REMITTANCES AND FINANCIAL DEVELOPMENT: SUBSTITUTES OR COMPLEMENTS IN ECONOMIC GROWTH?

II.1 Remittances and economic growth

The literature on workers’ remittances to developing countries has developed rapidly in recent years. Many empirical studies concentrate on the impact of remittance inflows on the living standards of recipient households, finding a significant positive influence on poverty alleviation, education and health improvements (Edwards and Ureta, 2003; Page and Adams, 2003; Hildebrandt and McKenzie, 2005; World Bank, 2006; Yang, 2008).¹ Others focus on the short-run macroeconomic impact of workers’ remittances, typically finding a positive correlation with aggregate income, investments and employment (Adelman and Taylor, 1990; Glytsos, 1993, 2005; León-Ledesma and Piracha, 2001; Bjuggren et al., 2010).

Although of great consequence, these results say nothing definite on the effects of an increase in remittance inflows upon the economic growth of the receiving country in the long run, which largely depends on how such financial resources are used, whether directly by recipients or indirectly, through the intermediation of financial institutions, by other people in the country. If remittances are channelled into investments, going to finance the start-up of small enterprises or the accumulation of human capital, or if they improve the creditworthiness of recipients and their access to external financial resources (collateral function), the impact on economic growth is positive. If, however, the prevailing end uses of remittances are on increasing consumption and expenditures on housing, land and other forms of second-hand non-financial assets the association with the economic growth is very feeble – depending on the type of purchased goods and on the existence of unexploited national productive capacity. In addition, if we consider the wealth effect of remittances on the labour force participation of recipient households and the effects on the competitiveness of national exports, the association between remittances and growth turns out to be negative. First, as Chami et al. (2005) argue, remittances take place in a context of asymmetric information due to the long distances between migrants and recipients. Therefore remitters do not have the chance to monitor the end-use of remittances, and moral hazard problems could induce recipients to consider remittances as a substitute for labour income and reduce their labour supply.² Second, the technology capacity of developing economies depends on the competitiveness of the tradable goods sector. In this view, workers’ remittances may negatively affect growth of recipient countries through the appreciation of their real exchange rate (Amuedo-Dorantes and Pozo, 2004; Lopez et al., 2007; Lartey et al., 2008).³

Given the contrasting channels through which remittances influence long-run growth, precisely what is the actual causal nexus between remittances and GDP growth is therefore a matter of empirics. The evidence is far from conclusive, results varying with time and the country sample analysed, the definition of remittances used, the econometric specification and estimation methodology adopted. However, a stylized fact that seems to emerge is that on average the effect of remittance inflows on economic growth is small in magnitude and statistically not very robust with regard to both the sign and significance of the estimated coefficients.

Chami et al. (2005) estimate cross-section and panel growth regressions to test whether and to what extent growth in real GDP per capita for 113 countries in the period 1970–98 is influenced by the remittance inflows of their emigrant workers. Together with other standard controls, they consider alternatively remittances as a share of GDP and the change in the remittance-to-GDP ratio as determinants of economic growth and show that the level of remittance inflows is statistically insignificant while a change in the remittance-to-GDP ratio negatively affects the growth of GDP. Clearly, endogeneity of remittances is a serious concern as higher growth rates in developing countries could stimulate more conspicuous transfers. In addition, both remittances and the rate of growth of the receiving economy might depend on some omitted factors. To overcome these problems, Chami et al. (2005) employ instrumental variable techniques, by using as external instruments for remittances the income and real interest gaps between the receiving country and the US (considered as a representative host country). Results from IV estimations confirm that the net effect of an increase in remittance inflows on the economic development of receiving countries is negative.

Two successive cross-section studies conducted by the IMF (2005) and Faini (2006) use as instruments for remittances the distance and the sharing of a common language between the home country and the main destination country for their emigrants. Moreover, like most subsequent studies, they adopt the broad definition of remittances recommended by the World Bank which add two other Balance of Payment items of ‘employee compensation’ and ‘migrant transfers’ to the item ‘workers’ remittances’. In both IMF (2005) and Faini (2006) the coefficient on the remittance-to-GDP ratio proves positive but statistically insignificant.

