THE IMPACT OF FINANCIAL DEVELOPMENT AND FINANCIAL
LIBERALIZATION ON THE POVERTY GAP
by
Fabian Weber
Student number: S-2353830 Email: f.weber.1@student.rug.nl
MSc International Economics & Business 2018
University of Groningen
Faculty of Economics and Business
Abstract
A sample of 123 countries for the period 1975 to 2015 have been analyzed with a fixed effects model. Our results suggest that financial development is poverty decreasing until a certain threshold is passed, after which it will increase poverty. A linear negative relationship has been found between financial liberalization and poverty. In addition, a significant positive relationship has been found between inflation, crisis, and the interaction effect between financial development and financial liberalization on the one hand on poverty on the other hand. Therefore, our study encourages financial deepening by making credit more widely available, but financial stability must not be forgotten.
Table of Contents
1. Introduction ... 1
2. Theoretical Framework ... 4
2.1 Financial Development and Poverty ...5
2.2 Financial Liberalization and Poverty ...7
2.3 Financial Development and Financial Liberalization on Poverty ...8
2.4 Poverty ...8
2.5 Control Variables ...9
3. Methodology ... 10
3.1 Data ...10
3.2 Variables...11
3.3 Panel Data Implications ...15
3.4 Empirical Model ...16
4. Results... 17
4.1 Main Results Equation (1) ...18
4.2 Main Results Equation (2) ...20
4.3 Robustness Check and Sensitivity Analysis ...24
5. Conclusion ... 28
References ... 32
Appendix A - Description and Sources of the Variables ... 35
Appendix B – Countries Included ... 36
1. Introduction
In the past few decades, rapid financial developments have changed the financial landscape of the world dramatically. The ratio of domestic credit to the private sector to GDP, a frequently used proxy to measure financial development, increased from 50% in 1950 to roughly 126% in 2017 (World Bank, 2019a). Despite the research that has been done on the effect of financial development on economic growth (Beck, Levine & Loayza, 2000; Beck & Levine, 2004; Arcand, Berkes & Panizza, 2015), little attention has been paid to the effect of financial development on the position of the poor (Jeanneney & Kpodar, 2008). Therefore, it could be possible that the poor will either disproportionally benefit from financial development or will be hurt by financial development.
In addition, it is still unclear as to how poverty can best be measured. For example, Sen (1985) uses a broader definition which also includes non-monetary measures of poverty. This multidimensional view describes poverty as a lack of education, utilities, health-care and security or as not being able to participate in social activities. However, very little data is available about non-monetary poverty since it is hard to measure. More in-depth surveys are necessary in order to gain more insight in the non-monetary measures of poverty (World Bank, 2018). Therefore, most research tend to quantify poverty based on monetary indicators such as income or consumption (Sehrawat & Giri, 2005; Donou-Adonsou & Sylwester, 2016). This research will use the poverty gap as indicator for poverty, this is the average shortfall of the population based on the poverty line (World Bank, 2018).
Recent research suggests that financial development will improve the poor’s access to financial services (Kaidi, Mensi & Amor, 2018) and therefore, is negatively correlated with the poverty gap. Also, Galor and Zeira (1993) and Galor and Moav (2004) claim that financial development will mitigate credit constraints, which makes it possible for the low-income group to accumulate physical capital and invest in human capital.
However, other researchers argue that the poor still cannot access the financial market because of a lack of credit history, information and transaction costs, and other set-up costs a poor household cannot afford. Therefore, they claim that financial development will be only beneficial for the people who already have access to the financial sector and will not benefit the poor. This would therefore result in an increase in income inequality (Banerjee & Newman, 1993; Aghion & Bolton, 1997).
healthier. They claim that after a certain level of financial development is reached, it becomes easier for the poor to get the credit needed to invest in physical and human capital. This is visualized by an inverted U-curve where financial development, at some point, will be beneficial for the poor.
Thus far, this section has shown that the current literature is not clear about to which extent financial development influences the position of the poor. An interesting question to ask is: if financial development does affect the poor, via which channels would the poor be affected? Jalilian and Kirkpatrick (2005) and Beck, Demirgüc-Kunt and Levine (2007) show that financial development may affect the poor indirectly through the income distribution channel and the economic growth channel. The effect of these two channels on poverty are influenced by the financial and monetary policies of an economy.
The main goal of these policies is to identify and mitigate risks in order to stabilize the financial market (Altunbas, Binici & Gambacorta, 2018). One of these policies concerns financial liberalization, which refers to a reduced role of the government relative to the role of the financial markets (Abiad, Oomes & Ueda, 2008). Financial liberalization regained a lot of attention since the financial crisis in 2007, but surprisingly very little attention has been paid to the influence of financial liberalization on the poverty gap (Beck et al., 2007; Arestis & Caner, 2010).
To date, the limited literature available offers contradictory findings about the effect of financial liberalization on poverty. Furceri and Loungani (2015) examine the effect of capital account openness on income inequality and find that financial liberalization will hurt the poor and lead to more income inequality. Also, Demirgüc-Kunt and Detragiache (2000) show that financial liberalization can be bad for the poor. In their research, they report of a significant relationship between financial liberalization and banking crises: financial liberalization can lead to over-borrowing and over-investing that can result in a banking crisis. A banking crisis can be particularly bad for the poor since they do not have enough capital to cope with the economic downfall. Thereby, the poor are more likely to get hit by the contractionary fiscal policy that will follow after the crisis (Arestis & Caner, 2004).
In contrast, Beck, Levine, and Levkov (2010) claim that financial restrictions will create and protect local monopolies that will decrease competition and raise fees that mainly hurt the poor. Therefore, they suggest that financial liberalization will be beneficial for the poor.
al., 2008; Furceri and Loungani, 2015). Therefore, little research has been done to the combined effect of financial development and financial liberalization on poverty (De Haan & Sturm, 2017). This study aims to contribute to the growing area of research on the relationship between finance and poverty, by providing an answer on the following research question:
What is the combined effect of financial development and financial liberalization on the poverty gap?
Understanding the link between finance and poverty can help the government to create an environment that is beneficial for the poor and ultimately, eradicate extreme levels of poverty.
To understand the relation between finance and poverty, it is necessary to add some control variables. Several studies suggest that financial development and financial liberalization can affect poverty directly or indirectly via the income distribution and crisis channels (Dollar & Kraay, 2002; Arestis & Caner, 2004). Therefore, a crisis dummy will be added to the model as control variable.
As already mentioned in this chapter, an environment with low institutional quality will disproportionally benefit the established interests and other insiders with private access to finance, to gain advantage of financial development which will hurt the poor. On the other hand, high-quality institutions will help the poor to invest in human and physical capital which is needed to escape from poverty (Law, Tan & Azman-Saini, 2014; Rashid & Intartaglia, 2016). For this reason, a variable measuring institutional quality will be included.
