F INANCIAL S ERVICES F IGHTING HUNGER AND DEATH WITH

(1)

1

Master Thesis

F

IGHTING HUNGER AND DEATH WITH

F

INANCIAL

S

ERVICES

A study on the relationship between health, and malnutrition, and

financial services

Arjan Kieneker

1

August, 2012

Abstract

This paper explores the possible relationships between financial services and mortality and malnutrition. After discussing the determinants of health and malnutrition, theoretical relationships are developed between health and finance. Indicators of mortality and malnutrition are used as dependent variables in fixed effects models. Measures of the availability and use of financial services are used as independent variables. The results are in support of the developed hypotheses. Especially the availability of financial services is significantly related with reductions in malnutrition and mortality. The indicators of use of financial services are possibly still subject to measurement shortcomings. The findings are robust for the inclusion of control variables and the use of developing countries only.

JEL: G21, I10, I15.

Keywords: Access, use, finance, financial services, malnutrition, mortality

1_{Affiliation: Student at the University of Groningen, Faculty of Economics & Business; Groningen; The}

(2)

2

C

ONTENTS

1. Introduction ... 3

2. Health and Malnutrition ... 4

3. Relationship between financial services and health ... 8

4. Methodology ... 14

5. Data description and descriptive statistics... 15

6. Results ... 22

7. Conclusion ... 40

References ... 41

(3)

3

1. I

NTRODUCTION

Many people in the world lack good health, in Sub-Saharan Africa 1 out of 6 children dies before the age of five in 2006. This is sixteen times the average of developed countries where 1 out of 100 children under the age of five dies. One of the major causes of bad health in developing countries is malnutrition (Skolnik, 2008). Malnutrition is one of the priorities of the Millennium Development Goals (MDGs). Unfortunately, least progress has been made on the matter of malnutrition (UN, 2012). And that while nutrition is of utmost importance. Proper nutrition at young age is necessary for the full development of physical and cognitive skills. In other words, if children do not receive the nutrients they need, they fail to study or work (Skolnik, 2008). Food aid and using innovative seeds and fertilizer can support in making more food available, but what if people do not have the money available to spend on food? There is the possibility that financial services can support in fighting hunger and malnutrition. If people are able to save or borrow in order to spread their income and receive nutrition, they remain able to work and gain income. Possibly, this could have large effects on malnutrition and even mortality (Morduch, 2012).

(4)

4

By employing fixed effect regression techniques I examine the effects of finance on health and malnutrition within countries. Using data from 2002 to 2010 I analyze the impact of financial services on health in order to determine the empirical validity of the developed theoretical framework. The evidence is mixed, but the results show that the availability and use of financial services improves health and decreases malnutrition. These findings are robust against the inclusion of control variables and the selection of developing countries only.

The remainder of the paper is organized as follows. The determinants of health and nutrition will be discussed in section 2. Section 3 describes the theoretical framework between health and finance and the hypotheses. Section 4 presents the methodology and estimated equations. The data description and descriptive statistics are covered in section 5. Section 6 presents the results and interprets the results. In section 7 conclusions are drawn and the limitations of this study as well as the future possibilities of research are presented.

2. H

EALTH AND

M

ALNUTRITION

There are several links between health and finance, but before discussing these I will briefly define health in the first paragraph. The second paragraph describes the determinants of health. An important determinant of health is nutrition. The subject of malnutrition is explained in the third paragraph, while the fourth paragraph discusses the determinants of malnutrition.

2.1 D

EFINITION OF

H

EALTH

Health is defined in the preamble of the constitution of the World Health Organization as ‘..a state of complete physical, mental and social well-being and not merely the absence of

disease or infirmity’ (WHO, 2006). Further the preamble states that: ‘[t]he enjoyment of the highest attainable standard of health is one of the fundamental rights of every human being without distinction of race, religion, political belief, economic or social condition.’

This statement that health is a human right is also recognized in article 25 of the Universal Declaration of Human Rights (UN General Assembly, 1948):

(5)

5

necessary social services, and the right to security in the event of unemployment, sickness, disability, widowhood, old age or other lack of livelihood in circumstances beyond his control.

(2) Motherhood and childhood are entitled to special care and assistance. All children, whether born in or out of wedlock, shall enjoy the same social protection.’

Thus, health is closely related to well-being and the quality of life and good health is globally recognized as important; however, global and local disparities remain. In the past 60 years, great progress has been made in the health sector, increasing the quality of life of a share of the global population. The expected duration of life increased with more than 50% and diseases such as smallpox and polio have (nearly) been eradicated (Skolnik, 2008). Academic studies show positive effects of preventive projects (Demombynes & Trommlerová, 2012) and the UN claims that already three of the Millennium Development Goals(MDGs) have been achieved, poverty has been reduced, a large proportion of the people have access to improved sources of drinking water and the share of the population living in slums has been reduced (UN, 2012). Unfortunately, there is also the well-known other side. Still large numbers of people die of easily preventable diseases. Within and between countries there are still large differences in health status and access to health services, especially between developed countries and a majority of the developing countries.

2.2 D

ETERMINANTS OF HEALTH

(6)

6

The characteristics can be divided into individual and external determinants. Individual

characteristics such as genetic make-up, sex and age are important determinants of one’s

health. There are four main external determinants of health which affect the impact of individual characteristics on health, shown in the squares of figure 2.1. Health behavior describes the consumption and life habits, such as nutritional value of food consumed, use of alcohol and driving style, as well as the knowledge about health phenomena and being able to identify illnesses. Access to health services can be limited by the physical availability of services as well as the costs and quality of these services. The social

environment contains elements such as socio-economic status, education, social capital

and culture; the physical environment elements like provision of water, sanitation, mosquito prevalence and air pollution (Skolnik, 2008).

Figure 2.1 Determinants of Health. Source: Skolnik, 2008.

In all countries the level of health is determined by these aspects, and this paper will study how finance can improve health by affecting these determinants. Unfortunately, health cannot directly be measured as the quality of life derived from these determinants. Therefore most studies on health often use life expectancy or mortality are as proxies for health (Deaton, 2003). The disadvantage of using such measures is that these are very broad, and provide ample insight in the real reasons of increases in life expectancy. Therefore, this study improves research in health by including data on malnutrition, one of the major risk factors for health in countries. Because it is such an important risk factor, malnutrition is closely related to the health level in a country (Skolnik, 2008). Additionally the measurement of malnutrition is more direct and does not need proxies.