Acosta et al. (2008), in the ambit of the far-reaching study on the determinants and consequences of remittances in Latin America conducted by the World Bank in 2006, present panel regressions on a sample of 67 countries in the period 1991–2005. Using external time-varying instruments (the average output per capita of the main destinations for migrants across the OECD countries weighted by the inverse of the distance between the sender and each destination and the share of migrants from the sender country in each destination), the authors of the report show that remittances positively influence economic growth even if their economic impact is quite modest in magnitude (the elasticity of GDP to remittances being 0.2) and broadly limited to capital accumulation.

The World Bank's findings are then confirmed in two successive studies on Latin American and Caribbean countries by Ramirez and Sharma (2008) and Mundaca (2009), the former using Pedroni's fully modified ordinary least squares (FMOLS) estimator and the latter the generalized method of moments (GMM) estimator.

Barajas et al. (2009) propose a new instrument as the remittance-to-GDP ratios of all countries other than the receiving one in question. To take into account the possibility that this instrument is correlated with the growth rate in the receiving countries through trade links, the authors include the trading-partner growth in the set of control variables. Regression results for a panel of 84 countries in the period 1970–2004 show that the higher the ratio of worker remittances to GDP, the lower is the five-year growth rate. However, Vargas-Silva et al. (2009), using the same instrument as in Barajas et al. (2009), but focusing on a smaller set of Asian countries, find that remittances have a positive, but non-significant effect on the economic growth of the receiving countries.

Other studies treat endogeneity problems by using internal instruments via dynamic panel techniques. Giuliano and Ruiz-Arranz (2009) analyse the five-year growth of GDP of a set of 73 developing countries in the period 1975–2002 using the system GMM estimator of Arellano and Bover (1995) and show that, on average, remittances have no significant influence on growth. Similarly, Jongwanich (2007) estimates a system GMM model for a panel of 17 countries in the Asia and Pacific region during the period 1993–2003, finding that the relationship between remittances and the three-year growth rate of GDP is statistically insignificant.

By contrast, Catrinescu et al. (2009) estimate a GMM growth model for a larger sample of 114 countries during the period 1991–2004, besides a more traditional IV model, and find that the coefficients on remittances are positive, albeit statistically not very robust and economically small. Also using the GMM estimator, Fayissa and Nsiah (2008) confirm that remittances exert a small positive effect on the GDP growth of African countries between 1980 and 2004.

II.2 Finance, institutions and growth

Although hotly debated in the past (Schumpeter, 1912; Robinson, 1952; Hicks, 1969), the idea that a well-functioning financial system is an essential requirement for a country's economic take-off and steady growth is nowadays considered as almost obvious by most scholars (Miller, 1998).⁴ The predominant view is that the existence of efficient banks provides savers with greater liquidity and risk diversification and mitigates asymmetric information problems with respect to investors arising both before and after the supply of finance. In this way, a highly developed banking sector in a sound regulatory environment ends up boosting saving and capital accumulation and, above all, enhances the efficiency of capital allocation and technological development within the economy.

The vast number of empirical studies covering the finance–growth nexus in the last 20 years is broadly consistent with this view: even after controlling for reverse causality and omitted variable problems, countries with better developed financial systems experience on average higher investments, output growth rates and more rapid technological advancements (Levine, 2005).