Moreover, the Gini coefficient will be added to control for the distributional effects in an economy. As will be explained in further detail in Section 3, this coefficient measures the income distribution in an economy. However, a distinction has to be made between poverty and income inequality, since the second does not say anything about the absolute income of the poor. Therefore, it is possible that an increase in income inequality is accompanied by a decrease in poverty which means that the absolute income of the rich increased relatively more than the absolute income of the poor. Chapter 2 and 3 will give a further explanation of the control variables.
financial liberalization on poverty. This will fill the gap in the literature and contribute to the ongoing debate about eradicating extreme levels of poverty and inform policy makers about the effect of financial reforms on poverty. (ii) This study will explore whether the effect of finance on poverty is affected by institutional quality, as suggested by Rashid and Intartaglia (2016.), or income inequality as suggested by De Haan and Sturm (2017). (iii) Different variables of poverty will be used. This will make it possible to compare the results of the regressions with different proxies for poverty as the dependent variable and investigate the sensitivity of dependent variable selection.
The results of the regression provide mixed results. On the one hand, we found evidence for the poverty reducing effect of financial development and financial liberalization, but on the other hand the squared function of financial development and our interaction term turned out to be positive related with the poverty gap. These outcomes have two implications. First, there is a non-linear relation between financial development and the poverty gap. Second, the combined effect between financial development and financial liberalization increase the poverty gap. Therefore, one should be careful when implementing financial policies in order to reduce poverty.
The remaining part of the paper proceeds as follows. The next chapter will discuss the related literature in more detail. Chapter 3 is concerned with the methodology and data used for this study. The results will be presented and analyzed in chapter 4. Finally, chapter 5 offers policy implications and conclusions.
2. Theoretical Framework
2.1 Financial Development and Poverty
According to Levine (2005), the function of the financial system is to reduce the transaction, information and enforcement costs made to allocate financial products between space and time. The allocation of resources will be naturally influenced by the financial system, in order to mitigate market frictions (Merton & Bodie, 1995). Financial development happens when the financial markets, the intermediaries, or the instruments evolve to adapt to the market frictions and decrease the transaction, information and enforcement costs (Levine, 2005).
Despite the large number of published studies about the relation between financial development and poverty, there is still no consensus on how finance may affect the poor. Some researchers attempt to establish a direct effect (Naceur & Zhang, 2016), while others claim that finance only affects the poor indirectly (Beck et al., 2007).
For example, financial development can directly affect the people who live in poverty if it improves the affordability and accessibility of the financial system for the poor (Naceur & Zhang, 2016). Especially the access to credit- and insurance services will help the poor to invest in human capital, increase the productivity, and increase their incomes (Galor & Zeira,1993; World Bank, 2001; Galor & Moav, 2004).
However, there are also some researchers who argue that financial development will not benefit the poor at all. One of the main arguments is that the poor are still not able to participate in the financial system and therefore, will not benefit from growth of the financial system. Papers like those of Banerjee & Newman (1993) and Aghion & Bolton (1997) argue that financial development will only be beneficial for the insiders, which will increase income inequality. This view is supported by Arestis and Caner (2009), who suggest that financial development will expand the formal sector at the expense of the informal sector. This could potentially be bad for the poorer segment of the society, since the informal sector is their main source of income.
A widely used concept that highlights the beneficial effects of financial development, is the trickle-down effect. This effect states that financial development will indirectly improve the situation of the poor through an increase in the economic growth rate (Arestis & Caner, 2009; Kaidi et al., 2018). However, the effect of financial development on growth is twofold. Some researchers claim that financial development will foster growth by boosting the real economy (Jalilian & Kirkpatrick, 2002), while other researchers (Atkinson & Morelli, 2011; Loayza & Ranciére, 2006) claim that financial development will increase the likelihood of a crisis which will have adverse effects on the average growth of GDP. Therefore, the relationship between financial development and growth of GDP could be non-linear where financial development has a negative effect on growth after a certain level is reached.
If poverty is affected by financial development via the growth channel, it is also possible that the effect of financial development on poverty is non-linear in a way that financial development decreases poverty till a certain level. After this level is reached, financial development may create economic instability which will have negative influences for the poor. Therefore, the first hypothesis is:
H1: Financial development will be negatively related with the poverty gap up to a
certain threshold. After this threshold is reached, financial development will be positively related with the poverty gap.
Empirical studies have used different variables to measure financial development. The most widely used variables are private credit to GDP, broad money (M2) to GDP, and stock market capitalization to GDP (Sehrawat & Giri, 2005; De Haan & Sturm, 2017). The literature does not consider M2/GDP as an appropriate proxy for financial development because it is more a facilitator of exchange than a facilitator of investment. Moreover, especially in many developing countries, M2 is treasured or kept outside the country for transaction purposes and is not circulating in the real economy. Therefore, using M2/GDP as a proxy for financial development will cause a serious underestimation of financial development (Dal Colle, 2010).
The variable for financial development used in this research will be domestic credit to
the private sector divided by GDP. This refers to the credit which is transferred from savers to
2.2 Financial Liberalization and Poverty
McKinnon (1973) and Shaw (1973) were the first researchers who focused on the effects of financial restriction on economic growth, income inequality, and financial crises. They argued against high reserve requirements and also claimed that real interest rates should be determined by the market in order to create an efficient financial system. According to McKinnon (1973) and Shaw (1973), financial liberalization will facilitate efficient capital allocation, contribute to growth and therefore, decrease income inequality and poverty.
In the footsteps of McKinnon and Shaw, many researchers have investigated the effect of financial liberalization on income inequality, but only a few studies have investigated the relationship between financial liberalization and poverty (Beck et al., 2007). Arestis and Caner (2004) showed that the effect of financial liberalization on poverty is ambiguous because there are too many channels and the underlying mechanisms are not clear. For example, financial liberalization may cause excessive risk taking which can result in a crisis.
Banerjee and Newman (1991) asses the effect of bank regulations on poverty and find that liberalizing banking markets and abolishing credit controls, will make it easier for the poor to access the credit needed for investments in order to increase their future income. Chigumira and Masiyandima (2003) agree with this conclusion by noticing that it is on average more expensive to lend to poor customers, because the monitoring costs are higher and there is a higher risk of default. As a result, the poor are generally left out by the financial institutions. Lowering the reserve requirements because of financial liberalization, will increase the credit supply. This, in combination with the increased competition caused by the deregulation, will give the financial institutions an incentive to seek new customers which can be beneficial for the poor borrowers who were normally rationed out (Arestis & Caner, 2004).
Despite the ambiguous effect of financial liberalization on poverty, studies in the past seem to suggest that financial liberalization is beneficial for the poor. Therefore, the next hypothesis will be:
H2: Financial liberalization will be negatively related with the poverty gap.
2.3 Financial Development and Financial Liberalization on Poverty
Besides the direct effect of financial development on poverty, the literature also suggests that the relationship between financial liberalization and poverty depends on the level of financial development in an economy (Beck et al, 2007). Therefore, financial liberalization can have a different effect on the poor in highly developed economies than in developing economies (Claessens & Perotti, 2007).
For example, Delis, Hasan, and Kazakis (2014) claim that the poor may only benefit from financial liberalization in economies with a substantial level of financial development, since in low financially developed economies, individuals with higher accumulated wealth still get the extra credit gained by financial liberalization (Greenwood & Jovanovic, 1990).