(7)

7

2.3 D

EFINITION OF

M

ALNUTRITION

Malnutrition is either overnourishment or undernourishment. The focus of this paper is on undernourishment, which is receiving poor nutrition. Although overnourishment is an important topic currently it is too different from undernourishment and too complicated to deal with in this paper. Undernourishment can happen because of a lack of food or due to suffering from deficiencies of micronutrients, such as iron or vitamin A, in daily meals (Skolnik, 2008).

Proper nutrition is important, especially in the first two years of life. In this stage the most important aspects of the body are being developed, such as the cognitive abilities and the immune system. In a majority of the cases of malnutrition, it creates chronic growth problems, preventing people to acquire full potential later in life (Skolnik, 2008). As parents are responsible for the nutrition of a child, the income, and health, of parents is closely related to the income and health of their children (Angelini & Mierau, 2012). At later stages in life proper nutrition is necessary to work and prevent illnesses; the immune system requires micronutrients, and with a weaker immune system people are no longer resistant to illnesses such as diarrhea, malaria or other treatable diseases. Additionally, a lack of iron consumption is associated with iron deficiency anemia, which leads to weakness and fatigue disabling people to work properly. Populations especially at risk of undernourishment are people living in rural areas, are uneducated or are poor (Skolnik, 2008).

There are two types of malnutrition, wasting and stunting. If height for age is more than two or three standard deviations below international reference it is respectively referred to as stunting or severe stunting. A too low height for age is caused by chronic undernutrition, a continuous lack of sufficient protein and energy at young age. Stunting is caused before the age of 2 years and is reason of most of the disabilities a person can face in later stages of life. If the weight is lower than two standard deviations below the international reference weight for height, this is referred to as wasting. If it is more than three deviations it is severe wasting. Wasting is caused by undernourishment, receiving not enough protein and energy. It causes lack of strength, preventing people to work and lowering the barriers to illnesses.

(8)

8

develop into a more healthy and economically-strong environment, with lower mortality, higher life expectancy and a decline of the burden of infectious diseases (Skolnik, 2008). Therefore, according to Skolnik (2008), investing in nutrition of poor people is one of the major policy areas in order to achieve important health gains. Additionally, the UN (2012) mentions malnutrition as one of the remaining challenges in developing countries and the Copenhagen Consensus in 2004 showed the importance of dealing with malnutrition, by ranking 4 malnutrition projects as very good, good and fair projects to address global health problems (Skolnik, 2008).

2.4 D

ETERMINANTS OF MALNUTRITION

Nutritional status depends on a number of factors, the direct, underlying and basic factors. The direct factors are food availability, diseases and quality of maternal and child care. Diseases can limit the nutritional status as they can prevent the good absorption of nutrients, as well as the ability of the person to reach food sources. Beyond these direct causes are the underlying factors, these are inadequate access to food, poor health services and environment, and inadequate care for children and woman. The access to food can be limited because of the availability, geographical situation and the costs. The other underlying factors are primarily external to the individual. Access to health services is determined by the costs as well as the availability of such services. But their quality can also be determined by the expenditure of the government in such facilities and the level of education of the nurses and doctors available. These underlying factors are affected by the basic factors of malnutrition: inadequate education, resources and control and political/economic factors (Skolnik, 2008). Income and education, next to availability of resources and the political situation are major determinants of health. Inadequate education leads to reduced income and knowledge about the threat of malnutrition.

3. R

ELATIONSHIP BETWEEN FINANCIAL SERVICES AND HEALTH

(9)

9

theoretical framework is developed, it discusses the possible relationships between financial services and the determinants of health and malnutrition.

When looking at the determinants of health, discussed in the previous section, it becomes clearer how finance can affect health through its main determinants. They are displayed in table 3.1 in which all of the underlined elements are elements possibly affected by finance. Individual characteristics are not mentioned in the table, as these are not affected by financial services. Table 3.2 summarizes the causes of malnutrition and underlines possible channels through which finance can have an effect on nutrition, also underlined. Factors not underlined are factors on which it is less likely that finance could affect them, because they are determined primarily by other aspects.

Health behavior Consumption behavior Other behavior Access to health services Availability Affordability Social environment Education Social-Economic Status Physical environment Provision of water Sanitation Mosquito nets Housing

Table 3.1 Determinants of health influenced by financial service access or availability

Direct causes Underlying causes Basic causes

Food availability Inadequate access to food Inadequate education

Disease Poor health services &

environment

Resources and control Quality Maternal & Child Care Inadequate maternal and

child care

Political/Economic factors

Table 3.2 Determinants of malnutrition influenced by financial service access or availability

The determinants of health and nutrition are almost alike, therefore I merge them and categorize them in the following categories, explained in the following sections:

- Consumption behavior: is about (choices in) food consumption, limited by the amount of food and money available.

(10)

10

- Physical environment: covers elements such as the quality of housing, sanitation and water supply.

- Socio-economic factors: examples are income, education of both the individual as parental factors.

Regarding all of the relationships, it is expected that the effect of finance on health is concave, as suggested by Deaton (2003). This implies that at lower levels of financial access or use the effect of increased access of use is larger than at high levels of financial access. As with income, a certain amount is enough to cover basic survival and health expenditures. After that, increases in income do not directly lead to improvements in health. This will be dealt with by using logarithmic transformations of the explanatory variables of interest.

3.1 C

ONSUMPTION BEHAVIOR

The access to food is constrained by the availability of money and food in the area. Farmers receiving the revenue of their harvest only one or two times a year need to save in order to spread their consumption pattern. In times of bad harvest, or in case of a weak labor market, people also need to be able to afford food. Especially as spreading consumption is very important for growth and to maintain good health. The ability to spread consumption, especially in times of crises is important, as a short term shock can have devastating effects in poor households, the impact can hurt several generations (Angelini & Mierau, 2012). This aspect can affect malnutrition and through that the level of health.

3.2 A

FFORDABILITY OF HEALTH SERVICES

(11)

11

child labor, worsening their health. These parents could also decide to work more or in worse jobs in these situations (Beck, Demirguc-Kunt, & Honohan, 2009; Karlan & Morduch, 2009). However, there is no empirical evidence of a lower burden of health costs when the use of financial services increases (A. Banerjee, Duflo, & Kinnan, 2009). In a study on slums in Hyderabad, India, there are no spill-over effects on health of increased access to or use of finance. Only households with a low intensity to become entrepreneurs do increase consumption in this study, but health expenditure does not change significantly.

