DO CONSTANT INFLOWS OF HIGH REMITTANCE AMOUNTS STIMULATE LONG-TERM ECONOMIC GROWTH IN LESS DEVELOPED ECONOMIES?

(1)

DO CONSTANT INFLOWS OF HIGH REMITTANCE AMOUNTS

STIMULATE LONG-TERM ECONOMIC GROWTH IN LESS

DEVELOPED ECONOMIES?

Evidence from the 20 largest remittance-receiving countries during 2001-2013

Master Thesis Finance

Tudor Crăciun 2934884

t.craciun@student.rug.nl

Date: 09/01/2020 Word count:

(2)

2 Abstract

This research paper aims to analyze the impact of constant large inflows of remittances to economic growth in a sample of 20 countries pertaining to mainly the African, South American and Asian regions. Through a methodology making use of ordinary least squares regressions and fixed and random effect estimations, it is made clear that an effect of remittances exists on economic growth, although albeit quite narrow. The amount of remittances received in an economy had a negative effect on the capacity of an economy to sustain long-term development.

Key words: remittances, economic growth, fixed effects estimations, random effects

estimations.

(3)

3

In the current economic and financial context, remittances have started playing a significant role in terms of capital outflows from developed countries to developing economies. Whilst remittances have started being investigated over twenty years ago, their effect on economic growth and financial development remains ambiguous and country-specific in most conducted studies. Moreover, as reported by the World Bank in their “Migration and Remittances - Recent Developments and Outlook” report, the remittance flows on a worldwide level have reached an all-time high level in terms of amounts transferred. Among these flows, it is already noticeable that some countries have constantly received remittance transfers in much higher amounts than others over the past years.

The intent of this research is to determine whether these sort of transfers carry on an economic growth effect into the highest-amount remittance-receiving countries. The intended period of observation for this study is 2001-2013, whilst the data encompasses the top 20 countries in the world in terms of amounts of remittances received, according to the World Bank.

(5)

5

It has become clear that remittances have a pivotal role on the worldwide financial and economic activities, however this role is not yet well-defined, as in the current economic context the effect of remittances on macroeconomic performance is still ambiguous. It is quite surprising that the literature on remittances is so polarizing considering the large amounts of such transfers. Therefore, this paper proposes an investigation into whether remittances do indeed stimulate economic growth for receivers of high amount of remittances.

The research question this paper proposes is the following: Have the largest remittance receiving countries in the world experienced economic growth due to the constant inflows of remittances?

Portrayed within Appendix A, multiple graphical representations show the evolution of Remittance flows over the recent years, starting with a top-down view on the countries receiving the highest remittance amounts. Naturally, countries like China and India lead in the total outstanding amount of remittances received due to their huge population and complementary number of migrants.

However, this study aims to investigate remittance flows in terms of Gross Domestic Product (GDP), hence the use of less developed countries is encouraged within this research paper. It is noticeable from Figure 1 of Appendix A that countries pertaining to an African region are more likely to be included within this research. As well as underdeveloped economies in Asia and South America.

As presented in Figure 2 of the same Appendix, it is made apparent that during the period of 2010-2015, the highest levels of growth in terms of remittance inflows occurred for Sub-Saharan African countries, whilst the Asia and Pacific region was also rather high.

(6)

6

Portfolio Debt and Equity Flows, etc. The role of remittances is best captured here, as it shows that their emergence in the 21st century is undeniable. Furthermore, it is clearly noticeable that Foreign Direct investment faces a decreasing path, whilst for remittances there is only room to grow.

(7)

7

Literature review

Adams and Page (2005) conduct research into whether remittance flows reduce poverty in developing countries and find strong, significant results that both international migration and remittances reduce poverty in the developing world. The authors do specify that their results might be subject to international migration and remittances being endogenous to poverty, however, the magnitude of this effect is seen as negligible, as the absolute value of the endogeneity bias is not large. This bears interesting informational content for this study, as multiple studies have tried in the past to study the interchangeable relationship between the effect of poverty and overall economic development. The common consensus on this topic is that economic growth reduces poverty in itself, however the vice-versa effect may not necessarily hold true, since GDP growth is impacted by multiple variables. However, since a reduction in poverty can be determined by economic growth, Adams’ and Page’s result could in fact have underlying implications about the remittance effect on growth, as this poverty-reducing effect could actually be a result of an unobserved economic growth outcome. It is therefore interesting to consider the results of this study for the remainder of this proposal.

