THE EFFECT OF REMITTANCES ON POPULATION HEALTH: AN ANALYSIS OF REMITTANCES, HEALTH OUTCOMES, AND HEALTH EXPENDITURES

THE EFFECT OF REMITTANCES ON POPULATION HEALTH: AN ANALYSIS OF REMITTANCES, HEALTH OUTCOMES, AND HEALTH EXPENDITURES A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy By Andrea Thoumi, M.Sc. Washington, DC April 10, 2016

Copyright 2016 by Andrea Thoumi All Rights Reserved ii

THE EFFECT OF REMITTANCES ON POPULATION HEALTH: AN ANALYSIS OF REMITTANCES, HEALTH OUTCOMES, AND HEALTH EXPENDITURES Andrea Thoumi, M.Sc. Thesis Advisor: Andreas Kern, Ph.D. ABSTRACT While the search for alternative health financing mechanisms to fund national health systems is not a new international development challenge, fiscal pressures to identify alternative sources of revenue for health are growing. Today, low- and middleincome countries are facing a dual burden of infectious and chronic diseases. Remittances, or the money that migrants send back to their households, have been found to positively impact certain health outcomes, including infant mortality and under-5 mortality; however, it is less clear if health spending is the channel leading to this outcome. This study uses data from the World Development Indicators database between 1995-2011 to explore the two-part question of whether remittances are associated with health spending and if health spending is associated with the positive health outcomes that result from higher levels of remittances. The study uses a multivariate analysis and a dynamic GMM approach. The key findings suggest that remittances are not significantly associated with a reduction in government health spending, but are positively associated with certain health outcomes, such as infant mortality rate and under-5 mortality rate. Policy implications include increasing public health spending to raise total health spending, implementing policy reforms to facilitate the transfer of remittances, and creating alternative routes to increase household consumption. iii

To Ines and Charlie iv

Table of Contents Introduction... 1 Chapter 1: Literature Review and Conceptual Framework... 4 Effect of Remittances on Health Outcomes... 4 Effect of Remittances on Health Spending... 5 Effect of Remittances on Public Goods Provision... 6 Theoretical Considerations and Hypothesis... 7 Chapter 2: Empirical Analysis... 9 Chapter 3: Results... 11 Chapter 4: Policy Implications... 14 Chapter 5: Conclusions and Limitations... 16 References... 24 v

List of Tables Table 1: Description of World Development Indicators... 19 Table 2: Descriptive Statistics (1995-2011)... 20 Table 3: Effects of Remittances (USD) on Health Outcomes... 21 Table 4: Effect of Public Health Spending (% of Government Expenditure) on Health Outcomes... 22 Table 5: Effect of Remittances on Health Spending Measures... 23 vi

Introduction While the search for alternative health financing mechanisms to fund national health systems is not a new international development challenge, fiscal pressures to identify alternative sources of revenue for health are growing. 1-3 As low- and middleincome countries (LMICs) continue to battle infectious diseases and preventable maternal, child and newborn mortality rates, these countries are undergoing epidemiological transitions. Driven by aging populations, improved technologies, and globalization, LMICs are facing a greater burden of non-communicable diseases, such as cardiovascular disease and diabetes, which are increasingly common causes of mortality and morbidity. 4 Along with the changing needs of populations, efforts to expand universal health coverage schemes continue to exert pressure on national health accounts, which are comprised of public and private health expenditures. 3 Furthermore, many factors beyond health expenditure determine health outcomes, higher health spending reduces adverse health outcomes, including infant mortality rate and under-5 mortality rate. 5,6 Thus, identifying innovative sources to cover rising health spending needs and to improve population health is relevant and timely. Theoretically, one funding source that has been identified as a potentially innovative health financing tool for domestic mobilization of health care expenditures is the transfer of remittances. 3 Remittances are the private transfers of money from a migrant to family members remaining in the country of origin. In 2015, remittance financial flows accounted for over $600 billion globally. 7 As these cash transfers occur formally through banks or informally through other channels, this estimate is an underestimated amount. Compared to other capital inflows, such as foreign direct 1

investment (FDI) and official development assistance (ODA) that are tied to specific projects, remittances offer relative flexibility and stability to remittance-receiving households by smoothing income and consumption and increasing the household s ability to allocate funding to their specific needs. 8-10 For example, for top remittance-receiving countries, remittances can represent 20 to 40 percent of national gross domestic product (GDP); 7 for remittance-receiving households, remittances can account for up to 50 percent of the household s income and act as a tool to reduce poverty and income inequality. 11,12 Nevertheless, remittances may also create inflation, increase corruption, lower governance and increase dependency on an external labor market. 13-17 Despite these potentially challenging macroeconomic effects, many countries and the international community have argued in favor of using remittances for fostering economic development. 16,18,19 Studies have shown the positive impact of remittances on health outcomes, including lowering infant mortality rates, increasing infant birth weight, household health spending, and health knowledge, and equalizing access to care. 10,20-23 Despite the positive effects on health indicators at the micro level, the global evidence on the effect of remittances on health spending at the macro level is mixed. Some evidence indicates that remittances can increase corruption and lower political demand for change, resulting in a government decreasing its social spending. 14,24,25 Other studies suggest that the government will increase accountability instead. 15 To date, few studies have analyzed the effect of remittances on public health expenditures or the exact channel between remittances and improvements in health outcomes. 24,26,27 Therefore, do remittances have 2