3.3 P

HYSICAL ENVIRONMENT

Financial services enable people to acquire education, housing or sanitation. These investments positively affect the level of health and welfare. Investment in education can contribute to long term additional income, better hygiene and better nutrition (in case of school meals).

Housing can prevent illnesses due to bad weather, provide a place of rest and enables people to store something at their own place (Deaton, 2003). Next to housing, customs within houses are important. People subject to conditions such as cooking within houses without proper ventilation often suffer from asthma or other lung illnesses, as they inhale the toxic air pollutants created by the stoves (Skolnik, 2008).

Additionally, these services can support people in investing in preventive matters such as sanitation, insecticide treated bednets against malaria or immunization against measles, DTP or malaria. Improper sanitation and contaminated water supply in several developing countries cause the spread of several illnesses (Skolnik, 2008).

3.4 S

OCIO

-

ECONOMIC FACTORS

There are several political and socio-economic factors which are influenced by finance and have a positive impact on health (Demirguc-Kunt & Levine, 2009). To a certain extent health is determined by income, which can be affected by finance (Beck et al., 2004). It is empirically shown that higher socio-economic status of the parents results in better health and that infant mortality rates are strongly dependent on income of the parents (Angelini & Mierau, 2012; Shandra, Shandra, & London, 2012).

(12)

12

on health improves and health behavior might improve as well (Angelini & Mierau, 2012). Education, especially female education has a high impact on infant mortality. It improves use of health facilities and improved access to information about nutrition, birth spacing, reproductive health and immunizations (Shandra et al., 2012).

Other possible indirect effects are the creation of jobs because of investment through available capital, which leads to increases economic growth and possibly in wages (Beck et al., 2004; Deaton, 2003). These higher wages make health costs relatively lower (Karlan & Morduch, 2009).

3.5 H

YPOTHESES

Based on the theoretical framework above several hypotheses can be derived regarding the relationship between finance and health. I expect that financial services improve health and reduce malnutrition. Regarding the determinants of health I expect that:

More use of financial services leads to better health (1)

Because of all of the above mentioned determinants. All of the mentioned determinants of health, consumption behavior, access to health services, the physical environment and socio-economic factors, are theoretically influenced by the use of financial products. The use of financial products increases the possibilities to cope with problems regarding health and therefore increase health.

Regarding nutrition, it is expected that all of the determinants above reduce the prevalence of malnutrition. Financial services create the possibility to affect the availability of food, support in the prevention of diseases and they shape the socio-economic environment. Regarding malnutrition I expect that:

More use of financial services reduces the prevalence of malnutrition (2)

However, regarding both hypotheses there are several mitigating factors. These factors might alter the impact of the financial services. Additionally, there are possibilities that the causality between finance and health and malnutrition is the reverse of the theory presented in this section. Both the mitigating factors and reverse causality are discussed in the following paragraphs.

(13)

13

3.6 M

ITIGATING EFFECTS

Most important mitigating effects are first, the income differences between countries and second, the quality of the health system.

Regarding the first effect, the relationship between health and finance can be different for developing and developed countries. According to World Bank definition developing countries are the countries in the low- and middle-income classes. Therefore this paper will check for these differences, by doing analysis on development countries separately in the robustness analysis and by adding the logarithm of GDP per capita as a control variable. (Deaton, 2003).

Regarding the second effect, the effect of health system and health insurance in one country might differ to the effects in another country. It depends on the quality of the health care system, availability of basic health services and insurances (especially for the poor) as well as the type of disease and knowledge about diseases and possible treatments. Better health care systems lead to a healthier population, which would alter the effects for countries with different types of health care systems. The higher the GDP per capita, the more money is publicly spent per person on health.2_{In most high-income}

economies this is approximately 10% of GDP, while low-income countries spent on average 3-6% of GDP (Skolnik, 2008). The public health expenditures as a percentage of GDP will be included as a control variable in the regressions.

Additionally, I will control for education and the percentage of the population living in rural areas. Both of these variables can be closely related to both finance and health. Education can improve health and the use of financial services (Skolnik, 2008; Demirguc-Kunt, 2012), while people in rural areas often suffer from worse health and do use financial services less often (Skolnik, 2008; Demirguc-Kunt, 2012).

3.7 R

EVERSE CAUSALITY

Regarding the level of savings and credits and the relationship with mortality and malnutrition, it is hard to draw conclusions on causality. It could be very well possible that healthy people earn more, have less expenditure on hospitals and other kinds of medical treatment and thus have more possibilities to save (Karlan & Morduch, 2009).

(14)

14

Another matter is the fact that higher education can lead to a higher reporting of illnesses, leading to an increase in reported morbidity, while actual morbidity has not been increasing. In certain culture specific diseases were not recognized as illnesses, but rather as regular phenomena and therefore not registered as illnesses (Shandra et al., 2012). Additionally, it is proven that education improves health (Cornelius, 2008), however better health, especially deworming, decreases school absenteeism as well. This latter could lead to higher financial literacy and higher use of financial services (Miguel & Kremer, 2004; Banerjee & Duflo, 2011).

When looking at the use of financial services, it could be that there is reverse causality. The use of services could increase because of a higher rate of disease prevalence. Because of the ambiguity of use of services, I will use different measures of use of financial products in order to prevent these endogeneity problems.

4. M

ETHODOLOGY

This section presents the methodology to test the hypotheses presented in the previous section. The data used in this paper will consist of country-level variables reflecting average health and average use of financial services in these countries for about 100 countries, covering the period 2002 to 2010. I will make use of fixed effect panel estimation on country-level data3_{with cluster-robust standard errors in order to test}

the hypotheses.

The advantage of fixed effects models is that it allows to estimate more complicated and realistic models. The availability of multiple observations per country makes it possible to control for country-specific time-invariant omitted variables. It assumes that there is an country-specific error term that is correlated with the independent variables. This error is removed by demeaning the data. It is probably the case that there are country-specific time-invariant variables which are omitted in this study because of the lack of available data. As I expect that these country-specific effects are correlated with the other dependent variables it is not valid to make use of a random effects model.