Catrinescu, Leon-Ledesma, Piracha and Quillin (2009) argue the importance of high quality political institutions and economic policies in determining the effects of remittances on economic growth. Within their study, the authors find that good institutions lead to more efficient channeling of remittances, which should as a result lead to a higher per capita income growth and better output.

(8)

8

seem to be intended as capital for economic development, but rather poor economic performance compensation. They show that remittance-receiving individuals are in fact using this inflow of funds to reduce their participation on the labor market, basically compensating thus poor economic performance. Furthermore, upon running their regressions, Chami et al have indeed found significant results of a negative correlation between immigrant remittance growth and per capita GDP growth. One of the main questions raised in the concluding part of the paper is actually intended to be addressed within this analysis, as Chami, Fullenkamp and Jahjah advise future research to be directed into the effects of remittances on economies with large remittance transfers. As a result, the paper suggested within this proposal aims to answer this question to some extent, as the issues regarding data availability (poor reporting of remittances) previously identified by Adams and Page seem to have persisted since the release of their paper.

Meyer and Shera (2016) conduct an economic growth study in connection with remittance impact and find significant positive effects of remittances on growth. They argue the importance of remittances for economic growth in general and warrant further research into the determinants of GDP growth and whether financial performance can be achieved through high levels of remittances. This paper relates strongly to this proposal as the methodology is similar to the one presented here. Their results directly contradict the findings of Chami et al (2005), and cast doubt over the overall effect of remittances. However, it may also be the case that the dataset employed by Meyer and Shera is more recent, therefore this contradiction could come off as a result of more actual data than available during the times of the Chami et al study. Further research is therefore warranted into this effect.

(9)

9

remittances in combination with financial development practices do have a higher joint effect on growth than individually. Furthermore, they show that past a certain threshold, these practices will have the opposite effect: they will hamper growth due to an over-expansion of the financial system.

Aggarwal, Demirguc-Kunt and Martinez Peria (2011) address the connection between remittances and financial development. They stress the importance of financial development in the context of growth-enhancing and poverty-reducing effects, and reach the conclusion that there is a positive relationship between remittances and financial sector development. To be more precise, remittances are positively correlated to bank deposit and credit activities. It is notable that Aggarwal et al find this result to be robust to different estimation techniques whilst at the same time their methods accounted for endogeneity biases. These findings support the results of Adams and Page (2005) and loosely align with the conclusions of Meyer and Shera (2016).

(10)

10

consensus has formed in certain areas of remittance research, such as the worldwide poverty-reducing effects.

The financial institution development topic is also tackled by Anzoategui, Demigurc-Kunt and Martinez Peria (2013), in their El Salvador World Bank study. As opposed to some of the previous conclusions analyzed within this proposal, the authors did not find a completely robust relationship between remittances and financial development. The researchers could not find a significant effect of remittances on demand for credit from formal financial institutions. The authors of the study do state that remittances aid in the overall financial inclusion process of a developing country, specifically through the use of deposit accounts by remittance-receivers. As a concluding piece to their research, Anzoategui et al theorize that remittances reduce the need for external formal institution financing while at the same time the demand for saving instruments (deposit & saving accounts) is increasing.

(11)

11

An IMF study conducted by Giuliano and Ruiz-Arranz (2005) posits that remittances drive economic growth by acting as an investment channel in the case of developing countries where the financial sector does not meet the credit needs of the population. The researchers show that remittances actually help relax the credit constraints of the poor and improve capital allocation. The key characteristic of their results is that remittances act in such a manner only in the case of less financially developed countries. The authors are quite careful in their assessment and make it clear that their analysis could be subject to the omitted variable problem, whilst also mentioning the lack of controlling activities for moral hazard and its effect and implications for the remittance flows. For higher financially developed countries, Giuliano and Ruiz-Arranz bring forth suggestive evidence that remittances are actually more likely to discourage labor supply. This is among the first macro-economic studies that actually provide evidence on how remittances and financial development impact growth.

(12)

12

openness impact the output growth volatility. The former may increase output growth volatility, whilst the latter, surprisingly, doesn’t appear to have a significant effect in this study. Remittances are the 3rd_{type of flow analyzed and the prevalent result is that increasing}

remittance levels decrease the output growth volatility, thus leading to positive effects on economic growth as a whole.