an effect on public health spending and through which channel do remittances affect health outcomes? The purpose of this paper is to expand the literature around remittances, public health spending and health outcomes. The two-part study includes an analysis of whether public health spending is the channel through which remittances are positively affecting infant mortality rate, under-5 mortality rate, and maternal mortality ratio and if remittances impact public health spending. The study analysis is based on static and dynamic (GMM) estimation techniques that are used to model the impact of remittances on health outcomes and health spending. The use of dynamic estimation techniques minimizes endogeneity bias especially reverse causality between remittances and health spending measures that might threaten the statistical validity of the analytical results. This study contributes to the literature by providing preliminary evidence that remittances will not significantly reduce public health spending suggesting that the results could mitigate concerns that remittances negatively affect social spending. While the literature suggests that remittances will likely decrease public health spending, the data show that there is no effect. Furthermore, the data confirm that remittances have a positive impact on health, though remittances do not affect maternal mortality ratio. This is likely because maternal mortality ratio reflects the functional status of the health care system whereas other health outcome measures, such as infant mortality rate and under-5 mortality rate, depend on many social determinants of health beyond health services. Lastly, the data show that raising overall health expenditure, not other health spending measures, will affect infant mortality rate and under-5 mortality rate. 3

Chapter 1: Literature Review and Conceptual Framework Global evidence shows the impact of remittances spans from reducing national poverty levels to improving economic and social development outcomes, including health and education. 11,18,19,28 Studies have shown that remittances generally have a positive effect on health outcomes, including reducing infant mortality rate and infant birth weight, improving child health and increasing overall health spending. 20,21,29 Moreover, remittances have been found to counter the initially negative effect on health of having a parent leave a household. 30 Other studies have suggested that there may be a negative effect on public health spending because the government may begin to engage in a public moral hazard knowing that households are now receiving more income from abroad, which could be invested in health through the private sector. 24 Effect of Remittances on Health Outcomes Overall, studies show the effect of remittances on infant mortality rate is positive. Kanaiaupuni and Donato (1999) found that migration in Mexico initially had a disruptive effect on community and family life, increasing infant mortality rate; however, overtime this effect positively impacted infant mortality rate as communities experienced longterm migration patterns. Other studies further support the positive association between remittances and health outcomes. Zhunio, Vishwasrao and Chiang (2012) found that a 1% increase in remittances per capita increased life expectancy by 0.03% and reduced infant mortality rate by 0.15% in Mexico. Duryea, Lopez Cordova, Olmedo (2005) posited that this positive association is partly due to improved housing conditions. Others have found a differential effect between migrant households compared to non-migrant households that are due in part to the influx of money to spend on more nutritional food, 4

medical care, and improved housing. For example, infants born in migrant households are 3% less likely to die than in non-migrant households and are approximately 5% less likely to be born underweight. 20 As more households receive remittances, infant mortality rate declines further. 27 Effect of Remittances on Health Spending Studies have shown a mixed effect of remittances on health care spending at the household- and government-level. On the one hand, Amuedo-Dorantes and Pozo (2009) estimated that a 100 Mexican peso increase in remittances in Mexico augments household health expenditure by 5 or 6 Mexican pesos, though the effect of remittances will differ depending on the amount of remittances received. Similarly, in Ecuador, an increase of $10 resulted in a private health spending increase of 5%. 22 Cen Camal (2012) found an additional household level effect. Specifically, Mexican households use remittances to pay off health related debt, suggesting that remittances can be used at the household level to prevent catastrophic health payments that can push a household further into poverty. In addition to the positive correlation between remittances and private health spending, the literature indicates differences depending on the level of remittances received. Amuedo-Dorantes, Sainz and Pozo (2007) found that higher income remittancereceiving households spend 7% of remittances on health compared to 4% in lower income remittance-receiving households. Others estimates suggest that households that do not have access to formal health insurance will spend 11% of remittances on health compared to 8% for households with formal health insurance. 31 5

Effect of Remittances on Public Goods Provision Despite the positive effects of remittances on household health spending, the evidence on the effect of remittances at the macro-level is mixed. While there is general consensus that remittances will change the local political economies in home countries, including altering the way that tax revenues are allocated, studies have found that remittances have a mixed effect on corruption and governance, key determinants of the provision of public goods. Ahmed (2012) noted that a 3 percentage point increase in remittances increases corruption by 1 point on a corruption index. Abdih et al (2008) found a similar negative association between remittances and corruption, indicating the potential for governments to free ride based on the availability of remittances to substitute their provision of public goods. 32 In contrast, Tyburski (2012) found that remittances have the potential to counter corruption by increasing government accountability. The relationship between remittances and governance is relevant to health outcomes given that the positive effect of higher health spending on health outcomes is dampened in low governance environments. 33 These results on overall government indicators suggest that the effect of remittances on health spending may be mixed. Indeed, while Naanwaab and Yeboah (2013) found that a 10% increase in remittances is associated with a 0.30% increase in health spending as percent of GDP, Ebeke (2012) argued that remittances reduce health spending in countries with poor governance structures. A recent study exploring the effect of remittances on social spending found that the stabilizing effect of remittances at the household level lowered the public s demand on the government to provide social 6

