Impact of Remittance on Enrollment and Health Care: The Case of Bangladesh

World Review of Business Research Vol. 8. No. 2. June 208 Issue. Pp. 56 66 Impact of Remittance on Enrollment and Health Care: The Case of Bangladesh Foqoruddin Al Kabir, Farhan Khan 2 and Sakib B. Amin 3 International remittance has become vital issue in this global era and for developing countries remittance is an imperative source of income. Impact of remittance on education enrollment and health care are the two important aspects in emerging countries like Bangladesh. However, to the best of our knowledge no studies have been conducted to explore the relationship between remittance inflow with education enrollment and heath care improvement in Bangladesh. Therefore, the core objective of the paper is to analyze the role of remittance in improving our concerned variables with time series data covering from 980-205. Johansen s cointegration test reveals that all of our variables are cointegrated and Granger causality test results reveal a unidirectional causality is running from remittance inflow to education enrollment and bidirectional causality is running between remittance inflow and health care improvement in the long run. Vector Error Correction Model (VECM) confirms that in the short run, there are no causalities. Our results might help policy makers to formulate proper combinations of policies for upcoming development. Field of Research: Economics. Introduction Remittance plays a vital role in developing countries like Bangladesh. It has wide impact on multi various perspectives like development, GDP growth, and employment etc. (ILO 204). From the socio structural perspectives, it is very important to test the impact of remittance on education and health care. This can be an interesting paper to test the impact of remittance on primary education enrolment and health care in Bangladesh which can influence further policy making in this ground. According to Shirin (206), Bangladesh is one of the largest remittance recipient countries through the export of its labor forces mainly to the Middle East and the Southeast Asian countries since the early 970s. Job openings in the Middle-east and the establishment of an overseas recruitment system in the mid-970s, pave the way open for Bangladeshi workers to go abroad for employment. Therefore, it is very obvious that those labor force send the remittance to back home to their family and they spend those money in education and health care. So, it is interesting that we may find some causal relationship between remittance and education and health care. BS Graduate, Department of Economics School of Business and Economics, North South University, Bangladesh, Email:foqoruddin.al@northsouth.edu 2 BS Graduate, Department of Economics, School of Business and Economics, North South University, Bangladesh, Email: farhan.khan008@northsouth.edu 3 Assistant Professor, Department of Economics, School of Business and Economics, North South University, Bangladesh, Email: sakib.amin@northsouth.edu 56

However, very few studies have been done in this issue as we find very limited studies. From the study done by Bureau of Manpower Employment and Training, it can be known that a major share of remittances (36 %) is used to meet consumption which includes education, health care and food and the study also presented a chart where it shows that 2.75% of the remittances are used on the purpose of child education. This is a good indicator that there might be an existence of causal Relationship between remittance flow with education and health care. In another study by Raihan et al. (2009), authors showed an analysis of the impact of remittance on the education of the recipient families, they found that about at micro level 9 % of the total sampled households of 2005 HIES are receiving remittances, and it s clearly seen that the migrants households spend more on education than the non-migrants. Another study was done by International Organization of Migration, Bangladesh (203), they also concluded that remittances might have a positive impact on the level of education in the migrant households. Likewise, it is also a matter of concern that if we can find any causal relationship between remittance flow and healthcare improvement in Bangladesh. Although few studies are available to examine the link between international remittances and household expenditure pattern but these studies are not focused to health expenditure and treatment where the samples were random and purposive. Here in this case, we get an opportunity to examine the causal relationship between remittance and healthcare improvements as well. The aim of this paper is to empirically investigate the causal relationship between remittance and education enrollment and improvement of health care in the context of Bangladesh. The following questions are particularly addressed in our paper: What are the short run long run causalities among our variables? The rest of the paper has the following structure. Section two discusses the literature review. Section three widely presents the methodology and data that have been used in this paper. Section four presents the results and related discussions and section five in conclusion is stated. 2. Literature Review For about two decades remittance has been contributing around 35% of export earnings. Moreover, it is greater than foreign aid and thus helps in lessening dependence on foreign aid. Remittance gets momentum in recent time in Bangladesh and is the second largest sector of foreign exchange earnings after the garments sector. If cost of imported raw materials is deducted from the foreign exchange earning of the garments sector, remittance becomes the single largest sector of foreign exchange earnings. Remittance earning is increasing day by day but at a lower rate than the increase in emigration from Bangladesh due to the increasing share of unskilled or semi-skilled labors than the professionals in international immigration. The share of remittance in GNI (Gross National Income) is increasing day by day. It affects almost all the macro-economic indicators of a country positively. Though there are also negative sides of remittance earning e.g. brain drain, its overall contribution to Bangladesh 57