Opperman and Adjasi (2018) investigate remittance volatility and financial sector development in sub-Saharan African countries. They find that, for countries in the researched region, remittances act as more of a substitute for a banking system and provide evidence that remittance volatility has negative effects on the development of banking systems. Moreover, it is determined that remittance volatility leads to extreme transfer costs imposed by banks and negative effects on the depth of the financial sector. The authors find similar results to Giuliano and Ruiz-Arranz’s study from 2009, in the sense that remittances in countries with poorly developed banking sectors act as an alternative means to finance investments and relieve constraints on liquidity.

(13)

13

Research method

In order to conduct this time series analysis over the period of interest 2001-2013 for 20 selected countries, multiple regressions and models are constructed, such as the fixed effects model, the random effects model and an Ordinary Least Squares (OLS) regression, similar to the methodology proposed by Meyer and Shera (2016). Since this study aims to investigate the countries with the highest remittance to GDP inflows, the selection procedure consisted of sorting through the World Bank Global Financial Database for the economies with the largest median during the established research period.

Regarding estimation methods, multiple options were explored in order to best fit the availability of data for variables of interest. The fixed effects technique is employed in order to determine the type of impact variables which change over time. As proposed by the previously mentioned authors, the fixed effects model fits this type of study since these sort of analyses are conducted to identify the factors of change within an entity. Moreover, this method has a unique characteristic in dealing with the potential endogeneity issues of a dataset.

The random effects estimation model is employed in order to check whether the differences between countries are random and uncorrelated to the independent variables in the model. Furthermore, this method also includes the random effects independent variables in modelling the time constraints.

(14)

14

In order to assess the collected information to completion, multiple separate research approaches are included within this methodological section. Three research models are constructed in order to determine whether remittance do indeed stimulate economic growth.

The three models for econometric estimations proposed within this analysis are constructed as follows:

I. 𝐺𝐷𝑃𝐺𝑅𝑂𝑊𝑇𝐻 = β0+ β1𝑅𝐸𝑀 + β2𝐹𝐷𝐼 + β3𝐺𝐹𝐶 + β4𝐸𝑋 + β5𝐺𝑂𝑉 + β6𝐷𝑐𝑟𝑖𝑠𝑖𝑠 + 𝜀 II. 𝐺𝐷𝑃𝐺𝑅𝑂𝑊𝑇𝐻 = β0+ β1𝑅𝐸𝑀 + β2𝐺𝐹𝐶 + β3𝐸𝐷𝑈 + β4𝐺𝑂𝑉 + β5𝐷𝑐𝑟𝑖𝑠𝑖𝑠 + 𝜀

III. 𝐺𝐷𝑃𝐺𝑅𝑂𝑊𝑇𝐻 = β0+ β1𝑅𝐸𝑀 + β2𝐺𝐹𝐶 + β3𝐺𝑂𝑉 + β4𝑀𝐼𝐺 + 𝜀

In order to avoid confusion, during the discussion part of this paper the above-mentioned models shall be referred to as “Model I”; “Model II” and “Model III”.

The dependent variable is defined as Gross Domestic Product growth (log of GDP/capita) and was collected from the World bank and OECD National Accounts data files. Gross-Domestic-Product is defined within this dataset as “the sum ofgross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products”. It is present across all estimations and follows to establish the relationship with all other variables.

(15)

15

The Global Financial Development database provides the following definition for remittances: “Remittances are classified as current private transfers from migrant workers resident in the host country for more than a year, irrespective of their immigration status, to recipients in their country of origin.”

FDI is a measure of the figures of financial direct investment across the studied countries, also related to GDP. This measure was collected using a dataset periodically updated by the International Monetary Fund containing also data from the World Bank. Foreign direct investments are defined as net inflows of investment to acquire a lasting management interest (10% of outstanding stock) in a company which operates in a different economy than that of the investor. Within the dataset provided, the IMF specifies that FDI is constructed as the sum of equity capital, the reinvestment of earnings, and other long-term and short-term capital as shown in the balance of payments.

GFC, or Gross Fixed Capital formation, relates to investment in public facilities, such as land improvements, public transport construction and investment and public institutions such as schools, hospitals, industrial buildings, etc. It is evident that this variable should carry statistical significance when explored in the context of economic (gross domestic product) growth. Furthermore, following 1993, a decision was made with regard to net acquisitions of valuables and their subsequent inclusion in the GFC dataset. Information regarding gross fixed capital formation was collected similarly from World Bank and OECD data files.

