Do transfer costs matter for foreign remittances?

Ahmed, Junaid; Martínez-Zarzoso, Inmaculada Article Do transfer costs matter for foreign remittances? Economics: The Open-Access, Open-Assessment E-Journal Provided in Cooperation with: Kiel Institute for the World Economy – Leibniz Center for Research on Global Economic Challenges Suggested Citation: Ahmed, Junaid; Martínez-Zarzoso, Inmaculada (2016) : Do transfer costs matter for foreign remittances?, Economics: The Open-Access, Open-Assessment E-Journal, ISSN 1864-6042, Kiel Institute for the World Economy (IfW), Kiel, Vol. 10, Iss. 2016-4, pp. 1-36, https://doi.org/10.5018/economics-ejournal.ja.2016-4 This Version is available at: https://hdl.handle.net/10419/126572 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/3.0/ Received January 30, 2015 Published as Economics Discussion Paper February 20, 2015 Revised December 31, 2015 Accepted January 7, 2016 Published February 1, 2016 © Author(s) 2016. Licensed under the Creative Commons License - Attribution 3.0 Vol. 10, 2016-4 | February 01, 2016 | http://dx.doi.org/10.5018/economics-ejournal.ja.2016-4 Do Transfer Costs Matter for Foreign Remittances? A Gravity Model Approach Junaid Ahmed and Inmaculada Martínez-Zarzoso Abstract Using bilateral data on remittance flows to Pakistan for 23 major host countries, this is the first study that examines the effect of transaction costs on foreign remittances. The authors find that the effect of transaction costs on remittance flows is negative and significant; suggesting that a high cost will either refrain migrants from sending money back home or make them remit through informal channels. They also find that remittances are facilitated by the existence of migrant networks and improvements in home and host country financial services. Distance, which has been used in previous studies as an indicator of the cost of remitting, is found to be a poor proxy. JEL F22 F30 O11 Keywords Remittances; geographical distance; transaction cost; financial services; Pakistan Authors Junaid Ahmed, Department of Economics, Georg August University Göttingen, Germany and Faculty of Management and Social Sciences, Capital University of Science and Technology Islamabad, Pakistan, [email protected] Inmaculada Martínez-Zarzoso, Department of Economics, University of Göttingen, Germany, and Institute of International Economics, Universitat Jaume I, Castellón, Spain Citation Junaid Ahmed and Inmaculada Martínez-Zarzoso (2016). Do Transfer Costs Matter for Foreign Remittances? A Gravity Model Approach. Economics: The Open-Access, Open-Assessment E-Journal, 10 (2016-4): 1—36. http://dx.doi.org/10.5018/economics-ejournal.ja.2016-4 www.economics-ejournal.org 2 1 Introduction One of the principal factors that encourage migration across national boundaries is the difference in expected real earnings adjusted for migration cost (Borjas, 1999; Stark and Taylor, 1991). The costs incurred during the migration process are considered to increase with distance from the migrant sending to the migrant hosting country, and decrease as social networks in the hosting country grow (Ozden and Schiff, 2006). As migration is often thought to be a family decision (Borjas, 1999), the resulting remittances should be a central element of familial arrangements (Rapoport and Docquier, 2006). As a result, the physical distance between the migrant and the staying-behind household can affect remittance patterns (Rapoport and Docquier, 2006). There are different arguments related to the way in which distance influence remittances. Three of them indicate that remittances might decrease with distance. Firstly, remittances might be motivated by altruism, which could decrease if distance rises and associated contact falls. Secondly, migration to far-off countries might reinforce strategic behavior, since greater distance from the family may reduce the enforcement capability of any family arrangement agreed prior to migration. Thirdly, remittances may decrease with distance if the latter would be a good proxy for transfer costs (Lueth and Ruiz-Arranz, 2008; Frankel, 2011). At the other end of the spectrum, the loan repayment hypothesis supports the view that remittances may increase with distance (De Sousa and Duval, 2010). An increase in physical distance between migrant home and host countries can result in an increase in remittances in return for the high migration cost paid by the family (De Sousa and Duval, 2010). Clearly, these interpretations are conflicting. Some authors even argue that the cost of transferring money might be unrelated to geographic distance. Portes and Rey (2005) claim that financial assets are “weightless” and are therefore not subject to transportation costs. Remittances cost could also reflect technological developments and degree of competition in the financial-services industry. These factors reduce the cost of sending remittances through the formal financial sector (Freund and Spatafora, 2008) and are unrelated to distance. Furthermore, distance is time invariant and is therefore unable to pick up technological changes. Beck and Pería (2011) also illustrate that corridors with a larger number of migrants and higher competition exhibit consistently lower costs than others, indicating that migration networks could also influence the cost of remitting. www.economics-ejournal.org 3 Remittances sent to developing countries via the official channel have increased more than tenfold over the last decade. The amount reached $404 billion in 2013, growing by 3.5 percent compared with 2012 (World Bank, 2014a). This overwhelming growth in remittances is partly due to increase migrant stocks and rising remittances per immigrant. It may also be attributed to better recording of data as well as to a shift from informal to formal channels induced by falling cost of remitting money. However, the prevalence of informal transactions is still likely to be substantial. Freund and Spatafora (2008) argue that informal remittances amount to about 35–75 percent of recorded remittances to developing countries. This is due to lower transaction fees generally charged by informal channels.1 Compliance with regulations to counter terrorism financing and anti-money laundering could be a major cost factor putting upward pressure on prices, thus leading sizeable flows to underground channels (World Bank, 2014a). None of the abovementioned studies have investigated the cost of remitting as a factor influencing remittances flows. Hence, the effect of implementing policies in the receiving country to facilitate the transfers and lower the cost remains an empirical question. For this reason, the main aim of this paper is to provide an estimate of the effect of transaction costs on remittances and to evaluate the magnitude of this effect. At the conceptual level, the contribution is also the comparison of the effect of distance with the effect of remittance costs. At the empirical level, this is the first country-study of this kind for the South Asian region. Given that the region accounts for the highest share of world wide remittances, the results could be helpful in better understanding these remittance flows. More specifically, we estimate a gravity model using panel data for remittances from 23 sending countries2 to Pakistan over the period from 2001 to 2013. The model is augmented with a new proxy for cost of remitting that, to the best of our knowledge, has never been used in previous studies. Moreover, we also include migration networks in the analysis as an important factor explaining the _________________________ 1 According to Sander (2004), in most cases the average cost of remitting is between 3 and 5 percent when using informal channel, whereas Orozco (2003) suggest that the costs is lower than 2 percent of the amount transferred when informal channels, such as hawala or hundi, are used. 2 The countries considered include: Australia, Bahrain, Belgium, Canada, Denmark, France, Germany, Greece, Ireland, Italy, Japan, Kuwait, the Netherlands, Norway, Oman, Qatar, Saudi Arabia, Spain, Sweden, Switzerland, UAE, the UK, and the US. www.economics-ejournal.org 4 variation of remittances over time. The cost of remitting has been constructed using the real cost of sending money for a number of countries (including a representative country in each regional area) for which the data are available for a period of 4 years. We have estimated a model of the determinants of the cost of remitting to extrapolate this information to our whole sample. The main variables used to predict the cost of remitting are proxies for the financial development in both home and host countries and migrant stocks. We focus on Pakistan because it is among the top ten remittance receiving countries in the world and relies heavily on international transfers. The development potential of these transfers is therefore of great importance. The rest of the paper proceeds as follows. Section 2 presents Pakistan’s migration and remittance main features. Section 3 discusses remittances cost. Section 4 reviews the literature, focusing particularly on bilateral remittance determinants. Section 5 employs a gravity model framework to examine the main determinants of remittance flows using bilateral data. Results are presented in Section 6. Section 7 concludes and outlines a number of policy implications. 2 Overview of bilateral migration and remittances to Pakistan The first major wave of migration from Pakistan began in the 1970s when thousands of Pakistani workers left for the states of the Persian Gulf to participate in the development of the newly-rich oil economies. By 2013, about 5.7 million Pakistani immigrants were estimated to reside abroad, compared with 3.7 million in 2000 and 3.6 million in 19903 (United Nations, 2014). This shows that 54 percent of this growth in migrants stock took place during the period 2000–2013. Factors driving this wave of migration include economic slowdown, increasing poverty, rapid population growth and substantial wage differentials (Ministry of Finance, 2013; Irfan, 1999). Among the immigrants’ destination countries, the Middle East is the most popular destination region accounting for more than half of Pakistani migrants, _________________________ 3 This corresponds to around 2.2 percent of the country population in 2013 residing abroad compared to 2.9 percent in 2000 and 5.9 percent in 1990. www.economics-ejournal.org 5 followed by North America, Europe, and Asia Pacific (UN, 2014). Saudi Arabia and the United Arab Emirates (UAE) host the largest Pakistani-migrant communities, possibly due to geographical proximity and cultural affinity. Moreover, the Gulf region also has attracted a large proportion of immigrants due to the availability of mediumand low-skilled jobs (Arif, 2009). The United States (US), Canada, the United Kingdom (UK), Italy, and Spain are also countries with sizeable Pakistani overseas communities. At present, rapidly growing Southeast Asian economies, such as Malaysia and Singapore and also Australia, are attracting an increasing number of Pakistani workers (UN, 2014). The presence of such a significant number of immigrants has not only accelerated the integration of Pakistan into the world economy, but has also translated into a large flow of remittances back home. This flow plays an increasingly important role in easing difficulties facing the country’s economy in terms of foreign exchange, balance of payments, and economic growth (State Bank of Pakistan, 2012). For many developing countries facing a weak balance of payments situation such as Pakistan, remittances have emerged as a large source of foreign exchange earnings. The flows reached about $14 billion in 2013, compared with $1 billion in 2001 (see Table 1). Similarly, this increase in remittances has outpaced that of net ODA and FDI, which accounted for only $2.17 billion and $1.31 billion in 