services. 25 These findings suggest the potentially downward effect of remittances on health spending. Theoretical Considerations and Hypothesis Remittance-receiving households use remittances for a variety of reasons spanning immediate consumption to long-term investments for human development, such as education and health. Often times, migrants cite increasing health care spending as a primary reason for sending remittances; however, households may use remittances to buy common household items, such as clothing, food or electronics, or to invest in physical infrastructure or social development, such as education. 8,29,34 The lag in spending on social development is likely due to effects that materialize in the long-term. In other words, a remittance-receiving household can receive immediate benefits, or higher utility, by purchasing durable goods, whereas the impact of a better education or better health access may occur in the long-term. Remittances are sent to households through two main mechanisms: formal and informal. The formal exchange of remittances is through banks, credit unions and the use of money transfer organizations (MTO), such as Western Union or MoneyGram. Formally remitting money can be expensive due to high transaction costs and asymmetric information between the bank and the sender. Other challenges include fear of interacting with the formal financing system or that the financial system in the home country is inadequate. 35,36 Assuming the migrant household can access a bank to receive the remittances, the decision on how to use the remittance will be the responsibility of the head of household in the home country. Often times, the primary reason a migrant cites for remitting may 7

not coincide with the primary use of the remittances. For example, nearly half of Mexican remitters have cited health care spending as a primary reason for sending money to families, which aligns with altruistic, insurance and investment theories that explain why immigrants transfer remittances home. 11 However, households are more likely to use remittances to buy common household items, such as clothing, food or electronics, or invest in physical infrastructure or social development, such as education. 10 Households will likely spend remittances on immediate basic consumption needs, followed by investments, which can include home construction or human capital. Therefore, while the implications on direct health spending are limited, household investments (e.g., improved housing conditions or social standing) due to the higher budgetary constraint could have a greater impact on health outcomes. The health implications of the informal exchange channel depend on whether the remittances are intended to support the community or the individual household. Migrants may form hometown associations (HTA) to send money back home to a community. Money pooled through HTAs could be used to build a health clinic for a local community with matching funds from the municipal government. 37 While the HTA is creating a public good that can positively impact the population s health, an individual household may benefit more if the informal mechanism is at the individual household level. As the money is directly tied to the migrant s household, instead of the community, the household could receive more direct benefits than if the informal channel is communitybased. An additional benefit to the individual household level mechanism is that knowledge may be transferred more easily, especially if the migrant is travelling back home and relaying health knowledge that may have been acquired abroad. 8

The determinants for health outcomes include social standing, income, education, living conditions, and access to health systems. 38 Regardless of whether the intended use of remittances aligns with the actual household expenditure, remittances can increase a household s standard of living by increasing their consumption budget line. It is through this expansion of the budget line that households may be able to invest more money in factors that affect health. This can include improvements in housing and living conditions, higher direct out-of-pocket spending on health services, or improved access to food and nutrition. It is possible that households may seek services through the private sector as they now have higher purchasing power to spend on health care. Based on the theory posited in the literature, the main hypotheses of this paper are that remittances will result in the government lowering public health spending and that the channel through which remittances is impacting health outcomes is indirect. The government may lower public spending on health care if they perceive this higher out-ofpocket spending as a substitute to government spending. Furthermore, as higher remittance levels implies that the country has a higher migrant stock, there may also be less public pressure to reform behaviors, such as corruption or lower social spending. 15,25 Moreover, as households receive remittances, they will likely increase their household consumption, which will have a greater impact on health outcomes, such as infant mortality rate and under-5 mortality rate, that are a function of social determinants of health. Chapter 2: Empirical Analysis The analysis follows a two-step approach. First, the effect of remittances on health outcomes and health spending are analyzed. Second, an empirical model studying 9

the effects of health spending on actual health outcomes is developed as a complementary tool. This second stage of analysis allows a more detailed exploration of the two key questions: whether and through which channels remittances impact selected health outcomes. The basic multivariate specification is the following: The central dependent variable in the model, HealthExpi,t, is health spending (all measures included in Table 1) for a country, i, at a time, t. The main independent variable, remittances (measures included in Table 1), is lagged, t-1, to minimize the threat of reverse causality and to capture the lagged adjustment of public health expenditures. Furthermore, remittances and health spending measures are logged. Controls are denoted through Xi,t and country and time fixed effects are captured using ut and Θt. All data are from the World Bank s World Development Indicators (WDI) dataset, a multisectoral, multi-country panel dataset that spans over 200 countries over 50 years. To ensure overlap between the independent variable, remittances, and the dependent variable, public health expenditure, the time period for this study is 1995-2011. Table 1 describes the variables, short form and source of the selected variables used in the model. Table 2 describes the number of observations and other summary statistics for the variables included in the model. Other controls included in the model were GDP per capita, GDP growth rate, urban population rate and life expectancy. As the selected model specification was relatively small, country and time fixed effects (fe) were used to eliminate the potential for omitted variable bias in a multivariate analysis. To control for the potential of reverse causation, a dynamic panel approach was used. Other studies have used instrumental 10