(16)

16

construct. The information regarding exports of goods and services related to GDP was obtained based on World Bank data.

The sixth variable included in the first regressive model is “Gov” and relates to the efficiency of the formal institutions for each of the countries included in the research. In order to determine whether formal institutions are efficient in these economies, a measure of government efficiency was included. This measure was collected from an institutional database created and updated by Daniel Kaufmann (“Worldwide Governance indicators”) and adjusted to fit the countries included in this dataset. Kaufmann defines his government effectiveness measure as an indicator of “perception of public quality services, civil services and the degree of independence from political pressures, the quality of policy formulation and implementation and the credibility of the government’s commitment to such policies.” Measures such as these are critical in the analysis of economic growth, as it is expected that a high level of institutional quality motivates individuals belonging to that specific economy to be more active and implicated in growth-enhancing activities.

(17)

17

Another variable included in this research is migration (MIG), the most volatile of all collected constructs. Migration data provided by the World Bank was collected from the United Nations Populations Division study: World Population Prospects: 2019 Revision. Within this dataset, net migration for an economy is determined based on the difference between the total number of immigrants and the annual number of emigrants, for both citizens and non-citizens of a country. It is expected that in a highly performing economy, the numbers of immigrants will exceed the number of emigrants, as individuals would prefer to reap the benefits from such an economy as opposed to emigrating to another country. The main limitation of this dataset is that the country-specific information is aggregated on five year estimates. As a result, the number of datapoints present for this construct are quite restricted, casting doubt over a conclusive interpretation of this variable.

Finally, a dummy variable was included in each of the above three models in order to control for the presence of the financial crisis of 2007-2008. Since the crisis period is indeed included in the timeline of this study, it seemed appropriate to construct this variable. Since according to many experts the effects of the crisis lasted through 2009 as well, the dummy variable takes the value 0 for the 2001-2006 and 2010-2013 time periods, whilst in the case of 2007-2009 dummy takes on a value of 1.

The Hausman test may be employed in order to determine which of the two main evaluation methods best fit the data (fixed effects or random effects model.) The hypotheses of the Hausman test assumes that individual effects are not correlated with the other model regressor. Should this hypotheses be rejected, the evidence would suffice to conclude that the random effects estimation is not an appropriate method for this dataset.

(18)

18

estimations were computed by taking into account potential autocorrelation and heteroskedasticity effects on the results of the regressions. Hence, robust standard errors were applied across all three estimation methods for the three proposed models.

(19)

19

Data breakdown

This study aims to investigate data from 20 developing countries with high inflows of remittances. Due to the availability of data, these countries could not be included across all estimations. Therefore, the first proposed model is the most comprehensive in terms of countries involved, as it is estimated across all high-remittance receiving economies. The 20 countries included in this study are: Albania, Armenia, Bangladesh, Bosnia and Herzegovina, Cabo Verde, Comoros, El Salvador, Gambia, Georgia, Guatemala, Honduras, Jordan, Lebanon, Lesotho, Moldova, Nepal, Philippines, Senegal, Tajikistan and Togo. As a result, the dataset contains five European economies, four Asian economies, two Middle-Eastern countries, three Central-American countries and no less than six African nations. It is evident that the dataset ranges across multiple geographical locations and regions, thus the effect of background noise in the data might be significant, as these countries face multiple political landscapes and much different economic conditions.

Whilst the first estimation model contains information from all the countries included in this study, the second estimated model differs, since data availability on education is restricted to only 11 out of the 20 countries that form the dataset. These 11 economies included in the education regression model are: Albania, Comoros, Gambia, Georgia, Honduras, Jordan, Moldova, Philippines, Senegal, Tajikistan and Togo.

(20)

20

data from 19 nations. Again, the latter model is expected to be rather weak, as migration only contains 3 datapoints per country during the period of interest (2002, 2007, 2012).

In the following, a thorough analysis of each estimation across all models will be conducted in order to determine whether remittance transfers do indeed have an effect on economic growth for the countries with the highest remittance to GDP inflows during the study period.

(21)

21

Results and Discussion

In order to best analyze and ponder the results of the estimations, the following section of this research paper is separated into three different areas, each describing the ouput of the three estimation methods proposed. The first analysis will relate to the OLS regression, whilst the following parts will discuss the fixed effect and random effect estimations. In the tables below, the output of the three proposed regressions for Model I are presented.

