Measuring Financial Inclusion in Southeast European Countries Using Multidimensional Index of Financial Inclusion

This paper provides an insight into measuring financial inclusion through a multidimensional index of financial inclusion in Southeast European countries (SEE). We used a two-stage principal component analysis to extract dimensions of financial inclusion. Data were obtained from two sources, the World Bank Global Findex Survey (WB-GFS) data base and the International Monetary Fund Financial Access Survey data (IMF-FAS), for twelve SEE countries for the years 2011, 2014, 2017, and 2021. The research confirms that financial inclusion can be measured using two dimensions in terms of access as one factor and usage and availability as the second factor. Practical implications of this research are in ensuring an adequate measure of the level of financial inclusion for SEE countries that can be used for further research related to understanding the underlying factors contributing to financial inclusion, barriers to financial inclusion as well as the impact of financial inclusion on economic growth and poverty alleviation.


Introduction
Financial inclusion and inclusive financial systems have been perceived as sine qua non for sustainable economic development, decrease in inequality, and poverty alleviation for several decades now.The World Bank (2018) declared that universal financial inclusion needs to be achieved by 2020, yet countries still struggle to create financial ecosystem that is accessible for all.According to the latest World Bank Global Findex Survey (WB-GFS) from 2021, account ownership around the world increased by 25% in the 10 years spanning 2011 to 2021, from 51% of adults to 76% of adults (Demirgüç -Kunt et al., 2022).From 2017 to 2021, account ownership in developing economies increased by 8%, but still there are more obstacles to achieving universal financial inclusion in developing economies due to a number of barriers and lower efficiency of financial system.
A precondition for fostering financial inclusion is understanding the driving factors of financial inclusion and providing adequate measures.There is extensive academic literature providing methodological approaches to measuring financial inclusion.The most prominent authors are Sarma 2008, 2012and 2014and Amidzic et al., 2014.Methodological approaches are based on using factor analysis or principal component analysis for index construction.Based on these approaches, this paper provides comprehensive literature review on existing methodologies for measuring financial inclusion, in order to develop the index of financial inclusion (IFI) for Southeast European (SEE) countries.The main research idea is to create the benchmark for future investigation on the level of financial inclusion in SEE countries and to better understand the driving factors of financial inclusion in SEE countries.
Based on the previous research, this paper lays out the methodological approach to constructing a multidimensional financial inclusion index that can be used in future research to track progress and impact of financial inclusion on sustainable economic development, energy efficiency, environment quality, and other Sustainable Development Goals SDG goals among SEE countries.
The rest of the paper is structured as follows.In the second part, the literature review on measurements of financial inclusion is provided focusing on the methods and variables (dimensions) used to measure financial inclusion.In the third part, the methodological framework is givendata source, the variable used and method for index construction.The fourth part presents the results and discussion while in the fifth part the conclusion is given as well as suggestions for further research.

Literature Review on Financial Inclusion Measurement
Financial inclusion is a rather broad socio-economic concept.Early attempts to define financial inclusion focused on financial exclusion as an antipode to financial inclusion, referring to the process where poor and disadvantaged are prevented to access financial system (Leyshon & Thrift, 1996) and the inability to access necessary financial services in an appropriate form (Sinclair, 2001).The World Bank (WB) (2022) provides the most comprehensive definition of financial inclusion in which financial inclusion means that individuals and businesses have access to useful and affordable financial products and services that meet their needstransactions, payments, savings, credit, and insurancedelivered in a responsible and sustainable way.
Recognizing the need for addressing the issue of financial inclusion, academic literature focused on measuring financial inclusion and identifying key drivers for building inclusive financial system has been growing in the last two decades.Early attempts to measure financial inclusion focused on defining a single indicator or a set of indicators to describe financial inclusion.The data measuring financial inclusion can broadly be divided into two groups: supply-side and demand-side data.
The supply-side data collected from regulatory banking authorities, such as the number of commercial bank branches per 100,000 adults, number of ATMs per 100,000 adults, number of depositors with commercial banks per 1,000 adults, and the aggregate number and value of bank loans and deposits were firstly used by Back et al. (2007).Today, these data are collected by the International Monetary Fund (IMF).The IMF Financial Access Survey (IMF FAS) covered the set of indicators focusing on the supply side, such as: number of bank accounts, number of commercial bank branches per 100,000 adults, number of ATMs per 100,000 adults, outstanding deposits with commercial banks (% GDP), outstanding loans from commercial banks (% GDP), and small and medium sized enterprise (SME) outstanding loans from commercial banks (% GDP).With the increase of electronic banking and development of FinTech companies, today the IMF (2019) also collects the data on the number of registered mobile money agent outlets per 1,000km 2 , number of registered mobile money accounts per 1,000 adults, and value of money transactions as % of GDP, as core indicators measuring financial access.
The most commonly used demand-side indicator of financial inclusion is a simple measure of the proportion of adult population (or households) within a country that have access to formal financial product(s)/service(s), usually a bank account at formal financial institution.Honohan (2008) introduced this indicator as the estimation of the proportion of households having access to financial services from secondary data.As Sarma (2012) suggests, the main limitation of this indicator is that it does not take into account the actual usage and quality of financial product(s)/services(s).
To further describe financial inclusion and to follow up on financial inclusion development, the set of new micro-level demand-side indicators was developed, which are systematically being collected nowadays by the WB.The WB initiated data collection in 2011 through its GFS.These indicators include demand side level data on account ownership and usage, borrowing and savings practices, quality of financial products/services, mobile payments, barriers to having account, etc.
Even though a significant number of both supply and demand side indicators have been developed over the years, there is still an ongoing debate if one single indicator or set can adequately capture financial inclusion complexity.Following that argument, Sarma (2008Sarma ( , 2012) ) was first to develop the methodology for creating the IFI.Sarma proposed creating a multidimensional index that would be based on both supply side datamacroeconomic parameters related to banking sector outreach and demand side datamicro level indicators related to access, usage and obstacles to using financial products/services by the general population.
Building on Sarma's work, methodological approaches to construct an IFI can be divided into two groups.One group follows the adapted methodology used by the United Nations Development Program (UNDP) for computation of Human Poverty Index, Human Development Index, Gender Development Index (Gupte, Venkataramani & Gupta, (2012) Most of the above-mentioned researches uses similar sets of variables grouped into two or three dimensions to construct the financial inclusion index.Grouping is done prior to deploying factor analysis.The majority of researchers (Sarma 2008, Sarma 2012, Amidzic, et al. 2014, Sarma 2016, Camara & Tuesta (2018), Goel & Sharma (2017) Park & Mercado (2018), Nguyen, (2021) use three dimensions: access (or penetration), availability and usage, Borhan et al., (2021) use two dimensions (access and usage) while Gupte, et al. (2012) use as many as four: outreach (penetration and accessibility), usage, ease of transactions, and cost of transactions.Camara & Tuesta, (2018) even introduce the barriers dimension.Approaches differ due to the availability of indicators and data.