3_{Country-level studies have the advantage of higher data availability, the possibility to analyze general}

(15)

15

robust standard errors are used to deal with possible violations of the assumption that

eit is i.i.d. (independently and identically distributed). By using clustered standard errors I assume the generalization that the standard errors are correlated within i but not between groups (Verbeek, 2012).

The models are defined as follows:

̅̅̅̅̅̅̅ ( ̅̅̅̅̅̅̅̅̅̅̅) ( ̅) ( ̅) (1)

̅̅̅̅̅̅ ( ̅̅̅̅̅̅̅̅̅̅̅) ( ̅) ( ̅ ) (2)

Both equation 1 and 2 show that the regressions use deviations from individual means. Through this manipulation, the country specific effects αi are removed from the regressions, and the residual remaining is only µ. β1 and β2 are the coefficients estimated by the fixed effect model. The variables with bars imply the country specific means of the variables.

All of the used variables will be discussed in more detail in the next section. Vector

morti,t contains the dependent variables related to mortality for country i in year t. Regarding the vector nutri,t there are several variables included measuring malnutrition. The vector financei,t includes the variables used to measure the availability and use of financial services. The vector Zi,t comprises the control variables suggested by the literature and discussed in the previous section and the vector π consists of the regression estimates for the control variables. Vector Zi,t contains indicators such as GDP per capita, educational attainment, the size of the rural population and public expenditure on health.

5. D

ATA DESCRIPTION AND DESCRIPTIVE STATISTICS

(16)

16

5.1 D

ATA DESCRIPTION

This paragraph first deals with the different dependent variables, those measuring health and malnutrition. The second part concerns the variables regarding finance. Finally, the control variables are explained in the third part. All of the variables used are summarized in table A2 in the appendix. In this table their sources and transformations are also given.

5.1.1 HEALTH AND MALNUTRITION

The difficulty in measuring health and malnutrition stems from the fact that both are quite broad. As discussed both concepts consist of different elements, all which could affect the level of health or malnutrition. In order to focus this study, for health I will include variables used in earlier literature, such as life expectancy and child mortality. These variables are discussed in related literature, such as studies on the effect of inequality and income on health (Deaton, 2003).4_{These first two variables will be used}

as dependent variables in equation 1.

Life expectancy is the average number of years someone is expected to live from birth in

that country. Although life expectancy has been increasing in the last thirty years, there are still significant differences between countries. The data is retrieved from the World Bank Development Indicators.

Child mortality measures the number of deaths before the age of five per 1,000 live

births. The measures are highly correlated with infant mortality, or the number of deaths before the age of one. The source of this variable is also the database of the World Bank Development indicators.

This study makes use of three variables measuring malnutrition: stunting, wasting and undernourishment. Stunting and wasting are measures of the prevalence of malnutrition for children. Undernourishment measures malnutrition in the entire population. These three variables will be used as dependent variables for equation 2.

4_{Other characteristics, such as prevalence of diseases or sanitation and availability of health services are}

(17)

17

Stunting is the most dangerous type of malnutrition. The too low height for age cannot

be reversed and is closely linked to lower levels of development of the cognitive and physical skills. It is expressed as the percentage of the children under 5 with a height for age lower than two standard deviations from the international reference height at that age (Skolnik, 2008). The source of this variable is the World Bank Health, Nutrition and Population Statistics database (WBHNPS).

Wasting is measured as the percentage of the children under 5 with a too low weight for

height. This variable is also retrieved from the WBHNPS. Although the consequences of wasting at young age are less severe, it is the cause of stunting and also contributes to worse educational attainment (Skolnik, 2008).

Undernourishment measures the percentage of the population who receive not enough

nutrients and proteins to stay healthy. These figures only available for the year 2008 as these are estimated once in every five or ten years from household-level studies by the WBHPNS.

5.1.2 FINANCE

In order to measure the availability and the use of services, different measures will be used as independent variables. There are different types of variables available regarding finance and I will focus on variables that are as close as possible to the measurement of use of financial services. The actual use of financial services is difficult to measure, as such as measure should include the number of users, the number of withdrawals and deposits and the corresponding amounts (Demirguc-Kunt, 2012). Although such data is not available, there are variables that touch these topics. I will use variables on geographical and demographical bank penetration as proxies for the availability of financial services. These variables provide a more clear insight in the distribution of the financial services, and are not limited to data only describing the amount of money deposited or the number of users. Additionally, I will use data on the number of depositors and borrowers as proxies for the use of financial services.5_{The paper covers}

the inclusion in terms of two major financial services: credits and savings. Insurances are ignored because of a lack of data (Demirguc-Kunt, 2012).

5_{The problem with the first two variables is that there is known measurement error. In some countries,}

(18)

18

All financial data is taken from the IMFs Financial Access Survey, which covers data on financial access between 2002 and 2010. It has been developed by several economists upon the framework of Beck et al. (2007). The variables included will be log-transformed in order to deal with the exponential relationship between finance and health (Deaton, 2003).

Geographical penetration measures the availability per 1,000 km2_{. There is data}

available on bank and Automated Teller Machine(ATM) penetration, which will both be used in this paper. Banks include the number of facilities in the country, where basic services regarding deposits and credits are delivered. Although the availability of these facilities is important, it could be that people are willing to travel, as long as there is the possibility to perform withdrawals. Therefore ATM penetration is used as an indicator.

Demographical penetration measures the availability per 100,000 adults. Also for this

indicator data is used for banks and ATM penetration. It extends the use of the previous indicator and deals with possibilities that the results are contaminated by large countries with smaller populations or the opposite.

Depositors is measured as the number of depositors at financial institutions per 1,000

adults. It is a direct measure of the use of financial deposit services. Savings and deposits available at these accounts enables people to receive and save money in order to deal with larger expenses and investments (Loayza, Schmidt-Hebbel, & Serven, 2000; Karlan & Morduch, 2009).

Borrowers is measured as the number of borrowers at financial institutions per 1,000

adults. Credits have long been regarded as the way to prosperity, as it was expected that the poor were too poor to save, but case study evidence showed that most of the poor do save (A. V. Banerjee & Duflo, 2007, 2011; Collins, Morduch, Rutherford, & Ruthven, 2009). Still, credits are an important source of capital in order to invest in business, education or health (Demirguc-Kunt, 2012).