In addition, financial liberalization in a financially developed economy can contribute to better access to credit for people who were first marginalized and can be beneficial for the poor (Arestis & Caner, 2010). Therefore, the beneficial effects of financial liberalization on poverty can be strengthened by an increase in financial development.
Thereby, existing literature has reached a consensus that financial liberalization can have serious consequences for the stability of a financial system if it is implemented in an underdeveloped financial system. Ultimately, this can lead to a crisis which can be bad for the poor (Brownbridge & Kirkpatrick, 2000). The next hypothesis is:
H3: The relationship between financial liberalization and poverty is positive in economies with low financial development and negative in economies with high financial development.
2.4 Poverty
Numerous studies have attempted to examine the extent of poverty across the world, thereby a plethora of proxies for poverty has been used. For example, Sehrawat and Giri (2015) used per capita consumption as proxy while others used the poverty gap or poverty headcount as proxy (Donou-Adonsou & Sylwester, 2016; Rewilak, 2017). To date, the drawbacks and advantages of the different proxies for poverty are still not clear (World Bank, 2018).
of residents and the poverty line, one will get the total money needed to increase the income of the poor up to the poverty line. One limitation of the poverty gap is that it does not give insight in the difference of severity in poverty, because the poverty gap is an average.
All the different proxies of poverty considered; this research will use the poverty gap to measure poverty. The definition of the poverty gap according to the World Bank (2019b) is “the mean shortfall in income or consumption from the poverty line (counting the non-poor as having zero shortfall), expressed as a percentage of the poverty line”. The poverty line is based on the price of a basket of food based on the national diet of the poor and is converted to a common currency based on the purchasing power parity of 2011. Therefore, the poverty gap is comparable across countries and does not depend on the nation’s policy and economy, unlike the national poverty line.
The World Bank reports on three different values of the poverty line: $1.90, $3.20 and, $5.50 a day. People who live below the $1.90 line live in extreme poverty, while the other two lines are sometimes used for middle- and high-income countries. Data of the World Bank suggests that the rapid decrease in extreme poverty according to the $1.90 line has not been matched by a reduction in the other two lines (World Bank, 2018). Therefore, in this research, the poverty line at the $1.90 level will be used for the main analysis and the other poverty lines will be used in the robustness section.
2.5 Control Variables
As pointed out in the introduction, poverty can also be affected by financial development and financial liberalization via other variables. Arestis and Caner (2004) pointed out three main channels through which poverty can be influenced. The first channel assumes that finance can reduce poverty via an increase in economic growth (McKinnon, 1973; Shaw, 1973; Jeanneney & Kpodar, 2008). This is also known as the trickle-down effect (Kaidi et al., 2018). However, economic growth has been omitted from the model in order to avoid multicollinearity. Chapter three gives more information about multicollinearity and other model implications.
negative financial shocks (Arestis & Caner, 2004). Financial crises will be added to the model, since there seems to be a relation between finance, poverty, and the occurrence of a crisis.
Finally, channel three states that financial development and financial liberalization will change the distribution of access to financial services and credit and thus, can affect income inequality. To date, the literature is not clear about the effect of finance on income inequality (De Haan & Sturm, 2017). Therefore, income inequality will be added to the model as control variable. This will provide more information about the effect of income inequality on the poverty gap.
As outlined in the introduction, without the implementation of adequate poverty reduction measures the basic structure of the banking system will remain unchanged which will still safeguard the influential position of the banks. High institutional quality facilitates a broad distribution of political and financial rights and may prevent politicians from using their power for their own interests (Lensink, 1996). Without regulation and supervision, the insiders and other people who are strategically positioned and therefore already have access to the financial markets, will still benefit from financial liberalization instead of the poor (Arestis & Caner, 2004; Rashid & Intartaglia, 2016). Therefore, a variable measuring the institutional quality will be included in this research.
Lastly, a control variable for inflation will be included. According to Easterly and Fischer (2001) inflation is expected to be worse for the poor than for the rich. The reasoning behind this is that the poor have a bigger share of their wealth in cash. Also, the poor are more dependent on state-determined income which is not fully indexed to inflation. Chapter 3 will give more information about the variables.
3. Methodology
This chapter is divided into four sections. The first section gives an overview of the database. The next section provides the sources of the data that have been introduced in the previous chapters and a further definition of the variables. After the variables have been defined, the third section explains the assumptions required for panel data analyses. Lastly, the fourth section explains the development of the empirical model that has been used to test the hypotheses.
3.1 Data
of the Fraser Institute (2018) while the data for the other variables is obtained from the World Development Database (2018). Table A1 of Appendix A shows the description and sources of all the variables These databases have been chosen because they are the most complete databases and contain the largest country data over a long time period. A list of countries is accessible in Table B1 of the Appendix B.
Given the fact that macroeconomic data can contain noise (Roine, Vlachos & Waldenstrom, 2009; De Haan & Sturm, 2017), a five-year-non-overlapping average will be used. This method has as advantage that it smoothens spikes in the data (Delis et al., 2014). In addition, some variables have not been measured annually, which means there are gaps in the data. A five-year-non-overlapping average will ignore these gaps and makes the data more balanced. Also, yearly data is not needed for this research since the short-term effects on poverty lies beyond the scope of this paper.
3.2 Variables
Dependent variables
As mentioned in the previous chapters, there are several ways to measure poverty. Our dataset contains different variables used to measure poverty. The first variable is the poverty headcount, which measures the percentage of the population who have to live with an income below a certain level. The second variable is the poverty gap and refers to the mean shortfall in income based on a certain poverty line, expressed as a percentage of the poverty line. Our database contains observations about the poverty lines at $1.90, $3.20 and the national poverty line. People are considered poor when they are consuming less than the poverty line a day. Table 1 shows the summary statistics of the poverty variables.
Table 1 Summary statistics and correlations of the dependent variables
VARIABLES Obs Mean SD Min Max (1) (2) (3) (4) (5) (6)
(1) Pov_hc_1.90 508 16.53 22.2 0 86 1 (2) Pov_hc_3.20 508 29.23 30.65 0 96.2 0.95 1 (3) Pov_hc_nat 254 32.39 17.23 .4 75.03 0.69 0.71 1 (4) Pov_gap_1.90 518 6.35 9.71 0 52.1 0.97 0.87 0.67 1 (5) Pov_gap_3.20 518 13.21 16.41 0 67.5 0.99 0.97 0.71 0.96 1 (6) Pov_gap_nat 116 12.24 8.74 .3 37.7 0.68 0.62 0.92 0.70 0.68 1
Note: Observations of Pov_hc_nat and Pov_gap_nat are lower due to the missing observations.
the national poverty lines will not be used for the main analysis in the next chapter. Another interesting result is the high correlation between the poverty headcount and the poverty gap variables with the same poverty line. These strong correlations mean that the poverty gap and the poverty headcount variables almost move in the same direction and therefore, have a strong relationship. Although, Rewilak (2017) argued that the poverty headcount variables are highly skewed and have many observations close to zero in comparison to the poverty gap. Therefore, this research will use the poverty gap at the $1.90 line as dependent variable for the main analysis. The other variables, listed in Table 1, will be included in the robustness analyses. This makes it possible to compare the outcomes of the regressions when using different proxies for poverty and test the importance of the dependent variable selection.