2013 respectively (WDI, 2014). Likewise, compared to FDI and foreign aid, remittances tend to be resilient and increase during periods of economic turmoil (Ahmed and MartinezZarzoso, 2013; Mughal and Makhlouf, 2011). Table 1 also indicates that Saudi Arabia, the USA, the UAE, and the UK represent Pakistan’s main remittance sending countries. The Middle East region (Gulf Cooperation States) accounts for more than 60 percent of overall remittances, which are mainly sent from Saudi Arabia and the UAE. Overall, the share from major remittances corridors has increased over the period 2001–2013. Remittances per immigrant, however, portray a somewhat different picture, with higher amounts coming from developed nations such as the USA, Australia, and the UK. www.economics-ejournal.org 6 Table 1. Remittance flows per immigrant by host country Host Countries Remittances by host country Host’ remittances over total remittances (percent) Remittances per immigrant 2001 2013 2001 2013 2001 2013 GCC 693.22 8462.78 63.83 60.79 251.08 1354.61 -Bahrain 23.87 282.83 2.20 2.03 367.23 2405.61 -Kuwait 123.39 619.00 11.36 4.45 1142.50 3453.78 -Qatar 13.38 321.25 1.23 2.31 243.27 3209.20 -Saudi Arabia 304.43 4104.73 28.03 29.48 178.87 1214.14 -UAE 190.04 2750.17 17.50 19.75 307.01 1423.61 -Oman 38.11 384.80 3.51 2.76 179.76 715.39 North America 139.71 2363.59 12.86 16.98 474.70 4763.44 -Canada 4.90 177.19 0.45 1.27 61.78 1127.75 -USA 134.81 2186.40 12.41 15.70 627.02 6448.11 Europe 112.87 2371.59 10.39 17.04 235.02 3013.14 -Belgium 1.10 3.34 0.10 0.02 275.00 256.92 -Denmark 3.83 25.03 0.35 0.18 900.33 1973.04 -France 2.22 36.26 0.20 0.26 222.00 1842.85 -Germany 9.20 83.18 0.85 0.60 262.86 2122.59 -Greece 0.00 11.18 0.00 0.08 0.00 455.42 -Ireland 0.20 90.07 0.02 0.65 66.67 11982.17 -Italy 0.55 35.74 0.05 0.26 27.50 499.80 -Netherlands 3.60 5.45 0.33 0.04 327.27 459.22 -Norway 5.74 37.84 0.53 0.27 410.00 1861.84 -Spain 0.06 53.44 0.01 0.38 4.00 709.87 -Switzerland 4.24 30.37 0.39 0.22 1060.00 6927.46 -Sweden 0.74 13.68 0.07 0.10 246.67 1248.40 -UK 81.39 1946.01 7.49 13.98 229.92 4087.02 Table 1 continued www.economics-ejournal.org 7 Table 1 continued Host Countries Remittances by host country Host’ remittances over total remittances (percent) Remittances per immigrant Asia Pacific 8.08 182.34 0.74 1.31 384.76 4133.76 -Japan 3.93 177.19 0.36 1.27 491.25 16681.42 -Australia 4.15 149.73 0.38 1.08 319.23 4471.15 Other 132.69 568.88 12.22 4.09 Total 1086.60 13921.70 305.52 1837.91 Notes: The figures in columns 1 and 2 are in current millions USD and in 5 and 6 in current USD. GCC denotes Gulf Cooperation Council states. Source: State Bank of Pakistan and author’s own calculations. 3 Cost of remitting to Pakistan Pakistani migrants use various channels to send money from the host country to their families back home. These include banks, money transfer operators such as Western Union and Money Gram, family members, and friends as well as the socalled “hawala or hundi.”4 Family, friends, and hundi are considered informal channels and are not recorded in the official statistics. In a study of remittances to Pakistan from Saudi Arabia, Arif (2009) points out that in 2009 about 38 percent of the remittances were transferred through the banking system, 28 percent through hundi, 17.9 through friends/relatives and 13.7 percent through migrants’ home visits. There is no difference in the reported cost of transfer money either through bank or hundi. However, the distance from the closest bank and the amount of time required for each transaction are the main factors pushing migrants and their families to use the hundi system. In another study, Amjad et al. (2013) mentioned that the time required to withdraw money from the nearest bank and the high transaction costs are the main barriers to using the banking channel. _________________________ 4 This is an informal method, which is comparatively cheaper than the formal transaction channel. The sender contacts a broker who acts as an intermediary and arranges the transfer. The sender remits a certain amount in Saudi Riyal and the broker contacts a counterpart in Pakistan, who makes the payment in Pakistani rupees to the family. Throughout the whole procedure, no money crosses the border, and no official records exist for this transaction. www.economics-ejournal.org 8 Therefore, the transaction costs of sending remittances and in particular the fees paid to intermediaries continue to be a significant concern for immigrants, development agencies, and other actors involved in the process. The World Bank has constructed a database that contains the cost of sending remittances to families back in the home country. The average cost for sending remittances from the major remittances corridor was 8.0 percent in 2011 and has fallen to 6.2 percent of the amount remitted in 2014. Figure 1 shows the cost of sending remittances with a significant heterogeneity across major remittances corridors. It reveals that it is significantly cheaper to send remittances to Pakistan from Saudi Arabia, UAE and the UK than from the other considered countries. Hence, the Middle East region was the least expensive corridor in 2013 with the cost being between 1.9 and 3.8 percent. Conversely, Singapore and Norway show the highest transfer costs. It is the most costly for a Pakistani resident to send money back home from Singapore with the cost being over 15 percent of the transfer. Sending money from Norway consistently costs more than twice, on average, than Figure 1. Average cost for sending remittances (as a share of funds sent) to Pakistan from major remittances corridors. Source: World Bank Remittances Prices Worldwide. All figures are percentages. 