variables to achieve this same goal. 24 The same approach was used for the analyses using health outcome data, including infant mortality rate, under-5 mortality rate and maternal mortality ratio. Sensitivity analyzes were also conducted by geographic region (i.e., Africa, Asia, Americas, Europe) and income group (i.e., Low, Middle). Chapter 3: Results Remittances and Health Outcomes: In a year and time fixed effect model including GDP per capita, the percentage of urban population, and GDP growth as controls, remittances (in USD) have a statistically significant effect on infant mortality rate and under-5 mortality rate (Table 3). On average, a one percent increase in remittances is associated with a reduction of 1.015 infant deaths per 1,000 live births. Similarly, including time-varying and country-varying controls, a one percent increase in remittances is associated with a 1.699 under-5 deaths per 1,000. In contrast, there is no statistically significant association between remittances and maternal mortality ratio. These relationships hold when robustness checks were conducted using a different measure of remittances (as percent of GDP) (Appendix A7) or regional and economic variation (Appendix A8). The GMM model shows the same statistical association; however, the relationship is somewhat smaller. The GMM model eliminates the problems due to reverse causality (e.g., improved health could create a larger migrant pool, which could, in turn, increase remittances) and serial correlation between health outcomes and remittances (e.g., health outcomes may remain relatively constant over time). Overall, the results indicate that remittances improve health the results indicate that remittances improve health outcomes. Separate analyses on the relationship between health spending and health outcomes are used to isolate the channels through which remittances operate. 11

Health Spending and Health Outcomes: In a year and time fixed effect model including GDP per capita, the percentage of urban population, and GDP growth as controls, public health expenditure (as % of government expenditure) has a statistically significant effect on maternal mortality ratio but does not have a statistically significant effect on IMR and under-5 mortality rate (Table 4). On average, a one-unit increase in the log of the percent of public health expenditure (% of government expenditure) is associated with a reduction of 81 maternal deaths per 100,000 live births. The GMM model shows a different statistical association where only under-5 mortality rate has a statistically significant effect. These relationships vary when robustness checks were conducted using different measures of public health expenditure (Appendix A3, A4, A5) or regional and economic variation (Appendix A9). For example, main findings from robustness checks using differing measures of health expenditure show the following: First, on average, a one percent increase in expenditure per capita is associated with a reduction of 2.335 infant deaths per 1,000 live births. There is no statistical association with under-5 mortality rate or maternal mortality ratio or in the GMM model (Appendix A3). Second, on average, a one unit increase in the log of the percent of public health expenditure (% of GDP) is associated with a reduction of 3.785 under-5 deaths per 1,000 and a reduction of 7.325 maternal deaths per 100,000 live births. There is no statistical association with infant mortality rate or in the GMM model (Appendix A4). Third, on average, a one unit increase in the log of the percent of public health expenditure (% of total health expenditure) is associated with a reduction of 4.628 under- 5 deaths per 1,000 and 122.2 maternal deaths per 100,000 live births. There is no statistical association with IMR or in the GMM model (Appendix A5). 12

Other robustness checks focused on the effect of household consumption, government consumption and development assistance show the following (Appendix A10). First, on average, a one-unit increase in the log of the percent of household consumption, as percent of GDP, is associated with a reduction of 5.603 maternal deaths per 100,000 live births and a reduction of 2.906 infant deaths per 1,000 live births. There is no statistically significant association between household consumption and under-5 mortality rate. Second, on average, a one-unit increase in the log of the percent of government consumption as percent of GDP is associated with a reduction of 5.425 under-5 deaths per 1,000 and reduction of 3.99 infant deaths per 1,000 live births. There is no statistically significant association between government consumption and maternal mortality ratio. Lastly, there is no statistically significant association between development assistance and the three selected health outcomes. Remittances and Health Spending: In a year and time fixed effect model including GDP per capita, the percentage of urban population, and GDP growth as controls, remittances (in USD) have a statistically significant effect on expenditure per capita (Table 5). On average, a one percent increase in remittances is associated with a reduction of $18 in health expenditure per capita. There is no other statistically significant effect on other health expenditure measures. The GMM model shows a statistically significant effect on public health expenditure as percent of government expenditure. On average, a one percent increase in remittances is associated with a 0.2 unit reduction in the log of public health expenditure as percent of government expenditure. There is no other statistically significant effect on other health expenditure measures. 13

Chapter 4: Policy Implications The key findings from this analysis suggest that remittances are not significantly associated with a reduction in government health spending, but are positively associated with certain health outcomes, such as infant mortality rate and under-5 mortality rate. Instead of inducing moral hazard, as the literature suggests, it appears that the government does not change their spending patterns based on remittance levels. While the results on remittances and health outcomes are consistent with the literature (e.g., remittances have an effect on health outcomes), remittances do not appear to have a significant association with health spending measures, including out-of-pocket payments or expenditure per capita. This indicates that the channel through which remittances impact health outcomes is not likely through direct health spending. Robustness checks analyzing household and government consumption strengthen the case that remittances primarily impact health outcomes through more stable and enhanced household consumption. Remittances increase the amount of disposable income at the household level, which may be invested in improving housing conditions or increase availability of food. These channels appear to have a stronger impact than direct health spending. Unlike infant mortality rate and under-5 mortality rate, which are associated with many social determinants of health, such as housing, food access, nutrition and family security, maternal mortality reflects direct access to health services. The results for maternal mortality support the hypothesis that maternal mortality ratio will not be driven by remittances, and instead, depend on the health system overall. 14

