Financial development and income inequality

The influence of the financial depth on income inequality was mainly analysed with the use of private credit, bank asset stock market capitalisation and portfolio debt and equity flows. The positive and significant influence of income inequality was reported by Roine et al. (2009), Clarke et al. (2006), Jauch et al. (2012), de Haan et al. (2017) and Maldonado (2017). Opposing results regarding the private credit to GDP were supported by Beck et al. (2007) and Naceur et al. (2016).

After the financial crisis of 2007–2008, financial stability has become one of the most significant components of financial development. In the presence of excessive financial access, financial stability was under pressure. ‘Rising inequality in developed economies caused redistribution in subsidised housing finance, which led to a housing boom and later to a crash with known consequences’ (Rajan et al., 2010, p. 5). Moreover, according to Kumhof and Ranciére (2010), when the income is maintained mainly by top earners, the poor are forced to borrow to sustain their level of consumption. If the gap is growing, then it may lead to financial instability and crisis. As a result of weak financial stability during the financial crisis, some countries increased the role of financial supervision and regulation. ‘Financial instability, in turn, can lower growth and worsen inequality (as evidenced, for example, in the United States following the global financial crisis), and thus undo the very objectives that were sought through inclusion’ (Barajas et al., 2015, p. 5).

Empirically, stability was researched by de Haan et al. (2017) and Delis et al. (2012). de Haan et al. (2017) found that the banking crisis lowers the income shares of the top income earners. Delis et al. (2012) assessed the influence of bank regulatory policies on income inequality and stated that banking deregulation causes lower income inequality.

One of the main difficulties in the evaluation of the impact of financial stability on income inequality is the endogeneity variables of banking regulations. As suggested by Delis et al. (2012), the income inequality has an impact on the choice of bank regulation and the policy must promote the creditworthiness of banks as well as the stability of the financial sector. The link between financial stability and bank regulations remains an unresolved issue that requires further research on the liberalisation of banking markets, primarily via efficient banking supervision and abolishing credit controls, which can help the poor get easier access to credit (Delis et al., 2012, p. 32).

Financial efficiency is related to the type of financial market that is deeply rooted in the legal origins of a given country. The main focus of discussions is about the financial efficiency of the market-based system and bank-based system. Maldonado (2017) and Roine et al. (2009) tried to include in the analysis the difference between the bank and market-oriented financial systems. The main results suggest that the financial development (relative share of the banking and stock market sectors in the economy) improves the income share of the top percentile. Maldonado (2017) in her research has shown that an increase in the market-based component of a financial system increases Gini coefficient. According to Maldonado (2017), the increase in capital market size and efficiency may increase income inequality and state that ‘Bank-based systems, where indirect finance dominates, and market-based systems, which rely more on financial markets, where direct trading of securities takes place. In the EU, households channel most of their savings in easily accessible deposits, while in the United States, households invest more in assets with maturity, such as shares, life insurances and pension funds’ (Maldonado, 2017, p. 7). The more market-based economies exhibit a higher premium for the rich and promote higher income as compared with more bank-based ones.

Financial access is a primary concern for the developing countries, and it is a policy tool that can help alleviate poverty and income inequality (Beck et al., 2007; Naceur et al., 2016). However, there are also some dangers associated with the broader financial access that can lead to financial crisis and widening income inequality. The uncontrolled increase in financial services caused instability and was one of the reasons for the U.S. subprime mortgage crisis of 2007. The Rajan's (2010) well-known phrase ‘let them eat credit’ means that with the increasing acceptance of credit default, the poorest could smooth their consumption at the cost of financial instability. Increasing financial access needs to be cautious and accompanied by the rise of the quality of financial supervision. ‘New evidence also shows that bank stability risks increase when access to credit is expanded, especially without adequate regulation and supervision’ (Barajas et al., 2015, p. 22). In contrast, people who have bank accounts are the nation's top income earners in the majority of developing countries. The poorest must rely on their savings to invest in better education or to start entrepreneurial activity. ‘In many EMDCs, low-income households and small-scale firms often face challenges in accessing financial services due to lack of financial knowledge, complicated processes, onerous paperwork, and other market failures’ (Dabla-Norris et al., 2015, p. 18).