0 2 4 6 8 10 12 14 16 18 Total Cost (Percent) Remittances Corridor 2010 2011 2012 2013 2014 www.economics-ejournal.org 15 ln (𝑅𝑅𝑅𝑖𝑖𝑖 =𝛼0+ 𝛼1ln (𝐺𝐺𝐺𝑖𝑖) + 𝛼2ln (𝐺𝐺𝐺𝑖𝑖) + 𝛼3ln (𝑅𝑅𝑅𝑀𝑀𝐷𝐷𝑖𝑖𝑖) +𝛼4ln (𝐵𝐷𝐵𝑅𝐵𝑀𝑖𝑖𝑖 ) + 𝛼5ln (𝑅𝐷𝑀𝐷𝐷𝑀𝑀𝑀𝑖𝑖𝑖) + �𝛼𝑘𝑍𝑖𝑖𝑘𝑖 + 𝐾 𝐾=1 𝜇𝑖 + 𝜀𝑖𝑖𝑖 (4) 𝑅𝑅𝑅𝑀𝑀𝐷𝐷𝑖𝑖𝑖 is the transaction cost of sending remittances from the host country to the home country. Since some variables are in natural logs (except dummies, exchange rate, financial development and exchange rate variable), the estimated coefficients can be interpreted as elasticities. 5.2 Data and variable definitions We collected data on remittances from 23 host countries to Pakistan. These countries account for about 90 percent of remittance flows to Pakistan during the examined period (see Table 1). The selection of countries depends on the availability of bilateral remittances data. As factors explaining bilateral flows, we use both country-specific and bilateral variables taken from different sources. In particular, bilateral remittances in USD millions come from the SBP. The limitation of the reported data is that they most likely underestimate the volume of remittances sent through informal channels (hawala or hundi). Data on informal remittance flows are indeed patchy and do not permit the construction of time series with any degree of reliability. A few estimates of informal flows exist for specific points in time and for specific remittance corridors. For example, Arif (2009) points out that more than half of the remittances sent to Pakistan from the Persian Gulf come through informal channels. This notwithstanding, the study is concerned with the effect of transaction costs on the amount of formal remittances received, for which officially available data of acceptable quality are used. In what follows, we describe the variables that are considered important factors in influencing remittance flows. The GDP for the host country in billions of USD comes from WDI and is the most obvious factor that influences higher remittances to home countries (Vargas-Silva and Huang, 2006). The second explaining factor is the income level (measured in term of GDP) in the home www.economics-ejournal.org 16 country, which has an ambiguous effect on remittances depending on the prevailing motive to remit. The migrants stock in the destination country is also considered a crucial factor in determining remittance volumes (Freund and Spatafora, 2005). The data of Pakistani migrants stock in the host countries are taken from the Bureau of Immigration and Overseas Employment (BIOE, 2013) and from the Organisation for Economic Cooperation and Development (OECD, 2013). For North America, Europe, and the Asia-Pacific region, where labor receiving countries are located, we use the OECD database for two main reasons. Firstly, the BIOE dataset only contains legal outflow per year of workers looking for employment, thus excluding migratory movements for education, family union as well as illegal migrants (Amjad et al., 2012). Secondly, it does not track returning workers, which makes it impossible to accurately estimate the country’s migrant stock. We estimate the stock of migrants for Middle Eastern countries using the BIOE dataset assuming that the returning workers represent around 4 percent of the total migrant stock. This figure is based on Iqbal and Khan (1981), who computed the share of returning migrants to be 3.4 percent of the Pakistani migrants stock in the Middle East. Geographical distance is measured as the distance from Islamabad, Pakistan’s capital, to the corresponding capital of the remittances-sending country. The variable comes from the CEPII database. The transaction cost variable is estimated using data from the World Bank Remittances Prices Worldwide for major sending corridors to Pakistan (World Bank, 2014b). To obtain data for each destination and time period, we formulate two assumptions. First, we assume that transaction costs of sending remittances from the UAE to Pakistan are similar to that of the neighboring countries Oman, Kuwait and Qatar. Similarly, the remittances cost from the US is also used for Canada. Moreover, the cost of remittances from Norway to Pakistan has been used to proxy for the cost from Germany, France, Italy, Sweden, Denmark, Greece and Switzerland. Secondly, we assume that the costs of remittances are determined by migrants stock in the remittance-host country as well as the financial development in both the home and host countries. Data for cost of remitting are available only for the years 2010 to 2013. We use data from these four year to estimate the transaction cost for each sending country by regressing the cost of remitting on migrants stock in the remittance-host country and financial development in both www.economics-ejournal.org 17 home and host countries as well as extrapolating the resulting predicted values for the missing time period (2001–2009).7 The study uses real exchange rate computed as the nominal exchange rate times the relative price of the respective countries, which is also an important determinant of remittances (Dakila and Claveria, 2007).The bilateral exchange rate of the PKR is obtained from DataStream. The relationship between remittances and the exchange rate is a priori ambiguous. Remittances could decrease or increase with home country currency depreciation depending on the motive to remit. With respect to the financial sector development for home and host countries, we use domestic credit to the private sector as a percent of GDP. The data come from the WDI. Financial development is another important factor that makes remittances easier and cheaper, hence stimulating the flows via official channels (Freund and Spatafora, 2008; Singh et al., 2011). We therefore expect that the overall financial-sector development might lead to greater availability and lower costs for remittance services. As a proxy for institutional quality, we use a political stability indicator from the World Governance Indicators from the World Bank. The improved political situation may encourage remittances, since such an environment favors investment in the home country (Singh, et al., 2011). On the other hand, weak institution may also