Based on the results from this study, governments should consider the following interrelated policy solutions, which weigh the positive effects of remittances against the potentially adverse macroeconomic effects: First, the data on health spending and health outcomes suggest that total health spending should increase. If migrants are spending more money on non-health services and durable goods, such as food or improvements to living conditions, then policy makers should respond by investing in public health expenditure to increase total health spending. This policy recommendation is also aligned with the World Health Organization s (WHO) efforts to increase universal coverage for health and reduce reliance on out-of-pocket spending, or the money that people spend on health care services directly. The data show that out-of-pocket payments do not have a statistically significant effect on health outcomes, which indicates that they should not be promoted as a measure to improve health outcomes. Second, governments should implement policy reforms to make it easier for migrants to send remittances back home. Remittances appear to have a sustained and consistent relationship on key health outcomes, such as infant mortality rate and under-5 mortality. This positive effect suggests that governments should open the channels through which remittances flow to increase the disposable income available to households, especially those who heavily rely on remittances as part of their household income. To account for potential negative effects, or the flow of illicit money without oversight, the international community should increase efforts toward improving government and regulatory frameworks. While making it easier for migrants to send money home can help at the household level, remittances are comprised of many sources 15

of funding, suggesting that more comprehensive financial channels should be in place to maximize the benefits of remittances and reduce the potentially negative consequences. Third, although governments should promote remittances for household consumption, national governments should consider alternative routes to increase household consumption. For example, governments can create or continue to invest in social protection programs that increase the availability of cash to program eligible households. In doing so, these types of programs can increase household consumption, which appears to be a driving force behind the reductions in mortality rates. This policy measure could be implemented in conjunction with the first recommendation regarding increasing total health spending. Chapter 5: Conclusions and Limitations As governments consider the policy recommendations described above, policy makers should consider the inherent trade-off regarding remittances. While increasing remittances and maintaining a stable flow of household consumption can increase health outcomes, remittances can also produce negative macroeconomic consequences, such as inflation or higher exchange rates. Facilitating the flow of remittances to benefit households could also inadvertently lead to the higher flow of illicit funds. Furthermore, the effect of remittances on chronic diseases, such as diabetes, remains unknown. A future study on the effect of remittances on chronic diseases, such as diabetes, which tend increase over time with lifestyle changes, would likely result in another negative outcome associated with remittances. In light of this policy trade-off, policy makers should reform remittance policy within a more comprehensive framework to mitigate the concerns. More work should be 16

directed toward improving financial regulation and governance to facilitate the growth of remittances in low- and middle-income countries, but maximizing the beneficial effects. Furthermore, governments should increase health spending to complement the higher household consumption to increase health outcomes, such as infant mortality rate and under-5 mortality rate, and improve health outcomes, such as maternal mortality ratio, that primarily depend on the existing health system instead of social determinants of health. The main limitations of this study were the following. First, the time period used was determined based on the availability of health spending data, which only began to be collected on an international level in 1995. Therefore, the effect of remittances and health outcomes used in this study was restricted to a 20-year time period. Second, remittances are underreported given that these data only capture formally recorded remittances. Others have also noted inconsistency in data reporting across countries. 21 Third, there is missing data across the three areas of interest: remittances, health outcomes and health spending. Therefore, it is possible that the results may change over time or when a more complete dataset is available. The international community and many national governments have been using remittances as a tool for development for some time. While this study supports that remittances should continue to be used to increase household consumption, it is important that governments do not rely on remittances as the sole mechanism for health improvements. It is evident that remittances can be a driving force to improve health outcomes, especially those that rely on social determinant of health, but remittances fall short of being a panacea for all health problems. Long-standing health issues such as 17

maternal mortality ratio and the growing burden of chronic diseases will require other solutions that go beyond increasing household consumption. Therefore, governments in developing countries, along with partners in international health funding institutions, should continue to identify innovative ways to use remittances and other funding sources more effectively. 18

19 Table 1: Description of World Development Indicators Variable Short form Type Source (as described in WDI) Personal remittances, received (current Remitusd Independent World Bank estimates based on IMF balance of payments data, US$) and World Bank and OECD GDP estimates Personal remittances, received (% of GDP) Remitgdp Independent World Bank estimates based on IMF balance of payments data, and World Bank and OECD GDP estimates Health expenditure, public (% of GDP) exppubgdp Dependent WHO Global Health Expenditure database Health expenditure, public (% of government expenditure) Health expenditure, public (% of total health expenditure) exppubgovexp Dependent WHO Global Health Expenditure database exppubthe Dependent WHO Global Health Expenditure database Health expenditure, total (% of GDP) expgdp Dependent WHO Global Health Expenditure database Health expenditure per capital (current $) expcap Dependent WHO Global Health Expenditure database GDP per capita (current US$) gdpcap Control World Bank, International Comparison Program database Urban population (% of total) urbanpop Control World Bank GDP growth (% annual) gdpgro Control World Bank Life expectancy at birth, total (years) Lifexpect Control/Dependent UNPD and others Mortality rate, under-5 (per 1,000) mortund5 Dependent UN Inter-agency for Child Mortality Estimation Maternal mortality ratio (modeled estimate, per 100,000 live births) Maternmort Dependent UN Inter-agency Mortality rate, infant (per 1,000 live births) Imr Dependent UN Inter-agency for Child Mortality Estimation