We will follow the approach of Sehrawat and Giri (2015) and use a log-linear model to examine the effect of finance on poverty. The original data for the poverty gap has many outliers and is skewed to the right. Taking the log of the poverty gap at the $1.90 line takes out the sharpness of the dataset and makes is more evenly distributed. The log-linear model provides more efficient results than a simple linear model.
Independent variables
The independent variables are financial development and financial liberalization. Financial development will be measured as domestic credit to the private sector as percentage
of GDP. As explained in the previous chapter, this refers to the financial resources provided by
financial corporations to the private sector (e.g. loans, nonequity securities, and other accounts receivable, that establish a claim for repayment). The data comes from the World Development Indicators database of the World Bank (2019).
In order to measure financial liberalization, the method of De Haan and Sturm (2017) will be followed. They used the Economic Freedom of the World (EFW) Index created by the Fraser Institute (2018). This index measures to what extent economic freedom is supported by the policies and institutions of a country. Since not all of the data is relevant, only four sub-indices of the EFW database will be used. The sub-sub-indices that are relevant to financial liberalization are 3D, 4C, 4D and 5A:
people (4D) and reflects the amount of capital controls and the freedom of foreign travelers to
visit this country. The last index is credit market regulations (5A) and measures to what extent the banking sector is privately owned, the extent of borrowing by the government relative to borrowing by the private sector and it measures the interest rate controls.
An average of the four indices will be calculated, where a zero refers to no financial liberalization and a ten means fully financial liberalized. To prevent risk of reversed causality with the dependent variable, a form endogeneity, both financial development and financial liberalization will be lagged with one period (five-year average) in the robustness section. The last section will give more information about endogeneity and other panel data implications. To make the data more balanced, variables with missing observations for the poverty gap, financial development or financial liberalization will be dropped.
Control variables
According to the literature, mentioned in the previous chapters, there are some variables which can affect the relationship between financial development, financial liberalization, and, poverty. Therefore, several country specific control variables will be added to the database.
The first variable is the Gini coefficient which controls for income inequality. The Gini coefficient is obtained from the Standardized World Income Inequality Database (Solt, 2019) and measures the income inequality before taxes and transfers. It is based on the Lorenz curve and measures the degree of inequality on a scale from 0 (perfect equality) to 100 (perfect inequality). Despite the fact that the Gini coefficient is still affected by government spending and taxes, it is less affected by government policies than the net Gini coefficient which is way more influenced by taxes and transfers.
Another variable which will be used to control for inequality is based on the income share per quintile. This measures the percentage share of income per subgroup of the population, where the first quintile measures the 20% of the population with the lowest income and the fifth quintile measures the 20% of the population with the highest income. Data of the income quintiles comes from the World Development Indicators of the World Bank (2019e). The income quintiles will be used in the robustness section to compare the outcomes with the other variable for income inequality, the Gini coefficient.
indicators of governance: Voice and Accountability (VA), Political Stability and Absence of Violence/Terrorism (PV), Government Effectiveness (GE), Regulatory Quality (RQ), Rule of Law (RL), and Control of Corruption (CC). The variables range between -2.5 (weak) and +2.5 (strong) government performance.
We will follow the approach of Bjørnskov (2006) and take the average of the six indicators of governance to measure institutional quality. This approach is supported by Langbein and Knack (2010) who show that the indicators are highly inter-correlated and measure the same broad concept of institutional quality.
The control variable crisis is a dummy variable, that will be one if a systemic banking crisis occurred in the five-year period prior to the observations on the poverty gap. Otherwise it will be zero. The crisis data comes from Laeven and Valencia (2013). Two conditions must be met to determine a systemic banking crisis. First, the banking system should be in significant financial distress which is indicated by significant losses, bank runs or bank liquidations. The second condition is met when there are significant banking policy measures and interventions in response to the financial distress (Laeven & Valencia, 2013).
The last control variable is inflation, which reflects the annual percentage change in the cost to the average consumer of acquiring a basket of goods and services. Data comes from the International Financial Statistics and Data Files of the World Bank (2019c). Hence, the final database consists of 518 observations from 123 countries with five-year averages over a period from 1975 to 2015. Table 2 provides a correlation matrix of all the variables.
Table 2 Pairwise correlations
Variables (1) (2) (3) (4) (5) (6) (7) (8) (1) Pov_gap_1.90 1.000 (2) Log_pov_gap_1.90 0.720 1.000 (3) Fin_dev -0.443 -0.540 1.000 (4) Fin_lib -0.450 -0.416 0.411 1.000 (5) Gini -0.041 0.102 0.227 0.307 1.000 (6) Inst. Qual. -0.440 -0.598 0.694 0.583 0.305 1.000 (7) Inflation 0.040 0.069 -0.021 -0.182 0.118 -0.339 1.000 (8) Crisis 0.023 0.023 -0.032 -0.122 -0.018 -0.012 -0.003 1.000
Note: Gini and Inst. Qual. are the original variables before the transformation to dummy variables.
A high correlation can be a sign of multicollinearity. This means that there is an exact linear relationship among independent variables, with as result that the regression model estimates cannot be uniquely computed. This can lead to skewed and misleading outcomes and can inflate the standard error. A test for Variance Inflation Factor (VIF) will be used to test for multicollinearity. This test measures how much the variance of a coefficient will increase because of multicollinearity. A VIF larger than the standard critical value of 10 is considered as inappropriate (Kennedy, 2003). Table C1 of Appendix C shows that there is no worrisome multicollinearity between the independent variables and therefore, it is possible to add financial development and financial liberalization simultaneously to the model
Unfortunately, it is not possible to add the Gini coefficient and institutional quality to the model. Table C2 of Appendix C shows a high VIF for financial development, the Gini coefficient and institutional quality. The VIF did not decrease when adding the control variables separately. Therefore, the variables, Gini coefficient and institutional quality, will not be included since there are high chances of multicollinearity with financial development.
One way to prevent multicollinearity and still control for institutional quality and income inequality, is to create dummy variables. These two dummy variables have a value of one for observations above the median (median is -.17 for institutional quality and 45.2 for the Gini coefficient) and a zero otherwise. Adding the dummy variables to the model makes it possible to control for high and low levels of institutional quality and the Gini coefficient. The VIF test in Table C3.1 of Appendix C shows acceptable levels for the dummy variable of the Gini coefficient, therefore we will include this dummy to the model.
Table C3.2 of Appendix C shows a VIF of 12.29 for financial development and a VIF of 5.9 for the dummy of institutional quality. A VIF of 12.29 is slightly above the threshold of 10, which means there are still signs of multicollinearity between the dummy and financial development. However, the VIF of the dummy variable is still way lower than the VIF of the original variable for institutional quality. Therefore, we will include this variable to the model, but we have to be cautious when interpreting the results.
3.3 Panel Data Implications
to determine whether to use fixed or random effects. The results of the Hausman test are shown in Table C4 of Appendix C. The significant chi-square implies that we can reject the null-hypothesis which states that the difference in coefficients is not systematic. Therefore, a fixed effects model will be used
Another econometric issue is related to the homoscedasticity of the residuals. The modified Wald test for groupwise heteroskedasticity in a fixed effects regression model, shows a significant p-value, which can be seen in Table C5 of Appendix C. For this reason, standard robust errors are necessary.