Empirically, the access to financial sector is mainly measured by the number of ATMs. Barajas et al. (2015) and Naceur et al. (2016) have shown that the increase in the number of ATMs has a positive influence on economic growth and might mitigate income inequality. Authors who support financialisation in their literature are Jaumotte et al. (2013), de Haan et al. (2017), Maldonado (2017) and Jauch et al. (2012). On the other hand, authors who advocate that financialisation has negative influence are Naceur et al. (2016), Beck et al. (2007) and Delis et al. (2012)

The difference between various aspects of financial sector has recently become important and one scientific paper considered all the dimensions was written by Naceur et al. (2016). This evolution in the approach helps us establish a better policy implication and prompts solution to decrease income inequality via financial development. According to Naceur et al. (2016), financial access, stability and efficiency can reduce income inequality. However, financial liberalisation increases income inequality.

Despite contradicting results, the banking sector improvements can substantially help in closing the income gap. ‘In addition, our evidence suggests that banking sector development has a better positive effect on income distribution rather than stock market development’ (Naceur and Zhang, 2016, p. 11). Further, recommendations for diminishing inequality and increasing financial development are proposed and they are as follows: reducing credit and interest controls and improving banking and securities market. An important policy implication will be to strengthen the regulatory system for financial institutions that would protect the individuals from the risk that banks are taking. The summary of the empirical literature is presented in Table A3 in Appendix.

Based on the theoretical considerations and previous studies, this section presents the econometric strategy for the empirical analysis in terms of methodology and data description.

This paper proposes a model that empirically estimates the impact of financial development dimensions on income distribution. We extended the model by including alternative and additional financial and inequality measures to better understand how they are interrelated. The analysis consists of three main statistical models, which together allow drawing robust conclusions of the influence of financial development on income inequality.

First, we verified the effect of financial development in terms of domestic credit to private sector (% of GDP) on income distribution, measured by Gini coefficient. It helps us to verify the hypothesis that financial development has an impact on inequality. Then to examine separately four of financial development the variables were extended with the list of financial indicators that was acquired from the Global Financial Development Database (GFDD). (MODEL 1)Ginii,t=crediti,t+control variablesi,t+Vi,t{\rm{Gin}}{{\rm{i}}_{i,t}} = {\rm{credi}}{{\rm{t}}_{i,t}} + {\rm{control}}\,{\rm{variable}}{{\rm{s}}_{i,t}} + {V_{i,t}}(MODEL 2)Ginii,t=four financial dimensionsi,t+control variablesi,t+Vi,t{\rm{Gin}}{{\rm{i}}_{i,t}} = {\rm{four}}\,{\rm{financial}}\,{\rm{dimension}}{{\rm{s}}_{i,t}} + {\rm{control}}\,{\rm{variable}}{{\rm{s}}_{i,t}} + {V_{i,t}} where i indicates the country, t is the year and V_i,t is an error term.

The primary considered variable private credit over GDP is describing how difficult is to get a credit in case of corporation or a household. According to Maldonado (2017), set of controls that are commonly used by economists such as the level of education, trade openness and the unemployment rate are used to avoid an omitted variable bias as these above factors are influencing income inequality and might be correlated with the explanatory variables. Our set of control variables contains the GDP growth (Jauch et al., 2012; Clarke et al., 2006; de Haan et al., 2017), general government final consumption expenditure (% of GDP) (Jauch et al., 2012; Naceur and Zhang, 2016), population growth (Maldonado, 2017; Beck et al., 2007) and trade as % of GDP (Beck et al., 2007; Clarke et al., 2006; de Haan et al., 2017; Naceur and Zhang, 2016).