Table 2: Descriptive Statistics (1995-2011) Variable N Mean SD Min Max Remitusd 2786 1.66 e+09 4.17e +09 0.00 6.88e+10 Remitgdp 2725 4.06 6.85 0.00 61.99 exppubgdp 3184 3.70 2.35 0.00 21.11 exppubgovexp 3180 11.09 4.63 0.00 42.38 exppubthe 3181 57.22 20.34 0.39 99.98 expgdp 3184 6.28 2.69 0.00 22.19 expcap 3182 674.344 1276.119 0 9120.81 gdpcap 3479 10727.47 18016.55 64.81 1.94e+05 urbanpop 3726 55.32 24.47 7.21 100.00 gdpgro 3434 4.076798 5.748358-47.553 106.2798 Lifexpect 3318 67.83558 10.10695 31.239 85.16341 Maternmort 517 163.9222 275.7378 0 2100 mortund5 3438 50.78365 53.0938 2.2 278.9 imr 3438 35.42693 32.18131 1.7 153.1 20

21 Table 3: Effects of Remittances (USD) on Health Outcomes Infant Mortality Rate Under-5 Mortality Rate Maternal Mortality (1) (3) (4) (5) (1) (3) (4) (5) (1) (3) (4) (5) ln_remitusd -3.960*** -1.435*** -1.052*** -0.143** -6.407*** -2.441*** -1.699*** -0.289** -16.33*** -3.191-0.268-0.847 (- (-0.561) (-0.41) (-0.293) (-0.0569) (-1.013) (-0.749) (-0.517) (-0.121) (-5.863) 5.549) (-5.072) (-1.528) gdpcap 0.000400*** 8.02e-06** 0.000697*** 2.93E-05 0.00229-0.000243 (-0.0000831) (-0.00000367) (-0.000151) (-0.0000208) (-0.00276) (-0.000315) urbanpop -0.462*** 0.00623* -0.789** 0.00915 4.469-0.0171 (-0.171) (-0.00332) (-0.357) (-0.00655) (-2.954) (-0.109) gdpgro -0.0369-0.00979-0.0176 0.00741-1.504-0.104 (-0.0338) (-0.0102) (-0.0638) (-0.0718) (-1.055) (-0.392) lifexpect -2.113*** -0.0166-4.237*** -0.236-29.78*** -0.676 (-0.287) (-0.0213) (-0.621) (-0.183) (-5.771) (-0.505) imr/mortund5/maternmort 0.970*** 0.921*** 0.904*** (-0.00599) -0.0366 (-0.0505) Year X X x X x x Country X X x X x x Constant 109.6*** 67.55*** 221.8*** 3.662** 170.6*** 104.6*** 408.0*** 23.27 531.9*** 200.3 1,890*** 0 (-11.74) (-7.826) (-21.29) (-1.78) (-21.22) (-14.45) (-44.34) (-15.06) -122.7-126.9-329.1 0 Obs 2,556 2,556 2,460 2,460 2,556 2,556 2,460 2,460 432 432 420 145 # countries 171 171 170 170 171 171 170 170 131 131 127 49 chi2 49.79 258234 40.03 129318 7.753 225657 hansenp 0.795 0.463 0.697 sarganp 0.468 1 0.714 R-squared 0.501 0.715 0.398 0.64 0.16 0.349 Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Models (1): Simple OLS; (2): OLS + year fixed effect; (3): OLS + year and country fixed effect; (4): Fixed effect; (5): GMM

22 Table 4: Effect of Public Health Spending (% of Government Expenditure) on Health Outcomes Infant Mortality Rate Under-5 Mortality Rate Maternal Mortality (1) (3) (4) (5) (1) (3) (4) (5) (1) (3) (4) (5) - ln_exppubgovexp -4.164** 0.289-0.364-0.147-8.116** -1.051-2.516-0.692* 83.54*** -80.14** -81.88** 5.591 (-1.932) (-1.298) (-1.132) (-0.174) (-3.602) (-2.503) (-1.962) (-0.418) (-29.63) (-35.11) (-35.67) (-5.875) gdpcap 0.000395*** 1.05e-05*** 0.000693*** 3.24e-05** 0.00483* -0.000315 (- (- (- (- (-0.0000683) 0.00000308) (-0.000125) 0.0000159) 0.00261) 0.000325) urbanpop -0.332** 0.00443-0.620* 0.00601 4.442 0.0503 (-0.158) (-0.00307) (-0.326) (-0.00592) (-2.909) (-0.0809) Gdpgro -0.02-0.0106* -0.0144-0.00616-1.127 0.0845 (-0.0483) (-0.00594) (-0.0825) (-0.0375) (-1.16) (-0.339) - Lifexpect -2.189*** -0.0370** -4.343*** -0.24 27.00*** -0.615 (-0.269) (-0.0171) (-0.558) (-0.164) (-5.539) (-0.544) imr/mortund5/maternmort 0.969*** 0.927*** 0.913*** (-0.00566) (-0.0324) (-0.0473) Constant 44.13*** 41.89*** 202.2*** 2.843** 68.01*** 64.16*** 380.9*** 19.66 412.3*** 315.7*** 1,871*** 0 (-5.078) (-3.046) (-18.24) (-1.377) (-9.463) (-5.944) (-36.66) (-12.75) (-73.93) (-102.2) (-332.3) 0 Obs 3,178 3,178 2,961 2,961 3,178 3,178 2,961 2,961 488 488 471 155 # countries 190 190 185 185 190 190 185 185 150 150 142 51 chi2 4.644 435108 5.078 186883 7.95 163396 Hansenp 0.0683 0.275 0.619 Sarganp 0 0.974 0.145 R-squared 0.486 0.699 0.395 0.642 0.182 0.337 Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Model 1: Simple OLS; Model 2: OLS + year fixed effect; Model 3: OLS + year and country fixed effect; Model 4: Fixed effect; Model 5: GMM