economy is very much effective. Appropriate and timely government policies and initiatives can boost up the amount of remittance and can rectify the problems related to it (Ali 204).Remittance has created a new dimension in the economic development of Bangladesh. We have to properly unlock the potentialities of remittances and utilize improperly to make it an indispensable tool of the economic development of Bangladesh. (Chatterjeeet al.200). Several studies have been undertaken about educational outcomes of family members and how they are affected by the remittances and migration experience. One stream of literature has shown that there is an increase of educational outcomes of family members due to the lift of the liquidity constraints. By relaxing the household s liquidity constraints, remittances allow an investment in education. An opposite stream, while agreeing for the potential for remittance transfers to alleviate credit constraints and thereby increase educational attainment of children, argues that migration of a family member, i.e. absence of the parent, may have negative effects on child schooling. Other literatures have looked at other aspects on the way the remittances impact the child schooling, including on incentives or disincentives provided by migration or remittances on the education of the family members left behind (Maraet al. 202). According to Tchantchaneet al. (203), a better understanding of the impact of remittances and expenditure on education on economic growth could help policy makers design strategic policies that could redirect remittances to areas in such a way as to maximize the positive impact of remittance flows for economic development of the country (Mara et al. 202). A number of researchers have analyzed the impact of remittance on economic growth rate with the help of macroeconomic variables like investment and human capital formation as well as physical capital formation (Burgess & Haksar 2005; Chamiet al.2005; Iqbal & Sattar 2005; Yang 2005; Jongwanich 2007; Ahortor & Adenutsi 2009; Mundaca 2009 and Matano & Ramos 203). Iqbal & Sattar (2005) established a strongly positive relationship between investment and growth and conclude that a higher the rate of investment, higher would be the rate of growth. Within this body or research Borraz (2005) finds that attainment in Mexico improves by 0.8 years on average for households that receive remittances, but that the effect is only statistically significant for households in small rural villages and in which the head of household has a low level of education. This finding suggests less of an impact than Hanson &Woodruff s (2003) research in Mexico which uses the same IV strategy to instrument for migration and determines a 0.73 to 0.89 additional years of schooling for girls (effect on boys were smaller and inconclusive) regardless of village size, though Borraz (2005) comments in his paper that the difference may stem from the methodology used that allowed him to identify with greater specificity the village size not possible in Hanson and Woodruff s study. After the wide discussion of different literatures, we can set our hypothesis relevant to our research question. We have two sets of hypothesis for both short run and long run analysis. The first set of hypothesis as follows, H : remittance does not cause education and H 2 : education does not cause remittance. The second set of null hypothesis as 58