encourage remittances to compensate for the loss of purchasing power of the family back home. Fragile institutions in the home country are among the main reasons behind the decision to emigrate (Collier et al., 2011). It has also been argued that common language and religious ties tend to affect the choice of destination countries. For instance, larger shares of Pakistan’s migrants reside in the Middle East and in the countries with a similar official language. We expect a positive sign for these two variables. The variables bilateral remittances, GDP (host), GDP (home) and bilateral exchange rate are at constant 2005 prices. Table2 provides descriptive statistics for the above-mentioned variables. _________________________ 7The predictions were estimated with OLS regression with a linear trend. www.economics-ejournal.org 18 Table 2. Descriptive statistics Variables and definitions Source Mean S.D Min Max Dependent variable Bilateral remittances million (USD) State Bank of Pakistan 177.17 342.34 .024 1717.62 Gravity variables Host GDP in billion (USD) WDI 1.40e+14 2.77e+14 1.27e+07 1.45e+15 Home GDP in millions (USD) WDI 1.17e+11 1.83e+10 8.75e+10 1.47e+11 Geographical distance CEPII 5436.47 2639.01 1801.39 11392.8 Common language CEPII 0.21 0.41 0 1 Transaction costs (percent) World Bank Remittances Prices Worldwide and author’s calculations 15.37 2.37 9.77 19.70 Other control variables Exchange rate DataStream 0.14 0.59 .002 4.45 Domestic credit to private sector as percent of GDP (home) WDI 23.30 4.46 15.65 28.74 Domestic credit to private sector as percent of GDP (host) WDI 104.65 51.44 27.26 232.10 Migrants stock BIOE, OECD, UN-DESA 0.2 0.54 .003 3.38 Institutional variables Political stability (home) World Wide Governance Indicator, World Bank 0.16 0.07 0 .10 0.40 Note: All the variables are in levels. Period 2001–2013. www.economics-ejournal.org 19 5.3 Estimation issues A variety of empirical techniques are employed in the study. The model is first estimated using a pooled OLS as a benchmark with standard errors corrected for heteroskedasticity. However, given the panel nature of the dataset, the pooled OLS is only consistent when unobserved fixed effect and explanatory variables are uncorrelated (Wooldridge, 2002). In order to take into account the resulting unobserved heterogeneity, we also use a panel data approach, using fixed and random effects models. Restricted F-statistics, Breusch and Pagan (1980) LM and Hausman (1978) specification tests are used in order to choose between pooled OLS versus fixed effects, pooled OLS versus random effects, and fixed versus random effects models. To choose between fixed and random effects, the Hausman test was used, which indicates that the country fixed effects are correlated with the regressors, and therefore, both OLS and random effects yield biased results. The inclusion of country fixed effects in this panel study controls for sources of endogeneity related to unobservable heterogeneity that are country specific and time-invariant. The fixed effect estimator, however, does not provide a direct estimation of the coefficients of time invariant variables. One solution for this is to use the Mundlak approach (Mundlak, 1978) who proposed approximating the country specific effects as a function of the mean of time-variant variables. This is an alternative procedure to the fixed effects model, which includes averages of time-varying explanatory variables (Wooldridge, 2002), instead of using dummy variables or the within transformation and will be used in order to obtain estimates for the distance variable, which is time invariant. Finally, given that some explanatory variables might be endogenous (GDPs, the cost of remitting as well as the migrants stock) we use a procedure proposed by Hausman and Taylor (1981) and also suggested by Baltagi et al. (2003) to tackle endogeneity issues in a panel data framework This Hausman-Taylor approach uses the means of the exogenous time-variant variables as instruments for the endogenous variables (Baum, 2006, p.229). Finally, in order to check for the quality of our estimations, we carry out several post estimation tests. The calculation of bivariate correlations between the explanatory variables helps us to identify collinearity between the explanatory variables. Variables that are highly correlated are used separately or are dropped from the regression. To test for autocorrelation, the Wooldridge test is used (the www.economics-ejournal.org 20 null hypothesis is that there is no first order autocorrelation while the alternative hypothesis is that there is a presence of autocorrelation) the Breusch-Pagan test is used to test for heteroskedasticity. 6 Empirical Findings In this section, we discuss our main empirical results. The benchmark estimates presented in Table 3 provide results for the baseline model using several estimation methods. The first column provides fixed effects estimates, the second column presents results using the Mundlak approach, and finally the third column presents Hausman and Taylor estimates. In the first specification, the log of remittances is regressed on the GDPs of host and home countries, geographical distance, the bilateral exchange rates, and migrants stock. Concerning the effect of economic activity in the home country on remittances, we find that the GDP of the home country has a positive and statistically significant effect on remittances (Columns 1–3 in Table 3). This shows that Pakistani migrants send more remittances when the economic conditions improve at home, which supports the portfolio investment motive. This result is consistent with the findings in Kock and Sun (2011), Lueth and Arranz (2008), and Docquier et al. (2012). However, remittance flows to Pakistan do not seem to respond to the host country’s economic conditions. This is in contrast to the findings of Schiopu and Siegfried, (2006), Vargas-Silva and Huang (2006) and Kemegue et al. (2011) who