Linear Regression Estimation Model 1

rem_GDP 0.0250 (0.0236) fdi_GDP 0.155** (0.0619) gfc 0.0560 (0.0379) ex_GDP 0.0391** (0.0187) gov -0.00776 (0.0112) D_crisis -0.00790 (0.00604) Constant 0.0164* (0.00844) Observations 247 R-squared 0.105

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(22)

22

efficiency. FDI was found to have a positive effect on GDP growth at a 5% level of significance. According to the least squares estimation, a marginal unit increase in foreign direct investment to GDP will translate in a 0.1548983 increase in overall gross domestic product growth. This is an expected effect, since available research outlines the importance of FDI in the context of financial development, technology transfers and human capital development. The only other estimate of significance (5% level) may be identified for the variable encompassing the overall exports to GDP figure. From a reasonable point of view, it is anticipated that an economy heavily involved in exporting goods and services is also constantly facing high inflows of funds, which if reinvested may trigger increases in gross domestic product growth.

The conclusion regarding Model I’s OLS estimation is that whilst it contains some data of relevance for the current study, the overall significance for key variables of interest is too low to offer valuable information for inference regarding economic growth. Below, the result of the Hausman test on Model I is provided.

Hausman test Model 1

(23)

23

It is observable that the Hasuman test results in a chi2 probability of 0.1604. This result points towards failing to reject the null hypothesis of the above-specified test. As a result, according to the definition provided to the Hausman test in the Research Methods section of this paper, the model that best fits the data in this specific case is the random effects model. The estimations of the fixed effects model are also presented within this section in order to provide context and contrast and compare between the two (similar) estimation methods employed.

The first observed statistic is the result of the F-test, which, similarly to the least squares regression, points towards a rejection of the null hypothesis of equal variance between variables. At a 5% level, the null hypothesis of the F-test can be confidently rejected.

Fixed Effects Estimation Model 1

rem_GDP 0.0496 (0.0519) fdi_GDP 0.163 (0.0976) gfc 0.0806 (0.0713) ex_GDP 0.118** (0.0441) gov -0.0697** (0.0279) D_crisis -0.00728 (0.00614) Constant 0.00400 (0.0179) Observations 247 Number of ID 19 R-squared 0.130

(24)

24

Furthermore, under fixed effects estimations, the foreign direct investment looses it’s significance as well. Exports to GDP stays significant at a level of 5%, and keeps its coefficient sign, although the magnitude of the coefficient differs significantly from the least squares estimation coefficient. Moreover, within the fixed effect estimation, government efficiency is estimated as a significant variable. What is mostly interesting about this point is that the sign of the coefficient is negative, which seems to suggest that improved government efficiency bears a negative effect on gross domestic product growth. Whilst this might not follow a strong sense of reasoning, it may be that in certain cases governments of poor developing countries actually overspend and implicate too heavily in private sectors, leading to slight underperformance with regard to GDP growth.

The final estimation of Model I. is presented in the output table below.

Random Effects Estimation Model 1

rem_GDP 0.0220 (0.0281) fdi_GDP 0.170* (0.0947) gfc 0.0702 (0.0457) ex_GDP 0.0686*** (0.0220) gov -0.0255* (0.0147) D_crisis -0.00779 (0.00681) Constant 0.0101 (0.0121) Observations 247 Number of ID 19

(25)

25

Recommended by the Hausman test as the estimation to best fit the data pattern for the proposed model, the random effects estimation offers considerable insight into the variables analyzed previously within the least squares and fixed effects estimations. The Wald chi2 test result is statistically significant, thus the null hypothesis of one or more coefficients of interest being 0 is rejected at all significance levels.

Whilst remittance to gross domestic product still remains of no significance within this model, both the foreign direct investment and the governance effectiveness variables are apparently less significant as well. At a 10% significance level both of the estimated coefficients of these variables are still relevant and the general effect of the coefficients remains the same as previously interpreted (positive sign for FDI to GDP and negative sign for the Gov coefficient.) Finally, the exports to GDP variable becomes the strongest estimate of the model, being significant at even a 1% level.

Following the results of Model I it is evident that the variables aggregated within these estimations are not strongly significant, with the exception of well-known economic growth factors such as foreign direct investment and exports to GDP. Although it was expected that the equation with most observations and strongest dataset would provide some statistical insight into how remittances affect GDP growth, that does not appear to be the case.