follows, H : remittance does not cause health care and H 2 : health care does not cause remittance. 3. Methodology and Data Set To check the stationary of the variables, existence of unit root has to be tested. Macroeconomic and financial data are well known because of their non-stationary. There are several ways to find out the existence of unit root of the variables. For example, Augmented Dickey Fuller (ADF) and Phillips-Perron (PP) test are broadly employed. For our study, we have performed the (ADF) test to test the existence of unit root and found that all of the variables are non- stationary at levels and thus cannot be regressed without making them stationary. After the ADF test, we performed cointegration test to investigate possible linear combination of the variables that can be considered stationary. If cointegration is established, then we ran the causality test to check the possible direction of causality between the variables of interest. Non stationary data may lead to specious regression in the context of time series analysis unless there is at least one cointegration relationship. The Johansen technique is employed to test for cointegration. A unified framework of estimation and testing cointegration relations are provided in the context of Vector Autoregressive (VAR) error correction models. Here one has to estimate Unrestricted Vector of Autocorrelation of the form: x = L + θ x + u () t α+ θ x t + θ2 x t 2 + θ3 x t 3 + L+ θk x t k+ k t k t In the equation above, is the difference operator, x t ( n ) is a vector of nonstationary variables (in levels) and is the vector of random errors. The information on long run relationship is articulated by the matrix θ k the variables are not cointegrated, if the rank of θ k =0. Nevertheless if rank (usually represented by r) is equal to one, there exists one cointegrating vector and in conclusion if, < r < n there are multiple cointegrating vectors. Johansen & Juselius (990) have derived two tests for cointegration, which are trace test and the maximum Eigen value test. The trace statistic assesses the null hypothesis that there are at most r cointegrating vectors while the maximal eigen value test, estimates the null hypothesis that there are r exactly cointegrating vectors in t x If two variables are cointegrated, then there is at least one direction of causality. Granger-causality is one of the important issue that has been enormously studied in empirical finance of macroeconomics. Granger-causality is introduced by Granger (969, 980 & 988). Engle & Granger (987) asserted that the presence of nonstationary can lead to distorted conclusions in Granger-causality test. In this test, we can only infer long run relationship between non stationary time series when the variables are cointegrated. 59

If x and y are variables of interest, they by applying Granger-causality test we can determine whether past value of y augment the explanation of present values of x given that by information in past values of x itself. y does not Granger cause x if changes past values of y does not explain changes in x values at present. Likewise, we can probe whether x Granger causes y. There are four probable outcomes in the Granger causalitytest: a. neither variable Granger cause each other b. y causes xbut not otherwise c. x causes y but not otherwise d. both x and y Granger cause each other Following two sets of equation will be estimated: x + t = α0 + αx t + L + αl xt l + β yt + L + βl yt l ut (2) t = α0 + αyt + LL + αl yt l + βxt + LL + βl xt l vt (3) y + It is for all possible pairs of (x,y) series in the group. The stated F-statistics are the Wald statistics for the joint hypothesis β β = β = L = β 0. = 2 3 l = Engle & Granger (987) asserted that a vector error correction model (VECM) is an appropriate method to model the long-run as well as short-run dynamics among the cointegrated variables. Causality inferences in the multi-variate framework are made by estimating the parameters of the following VECM equations. Y = m n 0 p s α + βi Y t i + γ j Xt j + δ M + ζ N + θzt + εt i= j= k = l= X = a + m n 0 p s bi Y + c j Xt j + d M + e N + fzt i= j= k= l= z t- is the error-correction term which is the lagged residual series of the cointegrating vector. Deviations of the series from the long run equilibrium relation is measured by the error-correction term. For instance, from equation (4), the null hypothesis that X does not Granger-cause Y is rejected if the set of estimated coefficients on the lagged values of X is jointly significant. Furthermore, in those instances where X appears in the cointegrating relationship, the hypothesis is also supported if the coefficient of the lagged error-correction term is significant. Changes in an independent variable may be interpreted as representing the short run causal impact while the error-correction term provides the adjustment of Y and X toward their respective long run equilibrium. Thus, the VECM representation allows us to differentiate between the short- and long-run dynamic relationships. The Chi-Square test statistic is used to determine the short run causalities between pairs of variables in the model. The paper is based on annual data covering the period of 980-205. Data on total remittance inflow ($US), primary education enrolment and life expectancy at birth (as a proxy of health care)are taken from World Development Indicator (WDI). It should be + ξt (5) (4) 60