23 Table 5: Effect of Remittances on Health Spending Measures exppubgovexp expcap exppubgdp expoopthe Model 4 Model 5 Model 4 Model 5 Model 4 Model 5 Model 4 Model 5 ln_remitusd 0.0543 0.260* ln_remitusd -18.87** 1.96 ln_remitusd -0.0166 0.0515 ln_remitusd -0.0666 0.064 (-0.115) (-0.139) (-8.798) (-5.555) (-0.0339) (-0.0324) (-0.456) (-0.313) gdpcap 2.36E-05 2.24e-05*** gdpcap 0.101*** 0.00223 gdpcap 1.50E-05 1.06e-05*** gdpcap 6.65E-05-3.04e- 05*** (- 0.0000211) (- 0.00000635) (- 0.00577) (- 0.00181) (- 0.00001) (- 0.00000232) (- 0.0000453) (- 0.00000924) urbanpop -0.0558-0.00496 urbanpop 1.155 0.260** urbanpop 0.00934 9.79E-05 urbanpop -0.284-0.00225 (-0.0646) (-0.00604) (-6.521) (-0.104) (-0.0201) (-0.0012) (-0.207) (-0.00685) Gdpgro -0.0411** -0.0115 gdpgro -1.061 1.220** gdpgro - 0.0121** 0.00207 gdpgro 0.0341-0.0202 (-0.0184) (-0.0152) (-1.03) (-0.48) (- 0.00486) (-0.00281) (-0.0399) (-0.0282) Lifexpect 0.058-0.00637 lifexpect -1.633 0.348 lifexpect 0.0122-0.00213 lifexpect -0.0706-0.0038 (-0.153) (-0.0152) (-2.992) (-0.396) (-0.0388) (-0.00358) (-0.176) (-0.0338) exppubgovexp 0.779*** expcap 1.022*** exppubgdp 0.890*** expoopthe 0.940*** (-0.0274) (-0.0196) (-0.0159) (-0.0133) Year X Year X Year x Country X Country X Country x Constant 8.433-1.681 Constant 237.3-89.38 Constant 2.245-0.503 Constant 55.85*** 1.827 (-11.34) (-1.821) (-411.5) (-79.66) (-2.961) (-0.43) (-19.12) (-3.32) Obs 2,303 2,303 Obs 2,303 2,303 Obs 2,305 2,305 Obs 2,305 2,305 # countries 168 168 # countries 168 168 # countries 168 168 # countries 168 168 chi2 2295 chi2 259681 chi2 13380 chi2 50334 Hansenp 0.495 hansenp 0.442 hansenp 0.0872 hansenp 0.227 Sarganp 1.97E-05 sarganp 0.18 sarganp 5.57E-06 sarganp 0.0122 R-squared 0.091 R-squared 0.844 R-squared 0.212 R-squared 0.059 Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Model 1: Simple OLS; Model 2: OLS + year fixed effect; Model 3: OLS + year and country fixed effect; Model 4: Fixed effect; Model 5: GMM