Lastly, this study tries to prevent endogeneity issues by using a one-period (five-year average) for financial development and financial liberalization. The results will be discussed in the robustness section of the next chapter.
3.4 Empirical Model
As explained in the previous section, this research will use a dynamic panel data approach by using a fixed effects model. An advantage of this model is that it controls for country- and time specific effects and for unobserved heterogeneity. Thereby, a panel data approach with fixed effects will generate more accurate predictions than an OLS regression. The model specification to test the first hypothesis will be:
Log_pov_190 (i, t) = β i + β1 Fin_dev (i, t) + β2 Fin_dev 2 (i, t) + β3 Fin_lib (i, t) + β4 X (i, t) + ni + ε (i, t) (1)
Equation (1) examines whether there is a threshold for financial development after which a further increase of financial development will increase the poverty gap. The equation is calculated with the fixed effects model. The left-hand variable is the log of the poverty gap based on the $1.90 poverty line for country i in period t, where the first period starts in 1975 and the last period ends in 2015. β i captures the intercept, Fin_dev is the linear function of
financial development while Fin_dev2 stands for the squared function. Fin_lib stands for
financial liberalization. X denotes the control variables, while n stands for country specific effects. Lastly, ε refers to the error term.
Log_pov_gap_190 (i, t) = β i + β1 Fin_dev (i, t) + β2 Fin_lib (i, t) + β3 Interactions + β4 X (i, t) + ni + ε (i, t) (2)
Equation (2) makes it possible to examine whether there is an interaction effect between financial development and financial liberalization on the poverty gap.
In the robustness section we will use time lags (t-1) for financial development and financial liberalization to prevent endogeneity. Table 3 provides the descriptive statistics for all the variables in this research.
Table 3 Summary Statistics
VARIABLES Obs. Mean St. Dev. Min Max
Pov_gap_190 518 6.35 9.71 0 52.1 Log_pov_gap_190 518 -.01 2.53 -4.605 3.953 Fin_dev 518 52.79 44.84 .186 246.58 Fin_lib 518 6.89 2.22 .367 9.833 0. Dummy Gini 250 39.82 3.92 21.54 45.19 1. Dummy Gini 249 50.21 4.33 45.25 68.52
0. Dummy Inst. Qual. 207 -0.68 0.35 -1.79 -0.17
1. Dummy Inst. Qual. 206 0.99 0.76 -0.17 2.35
Inflation 498 13.26 79.39 -1.08 1667.15
Crisis 518 .137 .344 0 1
Note: Crisis is 1 when there was a systemic banking crisis in the previous period.
Inst. Qual and Gini are 1 for observations above the median.
4. Results
In this chapter we will present and discuss the results of the equations as explained in the previous chapter. This chapter will proceed as follows. First, we will use equation (1) and regress financial development in a bivariate regression with the logged poverty gap as dependent variable. Thereafter, we will add the squared function of financial development and we will add control variables to the model. The results are shown in Table 4.
In the next section, we will use equation (2) to test the second and third hypotheses. First, the independent variables financial development and financial liberalization will be added to the model. Then the model will be expanded with the interaction effect and the control variables, as can be seen in Table 5.
regressions with other proxies for poverty as the dependent variable. This will give more information about the errors concerning the selection of the dependent variable.
4.1 Main Results Equation (1)
This section investigates the non-linearity theory of financial development. This theory claims that financial development will reduce poverty till a certain level. After this level, a further increase of financial development will increase poverty. To test this hypothesis, the squared function of financial development will be added to the model. The outcomes are shown in Table 4.
The first model shows the results of the bivariate regression with financial development as the only independent variable. As can be seen in Table 4, the coefficient of financial development is significant at p<0.01 in Model 1 and stays significant in the subsequent models. The coefficient of -0.0205 indicates a negative relationship between financial development and the poverty gap, where a 1%, one unit, increase in financial development, domestic credit to
the private sector to GDP, will decrease the poverty gap at the $1.90 line by 2.07%1. This is in line with the theory of Galor and Moav (2004) and Kaidi, Mensi and Amor (2018) who suggest that financial development will mitigate credit constraints and makes it possible for the poor to accumulate physical capital and invest in human capital. Different from their research, we have added a squared function of financial development to the model which gives interesting results. Model 2 shows a positive coefficient of the squared function which is significant at p<0.01. This result suggests that there is also a squared relationship between financial development and poverty such that an increase in the squared function of financial development will also lead to an increase in the poverty gap.
The negative coefficient of financial development in combination with the positive coefficient of the squared function confirms hypothesis 1 and shows that the linear effect of financial development will dominate at low levels, which will decrease the poverty gap, but that the effect of the squared function will dominate at higher levels of financial development, which will increase the poverty gap. As mentioned in the literature review, a possible explanation is suggested by Loayza and Ranciére (2006). They show that high levels of financial development can cause economic instability which will decrease the economic growth and therefore, will harm the poor as suggested by the trickle-down theory (Arestis and Caner, 2009).
Table 4 Dep. Var. Log Poverty Gap at $1.90 with non-linear effect of financial development
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Fin_dev -0.0205*** -0.0449*** -0.0432*** -0.0427*** -0.0367*** -0.0431***
(0.00507) (0.00734) (0.00816) (0.00813) (0.00798) (0.00831)
Fin_dev # Fin_dev 0.000129*** 0.000123*** 0.000120*** 9.87e-05*** 0.000118***
(2.95e-05) (3.13e-05) (3.14e-05) (2.69e-05) (3.18e-05)
Fin_lib -0.0342 -0.0201 -0.0151 -0.0290 (0.0527) (0.0526) (0.0603) (0.0526) Inflation 0.00122*** 0.0297*** 0.00123*** (0.000194) (0.00648) (0.000187) Crisis 0.219 0.156 0.174 (0.132) (0.143) (0.135) Inst. Qual. 0.317 (0.293) Gini 0.500* (0.289) Constant 1.073*** 1.743*** 1.914*** 1.747*** 1.000* 1.575*** (0.267) (0.284) (0.351) (0.391) (0.587) (0.401) Observations 518 518 518 498 402 479 R-squared 0.125 0.174 0.177 0.198 0.224 0.209 Countries 123 123 123 120 119 116
Note: Crisis is 1 when there was a systemic banking crisis in the previous period. Inst. Qual and Gini are 1 for
observations above the median. Fixed effects model with corrected robust standard errors for heteroscedasticity Robust standard errors in parentheses. Level of significance: *** p<0.01, ** p<0.05, * p<0.1.
Model 3 shows an insignificant effect of financial liberalization on poverty; the next section will examine the effect of financial liberalization on poverty based on equation (2). The last interesting result can be seen in Model 4 where the control variables inflation and crisis have been added, the inflation variable shows a significant positive result at p<0.01. This indicates that inflation will hurt the poor and is in line with our expectations based on the theory of Easterly and Fischer (2001) who showed that inflation will decrease the income and capital of the poor. The Gini coefficient is only significant atp<0.1 while the dummies for crisis and
institutional quality are not significant at all. We will further examine these variables in the next section.