(26)

26

The remittance to gross domestic product key independent variable of the study is shown, again, to not be significant at any levels. However, with the inclusion of the education term, it appears that gross fixed capital formation becomes significant at a level of 5%. Similarly, the education variable itself is now of importance to the study, as it is predicted that a marginal unit increase in education will have positive, significant effects on gross domestic product growth. It is surprising that the GFC variable is significant within this model for the first time in the current analysis. Model I would’ve been expected to contain informational content for GFC, due to the importance of infrastructure and public/private development to the overall level of development

within an economy.

As previously observed within Model I as well, the result of the F-test on the variance distributions of the independent variables points towards the rejection of the null hypothesis of equal variance, hence the results of the least squares estimations are interpretable, in the situations in which they are significant, naturally.

Linear Regression Estimation Model 2

(27)

27

rem_GDP -0.0664 (0.0440) gfc 0.159*** (0.0603) edu 0.000974** (0.000460) gov -0.00797 (0.0214) D_crisis -0.0105* (0.00548) Constant -0.0555 (0.0422) Observations 143 Number of ID 11 chi-square test 4.499 Prob > chi2 0.480

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

When assessing the results of the Hausman test (presented at the top of the page), it becomes evident that the preferred estimation model is, again, the random effects model. This occurs because the very high probability value returned points towards failing to reject the null hypothesis of the Hausman test. Therefore, we fail to reject the null and conclude that the differences in coefficients is not systematic.

(28)

28

rem_GDP -0.122*** (0.0346) gfc 0.189** (0.0840) edu 0.00164 (0.000929) gov -0.0135 (0.0208) D_crisis -0.00977 (0.00660) Constant -0.109 (0.0772) Observations 143 Number of ID 11 R-squared 0.097

(29)

29

alleviate their implication on the labor market, even leading to preference of short term financial stability as opposed to long-term investment and outlooks.

The implications of the fixed effect regression on global fixed capital also point towards the significance of this positive coefficient at a 5% significance level. Similarly to the description provided on the linear regression global fixed capital variable, it is expected that high investment levels in poor developing countries will have a positive overall effect on in-economy spending, hence stimulating economic growth. However, the Hausman test points towards the random effects model being preferred for this specific dataset, hence whilst the results of the fixed estimation have important informational content, it remains to be seen whether the random effect estimation of the second proposed model aligns with the fixed effects estimations.

The random effects model estimations presented at the beginning of the next page picture a promising connection with the fixed effects model. The first observable item on the estimation output is the value of the Wald chi2 test statistic. The result of this test is strongly significant at any significance level and seems to suggest that the coefficients tested within this estimation are in fact statistically different than 0.

(30)

30

Random Effects Estimation Model 2

rem_GDP -0.0664*** (0.0247) gfc 0.159* (0.0917) edu 0.000974** (0.000433) gov -0.00797 (0.0195) D_crisis -0.0105 (0.00666) Constant -0.0555 (0.0353) Observations 143 Number of ID 11

However, whilst it is true that the direction of the coefficient is the same, its magnitude is considerably lower, suggesting that remittances might indeed promote gross domestic product growth, should a change in receiver behavior change. The negative effect of remittances on economic performance does not come as a surprise. As mentioned previously, great insight is provided in the literature review into both the potential positive effect of remittances on growth and into situations where remittances act as a substitute for labor. The latter brings dire consequences in terms of worker productivity and hence, long-term growth is impacted negatively.

(31)

31

encountered in the fixed effect regression, the education variable becomes more significant in the actual context of the model. Similarly to the least squares regression, the education output is shown to have a positive, yet minimal effect on economic growth. This rather modest coefficient might be explained by the composition of the education data, which was mainly focused on primary school enrollments. It might be that in this situation, primary school enrollments might not have as huge of an effect on the growth of the gross domestic product, since even in most countries that build the analysis dataset it appears that this enrollment figure is rather high. As a point of further research, maybe some insight into higher levels of education (secondary school or higher education enrollments) could be provided so that the magnitude of these enrollments’ effect may be properly assessed and compared.

Model II has without a doubt provided compelling information with regard to the aim of the study. It is not only the first model to provide information regarding the overall effect of remittances in the economies included in the study, but it also provides the strongest statistically significant regression model.