Gini coefficient used by Clarke et al. (2006), Maldonado (2017), de Haan et al. (2017), Jauch and Watzka (2012) and Naceur and Zhang (2016) is one of the most popular measures of inequality. However, it has series of disadvantages and so it was decided to use alternative variables such as quintiles of income distribution, which were used by Delis et al. (2012), Roine et al. (2009) and Dollar and Kraay (2001). The regressions with quintiles of income distribution were in the form of the following model: (MODEL 3)Quintilesi,t=four financial dimensionsi,t+control variablesi,t+Vi,t{\rm{Quintile}}{{\rm{s}}_{i,t}} = {\rm{four}}\,{\rm{financial}}\,{\rm{dimension}}{{\rm{s}}_{i,t}} + {\rm{control}}\,{\rm{variable}}{{\rm{s}}_{i,t}} + {V_{i,t}}

Finally, as it was mentioned above, to compare the effect of financial development in the short and long run, we decided to build two datasets. The first one would have yearly observations with 165 observations from 2003 to 2014 and the second one would take the three-year averages also from the same time span of 2003–2014. This comparison was primarily suggested by Forbes (2001) that verified the effect of income inequality on GDP growth in this manner.

The comparison was performed having Gini coefficient as a dependent variable and the model exhibits model 2. This approach would show us whether the short-term fluctuations are significant for the explanation of the underlying phenomenon.

The main dependent variable in our empirical models is Gini coefficient that measures the distribution of income among individuals or households in economy and it is acquired from the World Bank. To assess the particular parts of income distribution, we have also used quintiles that express the share of total income going to each fifth of the population according to the size of their incomes (World Income Inequality Database (WIID), 2018). The first quintile is the poorest 20%, while the fifth quintile is the richest 20% and they are taken from the World Income Inequality Database (WIID).

In the early empirical analysis, researchers such as Clarke et al. (2006) and Beck et al. (2007) used private credit to GDP. We have also decided to test the significance of private credit to GDP taken from the World Bank.

To measure other dimensions of financial development, we decided to use GFDD (2018). To include each of the dimensions (efficiency, stability, depth and access), we selected bank noninterest income to total income (%), ATMs per 100,000 adults, Bank concentration (%) and Foreign banks among total banks (%). To verify the robustness of the result, we used several controlled variables that arrive from the World Bank as suggested by literature such as trade is the sum of exports and imports of goods and services measured (% of GDP); general government final consumption expenditure (% of GDP); GDP per capita growth (annual %) and annual population growth rate. The descriptive statistics, as well as definition of variables, are presented in Tables A1 and A2 in Appendix.

To examine the effect in the short and long run, we decided to build two separate datasets with the same number of variables, but with the difference in the number of observations. The first one would have yearly observations from 2003 to 2014 and the second one would take the three-year averages also from the same time horizon. This comparison was primarily suggested by Forbes (2001) that verified the effect of income inequality on GDP growth in this manner. As it was suggested by Forbes (2001), to assess the short-and medium-term relationship within individual countries, we decided to estimate the models using three-year average and then to evaluate short-term effect regression was run on the yearly observations. Due to data imperfection and missing observations analysing at the averages or yearly data might appear to be misleading.

We have to deliberate carefully on possible statistical difficulties when choosing the right econometric method to estimate the underlying relationship. In case of investigation financial development and income inequality, econometrical obstacles such as the problem of reverse causality and endogeneity might arise. Considering this in the examination the relationship between financial development and income inequality, one may refer to the literature discussing the political dimension in the inequality and finance nexus.

One of the most prominent theoretical works in this area was written by Rajan (2010) in which he stated that increased credit given to US American households was the direct result of the rising inequality trend over the last two decades. Moreover, according to Rajan (2010) to use traditional forms of redistributive taxation together with the political inability will make it less difficult for authorities to make easier access to credit for the poor.

Other proponents, such as Kumhof and Ranciére (2010), in their theoretical model, explain endogenously how high credit growth and financial crises can be a consequence of rising income inequality.

However, the hypothesis that rising inequality can cause a credit boom is empirically rejected by Bordo and Meissner (2012), who advocate that there is no statistical evidence that rising inequality leads to credit booms. ‘This is of course very important for our study because we ideally want to treat financial development as a variable that is reasonably independent from income inequality’ (Jauch and Waztka, 2012). The issue seems empirically settled by Bordo and Meissner (2012) who proved that this view is incorrect; we believe that our results are not affected by endogenous variable.