A key issue in this literature is how to measure financial development. Starting from Goldsmith (1969), the traditional approach refers to quantitative indicators of financial depth capturing the relative size of the financial sector and the banking industry, such as liquid liabilities of the financial system, bank deposits or bank credit to private enterprises all normalized by the country's GDP. Albeit to differing degrees, all these quantity-based indicators suffer from three main weaknesses concerning, first, the exogeneity of financial aggregates with respect to investments and growth, secondly, their strict correlation with price trends in financial markets and, thirdly, their capacity to capture the fundamental role played by banks in economic development (De Gregorio and Guidotti, 1995; Lucchetti et al., 2001; Honohan, 2004; FitzGerald, 2006). In this view, a number of papers have suggested using quality-based indicators of financial development built by aggregating some measure of the efficiency of banks in the economy (Lucchetti et al., 2001; Hasan et al., 2007). The idea is that the efficiency with which a single bank uses inputs and produces outputs is less influenced by the future rates of GDP growth than the quantity of funds intermediated and credit granted. In addition, the microeconomic efficiency of banks better captures their effectiveness at allocating capital, selecting the right projects and entrepreneurs and monitoring the use of money lent. Existing evidence confirms that the microeconomic efficiency of banks is positively associated with output growth at the level of country, region and industry, and that it has more significant effects than quantity-based indicators of financial development (Lucchetti et al., 2001; Berger et al., 2004; Hasan et al., 2007; Romero-Ávila, 2007; Inklaar and Koetter, 2008; Kessy, 2008; Hakenes et al., 2009).

Besides the quality of financial institutions, the development and growth literature has increasingly emphasized the role of appropriate political and governance institutions, capable of protecting property rights, enforcing laws and reducing transaction costs, as a causal factor fostering economic growth. This approach has a noble intellectual history, from the Baron de Montesquieu and Adam Smith to James Buchanan and Douglas North, further enhanced by modern statistical methods and the recent production of statistics on political institutions which have provided rigorous empirical legitimacy (Knack and Keefer, 1995, Hall and Jones, 1999; Acemoglu et al., 2001; Easterly and Levine, 2003; Glaeser et al., 2004; Rodrick et al., 2004).

II.3 Interactions between remittances, finance and institutions in promoting growth

The weak or insignificant average effect of remittances on growth has prompted many authors to look at different possible sources of heterogeneity in the remittance–growth nexus to test whether remittances have different long-run macroeconomic impacts on the receiving country in accordance with some of its structural characteristics. Two such characteristics are financial and institutional development.

Whether well-functioning financial markets and governance institutions magnify or weaken the impact of remittances on growth is a priori ambiguous. On the one hand, remittances can be viewed as substitutes for formal external finance and weak institutions. When domestic credit markets are poorly developed, a large number of households with potentially productive investment projects have no access to external finance or may borrow only at a large premium over the risk-adjusted interest rate. In this case, remittances may be used as an alternative source of finance allowing recipient households to fund productive activities, or may be pledged as collateral, helping recipients to access formal credit markets. Alternatively, they may be intermediated directly by recipients, enabling other village households to overcome credit constraints and start new businesses. On the other hand, well-functioning banks attract remittances in the formal credit circuit by lowering monetary and non-monetary costs of opening a deposit account (Freund and Spatafora, 2008) and allow financial resources to be funnelled towards more productive, growth-enhancing projects by improving selection and control of entrepreneurs.

Similarly, well-functioning political and governance institutions might encourage recipient families to use remittances to improve the education of their young members and establish entrepreneurial activities. However, the presence of weak, unreliable institutions can render saving and investing remittances productively more urgent.

Existing evidence seems to document that remittances are a substitute for external finance in boosting investments and growth. An initial influential attempt to empirically explore how finance and remittances interact in the growth process is the study by Giuliano and Ruiz-Arranz (2009). Consistent with the substitutability hypothesis, they find that the level of financial development in the receiving country weakens the impact of remittances on GDP growth and that remittances contribute to promote growth only in countries where the financial sector is sufficiently small. Similarly, the shallower the financial sector in the receiving country and the harder the access to external finance, the greater the marginal impact of remittances on investments becomes. Therefore, where credit markets are thin, remittances are additional resources that can be invested in productive activities; where financial systems can guarantee widespread access to credit, liquidity constraints are no longer an issue and remittances are used mainly to fund consumption expenditures.