(19)

19

5.1.3 CONTROL VARIABLES

As discussed in paragraph 3.6, several variables could affect health in a country, for as far as possible these variables are included in the analysis. Furthermore the panel-data analysis will incorporate time-invariant cross-country characteristics which cannot be controlled for. The vector of control variables consists of enrollment at primary school, GDP per capita at Purchasing Power Parity, the percentage of the population living in rural areas and public expenditure on health. These are included because they are closely related to the level of health in a country (Skolnik, 2008). All of the variables are retrieved from the World Development Indicators.

Enrollment of primary school is the percentage of the total cohort which is enrolled at

primary school. It is a measure of educational attainment, which can both affect the level of financial services used and the quality of health.

GDP per capita at purchasing power parity (PPP) is included to control for the possible

correlation between use of financial services, income and health. Income is a direct determinant of health, as it enables people to receive proper nutrition, treatment of illnesses and medication. Because of the nonlinear relationship between income and health, the log-transformation of the variable is used (Deaton, 2003).

The percentage of people living in rural areas is an important determinant of health as the availability of health services is often lower in rural areas, therefore these people are at risk. It is a measure of the population living in rural areas as a percentage of the total population (Skolnik, 2008).

The public expenditure on health is related to health as it is an important determinant of the level of public health services available. It measures the government expenditure on public health as a percentage of total GDP. According to literature, it is also nonlinearly related to health, therefore the logarithm of the variable is used in regressions (Skolnik, 2008).

5.2 D

ESCRIPTIVE STATISTICS

(20)

20

rates, but still the differences are remarkable. The data on stunting and wasting covers poorer countries primarily, however the differences are still relatively large, with minimums of 0.6% of the children under 5 being stunted and maximums of 59.3% of the children under 5 being stunted.

The statistics of the independent variables provide some interesting insights. ATM penetration is much higher than bank penetration, which could be expected as it is less costly to increase the number of ATMs than the number of banks. The number of depositors appears to be much higher than the number of borrowers in most countries and indication that more people make use of deposit accounts and possibly rather save than borrow. There are also some peculiarities. The maximum values for both the geographical bank and ATM penetration are very high. There are 25 countries with more than 1,000 banks per 1,000 km2_{, all of the datapoints are relatively small countries,}

which explains these extreme findings. Demographic penetration is more evenly spread, but is also skewed.

Description Years Obs Countries Mean Min Max

Life Expectancy 9 1761 207 68.465 41.489 83.159

Child Mortality 9 1728 194 47.666 1.900 219.600

Stunting 9 209 123 27.542 0.600 59.300

Wasting 9 203 121 6.923 0.300 26.000

Undernourishment 1 171 171 13.380 5.000 65.000

Geo. Bank Penetration 9 1003 151 93.793 0.010 5,992.900

Demo. Bank Penetration 9 881 133 22.877 0.339 590.206

Geo. ATM Penetration 9 894 138 198.729 0.000 23,156.030

Demo. ATM Penetration 9 802 122 72.442 0.000 1,009.320

Borrowers 9 462 82 216.294 0.291 2,753.560

Depositors 9 492 82 1,085.062 0.841 42,496.070

Education 9 1149 173 87.145 25,446.000 100.000

Income 9 1637 188 11,848.800 249.204 77,108.220

Rural 9 1890 214 43.317 0.000 91.220

Public Health Expenditure 9 1496 187 3.824 0.176 16.051

Table 5.1 Descriptive statistics of the variables measuring health, finance and the control variables

(21)

21

5.3 C

ORRELATIONS

It is interesting to see whether the mortality and malnutrition variables correlate. As table 5.2 shows, they actually do to a large extent. Life expectancy is negatively correlated with the other indicators, as it is the exact opposite of mortality and improved with better nutrition. Interestingly wasting is less strongly correlated with life expectancy. Possibly the consequences of wasting are often less severe on life expectancy, if compared to the consequences of stunting, it could possibly be that this shapes this primer difference. Regarding finance, there are some significant and large correlations displayed in table 5.3. Looking at this table, all of the access and use variables are highly correlated.

The underlying correlations between the control variables show that a lot of the development of a country impacts on different sides. Education and income appear highly correlated in table A3 in the appendix. The share of the population living in rural areas is significantly correlated to GDP as well, but it is not highly correlated to education, although the negative relationship was expected.

1 2 3 4 5 Life Expectancy 1.0000 Mortality under 5 -0.935 1.000 Stunting -0.696* 0.705* 1.000 Wasting -0.404* 0.513* 0.576* 1.000 Undernourishment -0.659* 0.630* 0.547* -0.010 1.000

Table 5.2 Correlations between health variables. (* = significant at 5% level)

1 2 3 4 5 6 Geo. Bank 1.000 Demo. Bank 0.744* 1.000 Geo. ATMs 0.892* 0.710* 1.000 Demo. ATMs 0.573* 0.770* 0.823* 1.000 Borrowers 0.717* 0.827* 0.845* 0.869* 1.000 Depositors 0.738* 0.796* 0.813* 0.805* 0.850* 1.000

Table 5.3 Correlations between finance variables. (* = significant at 5% level)

(22)

22

possible exponential relationship between finance and health. All the signs are as expected.

Lower mortality is associated with higher availability and use of financial services. Additionally, the prevalence of stunting and wasting is lower in areas where there are more financial services used and available. Not all of the different penetration measures and use measures are significantly correlated at high levels to malnutrition, with the exception of undernourishment, which is correlated to most of the finance variables. In table A4 in the appendix, the correlations between the dependent and independent with the control variables are presented, which are all as expected.

Logs Life

expectancy Mortality under 5 Stunting Wasting Undernourishment Geo. Bank 0.6470* -0.6318* -0.3584* -0.1325 -0.4338* Demo. Bank 0.7731* -0.7778* -0.6492* -0.4546* -0.6405* Geo. ATMs 0.6937* -0.7338* -0.6170* -0.3676* -0.6001* Demo. ATMs 0.7317* -0.8003* -0.7655* -0.5248* -0.7398* Borrowers 0.7460* -0.8332* -0.4742* -0.3127* -0.6967* Depositors 0.7093* -0.7862* -0.6018* -0.3908* -0.6026*

Table 5.4 Correlations between health variables and logarithm of finance variables. (*=significant at 5% level)

6. R

ESULTS

This section presents the results of the fixed effects models per hypotheses. The first paragraph will test the first hypotheses, while the second section test the second. The third paragraph will discuss the robustness of the findings and the fourth contains a discussion of the findings.