From a statistical point of view, GMM has become widely used as it is a convenient way to overcome potential endogeneity problems. Nevertheless, the method is not a perfect one and has as usual several downsides. First, according to Wooldridge (2009), it can be a complicated method that can easily give invalid estimates. Also, it is susceptible to external instruments and does not solve cross-sectional dependence. Moreover, it is not advisable in cases where panels have a long time span.

We have chosen to perform panel data to take into account the individual heterogeneity. The main reason for this is the fact that FE regression explores the relationship between the dependent and independent variables within an entity, such as country and corrects for individual effect. Each country has its own individual characteristics that can influence the explanatory variables. According to Green (2017), using FE technique allows for controlling for the factors within the individual that result from the existing correlation between an entity's error term and predictor variables.

To arrive to a valid statistical model, regressions are first estimated by ordinary least squares (OLS), i.e. pooled regression, followed by random effects models, which are tested against FE models, and then it is verified for whether time FE or first differences need to be introduced. In every model, statistically non-significant variables are dropped as suggested by the general to a specific approach.

In the analysis, a time span of 11 years from 2003 to 2014 is used that contained 52 countries. Estimation results are presented in Table 1. Results for each finance dimension individually (1–5) are similar to the outcomes of the model with variables taken all together in model 6 in terms of direction, size and significance of the coefficients with some small differences.

Variable Domestic credit to private sector (% of GDP) is one of the most commonly used in research about finance and inequality. In Table 1, variable Private credit to GDP alone is insignificant but in case of model 6 it becomes significant at 0.05. The direction of the sign near this variable is also interesting in comparison to other research because its increase leads to higher income inequality. This contradicts results of Beck et al. (2007) and Clarke et al. (2006) that advocate decreasing influence of financial depth (measured by Private credit to GDP) on income inequality. On the other hand, it appeared positive and significant to research conducted by Jauch and Watzka (2012) and de Hann et al. (2017). Our analysis reveals similar results. When Domestic credit to private sector (% of GDP) increases over time within countries by 10% of GDP, the Gini coefficient rises on average ceteris paribus by 0.11 points. This variable is however statistically insignificant at 1%, 5% and 10% level in regression where it is the only explanatory variable of the financial system. This positive significant effect on income inequality can confirm Greenwood and Jovanovic (1990) hypothesis partially regarding rising inequality due to higher financialisation.

An alternative variable for Domestic credit to private sector (% of GDP) for measuring financial depth is used. The Percentage of the number of foreign owned banks to the number of the total banks in an Economy is significant and increase in 1% point leads to decrease of Gini index by 0.12 points and it is statistically significant at the 1%.

Using this alternative measure reveals contradictory results to the previous variable that represented financial depth. However, it is more consistent with research conducted by Beck et al. (2007) and Clarke et al. (2006) that advocate decreasing influence of financial depth on income inequality. Also, it supports the theoretical hypotheses of Galor and Zeira (1993), which states that financial development has a negative impact on income inequality (decrease levels of income inequality). Nevertheless, this inconclusive effect of Domestic credit to private sector (% of GDP) requires looking at other variables that might reveal more in-depth analysis.

When we analyse all control variables, the only significant variable that is negatively correlated with income gap is government spending. It is not surprising since it captures the effects of redistributive policies. This variable in all previous research has the same direction and similar significance.

Claessens and Perotti (2007) and Beck et al. (2007) found evidence supporting the importance of access to finance in reducing poverty and inequality. Especially, Naceur and Zhang (2016) found the negative and significant influence of Number of ATMs per 100,000 adults. In our research, we also confirm its statistical significance and negative relation to Gini coefficient. Considering Number of ATMs per 100,000 adults as the only explanatory variable of the financial system, the coefficient of Number of ATMs per 100,000 adults stays statistically significant in all well-specified models. Increasing the number of ATMs per 100,000 by 100 leads to an average decrease in Gini coefficient by 5 points at ceteris paribus. The results are confirming the importance of financial access in battling with rising inequality because a higher financial inclusion reduces the participation costs of the poorest that is in line with the theory of Greenwood and Jowanovic (1990).