mentioned here that as Bangladesh got her independence in 97 and this research paper focusses over the period 980-205 for which 36 observations are available at most. Small sample size might be problematic in finding the long run relationship. We resorted to use of the E-Views 9.5 software for carrying out all econometric tests in our paper. All the econometrics results are available on request. 4. Results and Discussions Unit root tests are conducted to determine the order of integration of the data series. Optimal lag is chosen by Schwartz Information Criterion (SIC). Table shows the ADF statistics and corresponding critical values of all the variables in their level and first differenced forms. Variable Table: Augmented Dickey Fuller Unit Root Test for the Variables ADF Statistics (Only Constant) Panel : Levels ADF Statistics (Constant & Trend) Decision LNEDU 0.3362-3.934774 Non Stationary in constant but stationary in constant and trend LNHC -2.406242 0.24933 Non Stationary LNREM 0.407222.598605 Non Stationary Variable ADF Statistics (Only Constant) Panel 2: First Differences ADF Statistics (Constant & Trend) Decision LNEDU -3.976038-3.956340 Stationary LNHC -6.563288-5.49572 Stationary LNREM -5.04649-4.967504 Stationary Table 2: Mackinnon Critical Values for Rejection of Hypothesis of Unit Root Critical Levels First Differences Value No Trend With Trend No Trend With Trend % -3.632900-4.243644-3.639407-4.252879 5% -2.948404-3.544284-2.9525-3.548490 0% -2.62874-3.209699-2.64300-3.207094 Unit root tests have non-standard and non-normal asymptotic distribution. These distributions are extremely affected by the inclusion of deterministic terms such as constant, time trend etc. An extraneous regressor whose enclosure reduces the power of the test is called time trend. Yet if the true data generating process were trend stationary, failing to include a time trend also results in a decline in power of the test. Additionally, this loss of power from without a time trend when it should be present is more severe than the reduction in power associated with including a time trend when it 6

is extraneous. One of the main issues in unit root testing is lag length selection. Including a moderately long lag length and select the model by the usual t-test is one of the approach. When the t-statistics on lag p is insignificant at some stated critical value, the regression should be frequently assessed using a lag length (p-) until the lag is significantly different from zero. From the unit root test, it is clear that all the variables are found to be stationary at their first differences. From the table it is clear that the variables would yield spurious results unless the variables are cointegrated. The results, however, allow to proceed to the next stage of testing for cointegration. The Johansen cointegration test results indicate that our variables have cointegrating relationship. Maximum Eigen value test and the trace test (Table 3 and 4) both point out cointegrating relationships at 90%. After the Contegration test, we performed Granger Causality Test at lag 2. LNEDU, LNHC and LNREM Table 3: Johansen Test for Cointegration (Trace Test) Null Alternative Statistics 90% Critical Value Hypothesis Hypothesis (At Most One) 4.8634 39.75526 None (26.9928) (23.34234) At Most Two 5.425003 0.66627 Table 4: Johansen Test for Cointegration (Maximum Eigen Value Test) Null Hypothesis Alternative Hypothesis Statistics 90% Critical Value LNEDU, (At Most One) 87.87055 23.44085 LNHC and LNREM None (2.56780) (7.2340) At Most Two 5.425003 0.66627 From the Table 5 below, we can see two sets of null hypothesis. From the first set, we can reject the first null hypothesis but cannot reject the second null hypothesis as the probability value is above the desired level. Thus, we can say that there is a unidirectional causality running from remittance inflow to education enrollment in the long run but not vice versa. This answers our research question. It might happen that increased personal remittance inflow is being utilized in different fundamental household activities in rural or semi-rural or even in urban areas and education is one of such activities. Therefore, education enrollment has improved. On the other hand, from the second set of null hypothesis, we can reject both of them. Thus, we can assert that there is a bidirectional causality running between health care improvement and remittance inflow in the long run. It means both of the variables cause each other. This also answers our research question. Increase amount of remittance inflow can lead the households to take better health care if needed and better health care can lead to better health and workers with better health can be sent to abroad as human capital by which more remittance can be earned. After observing the long run causal relationship, we now move to investigate the short run causal relationship among the variables through VECM approach. (With the same 62