Figure. 1 Fitted values and scatterplot of the log poverty gap at $1.90 and financial
development
Note: Predicted Fitted values are based on the datapoints of Model 2 in Table 4.
The level after which financial development will increase poverty can be found by setting the derivate of Model 2 equal to zero2. Financial development will cause an increase of the poverty gap after the domestic credit to the private sector to GDP reached 174.03%.
4.2 Main Results Equation (2)
Table 5 on the next page shows the results of the panel fixed effects regression based on equation (2). The first two models show a bivariate regression with the independent variables separately included. As can be seen, both the independent variables are significant, although the coefficient of financial development on poverty is significant at p<0.01 while the coefficient of financial liberalization is significant at p<0.05. The negative sign means that an increase in financial development or financial liberalization will decrease the poverty gap. Therefore, both variables are poverty reducing.
Model 3 shows a regression with both independent variables included; an interesting result is the loss of significance of financial liberalization. In Model 4, the interaction effect of financial development and financial liberalization has been added to the model. Even though the interaction effect is only significant at p<0.1, both the independent variables have become significant at p<0.01. The addition of the interaction effect strengthens the main effect of the independent variables on the dependent variable. Moreover, the effect of the coefficient of
financial development increased from -0.0193 to -0.0419 and the effect of the coefficient of financial liberalization increased from -0.0638 to -0.143. The significant negative relationship between financial liberalization and the poverty gap supports hypothesis 2. This hypothesis is based on the literature that suggests that financial liberalization makes it easier for the poor to get the credit needed for investments, in order to increase their future income.
Another remarkable result is the positive coefficient of the interaction effect, this indicates that high levels of financial liberalization in combination with high levels of financial development increase the poverty gap. This result is the opposite of what we expected based on the literature and therefore, does not support the third hypothesis. A possible explanation can be found when combining the theory of Demirgüc-Kunt and Detragiache (2000), who suggest that financial liberalization will create instability, and the theory of Loayza & Ranciére (2006), who suggest that financial development can lead to instability, and eventually a crisis. Therefore, it is possible that high levels of both financial liberalization and financial development will lead to more financial instability, which will hurt the poor. However, as mentioned by Arestis and Caner (2004), the effect of finance on poverty is still unclear. Further research should be carried out to examine the relationship between finance and poverty.
Model 5 adds the control variable inflation to the model. The effect of inflation on poverty has not changed in comparison with equation (1). Inflation still has a positive significant effect on poverty which means that inflation will increase poverty. Controlling for inflation increases the significance level of the interaction effect from p<0.1 to p<0.05. In addition, the coefficients of the independent variables have become stronger. The coefficient of the crisis dummy is added in Model 6. The positive and significant coefficient indicates that the poverty gap increases with 28.4%3 in the period after a crisis. This is an interesting result since crisis was not significant in equation (1). It seems that the addition of the interaction effect does not only affect the independent variables but also strengthens the effect of the crisis dummy on the poverty gap.
Table 5 Dep. Var. Logged Poverty Gap at $1.90 with interaction effect between financial development and financial liberalization
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10
Fin_dev -0.0205*** -0.0193*** -0.0419*** -0.0471*** -0.0418*** -0.0497*** -0.0397*** -0.0469*** -0.0369*** (0.00507) (0.00514) (0.0155) (0.0147) (0.0157) (0.0147) (0.0151) (0.0148) (0.0120) Fin_lib -0.114** -0.0638 -0.143*** -0.163*** -0.132*** -0.183** -0.136*** -0.148*** -0.110 (0.0508) (0.0511) (0.0465) (0.0437) (0.0477) (0.0859) (0.0458) (0.0452) (0.0684) Fin_dev # Fin_lib 0.00284* 0.00347** 0.00284* 0.00376** 0.00247 0.00346** 0.00257** (0.00160) (0.00149) (0.00163) (0.00150) (0.00155) (0.00151) (0.00122) Inflation 0.00145*** 0.00151*** 0.0320*** (0.000335) (0.000340) (0.00597) Crisis 0.250** 0.269** 0.161 (0.124) (0.132) (0.145) Inst. Qual. 0.154 0.188 (0.275) (0.261) Gini 0.496* 0.188 (0.256) (0.334) Constant 1.073*** 0.773** 1.447*** 2.036*** 2.154*** 1.919*** 2.126*** 1.751*** 2.006*** 1.016 (0.267) (0.350) (0.363) (0.426) (0.414) (0.440) (0.700) (0.423) (0.432) (0.642) Observations 518 518 518 518 498 518 413 499 498 498 R-squared 0.125 0.029 0.133 0.154 0.176 0.161 0.150 0.164 0.185 0.185 Countries 123 123 123 123 120 123 122 119 120 120
Note: Crisis is 1 when there was a systemic banking crisis in the previous period. Inst. Qual. and Gini are 1 for observations above the median. Fixed effects model with
The results of the regressions including the control dummies for institutional quality and the Gini coefficient are shown in Model 7 and Model 8. Unfortunately, the institutional quality dummy is still not significant, and the Gini coefficient dummy is only significant at
p<0.1. Therefore, we will not include these variables in the other equations. Further research
is required to determine whether the poverty gap is affected by institutional quality and the Gini coefficient. The next section will use the income quintiles as another proxy for income inequality. The last column of Table 5 shows the full model with the interaction effect and the control variables.
A graphical representation of the interaction effect described in Table 5 can be seen in Figure 2. The graph shows that the effect of financial liberalization on the poverty gap is affected by the level of financial development, whereas higher levels of financial
development will cause a positive relationship between financial liberalization and the poverty gap.
Fig. 2The combined effect of financial liberalization and financial development on poverty.
Note: Figure 2 is based on Table 5, Model 4.
gap. Therefore, more research is necessary to the non-linear effect of financial development on poverty.
4.3 Robustness Check and Sensitivity Analysis
A risk of the regressions performed in the previous sections is the possible presence of endogeneity. Therefore, time lags will be used to prevent endogeneity. Both, financial development and financial liberalization will be lagged by one period (five-year average) prior to the observations of the poverty gap at the $1.90 line. Table 6 shows the results of the regressions.
Similar to Table 4, the first three models of Table 6 show the results of the bivariate regressions with the independent variable separately and the result of the regression with both the independent variables and the poverty gap. These results indicate that the effect of financial development on poverty is weaker when using a time lag, while the effect of financial liberalization on poverty is stronger when using a time lag. Another interesting result can be seen when looking at Model 4. Financial development has lost its significance, while it previously was significant at p<0.01 in the regression without a time lag. Also, the interaction effect between financial development and financial liberalization lost its significance, thereby the coefficient has become smaller. Therefore, there does not seem to be a relationship between the lagged variable of financial development and the poverty gap.
The results of Model 4 stay the same when adding more variables to the regression in Model 5 and Model 6. Another interesting result can be seen in Model 6, the coefficient of the crisis dummy has increased from .269 in Table 5 to .473 in Table 6. Thereby, the effect of a crisis on poverty became significant at p<0.01 which means that the occurrence of a crisis in the period preceding the period covered by the poverty gap, will cause a higher increase in the poverty gap when using a time lag for the independent variables. Inflation is still poverty increasing, even though the effect is smaller and only significant atp<0.1.