Another significant variable at 5% level is Bank noninterest income to total income (%) (efficiency). Increasing by 1% point Bank noninterest income to total income (%) leads to increase in Gini coefficient by 0.03 points. This is not surprising, since Naceur and Zhang (2016) found that financial efficiency leads to higher income inequality.

Variable Bank concentration (%) (financial stability) is statistically insignificant. This seems interesting since Jeanneney and Kpodar (2011), de Haan et al. (2017) and Naceur and Zhang (2016) establish that financial stability lowers income inequality. Also, it is suggested by the theory that financial crises can cause rising income inequality (Rajan, 2010).

Summing up results for each finance dimension taken separately (regressions [1–5]), It is found that better financial access and more financial depth to individuals leads to lower income gap. Financial stability appeared to be insignificant. However, increasing financial efficiency leads to higher income inequality.

The model comprising all four financial dimensions as explanatory variables gives us similar but not identical results. The best performing model contains country and time FE and the control variables such as government spending, education and unemployment, but the coefficients of the last two are statistically not significant. The primary difference is that Domestic credit to private sector % GDP became significant at 5% level. Increasing Domestic credit to private sector % GDP by 1% point lead to increase in Gini coefficient by 0.03 points.

Increasing the Number of ATMs per 100,000 by 100 leads to decrease on average ceteris paribus in Gini coefficient by 6 points and is statistically significant at 1% level. Percentage of the number of foreign owned banks to the number of the total banks in an Economy (Foreign banks [depth]) is significant and increase in 1% point leads to decrease of Gini index by 0.14 points and it is statistically significant at the 1%. Excluding the insignificant variables basically did not change anything, the significant variables stayed significant. Analysis of the missing values does not reveal any pattern, such that regressions of the unbalanced panel are likely to be consistent.

Substituting the Gini coefficient by another inequality measure, quintile of income distribution gives us more in-depth understanding of how the rich and poor are influenced financial development. Results of the best performing model that includes all four finance variables (model 1E) are shown in Table 2. Again, country FE is included, but introducing time FE does not improve the model.

The sub-sample covers 52 countries from 2003 to 2014 with a total of 361 observations. Estimations by quintile are presented in Table 2. All estimations include country-FE and all except the bottom quintile regression include time-FE. Education and population growth remain statistically non-significant and are dropped in the final parsimonious model as determined by information criteria. Government spending influences income growth at all levels with an impact of between on average ceteris paribus 0.006% and 0.009% points for a 1% point increase in government spending, significant at the 1% level.

Foreign banks and ATM have a positive impact on income growth for all except the bottom quintile and are statistically significant at 1% level. But for the top quintile Foreign banks and ATM have a negative and significant effect, respectively. This supports their decreasing effect on income inequality. On the other hand, for all but the bottom quintile, domestic credit to private sector % GDP has a negative impact on income growth and is statistically significant at 1% level. But for the top quintile it has a positive meaning, that is, it increases their share of income and has a significant effect. This advocate the results of Jauch and Watzka (2012) and de Hann et al. (2017) about increasing influence of domestic credit to private sector % GDP on income inequality. Bank concentration (stability) is significant for Second, Third and Fifth quintile at 1% level of significance.

Then we decided to verify whether in the long and short run effects are similar in terms of significance and direction of the relationship. In the first regression in Table 3, the only significant variables are foreign banks (depth), ATM (access), private credit to GDP and government spending. In the second regression in Table 3, the above mentioned variables are also significant but bank concentration (stability) and population growth are also significant. The size and the direction of the variables remain rather the same in both regressions. What is surprising is that bank concentration (stability) stood out as a variable, which increases income inequality. Increasing financial depth and access reveal the increase in income share for the poor and shrinkage of income for the rich.

It is observed that the most significant variables that lower income inequality appeared to be numbers of ATMs. This confirms the theoretical as well as previous empirical results that financial inclusion is crucial in fighting inequality. Based upon the financial development and the level of financial services, one may conclude from our results that financial deepening will be unfavourable for reducing the level of income inequality if financial services are available to only a few in a given country. Enhancing access to finance is crucial for lowering income inequality and for better quality financial deepening.