ritzena (2003) studies of remittances among Asian countries estimate unrecorded remittance flows as percentage of total remittances ranging from 13 percent (Sri Lanka) to 50 percent (Philippines). Our results show a value of 0 for Sri Lanka and 0.72 for Philippines.
The Remittance market
The importance of remittances as a means of development finance and household income in developing countries has sparked substantial interest in how the decision to remit is determined and what influences the volume and duration of remittances. This section reviews the literature on these issues.
What influences remittance flows?
It is important to underline that not all immigrants send home remittances, and equally, not all migrant households receive remittances. Remitting behavior varies depending, among other things, upon age, education, occupation, employment, motive for remitting, gender, size of the household, access to credit, and years since migration.
Microeconomic determinants of the decision to remit There is a growing body of literature which sheds some light on the microeconomic motives behind remittances (Stark, 1992; Brown 1997; Poirine, 1997; Smith 2003; Rapoport and Docquier 2004, 2005; Russell, 1986; and Solimano, 2003, 2004). These surveys list three basic motives for remittances: altruism (family obligations and assistance, inheritance), insurance (indemnifying the human and social development of the family left behind against income shocks), and investment (asset accumulation back home as part of migration life-cycle planning). Table 3.7 provides a summary of the different models and the variables proposed to explain why emigrants send part of their income to family and relatives in source countries. The table gives the predicted signs for the effects of explanatory variables on the decision to remit. For example, the altruistic model predicts that the migrant, is motivated by the well being of the family in the home country, would tend to decrease his remittances over time. This prediction is shown with a negative sign in table 3.7.
In fact remittance decisions are complex and respond to multiple motives. Van Dalen, Grownewold and Fokkema (2005) examine empirically the motives for remittances to households in Egypt, Turkey and Morocco with family members living abroad. Their results show that it is difficult to determine clearly which motive prevails. Motives vary across countries, and within a country different motives can explain remittance behavior among households and over time. As they state, “the inconclusive nature of empirical research is understandable. One cannot expect remittances to be driven by a single motive”.18 Data from similar surveys of Latin American migrants in the United States and Japan point to some interesting differences in behavior related to the age, education and job status of the remitter. In the United States, Orozco (2004) found that migrants from Central American and Caribbean send an average of US$ 200- 300 monthly to their countries of origin. A 2001 Survey performed by the IADB found similar results. The Inter-American Development Bank recently commissioned a study to gain a better understanding of the behavior of migrant workers from Brazil and Peru living in Japan, where more than 435,000 Latin American adults are live (IADB, 2005). Orozco (2004) finds that migrants living in Japan on average send US$ 600 monthly to their home country, compared to US$ 200 remitted by immigrants living in the USA.
Table 3.8 shows the differences across remitters from Latin America in the two destinations. Migrants to Japan are older, higher skilled, and earn substantially larger incomes than their counterparts in the United States. They also remit a larger portion of their incomes than migrants residing in the US. The channels through which the two groups remit are also quite distinct, with Latinos in the US using mainly cash to cash transfer mechanisms (such as Western Union) and those in Japan preferring account to account transfers through financial institutions.
These data largely confirm a number of hypotheses regarding migrant behavior. Temporary workers (or migrants planning to return) such as those in Japan tend to remit a larger share of their income. Less educated migrants and those with lower incomes tend to use informal financial channels and cash transactions. But, a number of empirical studies also find that unskilled workers have a higher propensity to remit than skilled workers, although the latter category earn larger incomes and hence may send larger nominal amounts of remittances. This is not borne out in the US-Japan comparison where skilled Latin American workers, remit both larger absolute amounts and a higher proportion of their income.
Macroeconomic influences on remittance behavior
Macroeconomic factors also appear to influence the volume of remittances, once the decision to remit has been made. Several studies have found that the flow of remittances is positively correlated with growth in the host countries. The income and employment situation in the remittance source country affect the migrant’s disposable income, as well as saving behavior, both of which affect the size of remittances. (IMF, 2005). The cost of living in the recipient country is also an important factor affecting a migrant’s remittance decision. Surveys suggest that the same remitter may reduce flows to destinations where the cost of living is lower.