Sample and Data
The initial set of variables was identified based on the literature review and data availability for SEE countries, namely: Albania, Bosnia and Herzegovina, Bulgaria, Montenegro, Greece, Croatia, Kosovo, Romania, North Macedonia Slovenia, Serbia, and Turkiye.While there are many indicators describing financial inclusion, data availability was the main limitation to include a larger number of financial inclusion indicators.The full list of variables is given in Table 1.The data sources are the WB-GFS data base and the IMF-FAS for twelve SEE countries for the years 2011, 2014, 2017, and 2021.Prior to performing two-stage PCA, the variables used for index construction were normalized using min-max normalization method, as proposed by Sarma (2008) and used by Amidzic et al. (2014), Nguyen (2021) and others, to scale data in the range between 0 and 1 using the following formula: where: The first-stage PCA was used for identifying and grouping the variables in relevant dimensions.Then, the weights of the indicators representing dimensions were extracted and estimated.The estimation of factors loading was obtained using rotation of the axes using the varimax technique.Based on the results of the first-stage PCA, two dimensions (factors) were extracted.
In the second-stage PCA, the weights for each dimension were calculated using unrotated matrix and the overall financial inclusion index by using the dimensions as explanatory variables were created.
IFI was then created based on the following formula: where : IFI icomposite index of financial inclusion of country i, w irelative weights of each dimension, e ivariation due to error, and   1 and   2two dimensions of financial inclusion of country i.
The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was tested before conducting Factor Analysis to further verify adequate correlations between the indicators.Source: Authors' calculations

Results and Discussion
Table 5 shows the results of the first-stage factor analysis -weights derived from matrix of rotated factor loadings.In contrast to Sarma (2008, 2012& 2016), Amidzic, et al. (2014), Camara & Tuesta, (2018), Goel & Sharma, (2017) Park & Mercado, (2018), Nguyen, (2021) who extracted three dimensions of financial inclusion, based on the data for SEE countries, two sub-dimensions were estimated in our study.Dimensions extracted for SEE countires, "access" and "usage and availability", are consistent with the dimensions extracted by Borhan et al., (2021).