Substitutability between remittances and finance appears to be robust to the sample of countries analysed and the econometric method employed. Ramirez and Sharma (2008), for example, after checking for the presence of stochastic and/or deterministic time trends in growth rates, remittances, finance and other control variable series, find that in Latin America the positive impact of remittances on economic growth is more pronounced in less financially developed countries. Calderón et al. (2008) consider Latin American countries and use external instruments (distance and migration flows) for remittances. Their findings confirm the hypothesis of substitutability for which the promoting effect of remittances on the investments and economic development of the receiving country declines as the domestic financial sector becomes deeper. Similar results are reported by Barajas et al. (2009), who analyse a larger set of countries by using fixed-effect IV methods, and by Bjuggren et al. (2010), who focus on the impact of remittances on investments.

The only study that seems to find a complementarity relationship between remittances and finance in boosting economic growth is Mundaca (2009). In particular, in this study the author reports that when domestic credit from banks is included in the model the marginal impact of remittance-to-GDP ratio on the growth rate of per capita GDP is 0.54, 15% greater than the estimated effect when omitting control for financial development. However, this result does not prove the complementarity of remittances and other financial resources for growth, as evidenced by the fact that Ramirez and Sharma (2008), for the same set of countries, confirm that the coefficient for remittances is greater when controlling for the level of financial development, like in Mundaca (2009), but also document that the impact of remittances on growth is lower when the financial sector is highly developed (i.e. the coefficient on the remittances × finance interaction term is negative).

All these studies, however, proxy financial development with quantity-based indicators of the country's financial depth. Apart from the weakness inherent in such indicators (see Section II.2 above), to the extent that remittances enter the financial circuit and increase the size of the financial sector, the remittances × finance term could capture the non-linear effect of the financial sector's size on growth (Rioja and Valev, 2004).

With regard to the role of political and governance institutions, Calderón et al. (2008) interact remittances with the ICRG indexes of quality of the political and socioeconomic environments as published by the Political Risk Service Group and find that in Latin American countries with a high quality of institutions remittances have a greater impact on investments and economic growth. Catrinescu et al. (2009) reproduce a similar exercise on a broader set of countries. Like Calderón et al. (2008), they show that good governance, sound socioeconomic conditions, a low level of ethnic tensions and the prevalence of law and order are essential conditions for a positive developing impact of remittances.

III.1 Models and methodology

Following Giuliano and Ruiz-Arranz (2009), the basic model we estimate is a growth regression including the ratio of remittances to GDP, the level of financial development and an interaction term between these two variables, plus the set of control variables X described hereinafter:

(1)

The sign and significance of the coefficient for the interaction term, α₄, allows us to discriminate between the substitutability and complementarity hypotheses. If α₂ > 0 and α₄ < 0 remittances promote growth only in receiving countries whose financial system is poorly functioning. By contrast, when α₂ < 0 and α₄ > 0 or when they are both positive and jointly significant, this means that a better functioning financial sector and more efficient banks would lead remittances towards growth-enhancing activities.

However, financial development and the efficiency of domestic banks could capture a more general effect of the quality of political and governance institutions. This is especially true in developing countries where the degree of integrity of politicians and bureaucrats, the design of regulation and the application of rule of law can be so poor as to make the presence of sound financial institutions unlikely. To take these aspects into account we augment Model (1) with institutional quality indicators:

(2)

Models (1) and (2) are estimated by using OLS and SGMM techniques to address the issue of the endogeneity of remittances, financial development and institutional quality to GDP growth. While the growth literature has suggested many instruments for these variables, such instruments are often time-invariant country characteristics which cannot be used in a panel framework. Moreover, economic growth is likely to be a persistent series, and a misspecification of GDP dynamics can lead to autocorrelation in the error term and biased estimates. For such reasons, and for the sake of comparison with Giuliano and Ruiz-Arranz (2009), we estimate a dynamic panel model using the Arellano and Bover (1995) system GMM estimator, where lagged values of remittances, financial development, institutional quality, interaction terms and dependent variables act as internal instruments.