argue that remittances are more responsive to the host country’s economic conditions than to the economic conditions of the home country. The results can be explained by considering the extent to which the migrant is integrated into the formal sector of the host economy. It could also be explained by the loan repayment hypothesis stating that remittances are fixed loan payments made by the emigrants to the households (Vargas-Silva and Huang, 2006). These reasons could also explain why the recent economic crunch has not adversely affected remittance flows to the country. www.economics-ejournal.org 21 Table 3. Baseline panel gravity model estimates (1) (2) (3) VARIABLES Fixed Effects Mundlak Approach Hausman and Taylor Approach GDP (host) -0.014 -0.014 -0.015 (0.015) (0.015) (0.026) GDP (home) 1.205** 1.205** 1.184*** (0.527) (0.532) (0.340) Migrants stock 1.534*** 1.534*** 1.542*** (0.367) (0.370) (0.167) Geographical distance -0.384 1.090 (1.017) (0.813) Common official language 1.006 -0.450 (0.745) (0.904) Bilateral exchange rate 0.472*** 0.472*** 0.451*** (0.056) (0.057) (0.120) Number of observation 299 299 299 R-squared 0.547 0.547 Hausman test (Fixed Vs Random effects) Prob>chi2 = 0.0404 Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenthesis. All the variables except dummies and the exchange rate are in natural logs. The endogenous variables in the Hausman and Taylor approach are GDP (host country), GDP (home country) and migrants stock. The geographical distance is not statistically significant in any of the estimated models (see Table 3 and Table 4).The mixed results in the previous literature for geographical distance indicates that distance is not always an important driver of remittance flows. The estimated results corroborate the graphical illustration in Figure indicating that the cost of transferring money to Pakistan is unrelated with geographical distance. Another possible interpretation of why distance is a poor proxy for remittance costs is that the cost of sending money from a developed to a developing country is significantly lower than the cost of remitting in the opposite www.economics-ejournal.org 22 direction, whereas distance is the same. Evidence shows that remittance cost is high in the same bilateral corridor depending on the direction of the flow (Ratha and Shaw, 2007). As a result, the cost of remitting money is more related to technological developments and increased competition in the financial-services than to geographical distance. In regard to the effects of migrants stock on remittances, our results expectedly show that remittances depend positively and significantly on migrants stock. This means that countries with an increasing size of migrants stock attract a higher volume of remittances (Freund and Spatafora, 2008). The results are robust and consistent with our expectations. Concerning the exchange rate variable, our findings indicate that it has a positive effect on remittances. This suggests that in case of appreciation of the home currency, migrants tend to send more money in foreign currency to insure the same amount of income in the domestic currency. Another possibility could be that migrants send more remittances in order to keep the same utility level of their family compared with their own personal utility level. Now, we turn to the extended estimated model that includes other important control variables that are likely to have an impact on remittance flows, namely, domestic credit to the private sector as a percent of GDP for host and home countries and political stability.8 The results for the augmented model are presented in Table 4. The inclusion of additional control variables does alter the magnitude and significance of GDP (home) in some of the estimated models. We also take into account the financial sector development (the driving factor of transfer cost) for both host and home country. As expected, remittances are positively and significantly related to financial sector development. The findings reveal that better financial development in the host and home countries turn into higher flows of remittances. In addition, financial improvement in the home country would enhance the availability of low cost remittance services that could then direct large amount of remittances through official channels (Freund and Spatafora, 2008; and Wahba, 1991). _________________________ 8The correlation matrix of the variables indicates that common religion and geographical distance are highly correlated. We dropped common religion as this might affect the direction and significance of the effect of other variables on the dependent variable. www.economics-ejournal.org 23 Table 4. Augmented gravity model (1) (2) (3) VARIABLES Fixed Effects Mundlak Approach Hausman and Taylor Approach GDP (host) 0.013 0.013 0.012 (0.010) (0.010) (0.023) GDP (home) 0.618 0.618 0.601* (0.623) (0.629) (0.341) Migrants stock 1.377*** 1.377*** 1.466*** (0.321) (0.324) (0.144) Geographical distance -0.426 -0.091 (0.950) (0.768) Common official language 1.108 0.063 (0.776) (0.838) Bilateral exchange rate 0.356*** 0.356*** 0.348*** (0.058) (0.059) (0.108) Credit to private sector (host) 0.013** 0.013** 0.012*** (0.006) (0.006) (0.002) Credit to private sector (home) 0.040*** 0.040*** 0.041*** (0.010) (0.010) (0.012) Political stability (home) -1.293** -1.293** -1.251*** (0.547) (0.553) (0.484) Observations 299 299 299 R-squared 0.647 0.647 Hausman test (Fixed Vs Random effects) Prob>chi2 = 0.0449 Note: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenthesis. All the variables except dummies and exchange rate are in natural logs. The endogenous variables in the Hausman and Taylor approach are GDP (host), GDP (home) and migrants stock. www.economics-ejournal.org 24 Countries with improved financial markets thus have more opportunities to attract remittances through formal channels and are thereby more likely to channel it into more productive uses. The coefficient