(32)

32

The ordinary least squares regression is presented in the following table:

Linear regression estimation Model 3

rem_GDP 0.0447 (0.0521) gfc 0.0506 (0.0963) gov 0.0452 (0.0277) mig -5.00e-09 (3.26e-09) Constant 0.0147 (0.0229) Observations 57 R-squared 0.084

The failure to reject the null hypothesis of the F test, as well as the significantly modest value of the R squared statistic highlights the uncertainty with respect to interpreting the data of the OLS regression. Furthermore, the probability values available for the 4 independent variables tested are too high, hence pointing towards the rejection of coefficients having a significant effect on the economic growth dependent variable.

Since OLS could not offer any valuable insight into how migration shapes the prospects of economic growth in the same model as remittances to GDP, we turn to the Hausman test in order to determine whether the fixed effects or the random effects estimation methods could be more suitable in shaping an interpretable relationship.

Introduced on the beginning of the next page, the results of the Hausman test seem to still indicates that the random effects model is most suitable for the limited number of

(33)

33

analysis of the probability value associated with the test. It is worth noting that out of the 3 Hausman tests conducted on the 3 estimation models, this specific test resulted in the highest significance level.

rem_GDP 0.0213 (0.0674) gfc 0.0798 (0.0734) gov 0.0369 (0.0304) mig -6.93e-09 (8.12e-09) Constant 0.0144 (0.0197) Observations 57 Number of ID 19 chi-square test 5.448 Prob > chi2 0.142

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

The fixed effects estimation in the case of the last proposed model provides compelling information regarding a couple of the dimensions included, out of which the most important would clearly be the newly introduced migration variable. The result of the F test shows that the output of the fixed effects estimation is significant at a level of 5%.

(34)

34

The migration coefficient is also significant at the same significance level as the GFC coefficient. Naturally, migration has a role to play in terms of enhancing growth within an economy, as less developed countries are expected to have stronger figures of migrant outflows than migrant inflows. As more and more individuals choose to emigrate from their home economies, the labour supply of these nations is expected to become more and more restricted, impacting the overall development in the economy.

rem_GDP -0.132 (0.0783) gfc 0.259** (0.102) gov -0.0196 (0.0676) mig -2.98e-08** (1.08e-08) Constant 0.0136 (0.0318) Observations 57 Number of ID 19 R-squared 0.179

(35)

35

To conclude this section, a final analysis of the random effects estimations on Model III are introduced into this discussion. Again, as in the previous three occasions, the Hasuman test indicated the random effects regression to actually best fit the proposed model. Hence, the results of the last estimation table output should be of relevance in the concluding part of the Results & Discussion section.

Random Effects Estimations Model 3

rem_GDP 0.0213 (0.0448) gfc 0.0798 (0.0946) gov 0.0369 (0.0301) mig -6.93e-09** (3.00e-09) Constant 0.0144 (0.0239) Observations 57 Number of ID 19

(36)

36

Conclusion

This paper set out to investigate whether economies receiving high inflows of remittances related to their gross domestic product distinguish a significant impact of these financial flows on potential economic growth. After cautiously selecting the 20 countries with the highest median of remittance to GDP during the study period of 2001-2013, three estimation models were built.

In total, no less than 260 individual yearly observations were collected with respect to the ratio of remittances to gross domestic product. Next to the remittance variable, a number of seven other commonly-used economic growth measure variables were employed. Only a handful of these variables were considered significant across the different models to confidently interpret as significant. Among these, we may distinguish between: foreign direct investment, gross fixed capital formation, exports related to gross domestic product, education and the measure of institutional quality.

Three different estimation methods were employed as part of this study: the common practice ordinary least squares method, the fixed effects method and the random effect method. Additionally, the use of the Hausman test was employed, as well as the application of F distribution and Wald chi2 tests in order to validate and select the optimal estimation methods.

(37)

37

The 20 countries included in the study were specifically selected based on the highest median values of remittances from all the nations included in the World Bank Global Financial Development database during the period of interest. The underlying line of reasoning for this was based on the notion (briefly tackled in other studies) that high inflows of remittances would in fact motivate the receiving individuals to invest or save these funds or use them as means to accelerate their productivity. However, the situation encountered within this dataset appears to support the results of an entire other line of thought on remittances, more specifically: Individuals receiving remittances in less developed economies will use the funds to either alleviate their labor supply participation or for short-term consumption in detriment to long-term growth investment.

In an attempt to find a sound reasoning for this effect, variables such as education, gross fixed capital formation and institutional quality were included in these estimations. Above all, gross fixed capital formation is found to be strongly significant and positively related to economic growth. As such, it may be that the lack of public and private infrastructure actually forces individuals receiving remittance flows to spend this additional income in order to compensate for this absence.