The sample period is 1970–2005 in model specifications with quantity-based indicators for financial development, and 1991–2005 when we consider the inefficiency of the domestic banking system. For the sake of comparison, we estimate the specifications with quantity-based indicators also on the subperiod 1991–2005. The sample period 1970–2005 has been split into seven non-overlapping five-year periods and Model (1) is estimated using the variable mean values. With regard to the period 1991–2005, estimates are run both with five-year averages and annual data.

III.2 Data and variables

We consider a panel of 66 developing countries for the period 1970–2005.⁵ The dependent variable is the growth rate of real per capita GDP in constant dollars as drawn from the World Development Indicators database. Data on remittances come from World Bank staff estimates based on the IMF's Balance of Payments (BoP) Statistics Yearbook 2006. Remittance inflows are defined as the sum of three items: workers’ remittances, compensation of non-resident employees for work performed for residents of other countries, and migrants’ transfers.⁶ As is well known, due to the high transaction costs of using the formal banking sector, a large part of migrants’ remittances go through informal channels like ‘hawala’ or ‘hundi’ and are not recorded in the BoP statistics.⁷ Obviously, our estimation of the growth effect is based on the official data available on remittance flows and excludes the amount of money transferred through informal channels which, according to the figures reported by El Qorchi et al. (2003) and Freund and Spatafora (2008), ranges from 20% to 80% of the recorded flows, reaching peaks of 250%. To be sure, the exclusion of non-official remittances, which is a common problem for studies on the macroeconomic impact of remittances, introduces a bias in our findings in an unpredictable direction. However, to the extent that non-official remittances compensate for low official remittances and the latter are lower where the deficiencies of the formal banking channel are greater, the global impact of remittances on growth should be higher or lower than that estimated using official remittances according to the development of the financial sector be a substitute or a complement of remittances in the growth process (Giuliano and Ruiz-Arranz, 2009).

With regard to financial development, we use the four traditional quantity-based indicators employed in the study of Giuliano and Ruiz-Arranz drawn from the World Development Indicators database: the ratio of liquid liabilities of the financial system to GDP (M2), the ratio of domestic credit provided by the banking sector to GDP (CREDIT), the ratio of bank deposits to GDP (DEPOSIT) and the ratio of claims on the private sector to GDP (LOAN).

In addition, we build an inefficiency index of the national banking system on the basis of the widely employed cost to income ratio⁸:

where B_i is the number of banks headquartered in country i and w_bt the market share of bank b in terms of total assets. We use data for 53,820 banks available from the Bankscope database (Fitch-IBCA) for the period 1991–2005. Among those included in our sample, the country with the highest number of banks surveyed in Bankscope is the Russian Federation with 454 banks, while the minimum number of banks is 4 and corresponds to Guyana, Seychelles, Sierra Leone, St Kitts and Nevis, St Lucia and Togo. Importantly, the Bankscope database contains information on domestic banks and subsidiaries of foreign banks, but does not provide information on either foreign branches or international loan flows. However, direct lending from abroad and the activity of foreign branches usually concern large borrowers and are scarcely affected by the flow of remittances.

For the sake of comparison, the set of explanatory variables we include in Model (1) is similar to those used by Giuliano and Ruiz-Arranz (2009).⁹ First, as the neoclassical growth theory suggests, we control for the initial level of real per capita GDP (GDP_t–1), to test for the convergence hypothesis predicting that poorer economies grow faster than richer ones (Barro, 1991), and for capital accumulation (INVESTMENT, defined as the log of gross fixed capital formation on GDP) and population growth (POPULATION GROWTH, measured by the log difference of total population) expected, respectively, to be positively and negatively correlated with per capita GDP growth. Then we control for countries’ inflation rates (INFLATION, defined as the annual percentage change in CPI), expected to reduce returns on physical and human capital and hence investments and total factor productivity (De Gregorio, 1993; Roubini and Sala-i-Martin, 1995); the size of public expenditures (GOVERNMENT EXPENDITURES, defined as the central government final expenditure on GDP), to test whether productive or unproductive government spending prevails (Barro, 1990; Herrera, 2007); and openness to international trade (OPENNESS, defined as the log of imports plus exports on GDP), in order to contrast import substitution and export-led hypotheses (Baldwin, 2003). The series of all the control variables are drawn from the World Development Indicators of the World Bank.