6.1 H

YPOTHESIS

1:

F

INANCE AND

M

ORTALITY

The results of the two dependent variables measuring health are presented in table 6.1 to 6.4. Both are separately discussed in the following subparagraphs.

6.1.1 LIFE EXPECTANCY

(23)

23

both are significantly related at the 1% level. After adding the control variables, the significance levels and magnitudes of the coefficients change. Bank penetration is no longer significantly related at the 10% level, while ATM penetration remains significantly related at the 5% level. The coefficient of 0.284 implies that with a 10% increase of the geographical ATM penetration, life expectancy will increase with 0.0270 years. The impact of a 10% change in demographical ATM penetration has an impact of 0.026 years on life expectancy.

The number of borrowers remains significantly related at the 10% levels, but the number of depositors doesn’t. The magnitudes of all of the betas decreases after adding the control variables. All of the control variables are correlated to life expectancy as expected in table 6.1 and 6.2.

The regressions in table 6.1 and 6.2 show that especially better availability is related to

better life expectancy, and thus better health. The evidence of the relationship between

the variables on use of financial services and health is mixed, but it appears that the number of borrowers has a positive impact on life expectancy.

6.1.2 CHILD MORTALITY

Table 6.3 and 6.4 show the results for child mortality. Both demographical and geographical ATM penetration are negative and significant. A 10% increase in ATM penetration leads to a 0.21 point lower child mortality.

None of the use of financial services is significantly correlated to child mortality after adding the control variables. The results show that better availability of financial services

is related to lower mortality, but that the evidence concerning use of financial services is

again mixed.

(24)

24

Table 6.1 Life Expectancy and availability of financial services

All of the regressions are fixed effect models with a constant included(not reported). The standard errors reported under the estimates are heteroskedasticity consistent. The regression estimated in column 1, 4, 7 and 10 is the dependent variable regressed on the indicated financial services variable. Columns 2, 5, 7 and 11 add the control variables log of GDP and log of public health expenditure as a percentage of GDP. The regressions in column 3, 6, 8 and 12 include all of the control variables, the variables enrollment at primary school and the percentage of the population living in rural areas are added.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) VARIABLES life_exp life_exp life_exp life_exp life_exp life_exp life_exp life_exp life_exp life_exp life_exp life_exp

(25)

25

Table 6.2 Life Expectancy and use of financial services

All of the regressions are fixed effect models with a constant included(not reported). The standard errors reported under the estimates are heteroskedasticity consistent. The regression estimated in column 1 and 4 is the dependent variable regressed on the indicated financial services variable. Columns 2 and 5 add the control variables log of GDP and log of public health expenditure as a percentage of GDP. The regressions in column 3 and 6 include all of the control variables, the variables enrollment at primary school and the percentage of the population living in rural areas are added.

(1) (2) (3) (4) (5) (6) VARIABLES life_exp life_exp life_exp life_exp life_exp life_exp

Borrowers 0.874*** 0.570*** 0.217* (0.135) (0.109) (0.124) Depositors _0.782*** _0.436** _0.0950 (0.253) (0.191) (0.0906) GDP 2.226*** 1.856*** 2.872*** 2.292*** (0.530) (0.644) (0.683) (0.520) Pub Health Exp 0.717* 0.294 1.085*** 0.251

(26)

26

Table 6.3 Child mortality and availability of financial services

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) VARIABLES mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive

(27)

27

Table 6.4 Child mortality and use of financial services

(1) (2) (3) (4) (5) (6) VARIABLES mortundfive mortundfive mortundfive mortundfive mortundfive mortundfive

Borrowers -5.975*** -3.778*** -2.400 (1.265) (1.201) (1.760) Depositors _-5.073*** _-2.731* _-1.292 (1.905) (1.550) (1.547) GDP -18.45*** -15.05** -22.48*** -18.81*** (4.140) (7.240) (4.457) (5.142) Pub Health Exp -2.695 0.293 -5.241* 0.658

(28)

28

6.2 H

YPOTHESIS

2:

F

INANCE AND

M

ALNUTRITION

The results regarding hypothesis 2 are presented in table 6.5 to 6.10. The three dependent variables will also be discussed separately in subparagraphs.

6.2.1 STUNTING

Regarding stunting, table 6.5 and 6.6 provide the results for respectively availability and use of financial services. The results are slightly different from the mortality indicators. Again, the penetration of ATMs is significantly and negatively related to the prevalence of stunting, but this time also the geographical penetration of banks is significant at the 5% level and negatively related to stunting, or too low height for age. A 1% increase in bank penetration is associated with a 0.38 percentage point reduction in the prevalence of stunting under children. The finding that the availability of financial services is

associated with lower prevalence of stunting is in support of the second hypothesis.

The use of financial services is correlated with stunting to some extent. The number of borrowers appears to be negatively related to stunting without control variables, but with control variables it is positively and significantly related to the prevalence to stunting, which is opposite to what is expected. The number of depositors is significantly and negatively correlated to stunting, which is in support of the second hypothesis. However, looking at the R2 _{and the number of observations these two models are}

over-fitted.

(29)

29

6.2.2 WASTING

The results for wasting, summarized in table 6.7 and 6.8, are different from stunting to some extent. Before the inclusion of control variables none of the relationships is significant and after the inclusion of control variables both demographic bank and ATM penetration are significant but only respectively at the 5% and 10% level. An increase of 10% in the demographic penetration of banks is associated with a decrease of 0.55 percentage points in the prevalence of wasting, or too low weight for height.

It appears that the number of depositors is significantly and negatively correlated to the prevalence of wasting. The last model is over-fitted, as the R2 _{is too high, but regressions}

1 and 2 in table 6.8 show a reasonable R2_{with significant results. These findings are in}

support of hypothesis 2, that financial services are associated with a lower prevalence of

malnutrition, unfortunately not controlling for education and the rural population.

Regarding availability and use of financial services, the control variables become significant in regression 3 and 6 of table 6.8, but are not significant in table 6.7. Again in both of the tables the signs and coefficients of the control variables are different from what was expected, but also these models are over-fitted and should not be interpreted.