Recent studies support the hypothesis that remitters may respond to homeland crises. (Hysenbegasi and Poza, 2002). As a country situation deteriorates, emigration numbers may rise and remittances increase. Kapur (2003) reviews the effects of the economic downturn in Ecuador in the late 1990s on the increase in remittances received. Remittances more than doubled, from US$ 643 million in 1997 to more than US$ 1.4 billion in 2001. The experience of the Philippines during the Asian Crisis demonstrates how exchange rate movements can affect the amount of remittances. Yang (2004) finds that appreciation of a migrant’s currency against the Philippine peso leads to an increase in the amount of remittances received from abroad.
Moving the money: market structure and costs
Surveys reveal that migrants use a wide array of mechanisms to send money to their home countries: banks, credit unions, small and large money transfer companies, postal services, hand delivery and other mechanisms such as hawala (Pakistan and Bangladesh)or hundi (India). A number of surveys of migrants have been undertaken to gauge the extent of difficulty encountered by migrants in using these channels (Citizenship and Immigration of Canada, CID 2004, DFID, 2005). The surveys reach similar conclusions, finding a series of problems in both sender and the receiver institutions and markets.
The choice of the intermediary is affected by, among other things, costs, trust in the intermediary, and convenience factors --such as location, hours of operation and language-- and identification requirements. Among these, high remittance costs stand out as the most important factor affecting the choice of service provider, instrument (check, money order, electronic wire, pre-paid card, debit card, and hand-carry), and amount of remittance flows. The fee for sending remittances generally reflects two components: a fee to send the money and the commission on the exchange rate of the quantity converted into local currency.
The cost of sending money home varies significantly by country, transfer channel, and method of transfer. In the US, for example, the cost of sending money to Latin America from the US ranges from, 4.94% in El Salvador to 11.75% in Venezuela (Orozco 2004). Most market analysts believe that the main reason for high transfer costs is lack of competition. For example, transfer companies to Mexico, El Salvador, and Guatemala charge lower fees than companies sending money to Jamaica and the Dominican Republic, where there is no competition. Countries having market restrictions such as Cuba and Haiti face higher charges. Venezuela for which Western Union conducts nearly 50 percent of the remittances has higher costs than for Colombia, Ecuador, Peru and Bolivia to which Western Union transfers only around 10 percent of total remittances (Solimano 2004). The industry in the Andean Markets is dominated by a small number of money transfer firms that generally charge higher fees than banks, which have a small participation in the remittance market.
In recent years, the remittance market for Latin Americans in the US has become more transparent and competitive and the costs of sending money have been slowly decreasing. Charges have decreased with greater competition and use of technology (e-remittances, debit card transfers, etc).19 But according to Solimano (2004) the costs of sending money from the United States to Latin America are double those of sending it to India or the Philippines. The foreign exchange spread is also higher. This in part is explained by higher concentration in the operators markets in Latin America.
In the U.K, a flat money transfer fee is charged by service providers. The fee declines as a percentage of the amount transferred, and as in the case of the United States, it varies with the destination. The fee for £100 can range between 3% and 35% of the value sent, while it ranges between 2% and 6% for sending £500. (Table 3.9) Major money transfer operators charge from £3 to £14 for sending £100 depending on the country of destination. The DIFD Survey (2005) finds that money transfer operators (MTOs) in the U.K tend to offer lower rates than banks, as well as more convenient services, such as longer opening hours.
4. Migration, Growth and Welfare
While there is general agreement that remittances channel billions of dollars in money and goods from immigrants back into developing countries, there is less consensus on the welfare implications of the movements of money for developing countries. Controversy also remains as to migration’s overall impact on labor exporting countries and their migrant-producing communities. This section looks at the growing body of evidence on how migration and remittances impact economic development and poverty reduction.