In Model (2) we add a set of controls INSTITUTIONAL QUALITY which includes well-known indicators collected by Kaufman et al. (2006) and the International Country Risk Guide. In particular, from Kaufman's dataset we consider three types of variables. First we consider three indicators capturing the transparency and efficacy of the political process: POLITICAL STABILITY, which measures the risk of sudden and violent changes in government; GOVERNMENT EFFECTIVENESS, which measures the ability of the government to implement good policies; and VOICE AND ACCOUNTABILITY, which measures citizens’ participation in politics. Second, we consider two indicators for protection of civil liberties and legal rights: CONTROL OF CORRUPTION, which measures people's perceptions of corruption in the country, and RULE OF LAW, which quantifies the confidence of the agents in the rules of society and in the protection of property rights. Finally, we take into account the effect of the quality of the market's regulatory environment by including the variable REGULATORY QUALITY which measures the incidence of market-unfriendly regulation like price controls or inadequate bank supervision. All these indicators vary between –2.5 and 2.5, with higher scores corresponding to higher quality of institutions.

As we stated above, political, legal and regulatory institutions should affect decisions to consume, hoard or invest remittances by the influence they exert on the level and uncertainty of returns on saving in the receiving country. In this perspective, we alternatively use the risk rating indicators from the International Country Risk Guide (ICRG) which aim to measure exactly the riskiness of the political, economic and financial environment for international investors. The POLITICAL RISK RATING includes 12 weighted variables covering both political and social aspects, like government stability, ethnic tension and democratic accountability, assuming values between 0 (high risk) and 100 (low risk). The ECONOMIC RISK RATING includes the risk associated, according to a certain scale, to five economic variables (GDP per capita, GDP growth, inflation rate, government budget and current account balance as a percentage of GDP). The FINANCIAL RISK RATING measures a country's ability to pay back its national and foreign debt obligations and aggregate risk grades from five variables like foreign debt services and official reserves. Both economic and financial risk ratings are based on 50 points, with a higher rating indicating lower risk. The total points from the three indices are then weighted (0.5 to the political risk and 0.25 to economic and financial risk) to produce the COMPOSITE RATE INDEX. With respect to the Kaufman indicators, that are available only after 1996, ICRG indicators cover a longer time span (1991–2004), even if they are available for only 57 of the 66 countries in our sample.

Although widely used in the literature, it is important to bear in mind the fundamental criticisms raised by Glaeser et al. (2004) and Rodrik et al. (2004) against these indicators, pointing out the fact that they are outcome measures, based on the subjective perceptions of business people who tend to be influenced more by the general soundness of the economic life in their country rather than by specific permanent characteristics of local institutions. However, while these shortcomings make it difficult to draw precise policy indications on the type of institutional improvements to pursue, they should not invalidate the clues on substitutability/complementarity of institutions and remittances for economic development.

IV.1 Descriptive statistics

As shown in the summary statistics and correlations for the variables of interest (Tables 1 and 2), on average, remittances correspond to more than 3% of GDP in our sample, but in some countries like Jordan, Haiti, El Salvador and Jamaica they account for more than 15% of GDP. Pairwise correlation between remittances and GDP growth is negative, but extremely weak and non-significant.

As expected, the efficiency of banks, M2, CREDIT, DEP and LOAN are all positively correlated to GDP growth, even if the relationship is statistically significant only for BANKS’ INEFFICIENCY, M2 and LOAN. Financial development indicators are positively correlated to each other, although the correlation between different quantity-based indicators is much higher than the correlation between quantity- and quality-based indicators. However, what emerges is also that developing countries display a great variability of financial development, especially with regard to the efficiency of the domestic banking system. The lowest value of BANKS’ INEFFICIENCY is 7.1 for Ethiopia in 2001, while the highest is 460.2 for Thailand in 2000.¹⁰ Giving more detail, Table 3 reports country ranking on the basis of the mean value of banks’ inefficiency over the period 1991–2005. The most efficient banking systems are those of Syria, the Seychelles and Ethiopia, whose banks display a cost-to-income ratio that is much lower compared to banks in other countries. In contrast, the worst performing countries are Pakistan, Argentina and Thailand where banks, on average, report operating costs that are 90 times their income. Such results echo some of the evidence emerging from the World Bank's report, Banking the Poor (2009) where, for example, Thailand's banking system appears to be the most inefficient in handling loan applications (on average, the processing time for a new application is 21 days), also applying the highest fees for business start-up loans. In stark contrast, Ethiopian banks are those where it is easiest to apply for a business loan, required fees are low and decisions are taken in less than 7 days.