of the political stability variable representing institutional quality in the home country is negative and significant, implying that an unstable political environment (associated with lower growth) may encourage larger amounts of remittances. This result supports the notion that the altruistic behavior of the migrant encourages sending more remittances when the earning prospects of the migrants home country income decreases, in order to assure the same level of satisfaction.9 Similarly, the money transfer could also increase by higher outflows of emigrants to other economically well-off destinations due to political turmoil at home. This stabilization role of remittances to compensate for the loss of purchasing power due to political instability indicates that remittances are used to hedge against political disorder. We also included initially political stability in the host countries as an additional regressor, however due to its high correlation with GPD it was dropped from the reported estimations.10 Finally, Table 5 reports the estimates of equation (4), which include transaction cost. Financial development is not included because it is highly correlated with the predicted transaction costs. Results in column (1) of Table 5 indicate that high transaction costs significantly reduce remittances. For instance, a one percent decrease in the transaction cost would yield about 1.6 percent increase in remittances flows. This seems to suggest that higher transfer costs deter transferring money back home. As discussed, variation in transfer costs has a large impact on remittances. In particular, a 20 percent reduction in the average cost of remitting will increase the average bilateral remittances (slightly below 200 USD, according to Table 2) by 32 percent (taking the coefficient of column 1 in Table 5, 32=1,6*20), which mean around 64 USD. This result is consistent with the two hypotheses stated in the paper, for instance, the notion that higher transfer costs discourages remitters or _________________________ 9 In unpredictable political situations, the cost of capital would increase and consequently, investors will look for more stable investment destinations. Therefore, political instability deters economic growth (Aisen and Veiga, 2013). 10 When excluding GDP host from the regression, the coefficient of political stability in the host country was negative but only statistically significant at the 10 percent level, with a magnitude of (-0.84). www.economics-ejournal.org 31 Demekas, D.G., Horváth, B., Ribakova, E. and Wu, Y. (2005). Foreign direct investment in Southeastern Europe: How (and how much) can policies help? IMF Working Paper No. 05/110. https://www.imf.org/external/pubs/cat/longres.aspx?sk=18247.0 Docquier, F., Rapoport, H. and Salomone, S. (2012). Remittances, migrants education and immigration policy: Theory and evidence from bilateral data. Regional Science and Urban Economics, 42(5), 817–828. http://www.sciencedirect.com/science/article/pii/S0166046211001219 Frankel, J. (2011). Are bilateral remittances countercyclical? Open Economies Review, 22(1), 1–16. http://link.springer.com/article/10.1007%2Fs11079-010-9184-y Feenstra, R. C., Markusen, J.R. and Rose, A.K. (2001). Using the gravity equation to differentiate among alternative theories of trade. Canadian Journal of Economics, 34(2), 430–447. http://onlinelibrary.wiley.com/doi/10.1111/0008-4085.00082/abstract Freund, C. and Spatafora, N. (2008). Remittances, transaction costs, and informality. Journal of Development Economics, 86(2), 356–366. http://www.sciencedirect.com/science/article/pii/S0304387807000818 Gibson, J., McKenzie, D.and Rohorua, H. (2006). How cost elastic are remittances? Estimates from Tongan migrants in New Zealand. University of Waikato, Department of Economics, Working Paper No. 20/6. http://researchcommons.waikato.ac.nz/handle/10289/1639 Hattari, R. and Rajan, R.S. (2008). Sources of FDI flows to developing Asia: The roles of distance and time zones. ADBI working paper series, No. 117. http://www.econstor.eu/bitstream/10419/53582/1/604626827.pdf Hausman, J.A. (1978). Specification tests in econometrics. Econometrica, 46(6), 1251– 1271. http://www.jstor.org/stable/1913827 Hausman, J.A. and Taylor, W.E. (1981). Panel data and unobservable individual effects. Econometrica 49(6), 1377–1398. http://www.jstor.org/stable/1911406 Iqbal, M. and Khan, M.F. (1981). Economic implications of the return flow of immigrants from the Middle East: A preliminary study. Pakistan Institute of Development Economics, Research Reports No. 132. http://opendocs.ids.ac.uk/opendocs/handle/123456789/2543#.VqCUg0bCoaA Irfan, M. (1999). Emigration from Pakistan – 1947–97. MPRA Paper No. 38623, University Library of Munich, Germany. https://mpra.ub.uni-muenchen.de/38623/ Karemera, D., Oguledo, V.I., and Davis, B. (2000). A gravity model analysis of international migration to North America. Applied Economics, 32(13), 1745–1755. http://www.tandfonline.com/doi/abs/10.1080/000368400421093 www.economics-ejournal.org 32 Kemegue, F., Owusu-Sekyere, E., and van Eyden, R. (2011). What drives remittance inflows to Sub-Saharan Africa? A dynamic panel approach. Ersa – economic research southern africa, Working Paper No. 262. http://www.econrsa.org/publications/working-papers/what-drives-remittance-inflowssub-saharan-africa-dynamic-panel-approach Kock, U. and Sun, Y. (2011). Remittances in Pakistan: Why have they gone up, and why aren't they coming down? International Monetary Fund, Working Paper No. 11/200. https://www.imf.org/external/pubs/ft/wp/2011/wp11200.pdf Lewer, J.J. and Van den Berg, H. (2008). A gravity model of immigration. Economics Letters, 99(1), 164–167. http://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1023&context=manageme ntfacpub Lueth, E. and Ruiz-Arranz, M. (2008). Determinants of bilateral remittance flows. The BE Journal of Macroeconomics, 8(1). http://www.degruyter.com/view/j/bejm.2008.8.1/bejm.2008.8.1.1568/bejm.2008.8.1.1 