set of hypothesis) The results from the VECM approach are given in the Table 6. According to the VECM results, no causality was found between our variables of interest (this answers the research question as well). One of the possible reasons for this is that the effect of remittance utilization might be subject to time lag. Variable LNEDU LNREM LNHC LNREM Table 5: Granger Causality Test Results Null F- P- Conclusion Hypothesis Statistic Value Granger Causality Test Statistics between LNEDU and LNREM LNREM does not 3.36623 0.0987 LNREM Granger Causes Granger Cause LNEDU LNEDU LNEDU does not 0.0973 0.9805 Granger Cause LNREM Granger Causality Test Statistics between LNHC and LNREM LNREM does not 2.40835 0.077 Bidirectional causality Granger Cause LNHC between LNHC and LNREM LNHC does not Granger 4.02003 0.0288 Cause LNREM Table 6: VECM Test Results Variable Null-Hypothesis Chi-Square Statistic P- Value Conclusion Causality Test Statistics between LNEDU and LNREM LNEDU LNREM does not cause LNEDU.50036 0.2390 No Causality LNREM LNEDU does not cause LNRY 3.020072 0.2209 Causality Test Statistics between LNHC and LNREM LNHC LNREM does not cause LNHC 0.020428 0.9798 No Causality LNREM LNHC does not cause LNUR 0.04856 0.9798 5. Conclusion In this paper we have investigated the role of remittance on education enrollment and health care improvement for Bangladesh with the help of annual data covering from 980 to 205. For conducting our investigation, we have used different econometric techniques. We have found that there is a unidirectional causality running from remittance inflow to education enrollment the long run but not vice versa and a bidirectional causality is running between remittance inflow and health care improvement. However, no causality has been found in the short run. The results are new contribution in the existing body of knowledge. 63

As mentioned earlier that remittance has multidimensional effect in developing countries like Bangladesh, proper policy framework is needed so that remittance earned can be used effectively. As can be seen from our results that remittance influences education enrollment, a policy framework can be done for expansion of primary schools both public and private in the regions where number schools are not significant. On the other hand, more hospitals or health care centers should be built which can provide proper health care. Furthermore, promotional activities can be undertaken so that remittance users can gain knowledge how and where the money can be used as well. One of the limitations of this paper is lack of data. Results could have been more empirically underpinning if we could incorporate more control variables. This paper can be further extended by all types of education enrollment as we have only used primary school enrollment. On the other hand, a panel study of South Asian countries can be done to gain knowledge how remittance impact education enrollment and health care improvement. This can provide a wide range of options for policy makers to come up with general policy for the South Asian region. Reference Ahortor, CRK, & Adenutsi, DE 2009, The impact of remittances economic growth in small-open developing countries, Journal of Applied Science, Vol.9, No.8, Pp.3275-3286. Ali, MA 204, Socio-economic impact of foreign remittance in Bangladesh, Global Journal of Management and Business Research: C Finance, Vol.4, No.5, Pp.-20 Bangladesh Expatriate Workers and their Contribution to National Development (Profile of migration, remittance and impact on economy), Bureau of Manpower Employment and Training, Bangladesh, <http://www.bmet.org.bd/bmet/resources/static%20pdf%20and%20doc/publi cation/remittance%20and%20its%20impact.pdf> Borraz, Fernando, 2005, Assessing the impact of remittances on schooling: the Mexican Experience, Global Economy Journal,Vol.5, No., Pp.-30 Burgess, R, & Haksar, V 2005, Migration and foreign remittances in the Philippines: IMF WP/05/, International Monetary Fund, Washington D.C., United States, viewed 20 October, 207, <http://unpan.un.org/intradoc/groups/public/documents/apcity/unpan049760.pdf> Chami, R, Fullenkamp, C, & Jahjah, S 2005, Are immigrant remittance flows a source of capital for development? International Staff Papers, Vol.52, Pp. 55-8. Chowdhury, Hamid & Chatterjee 200, Remittance as a tool of economic development: Bangladesh perspective, Bangladesh Research Publication Journal, Vol.4, No.3, Pp. 286-296. Engle, RF & Granger, CWJ 987, Co-integration and error correction: Representation, estimation and testing, Econometrica, Vol. 55, Pp. 25-76. Granger, CWJ 969, Investigating causal relations by econometric models and crossspectral methods, Econometrica, Vol. 37, Pp. 424-438. Granger, CWJ 980, Testing for causality, a personal viewpoint, Journal of Economic Dynamics and Control, Vol. 2, Pp. 329-352. 64