In order to help ascertain whether financial development increases or decreases the responsiveness of growth to remittances, in Figure 1 we plot the relationship between average remittances over the period 1970–2005 and the average growth rate of GDP, splitting the sample of developing countries according to the median value of M2, CREDIT, DEP, LOAN and BANKS’ INEFFICIENCY. With the traditional quantity-based indicators, a sort of substitutability relationship between remittances and financial development in fostering economic growth seems to exist. In countries with levels of financial depth under the median, remittances and growth appear to be uncorrelated, while the relationship becomes strongly negative at high levels of financial development, especially if we exclude Jordan. When we consider the inefficiency of the domestic banking system, however, financial development and remittances seem to be complements to economic growth. When the inefficiency of local banks is low, growth and remittances are positively correlated (once again, Jordan is an outlier), while for high levels of BANKS’ INEFFICIENCY the relation between growth and remittances reverses.

IV.2 Regression results

In Table 4 we report regression results for the basic model with financial development measured by standard quantity-based indicators using OLS and SGMM estimators.¹¹ As far as control variables are concerned, we find evidence of a convergence process and a positive correlation between investments and GDP growth, as did Chami et al. (2005), Acosta et al. (2008), Giuliano and Ruiz-Arranz (2009) and Catrinescu et al. (2009). By contrast, population growth and government expenditures appear to be negatively and highly significantly associated with economic growth (Jongwanich, 2007; Acosta et al., 2008) and the adverse impact of inflation is only rarely significant (Chami et al., 2005). Finally, the effect of the economy's openness is statistically non-significant. These results tend to be preserved both in sign and significance, even if in some specifications the statistical significance slightly changes.

Moving on to our key variables, we find that all four quantity-based measures of financial development are positively correlated with the country's economic development, regardless of the estimation method. In contrast, the average effect of workers’ remittances on growth is statistically not different from zero (Barajas et al., 2009; Giuliano and Ruiz-Arranz, 2009). This is true whether we consider REMITTANCES as exogenously or endogenously determined with respect to GDP growth, even if the coefficient changes sign from OLS (–) to SGMM (+) estimates. However, inclusion of the interaction term between remittances and financial development clearly shows that the insignificant average effect of remittance on growth hides the existence of significant effects varying with the level of financial development. In particular, consistent with the substitutability hypothesis, remittances appear to promote growth in countries where the size of the financial sector is small and access to external credit for receiving families is very limited, while they restrain GDP growth if the financial sector is large and funding investments is no longer an issue.

Things change when we consider the quality of domestic banks and take BANKS’ INEFFICIENCY as a measure of financial development (Table 5). Financial development is positively correlated with economic growth and, what is more interesting, it seems to be complementary to the effectiveness of remittances in boosting the growth of GDP. These results hold regardless of the estimation method (Table 5, columns 3 and 4)¹² and even after controlling for the size of the financial sector in the country (columns 5–8). In addition, similar findings hold with annual data (Table 6), although in this case coefficients are significant only in SGMM regressions.