6.2.3 UNDERNOURISHMENT

Regarding undernourishment, OLS is used, as there is only data available for 2008. Tables 6.9 and 6.10 contain the results from the regression analysis. The geographical penetration of banks is significantly and negatively related to undernourishment at the 5% level. The effect of geographical ATM penetration is significant at the 1% level and has a larger estimate than the bank penetration variable. Demographical ATM penetration has the largest effect, at a 1% level a 10% increase reduces undernourishment with 0.273 percent points. Regarding variables measuring use of financial services, both borrowers and depositors remain significant with the inclusion of control variables. At the 10% level a 1% increase in the number of borrowers is associated with a 0.248 decrease in the prevalence of undernourishment. At the 1% level, depositors have approximately the same effect. Interestingly, almost none of the control variables in these regressions is significant (column 2, 3, 5 and 6 in table 6.10) These findings are in support of hypotheses 2, both the availability and the use of

(30)

30

Table 6.5 Stunting and availability of financial services

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) VARIABLES stunting stunting stunting stunting stunting stunting stunting stunting stunting stunting stunting stunting

Geo. Bank -2.631* -2.341 -4.002** (1.359) (2.156) (1.841) Demo. Bank -2.554* -1.629 -2.251 (1.438) (2.269) (2.148) Geo. ATM -1.589*** -3.494** -4.180*** (0.533) (1.529) (1.451) Demo ATM -1.986*** -3.917** -4.507*** (0.491) (1.576) (0.966) GDP 2.200 14.50 -3.255 -2.615 15.29 13.62 11.24 1.879 (14.93) (16.43) (15.52) (17.44) (13.01) (14.67) (11.89) (2.979) Pub Health Exp 5.461* 2.479 5.508* 3.522* 4.005 -0.00856 6.025 2.965

(31)

31

Table 6.6 Stunting and use of financial services

(1) (2) (3) (4) (5) (6) VARIABLES stunting stunting stunting stunting stunting stunting

Borrowers -2.487* 1.544 14.15*** (1.287) (3.114) (0.660) Depositors _-1.867 _0.0577 _-4.982*** (1.276) (1.924) (0.647) GDP -23.65* -52.97*** -8.996 8.362* (13.68) (3.191) (12.69) (4.448) Pub Health Exp 4.839 -19.14*** -4.817 -18.37***

(32)

32

Table 6.7 Wasting and availability of financial services

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) VARIABLES wasting wasting wasting wasting wasting wasting wasting wasting wasting wasting wasting wasting

Geo. Bank -1.047 -3.527 -4.181 (1.288) (2.642) (3.029) Demo. Bank -1.085 -4.017 -5.836** (1.470) (2.626) (2.587) Geo. ATM 0.0921 -0.114 -0.347 (0.338) (0.624) (0.901) Demo ATM 0.166 -0.337 -1.328* (0.375) (0.781) (0.749) GDP 15.70 22.87 18.46** 36.56* 2.444 -6.805 5.353 2.911 (9.481) (19.78) (8.777) (18.45) (4.917) (8.783) (6.113) (3.231) Pub Health Exp -0.0164 -0.0635 0.113 -0.0982 -0.437 -4.400** -0.839 -8.688***

(33)

33

Table 6.8 Wasting and use of financial services

(1) (2) (3) (4) (5) (6) VARIABLES wasting wasting wasting wasting wasting wasting

Borrowers -0.0468 -0.920 -0.236 (0.591) (1.178) (0.354) Depositors _-5.118* _-9.089*** _-12.88*** (2.719) (0.602) (0.238) GDP 8.867 -0.516 21.46*** 29.23*** (6.415) (1.672) (3.923) (1.654) Pub Health Exp -0.520 -10.31*** 1.618 -11.95***

(34)

34

Table 6.9 Undernourishment and availability of financial services

All of the regressions are estimated using ordinary least square with a constant included(not reported). The standard errors reported under the estimates are heteroskedasticity consistent. The regression estimated in column 1, 4, 7 and 10 is the dependent variable regressed on the indicated financial services variable. Columns 2, 5, 7 and 11 add the control variables log of GDP and log of public health expenditure as a percentage of GDP. The regressions in column 3, 6, 8 and 12 include all of the control variables, the variables enrollment at primary school and the percentage of the population living in rural areas are added.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) VARIABLES undern. undern. undern. undern. undern. undern. undern. undern. undern. undern. undern. undern.

Geo. Bank -2.214*** -0.587 -0.856** (0.392) (0.379) (0.415) Demo. Bank -5.915*** -2.221** -1.829 (0.823) (0.956) (1.126) Geo. ATM -2.556*** -0.855** -1.162*** (0.360) (0.372) (0.417) Demo ATM -4.514*** -2.426*** -2.867*** (0.460) (0.758) (0.834) GDP -6.041*** -3.421*** -5.195*** -3.036* -5.544*** -2.681** -4.043*** -0.969 (0.874) (1.204) (1.138) (1.587) (0.846) (1.243) (1.207) (1.850) Pub Health Exp -0.756 -0.303 -0.315 -0.603 0.292 0.333 1.089 0.439

(35)

35

Table 6.10 Undernourishment and use of financial services

All of the regressions are estimated using ordinary least squares with a constant included(not reported). The standard errors reported under the estimates are heteroskedasticity consistent. The regression estimated in column 1 and 4 is the dependent variable regressed on the indicated financial services variable. Columns 2 and 5 add the control variables log of GDP and log of public health expenditure as a percentage of GDP. The regressions in column 3 and 6 include all of the control variables, the variables enrollment at primary school and the percentage of the population living in rural areas are added.

(1) (2) (3) (4) (5) (6) VARIABLES undern. undern. undern. undern. undern. undern.

Borrowers -5.181*** -2.043* -2.602* (0.773) (1.065) (1.467) Depositors -5.040*** -1.705** -2.540*** (0.794) (0.841) (0.907) GDP -6.454*** -3.421 -7.257*** -3.598 (1.886) (2.990) (1.445) (2.162) Pub Health Exp 0.395 -0.328 0.947 1.045

(36)

36

6.3 R

OBUSTNESS

In order to determine the robustness of the results presented in the previous paragprah, several robustness checks will be conducted. I will control for the possibility that the results will be different for developing countries by performing separate regressions on these countries. Additionally I will check for the possibility of omitted variables, measurement error and reverse causality by instrumental regression. Finally I discuss the possibility that the changes in significance levels are driven by the reduction in the number of observation or because of multicollinearity.