Table 6 Dep. Var. Log Poverty Gap at $1.90 with lagged independent variables
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
Fin_dev -0.0120*** -0.00929** -0.0170 -0.0228 -0.0249 -0.0233*** (1-period lag) (0.00414) (0.00382) (0.0165) (0.0186) (0.0177) (0.00813) Fin_lib -0.179*** -0.155*** -0.177*** -0.191*** -0.166*** -0.112** (1-period lag) (0.0438) (0.0437) (0.0478) (0.0522) (0.0492) (0.0494) Fin_dev # Fin_lib 0.000956 0.00163 0.00166 (1-period lag) (0.00178) (0.00201) (0.00189)
Fin_dev # Fin_dev 6.10e-05**
(1-period lag) (2.83e-05)
Inflation 0.000997 0.00115* 0.000909*** (0.000610) (0.000584) (0.000239) Crisis 0.473*** 0.438*** (0.139) (0.143) Constant 0.683*** 1.265*** 1.560*** 1.730*** 1.821*** 1.684*** 1.607*** (0.200) (0.284) (0.318) (0.433) (0.470) (0.436) (0.376) Observations 472 472 472 472 453 453 453 R-squared 0.062 0.095 0.131 0.133 0.148 0.178 0.184 Countries 119 119 119 119 116 116 116
Note: Fin_dev and Fin_lib are lagged by 1 period (five-year average). Crisis is 1 when there was a systemic
banking crisis in the previous period. Fixed effects model with corrected robust standard errors for heteroscedasticity. Robust standard errors in parentheses.
Level of significance: *** p<0.01, ** p<0.05, * p<0.1.
The results of the regression including the income shares per quintile can be seen in Table 7. The income share per quintile is, besides the Gini coefficient, another way to measure inequality. An advantage of the quintile variables is that they do not have multicollinearity with the other variables. The first quintile shows the income share of the bottom 20% in an economy. Evidently, an increase in the income share hold by the first 20% will decrease the poverty gap. A more interesting result is found when comparing the fourth quintile in Model 4 with the fifth quintile in Model 5. The negative coefficient of the fourth quintile variable, shows that even an increase in income of this group will decrease the poverty gap, while the positive coefficient of the fifth quintile variable indicates that an increase of the income share of the richest 20% will increase the poverty gap.
more income inequality, it can be bad for the poor. These findings are consistent with the study of Sehrawat and Giri (2018) who claim that an increase in income inequality will aggravate poverty. Further work is required to develop a full picture of the effects of income inequality on the relationship between finance and poverty.
Table 7 Dep. Var. Log Poverty Gap at $1.90 with income quintiles
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5
Fin_dev -0.0419*** -0.0452*** -0.0470*** -0.0470*** -0.0454*** (0.0120) (0.0126) (0.0132) (0.0143) (0.0126) Fin_lib -0.156*** -0.161*** -0.172*** -0.156*** -0.168*** (0.0510) (0.0465) (0.0433) (0.0426) (0.0433) Fin_lib # Fin_dev 0.00291** 0.00332** 0.00355** 0.00351** 0.00336** (0.00129) (0.00135) (0.00138) (0.00146) (0.00133) First Quintile -0.586*** (0.100) Second Quintile -0.492*** (0.0970) Third Quintile -0.402*** (0.0977) Fourth Quintile -0.235** (0.0936) Fifth Quintile 0.130*** (0.0283) Inflation 0.00113*** 0.00116*** 0.00124*** 0.00146*** 0.00121*** (0.000255) (0.000262) (0.000287) (0.000325) (0.000270) Crisis 0.174 0.198* 0.201 0.229* 0.183 (0.111) (0.118) (0.122) (0.130) (0.118) Constant 5.749*** 7.328*** 8.158*** 7.051*** -3.943*** (0.800) (1.166) (1.598) (2.130) (1.291) Observations 495 495 495 495 495 R-squared 0.356 0.319 0.279 0.216 0.307 Countries 120 120 120 120 120
Note: Quintiles correspond to the income quintiles, where the first quintile is the 20% people with the lowest
income. Crisis is 1 when there was a systemic banking crisis in the previous period Fixed effects model with corrected robust standard errors for heteroscedasticity. Robust standard errors in parentheses
Level of significance: *** p<0.01, ** p<0.05, * p<0.1
financial development and financial liberalization are slightly larger than the coefficients at the $3.20 line. This indicates that finance is more beneficial for the extremely poor who live below the $1.90 line than for the people between the $1.90 and the $3.20 line. This is in accordance with the theory of Galor and Zeira (1993) who claim that finance will disproportionally benefit the poor. However, to support their claim, more research is necessary since the people who live between the $1.90 and $3.20 line can still be considered as poor according to the World Bank (2018).
Despite the fact that the crisis dummy was not always significant when using the poverty gap at the $1.90 line as dependent variable, it is still worth mentioning that the coefficient of the crisis variable has become smaller and insignificant when using the $3.20 as dependent variable. This result is interesting and seems to be in line with the prediction of Arestis and Caner (2004). They claim that the occurrence of a crisis is more harmful for the extreme poor since they do not have the capital to mitigate the effects of a crisis and are more likely to be harmed by the contractionary fiscal policy followed after a crisis.
The last two columns are slightly harder to interpret since the dependent variables are based on the national poverty line, which differs per country and year. Therefore, they cannot be compared as easily as the poverty line at $1.90 and $3.20. However, the effects of financial development and financial liberalization on poverty do not seem to deviate much since the signs of the coefficients of the independent variables are still negative and the interaction effect is still positive.
However, we have to be extra cautious when interpreting the results of the last two columns since there were only 112 observations for the poverty gap at the national poverty line and 247 observations for the poverty headcount at the national poverty line.
Table 8 Different dependent variables for poverty
VARIABLES Gap $1.90 HC $1.90 Gap $3.20 HC $3.20 Gap Nat. HC Nat.
Fin_dev -0.0469*** -0.0533*** -0.0454*** -0.0414*** -0.0252* -0.0241*** (0.0148) (0.0161) (0.0147) (0.0143) (0.0147) (0.00662) Fin_lib -0.148*** -0.130*** -0.123*** -0.0926*** -0.118** -0.0904*** (0.0452) (0.0429) (0.0377) (0.0313) (0.0528) (0.0337) Fin_dev # Fin_lib 0.00346** 0.00425** 0.00373** 0.00345** 0.00241 0.00166* (0.00151) (0.00164) (0.00154) (0.00151) (0.00208) (0.000898) Inflation 0.00151*** 0.00169*** 0.00151*** 0.00138*** 0.00788 -0.00450 (0.000340) (0.000366) (0.000343) (0.000311) (0.0126) (0.00731) Crisis 0.269** 0.262* 0.169 0.130 0.319 -0.0313 (0.132) (0.140) (0.118) (0.106) (0.208) (0.117) Constant 2.006*** 2.952*** 2.739*** 3.401*** 3.170*** 4.431*** (0.432) (0.446) (0.380) (0.319) (0.392) (0.255) Observations 498 498 498 498 110 234 R-squared 0.185 0.189 0.189 0.175 0.177 0.243 Countries 120 120 120 120 57 94
Note: Crisis is 1 when there was a systemic banking crisis in the previous period. Fixed effects model with
corrected robust standard errors for heteroscedasticity. Robust standard errors in parentheses. Level of significance: *** p<0.01, ** p<0.05, * p<0.1.