References 1. Jamison DT, Summers LH, Alleyne G, et al. Global health 2035: a world converging within a generation. The Lancet. 2013;382(9908):1898-1955. 2. World Bank. World Development Report 1993: Investing in Health. Washington, D.C.: The World Bank;1993. 3. World Health Organization. The World Health Report: Health System Financing: The Path to Universal Coverage. Geneva, Switzerland: World Health Organization;2010. 4. Daniels ME, Donilon TE, Bollyky TJ. The Emerging Global Health Crisis: Noncommunicable Diseases in Low- and Middle-Income Countries. New York, NY: Council on Foreign Relations;2014. 5. Farag M, Nandakumar A, Wallack S, Hodgkin D, Gaumer G, Erbil C. Health expenditures, health outcomes and the role of good governance. Int J Health Care Finance Econ. 2013;13:33-52. 6. Nixon J, Ulmann P. The relationship between health care expenditure and health outcomes. Evidence and caveats for a causal link. Eur J Health Econom. 2006;7:7-18. 7. World Bank. Migration and Remittances Factbook 2016. Washington, D.C.: The World Bank;2016. 8. Brown S. Can Remittances Spur Development? A Critical Survey. International Studies Review. 2006;8:55-75. 9. Vassall A, Martinez-Alvarez M. The health system and external financing. In: Smith RD, Hanson K, eds. Health Systems in Low- and Middle-Income Countries: An Economic and Policy Perspective. Oxford, England: Oxford University Press; 2012. 10. Amuedo-Dorantes C, Sainz T, Pozo S. Remittances and Healthcare Expenditure Patterns of Populations in Origin Communities: Evidence from Mexico. Vol 112007. 11. Amuedo-Dorantes C, Pozo S. Remittances and Income Smoothing (Discussion paper no. 5568). Bonn, Germany: Institute for the Study of Labor;2011. 12. Acevedo C, Cabrera M. Social policies or private solidarity? The equalizing role of migration and remittances in El Salvador. In: Cornia G, ed. Falling inequality in Latin America: Policy Changes and Lessons. Oxford, United Kingdom: Oxford University Press; 2012:164-187. 13. Mughal M. Remittances as development strategy: Stepping stones or slippery slope?. Journal of International Development. 2013;25:583-595. 14. Ahmed FZ. Remittances deteriorate governance. The Review of Economics and Statistics. 2013;95(4):1116-1182. 15. Tyburski MD. The resource curse reversed? Remittances and corruption in Mexico. International Studies Quarterly. 2012;56:339-350. 16. Mundaca G. Remittances, Financial Market Development, and Economic Growth: The Case of Latin America and the Caribbean. Review of Development Economics. 2009;13(2):288-303. 24

17. Nyamongo EM, Misati RN, Kipyegon L, Ndirangu L. Remittances, financial development and economic growth in Africa. Journal of Economics and Business. 2012;64:240-260. 18. Ratha D. The Impact of Remittances on Economic Growth and Poverty. Washington, D.C.: Migration Policy Institute;2013. 19. Chami R, Barajas A, Cosimano T, Fullenkamp C, Gapen M, Montiel P. Macroeconomic Consequences of Remittances. Washington, D.C.: International Monetary Fund;2008. 20. McKenzie D. Beyond remittances: the effects of migration on Mexican households. International migration. 2005. 21. Zhunio MC, Vishwasrao S, Chiang EP. The influence of remittances on education and health outcomes: a cross country study. Applied Economics. 2012;44(35):4605-4616. 22. Ponce J, Olivie I, Onofa M. The role of international remittances in health outcomes in Ecuador: prevention and response to shocks. International Migration Review. 2011;45(3):727-745. 23. López-Cevallos DF, Chill C. Migration, remittances, and health care utilization in Ecuador. Rev Panam Salud Publica. 2012;31(1):9-16. 24. Ebeke CH. Do Remittances Lead to a Public Moral Hazard in Developing Countries? An Empirical Investigation. Journal of Development Studies. 2012;48(8):1009-1025. 25. Doyle D. Remittances and Social Spending. American Political Science Review. 2015;109(4):785-802. 26. Naanwaab C, Yeboah OA. Migrant remittances and human capital investments. Review of Applied Socio- Economic Research 2013;6(2):191-203. 27. Duryea S, Lopez Cordova JE, Olmedo A. Migrant remittances and infant mortality: evidence from Mexico. Washington, D.C.: Inter-American Development Bank;2005. 28. Adams RH. Evaluation the Economic Impact of International Remittances on Developing Countries Using Household Surveys: A Literature Review.. Journal of Development Studies. 2011;47(6):809-828. 29. Amuedo-Dorantes C, Pozo S. New evidence on the role of remittances on healthcare expenditures by Mexican households. Rev Econ Household. 2009;9:69-98. 30. Kanaiaupuni S, Donato K. Migradollars and mortality: The effects of migration on infant survival in Mexico. Demography. 1999;36(3):339-353. 31. Valero JN. Remittances and the household s expenditures on health. Munich Personal RePEc Archive;2008. 32. Abdih Y, Chami R, Dagher J, Montiel P. Remittances and institutions: Are remittances a curse? Washington, D.C.: International Monetary Fund;2008. 33. Makuta I, O'Hare B. Quality of governance, public spending on health and health status in Sub Saharan Africa: a panel data regression analysis. BMC Public Health 2015;15:932-943. 34. Cen Caamal M. Remesas y estructura del gasto en los hogares receptores: Caso de estudio de una comunidad maya contemporanea. (Remittances and Expenditure 25

Structure of Recipient Households: A Study Case of a Contemporary Mayan Community. With English summary.). Analisis Economico. 2012;27(64):5-21. 35. Hernandez-Coss R. The U.S.-Mexico Remittance Corridor: Lessons on Shifting from Informal to Formal Transfer Systems.. Washington, DC: World Bank;2005. 36. The World Bank. Mexico s Financial Inclusion Strategy. 2010; http://siteresources.worldbank.org/financialsector/resources/282044-1260476242691/mexico_financial_inclusion.pdf Accessed December 1, 2015. 37. Orozco M. Mexican Hometown Associations and Development Opportunities. Journal of International Affairs. 2004;57(2):1-21. 38. Loewenson R, Gilson L. The health system and wider social determinants of health. In: Smith RD, Hanson K, eds. Health systems in low- and middle-income countries. New York, NY: Oxford University Press; 2012. 26