First, I conduct the same regressions on developing countries only. These countries are the countries defined by the World Bank as low- or medium-income countries. Most of the results are very close to the results already shown. The results on the impact of availability of financial services are supported by the findings, the others did not have enough observations to analyze. To save space, the tables are not printed in this paper. A two-stage panel least square is used in order to deal with possible reversed causality, omitted variables and measurement error. The financial reform index of Abiad et al. (2009) is used as the instrument for financial services. This index measures the degree of liberalization within a country. Liberalization could lead to higher financial development and availability and is not likely to be related to health, therefore it is a suitable instrument (Verbeek, 2012). Other instruments often used in the finance literature are the latitude of the government city and the historical origin of the law system (Beck et al., 2004). Theoretically these variables are possibly linked to the health indicators, making them less suitable as instrumental variables in this study. The law system could be developed together with the health system, and laws cover topics regarding the health system, creating a relationship between the two variables. The latitude of a country is often used in health studies in order to measure the prevalence of certain diseases, such as malaria. Because of that, it is likely that the latitude is correlated with other latent variables measuring health.

(37)

37

variables, all of the relationships which were significant in table 6.1 to 6.10 remain significant.

Unfortunately, in the first stage regressions the measure of financial reform is not significantly related to any of the measures when the other control variables are included as well. Without control variables, financial reforms are significantly and positively related at the 1% level to demographic and geographic ATM penetration. At the 5% level, financial reform is also significantly related to geographic bank penetration. All of the F-values of the first-stage regressions are below 10 indicating that the instrument is not strong. The same result occurs when two-stage-least-squares is used with country dummy’s rather than the fixed effects estimation.

(38)

38

to identify the individual impact of set of variables, due to high correlations. This problem causes unreliable regression estimates. This relationship is not restricted to two variables only, but can involve more or even all regressors (Verbeek, 2012). Although table A4 in the appendix does not provide direct evidence of collinearity, multicollinearity can still exist if multiple independent variables are included in a regression. Multicollinearity can often be detected by high standard errors, strange signs and in cases with high R2_{, while none of the variables appears highly significant.}

Additionally it can be determined by using auxiliary regressions with the independent variables to determine the variance inflation factor (VIF)6_{(Verbeek, 2012). The}

calculated VIFs for the independent variables are lower than 10 and thus do not provide any sign of multicollinearity when using all the separate groups of observations on which the regressions were conducted in table 6.1 to 6.10 or all of the observations available. The VIFs of GDP are sometimes larger than 4, which is another threshold used by other researchers. Possibly GDP is highly correlated to all the other control variables. Conducting all of the regressions without GDP provides interesting results. All the estimates for the different financial services become significant at the 10% level when regressed on life expectancy, child mortality and undernourishment and the signs are correct. Furthermore, most of the control variables are significantly related at the 10% level to the dependent variable and have the right sign and in a majority of the regressions the R2_{reduces with only 0.05 percentage points. For wasting and stunting}

the results remain ambiguous, probably due to the low level of observations. Unfortunately, the interpretation of the estimates becomes ambiguous, because of the omitted variable GDP. It is no longer clear whether the estimate of interest is measuring the effect of financial services or of income on health and malnutrition. This can imply that the results do not actually show a relationship between financial services and health and malnutrition, but between income and health and malnutrition. Fortunately, table 6.1 to 6.10 provide enough evidence to state that the effect of finance on health and malnutrition is significant after controlling for income.

6 The variance inflation factor is given by: ( )

. It indicates the factor by which the variance of

bk is inflated. A VIF > 10 is an indication of possible multicollinearity, but does not imply that it is

(39)

39

6.4 D

ISCUSSION

The first results of the analysis on the relationship between finance and mortality are positive and in support of the developed framework. With control variables included, geographical and demographical ATM penetration are significantly related at the 1% with life expectancy and child mortality. It is interesting to see that bank penetration is not significantly correlated to these variables, which can be caused by multicollinearity. Additionally, the variables measuring the use of financial services do not remain significant after the inclusion of control variables in the regressions on life expectancy and child mortality. These ambiguous findings are probably due to measurement error or multicollinearity. The number of borrowers and depositors does not actual indicate the real use of these services, unfortunately this could not be tested. Summarizing, it appears that financial services have the expected effect on mortality.

Regarding malnutrition, the results are not as strong as expected. What is clear is that the geographical bank penetration and both the ATM penetration variables appear significantly correlated to stunting and undernourishment of the population. Regarding wasting, bank penetration appears to have a significant effect, while ATM penetration is not significant. The number of depositors is significant in all of the nutritional relationships. However, most of the evidence is subject to the low level of observations and possible over-fitted models. Therefore the evidence is too weak to conclude that the relationship between use of financial services and stunting and wasting is positive and significant. Regarding undernourishment it appears that both availability and use of financial services are significantly related.

(40)

40

7. C

ONCLUSION

This paper started by discussing health, malnutrition and their determinants in order to determine whether there are possible connections with financial services. After establishing a theoretical framework, a panel data analysis was conducted in order to find evidence for the framework. The empirics show that increased availability of financial services is associated with lower mortality and improved nutrition. It is shown that geographical and demographical ATM penetration has an economically significant impact on life experience, mortality rates and malnutrition. The wide availability of these services possibly is the explanation for this finding.

However, regarding the stunting and wasting, two measures of nutrition, the evidence is rather weak and possibly subject to measurement error or outliers. Some of the included variables show up with unexpected signs or remain insignificant.

Additionally, it remains hard to test and assess the real impact of the use of financial variables on health. Although the number of users is an indicator of use, it does not measure the number of withdrawals nor the amount used. As it appears the use of financial services, as measured by the number of depositors and borrowers does not have a significant impact on mortality, which can be due to measurement error.

This paper does not cover the effects of all of the other services offered in countries, such as through non-governmental organizations, microfinance banks and village level saving groups such as ROSCA’s. Kendall et al. (2010) suggests the separation between the effects of different institutions when measuring access to and use of financial services. Neither does this paper determine the actual impact of the use of financial services on household or individual health. Regarding health and malnutrition, only a small part is covered, topics such as diseases and obesity are missing in the analysis, but there are possible interesting relationships. Additionally, there could be omitted variables or severe measurement error, which unfortunately could not be dealt with in this paper. The lack of data available and a proper instrument prevented me from determining the real causality between financial services and health.