Therefore, consistent with the information gathering and monitoring functions of banks, our findings indicate that remittances have a positive effect on economic growth only if the domestic banking system is sufficiently sound. To be precise, considering the value of coefficients in column 4 of Table 5, the value of BANKS’ INEFFICIENCY for which the remittance impact becomes zero is 52.72, while the sample mean (median) is 63.25 (61.8). This means that only a small group of countries can concretely benefit from remittances: at the 10 percentile of BANKS’ INEFFICIENCY distribution (40.93) an increase in the remittance over GDP ratio of 1% would cause a variation in GDP of around 1.4%. By contrast, at the median level of BANKS’ INEFFICIENCY, the impact of remittances would be negative and around 0.9% while at the 75 percentile the harmful effect can be estimated at 1.8%.

Figure 2 shows the impact of remittances on GDP computed for each country at the mean level of the inefficiency indicator over the period 1991–2005 as reported in Table 3. Out of 66 countries considered in the analysis, only 21 seem to benefit overall from remittances and, consistent with the previous literature, the effect is on average rather small in magnitude. Syria, the Seychelles and Ethiopia are the countries that gain most from remittances in terms of economic growth, their GDP growth increasing, respectively, by 10% and 6% thanks to a 1% increase in the remittance over GDP ratio. By contrast, Thailand, Argentina and Costa Rica, which are at the right-hand side on BANKS’ INEFFICIENCY distribution, are the most damaged by higher remittance inflows.

To test whether the diverging results between quantity- and quality-based indicators depend on the different time span, we also re-estimate all the specifications in Table 4 for the subperiod 1991–2005. With all four indicators (Table 7), results seem to hold compared to the 1970–2005 estimates; the coefficient of the interaction is negative and significant, although the remittance variable itself does not seem to exert any significant effect.¹³

Finally, in Table 8 we report SGMM estimates of Equation (2) with annual data. As far as the World Bank governance indicators are considered, CONTROL OF CORRUPTION, REGULATORY QUALITY and RULE OF LAW (columns 1–3) contribute significantly to economic growth and the positive sign on the interaction term shows that the impact of remittances is stronger in the presence of sound institutions. Specifically, a lower level of corruption, adequate bank supervision and protection of property rights are directly connected to the effectiveness of remittances in stimulating investments. The other three dimensions of institutional quality concerning the transparency and efficacy of the political process do not seem to influence the effectiveness of remittances in boosting growth, while only GOVERNMENT EFFICIENCY exerts a positive impact on the growth of a country's GDP. When we include in the regression all the political and governance variables (column 7), only REGULATORY QUALITY maintains its significance, reinforcing the positive effect of remittances on GDP. That said, what is important to note is that, after controlling for the quality of the institutional environment, the complementarity between remittances and the efficiency of domestic banks in economic growth is not rejected at a 1% level of confidence, regardless of the specification considered.

The last four columns of Table 8 report results of the estimates with ICRG indicators also employed by Catrinescu et al. (2009). While complementarity between the quality of the domestic banking sector and remittances and the positive impact of sound institutions on growth are confirmed,¹⁴ the complementarity between institutions and remittances is statistically weaker. Only the interaction with the ECONOMIC RISK RATING is negatively significant, meaning that the impact of remittances on growth is stronger where economic risk is higher and the productive use of remittances is more pressing. These results conflict with those provided by Catrinescu et al. (2009) for longer periods and a larger set of countries. Like us, they find that the political risk indicator has no direct impact on growth while it boosts the effect of remittances. However, when they split this indicator into its components they find that many such components, like ‘low ethnic tension or government stability, are preconditions for a successful use of migrant remittances’ (Catrinescu et al., 2009, p. 90).

Summing up, our results indicate that the size of the domestic banking sector and its efficiency have opposite effects on the development potential of remittances. Confirming the previous literature, the ease of access to credit, captured by financial depth indicators, seems to set in motion moral hazard behaviour on the part of recipient families who, not being financially constrained, are more inclined to direct remittances towards unproductive uses. However, we cannot exclude that the negative sign of the coefficient on the interaction term remittances × finance captures the decreasing marginal impact of financial sector size on growth. By contrast, when domestic banks are able to efficiently intermediate savings and allocate credit within the economy, remittances contribute positively to the economic development of the receiving country, going to finance productive activities run by the migrant's family or by other small potential entrepreneurs in the economy.