All the small differences in the models in Table 8 considered, it still seems that the selection of different variables for poverty does not affect the results to a large extent. This confirms the conclusion, made in the method section, which states that all the proxies for poverty are highly correlated and generate similar results.
5. Conclusion
financial development and financial liberalization has been added to the model to investigate the combined effect of financial development and financial liberalization on the poverty gap.
Based on a fixed effects model with panel data for the period 1975-2015, we have proven the first hypothesis which states that the poor will benefit from financial development till a threshold, after this threshold is reached, a further increase of financial development will cause more poverty. Furthermore, the results showed a negative relationship between financial liberalization and the poverty gap, which is in line with our second hypothesis. Despite the negative bivariate relationship between financial development and financial liberalization on poverty, no evidence was found for our third hypothesis, which suggested a negative interaction effect between the two variables and poverty. In fact, our results showed a significant positive relationship of the interaction between financial development and financial development on the one hand and the poverty gap on the other hand. This indicates that more financial liberalization in economies with a high level of financial development will lead to more poverty and is exactly the opposite of our third hypothesis.
Our clustered standard errors are robust for heteroskedasticity and our results are holding up when including inflation and crisis as control variables. Both the coefficients of the control variables were positive and significant when combining them with the interaction effect. This indicates that both inflation and crisis will increase poverty, which is in line with our expectations. Furthermore, hypotheses one and two are holding up when using a one-period time lag for the independent variables. However, financial development and the interaction effect between financial development and financial liberalization turned out to be insignificant when using time lags. Therefore, there does not seem to be a relationship between the lagged variable of financial development and the poverty gap.
Our robustness section also offered interesting results of the effect of the income quintiles on the poverty gap. It showed that an increase in the income share of the top 20% will lead to more poverty, while an increase in the income share of the other quintiles will cause a decrease in poverty. An interesting avenue for further research will be to investigate whether the effect of finance on poverty is conditioned by inequality.
within this thesis. More in-depth consulting with the people who live in extreme poverty is crucial to get a better understanding of the multidimensional view of poverty.
The second limitation concerns the multicollinearity between financial development and the control variables institutional quality and the Gini coefficient. Adding these variables could give more insight in the effect of institutional quality and inequality on the relation between finance and poverty. Therefore, using different proxies for institutional quality, inequality, or financial development could be an interesting path for future research.
Third, one can suggest that several important variables have not been included in the thesis. For example, literature suggest that economic growth can be the link between financial development and poverty (Demirgüc-Kunt & Levine, 2007). However, this research did not account for economic growth due to multicollinearity. Therefore, a fruitful area for further research would be to extend our model with several other variables that can affect the relationship between finance and poverty according to the literature.
Fourth, due to the research set-up of this thesis we were not able to jointly include the interaction effect and the squared poverty gap. Further research on the interaction effect of financial development and financial liberalization on poverty should take the non-linear relationship between financial development and poverty into account. It could be possible that the interaction will be affected by the non-linearity of financial development.
Lastly, it is worth mentioning that this research could be replicated on a smaller scale. The database for this research is created with observations from many different countries across the world, therefore the findings from the thesis are very general. In order to be more useful for local policies, further research needs to be carried out in order to explore the specific effects of finance on poverty in different economic systems or continents.
Up to now, the world has made enormous progress in the reduction of extreme poverty. Surveys from the World Bank (2018) show that the percentage of people who live in extreme poverty globally decreased from 36% in 1990 to a new low of 10% in 2015. However, the average decrease in poverty globally is stagnating and in some low-income countries it is increasing again.
widely available and accessible for the people, especially the poor. However, we must ensure that this does not lead to more inequality and instability.
References
Abiad, A., Oomes, N., & Ueda, K. (2008). The Quality Effect: Does Financial Liberalization Improve the Allocation of Capital? Journal of Development Economics, 87(1), 270– 282.
Abiad, A., Detragiache, E., & Tressel, T. (2010). A New Database of Financial Reforms.
IMF Staff Papers, 57(1), 281–302.
Aghion, P., & Bolton, P. (1997). A trickle-down theory of growth and development with debt overhang. Review of Economic Studies, 64(2), 151–172.
Altunbas, Y., Binici, M., & Gambacorta, L. (2018). Macroprudential policy and bank risk.
Journal of International Money & Finance, 81(1), 203–220.
Alderson, A., & Nielsen, F. (1999). Income inequality, development, and dependence: A reconsideration. American Sociological Review, 64(4), 606-631.
Arcand, J., Berkes, E., & Panizza, U. (2015). Too much finance? Journal of Economic
Growth, 20(1), 105-148.
Arestis, P., & Caner, A. (2004). Financial liberalization and poverty: channels of influence. Basingstoke, UK: Palgrave Macmillan.
Arestis, P., & Caner, A. (2009). Financial liberalization and the geography of poverty.
Cambridge Journal of Regions Economy and Society, 22(1), 229–44.
Atkinson, A., & Morelli, S. (2011). Economic crises and inequality. UNDP-HDRO
Occasional Papers, 1(1), 1-49.
Banerjee, A., & Newman, A. (1993). Occupational choice and the process of development.
Journal of Political Economy, 101(2), 274–298.
Beck, T., Demirguc-Kunt, A., & Levine, R. (2005). Finance, Inequality, and Poverty: Cross-Country Evidence. Journal of Economic Growth, 10(3), 199-229.
Beck, T., Demirgüc ̧-Kunt, A., & Levine, R. (2007). Finance, inequality and the poor.
Journal of Economic Growth, 12(1), 27–49.
Beck, T., & Levine, R. (2004). Stock markets, banks, and growth: Panel evidence. Journal of
Banking and Finance, 28(1), 423-2.
Beck, T., Levine, R., & Loayza, N. (2000). Finance and the Sources of Growth. Journal of
Financial Economics, 58(1), 261-300.
Beck, T., Levine, R., & Levkov, A. (2010). Big bad banks? The winners and losers from bank deregulation in the United States. Journal of Finance, 65(1), 1637–1667. Bjørnskov, C. (2006). The multiple facets of social capital. European Journal of Political
Economy, 22(1), 22–40.
Brownbridge, M., & Kirkpatrick, C. (2000). Financial regulation in developing countries.
Journal of Development Studies, 17(1), 1–24.
Chigumira, G., & Masiyandima, N. (2003). Did Financial Sector Reform Result in Increased Savings and Lending for the SMEs and the Poor? IFLIP Research Paper, 3(7), 1-69. Claessens, S., & Perotti, E. (2007). Finance and Inequality: Channels and Evidence. Journal
of Comparative Economics, 35(4), 748-773.
Dal Colle, A. (2010). Finance-growth nexus: Does causality withstand financial