Moreover, education may also be a key reason for remittance inflows not contributing to overall economic growth. Education was shown to be positively correlated with GDP growth in Model II, hence it may be the case that poorly educated individuals within this 20 country dataset are spending their remittance flows expeditiously, providing an explanation for the adverse effects of these financial flows.

(38)

38

In terms of limitations encountered during the development of this study the main issues concerned data collection and data availability. Specifically, multiple variables were constructed on restricted sets of available data. The collection of migration data comes to mind, as this dataset contained observations adjusted for 5 years periods, rendering the results of the estimations for Model III rather hard to interpret, in the few cases of statistical significance. Another variable of issue was education, since specific data on primary school enrollments wasn’t available for a number of countries included in the remittance to GDP dataset, foreign direct investment to GDP dataset and gross fixed capital formation dataset.

Further research should take into account the data restrictions and limitations. As it currently stands, the conclusion this paper draws is based mainly on a dataset of 143 observations on remittances. For future topical studies, a larger dataset is strongly recommended. Furthermore, a more comprehensive research design may be outlined, in order to attempt and group multiple variables of significance to GDP growth within one single estimation. Furthermore, a regional analysis may prove to be more effective, as geographical and regional differences between European, African, South American and Asian economies are quite substantial. Lastly, since this paper aimed to investigate the largest inflows of remittances to less developed countries, maybe some research is more warranted into the frequency of these remittance flows reported to their overall value.

(39)

39

References

Adams, R. (2010). Evaluating the Economic Impact of International Remittances on Developing Countries Using Household Surveys: A Literature Review. Journal of Development studies, 47(6), 809-828.

Adams, R. H. JR., & Page, J. (2005). Do international migration and remittances reduce poverty in developing countries? World Development, 33(10), 1645-1669.

Aggarwal, R., Demirguc-Kunt, A., & Martinez Peria, M. S. (2011). Do Remittances Promote Financial Development?. Journal of Development Economics, 96(2), p. 255-264.

Anzoategui, D., Demirguc-Kunt, A., & Martinez-Peria, M. S. (2014). Remittances and financial inclusion : Evidence from El Salvador. World Development, 54, 338-349.

Bugamelli, M., & Paterno, F. (2011). Output growth volatility and remittances, Economica, 78(311), 480-500.

Catrinescu, N., Leon-Ledesma, M., Piracha, M. and Quillin, B. (2009). Remittances, Institutions and Economic Growth. World Development, 37, 81-92.

Chami, R., Fullenkamp, C., & Jahjah, S. (2005). Are immigrant remittance flows a source of capital for development?. IMF Staff papers, 52(1), 55-81.

Giuliano, P., & Ruiz-Arranz, M. (2009). Remittances, financial development, and growth. Journal of Development Economics, 90(1), 144-152.

(40)

40

Meyer, D., & Shera, A. (2016). The impact of remittances on economic growth: An econometric model. Economia, 18(2), 147-155.

Opperman P., Adjasi, C.K.D. (2019). Remittance volatility and financial sector development in sub-Saharan African countries. Journal of Policy Modeling, 41, 336-351.

Siddique, A., Selvanathan, E.A., Selvanathan, S. (2012). Remittances and Economic Growth: Empirical Evidence from Bangladesh, India and Sri Lanka. Journal of Development Studies, 48:8, 1045-1062.

World Bank. (2018). Global Financial Development. Retrieved from

https://datacatalog.worldbank.org/dataset/global-financial-development.

World Bank. (2019). Migration and Remittances - Recent Developments and Outlook. Retrieved from https://migrationdataportal.org/tool/migration-and-remittances-recent-developments-and-outlook.

(41)

41

Appendix A. Remittance Data Worldwide

Figure 1. Top Remittance Recipients in 2018

(42)

42

(43)

43

Figure 3. Remittance Flows to Low- and Middle-Income Countries, 1990–2019

(44)

44

Figure 4.Top Remittance Recipients in the East Asia and Pacific Region

(45)

45

Appendix B. Dataset Specifics

Average Growth (GDP)

Average Remittance (GDP)

DO CONSTANT INFLOWS OF HIGH REMITTANCE AMOUNTS STIMULATE LONG-TERM ECONOMIC GROWTH IN LESS DEVELOPED ECONOMIES?