568.xml Mayda, A. M. (2010). International migration: A panel data analysis of the determinants of bilateral flows. Journal of Population Economics, 23(4), 1249–1274. http://www.jstor.org/stable/40925859 Ministry of Finance, (2013). Pakistan Economic Survey 2012–13.Government of Pakistan, Islamabad. http://www.finance.gov.pk/survey_1213.html Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica, 46(1), 69–85. https://www.econometricsociety.org/publications/econometrica/1978/01/01/poolingtime-series-and-cross-section-data Mughal, M.Y. and Makhlouf, F. (2011). Volatility of remittances to Pakistan: What do the data tell? Economics Bulletin, 31 (1), 605–612. http://www.accessecon.com/Pubs/EB/2011/Volume31/EB-11-V31-I1-P58.pdf OECD (2013). International migration database. OECD Publishing. https://stats.oecd.org/Index.aspx?DataSetCode=MIG Oak, C (2014).Country Viewport: Digital Money in Pakistan 2014, Shift Thought Ltd. http://digitalmoney.shiftthought.co.uk/sage/View_Viewports/Viewport_Pakistan_201 4/ Orozco, M. (2003) Worker remittances in the international scope. Inter-American Dialogue, Inter-American Development Bank Working Paper, commissioned by the Multilateral Investment Fund. http://www.iadb.org/wmsfiles/products/publications/documents/35076501.pdf www.economics-ejournal.org 33 Ozden, C., and Schiff, M.W. (Eds.). (2006). International migration, remittances, and the brain drain.World Bank Publications, Report No. 33988. http://documents.worldbank.org/curated/en/2006/01/6360661/international-migrationremittances-brain-drain Portes, R., and Rey, H. (2005). The determinants of cross-border equity flows. Journal of International Economics, 65(2), 269–296. http://www.sciencedirect.com/science/article/pii/S0022199604000716 Rapoport, H., and Docquier, F. (2006). The economics of migrants’ remittances. Handbook of the Economics of Giving, Altruism and Reciprocity, 2, 1135–1198. Ratha, D.and Shaw, W. (2007). South-South migration and remittances, World Bank, Washington DC. Schiopu, I. and Siegfried, N. (2006). Determinants of workers’ remittances: Evidence from the European neighbouring region. European Central Bank, Working Paper No. 688. https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp688.pdf?f31c4dfff2b68deec59ef90 384bc8e4f Singh, R.J., Haacker, M., Lee, K.W. and Le Goff, M. (2011). Determinants and macroeconomic impact of remittances in Sub-Saharan Africa. Journal of African Economies, 20(2), 312–340. http://mh.devgro.com/papers/singhhaackerleelegoff2011.pdf Sander, C. (2004). Capturing a market share? Migrant remittances and money transfers as a microfinance service in Sub-Saharan Africa. Small Enterprise Development, 15(1), 20–34. Stark, O., and Taylor, J.E. (1991). Migration incentives, migration types: The role of relative deprivation. The Economic Journal, 101(408), 1163–1178. http://www.jstor.org/stable/2234433 State Bank of Pakistan (2012). Annual performance review, 2011–12. http://www.sbp.org.pk/reports/annual/arFY13/Vol2/anul-index-eng.htm State Bank of Pakistan (2014). Country-wise workers’ remittances. http://www.sbp.org.pk/stats/survey/index.asp Vargas-Silva, C., and Huang, P. (2006). Macroeconomic determinants of workers' remittances: Host versus home country's economic conditions. Journal of International Trade and Economic Development, 15(1), 81–99. http://www.tandfonline.com/doi/pdf/10.1080/09638190500525779 United Nations (2014). Trends in international migrants stock, Department of Economic and Social Affairs. UNCTAD (2013). UNCTA Dstat. New York: United Nations. www.economics-ejournal.org 34 Wahba, S. (1991). What determines workers' remittances?.Finance and Development 28 (4), 41–44. Wooldridge, J.M. (2002). Econometric analysis of cross section and panel data. MIT press. https://mitpress.mit.edu/books/econometric-analysis-cross-section-and-paneldata WDI (2013).World Development Indicators. Online Database, World Bank. World Bank. (2014a). Migration and remittance: Recent developments and outlook. Migration and Development Brief 22. Migration and Remittances Unit, Development Prospects Group. Washington DC. https://siteresources.worldbank.org/INTPROSPECTS/Resources/3349341288990760745/MigrationandDevelopmentBrief22.pdf World Bank. (2014b). Remittance prices worldwide: Making markets more transparent. Washington DC. www.economics-ejournal.org 35 Appendix Table A.1. Correlations matrix 1 2 3 4 5 6 7 8 9 10 1 1 2 0.05 1 3 -0.23 0.13 1 4 0.73 0.00 -0.41 1 5 0.28 0.00 0.13 0.62 1 6 0.15 -0.12 -0.16 0.07 -0.10 1 7 0.45 0.24 -0.48 0.53 0.22 -0.03 1 8 -0.05 -0.41 -0.07 -0.00 -0.00 0.07 -0.08 1 9 -0.01 -0.32 -0.05 0.00 0.00 0.03 -0.07 -0.11 1 10 -0.76 0.00 0.65 -0.88 -0.31 -0.10 -0.73 0.00 0.00 1 Note: Number of observations: 299. 1. GDP (host). 2. GDP (home). 3. Migrants Stock. 4. Geographical distance. 5. Common language (official) 6. Bilateral exchange rate 7.Domestic credit to private sector (host) 8.Domestic credit to private sector (home) 9.Political stability (home) 10. Common religion. www.economics-ejournal.org 36 Table A.2. Gravity model for 2008-2013 with actual transaction cost VARIABLES (1) GDP (host) -0.0282 (0.0202) GDP (home) - - Migrants stock 0.752*** (0.205) Common language 1.159 (0.724) Transaction cost -0.199 (0.158) PRI_ dummy_2011 0.186** (0.078) Bilateral exchange rate 0.443*** (0.132) Political stability (home) - - Observations 116 Note: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenthesis. All the variables except dummies and exchange rate are in natural logs. Hausman and Taylor approach used. The endogenous variables are GDP (host) and GDP (home) and migrant stock. Please note: You are most sincerely encouraged to participate in the open assessment of this article. You can do so by either recommending the article or by posting your comments. Please go to: http://dx.doi.org/10.5018/economics-ejournal.ja.2016-4 The Editor © Author(s) 2016. Licensed under the Creative Commons Attribution 3.0.