Source: Authors' calculations
As it can be observed, the access dimension is represented by three indicators: percentage of individuals with an opened account at financial institutions, number of deposit accounts per 1,000 inhabitants, and outstanding loans.The other dimension is comprised from the set of two indicators related to availability (the number of branches and the number of ATMs per 100,000 inhabitants) and usage (usage of debit account, borrowing, and saving).
In the second-stage PCA, the weights for two obtained factors were calculated in order to construct the index.
Table 6 shows the results of the second-stage factor analysis and obtained factor loadings.It was expected to obtain equal wights since the first-stage PCA analysis extracted two factors.This result also shows that Sarma's (2008) approach to equal weighting is appropriate, since the load factors obtained for two dimensions were equally weighed as well.The KMO statistics for the first-stage PCA was 0.573, which was more than the minimum required threshold (0.50).This confirms that the indicators were correlated and suitable for structure detection using factor analysis.owest, with the value of 0.084.Sarma (2008), the country membership to one of these three categories depends on the value of the IFI and it is given in a scale below: 1. 0 ≤ IFI < 0.3 -low financial inclusion, 2. 0.3 ≤ IFI < 0.5 -medium financial inclusion, and 3. 0.5 ≤ IFI ≤ 1 -high financial inclusion.
Based on the scale provided above, in the observed period Slovenia, Greece, Croatia and Bulgaria on average had a high level of financial inclusion, but also these countries retained the high level of financial inclusion during the whole-time period.Turkiye, Montenegro, Serbia, and North Macedonia on average had a medium level of financial inclusion, but in 2021, three countries -Turkiye, Montenegro and Serbia had a high level of financial inclusion.On average, Bosna and Herzegovina, Albania, Romania, and Kosovo had a low level of financial inclusion, but Bosnia and Herzegovina and Romania scored over 0.3 on the financial inclusion scale, which put them among the countries with a medium level of financial inclusion.
Overall, the improvement of financial inclusion in SEE countries can be observed during the ten-year period.Figure 1.shows that the number of countries with high financial inclusion increased from 4 to 7, while at the same time, only two SEE countries in the SEE region had a low level of financial inclusion in 2021.These results are consistent with the previous research on financial inclusion worldwide (Sarma, 2016;Park & Mercado, 2018;Bohran et al, 2021), which also found that the number of countries with a high level of financial inclusion is on the increase while at the same time, the number of countries with low financial inclusion is on the decrease.Source: Authors' calculations

Conclusions and Future Research
There is a consensus among scholars and policy makers that financial inclusion is one of the key drivers of economic development.In that respect, efforts to increase financial inclusion need to be in line with the empirical evidence of the underlying factors driving the increase of financial inclusion.During the last several decades, many different indicators, both supply and demand side, were collected.These data need to be analyzed so as to provide actionable guidelines for financial inclusion policies.Therefore, the IFI has proven to be an adequate measure of financial inclusion, but academic papers provide different methodologies for its calculation.
Based on the existing methodological framework, using the WB and IMF collected data for four time points (2011, 2014, 2017 and 2021) and the weights extracted from a two-stage PCA method, we created an overall IFI for SEE countries.This index is comprehensive as it uses demand and supply side data to track the progress of financial inclusion in SEE countries.To our knowledge, this paper presents the first IFI calculated for SEE countries.
Two-stage PCA method proves to be an adequate statistical approach for the construction of IFI, showing that financial inclusion in SEE countries is determined by two factors, one related to access as the dimension of financial inclusion and the other related to availability and usage of financial products/services.Furthermore, PCA confirms that equal weighting of the factors for constructing the index can be used, as previously proposed by Sarma (2008).
The most obvious contribution of this paper is that it uses both supply side and demand side data for computing the IFI.Also, this research contributes to existing literature on the development of IFI, confirming earlier studies that financial inclusion is determined by access, availability and usage of financial products/services offered by formal financial institutions.It contributes to better understanding of the efforts to increase financial inclusion in the SEE region in terms of simply capturing the progress measured by IFI or investigating the relationship between IFI and the relevant macroeconomic variables such as GDP, GDP per capita, unemployment or inflation.
It is also a useful tool for policymaking and policy evaluation of financial inclusion initiatives, but also for financial institutions and Fintech companies to improve their efforts in promoting financial inclusion.
Building on the developed IFI, future research will use the constructed index to analyze contribution of financial inclusion to economic development, unemployment rates, inflation and poverty alleviation among SEE countries.Furthermore, the index will be used to better understand the barriers to financial inclusion in order to provide recommendation to policymakers for building more inclusive financial systems in SEE countries.
The main limitation of this research and, in general, the researches on construction of IFI are limited data series and limited number of indicators used for the construction of the index over several years.
d ithe normalized value of indicator i, A ithe actual value of indicator i, m ithe observed minimum value of indicator i, M ithe observed maximum value of indicator i.

Figure 1 .
Figure 1.Countries in respect to the level of financial inclusion index per years (number of counties) Note: The total number of SEE countries = 12

Table 2 .
Table2provides descriptive statistics of the indicators used for constructing the index for the overall sample.The data show that on average, 64% of adults in SEE countries have an open account at financial institutions, over 39% of them use debit cards, while just above 20% borrow and 14% save money.The average number of deposit accounts for SEE countries per 1,000 adults is 1,845 accounts, which means that on average one person has more than one account opened at financial institutions.There are approximately 29 bank branches and 66 ATMs per 100,000 inhabitants in SEE countries.Descriptive statistics of indicators used for the construction of IFIoriginal data Table3provides descriptive statistics of the indicators used for the construction of the IFI per years, aggregated for all SEE countries.As it can be observed, the value of the indicators increases over the years for all selected indicators, except for the number of bank branches which is consistent with a general trend in the banking industry related to a shift towards electronic banking.

Table 4 .
Descriptive statistics of indicators used for the construction of IFI per years for SEE countries aggregated

Table 7
presents the computed IFI values for SEE developing economies for the years 2011-2021.As it can be observed from the Table, Slovenia has the highest level of financial inclusion with the average value of 0.903 and Kosovo's level is the lowest, with the value of 0.084.

Table 7 .
Values of the IFI for SEE countries, 2011-2021