Granger, CWJ 988, Some recent developments in a concept of causality, Journal of Econometrics, Vol. 39, Pp. 99-2. Hanson, G & Woodroff, C 2003, Emigration and education attainment in Mexico viewed 20 October 207, <http://www.childmigration.net/files/hanson_2003.pdf> ILO 204, Reinforcing ties: Enhancing contribution from Bangladeshi diaspora member, International Labour Organization, ILO, Country Office for Bangladesh viewed 3 December 204, <http://www.ilo.org/wcmsp5/groups/public/---asia/--- ro-bangkok/---ilo-dhaka/documents/publication/wcms_308803> Iqbal, Z., and Sattar, A, 2005, The contribution of worker's remittances to economic growth in Pakistan: PIDE research report 87, Islamabad, Pakistan: Pakistan Institute of Development Economics, <https://www.pide.org.pk/research/report87.pdf > Johansen, S &Juselius, K 990, Maximum likelihood estimation and inference on cointegration with Applications to the Demand for Money, Oxford Bulletin of Economics and Statistics, Vol. 52, Pp. 69-20. Jongwanich, J, 2007, Worker s remittances, economic growth and poverty in developing Asia and Pacific countries: UNESCAP Working Paper 07/0. Bangkok, Thailand: Economic and Social Commission for Asia and Pacific, viewed 20 October 207, <http://www.unescap.org/sites/default/files/wp-07-0.pdf.> Mara, I, Narazani, E, Saban, N, Stojilovska, A, Yusufi, I & Zuber 202, Analysis of literature on the effects of remittances on education and health of family members left behind, Skopje/Tirana, viewed 26 March 202, <http://analyticamk.org/~analytic/images/stories/files/analysis_of_literature_on_t he_effects_of_remittances_on_education_and_health_of_family_members_left_ behind.pdf> Matano, A & Ramos, R 203, Remittances and educational outcomes: Evidence for Moldova, AQR Research Group-IREA, Universitat de Barcelona, <http://www.ub.edu/searchproject/wp-content/uploads/203/05/search-wp- 3.0.pdf> Mundaca, BG 2009, Remittances, financial market development, and economic growth: The case of Latin America and the Caribbean, Review Development Economics, Vol.3, No.2, Pp. 288-303. Raihan, S, Khondker, BH, Sugiyarto, G&ShikhaJha2009, ADB Economic Working Paper Series No.89, viewed 20 October, 207, <https://www.adb.org/sites/default/files/publication/28402/economics-wp89.pdf> Shirin, AK 206, Remittance inflows and its contribution to the economic growth of Bangladesh, China Business Review, Vol. 62, Pp. 25-245. Tchantchane, A, Rodrigues, G & Fortes, PC 203, An empirical study of the impact of remittance educational expenditure and investment on growth in the Philippines, Applied Econometrics and International Development, Vol. 3-, Pp. 73-86. International Organization for Migration 203, Understanding the contribution of remittances at the macroeconomic and household levels and exploring how these transfers could be better leveraged for development in Bangladesh, Dhaka Bangladesh, <http://publications.iom.int/system/files/pdf/migrants_remittances.pdf> 65

Yang, D 2005, International migration, human capital, and entrepreneurship: Evidence from Philippine, Migrant's Exchange Rate Shocks: World Bank Policy Research Working Paper 3578, <http://fordschool.umich.edu/rsie/workingpapers/papers526-550/r53.pdf> 66