1. Measuring Financial Inclusion for Asian Economies

See, for example, Levine (1997) for a survey of this debate.

See, for example, Kempson et al. (2004).

The UNDP classification of countries is available at http://​www.​unep.​org/​tunza/​tunzachildren/​downloads/​country-Classification.​pdf, last accessed in May 2015.

Some of the countrywide surveys of access to financial systems are the Finscope surveys for African and some Latin American countries and the Eurobarometer surveys for the European countries. Honohan (2008) gives a list of the countries for which such surveys are conducted.

While its worldwide coverage of Global Findex database is impressive, the sample for individual countries is small; for example, the samples for India and China are only about 0.0004 % of their respective adult populations. For more on this database, see Demirguc-Kunt and Klapper (2012).

Honohan (2008) uses a regression-based method to estimate these measures for countries where survey-based information is not available; in countries where survey-based information on percentage of adults/households with access to financial services is available, that information is taken directly. These estimates suffer from several limitations some of which are mentioned by the author himself. The first limitation regards the inconsistencies of the survey dates and survey units. The country surveys used in the estimation pertain to different points of time, so there is an inconsistency regarding the date. Further, some of these surveys have adult individuals as the unit (such as the Eurobarometer surveys) while others have households as their unit (like the Finscope surveys). Honohan (2008) uses both interchangeably, simply by stating that “the difference may not be all that great”, although there are reasons to believe otherwise. While estimating the proportion of adults/households with access to financial services, the author uses a log-linear relationship between proportion of financially included adults/households and the number of bank accounts (including number of microfinance accounts). This log-linear relationship is justified by a good fit of this relationship for only 13 countries for which both survey-based proportion of financially included adults/households and number of accounts (bank and microfinance institutions) data are available. However, as in any econometric exercise, such a relationship may not hold true if the data set changes due to a change in the period and/or a change in the number of countries. Thus, these estimates are not easily amenable to computing on a periodic basis to compare financial inclusion over time and across countries.

The Reserve Bank of India reports population per bank branch, population per ATM, percentage of population having bank deposit accounts, credit to GDP ratio etc. to report on the progress of financial inclusion in India. In 2010, the Superintendence of Banking, Insurance Companies and Private Pension Funds of Peru began to develop a set of financial inclusion indicators, with an objective of providing information on access and use of financial products and services. These indicators include number of branches, ATMs and agents per 100,000 adults and per 1000 sq. km., number of depositors and borrowers per 100,000 adults, average size of deposit and credit as a ratio of GDP per capita etc.

The IFI was first proposed in Sarma (2008). Sarma (2010) modified the methodology. In Sarma and Pais (2011) the modified IFI was used to identify country specific factors associated with financial inclusion. Subsequently, the methodology was further improved in Sarma (2012), replacing all previous versions of the IFI. For a discussion on the improved IFI, see Sarma (2015).

This section is largely drawn from Sarma (2012).

Until 2011, UNDP used a simple arithmetic average to compute Human Development Index (HDI), Gender-related Development Index (GDI), and Gender Empowerment Measure (GEM) and a geometric average for computing Human Poverty Index (HPI). In 2011, it revised the methodology for HDI by using geometric mean instead of an arithmetic mean. The Human Development Report 2011 (UNDP 2011) also computes other indices like Inequality-adjusted HDI (IHDI), Gender Inequality Index (GII) and Multidimensional Poverty Index (MPI) that adopt combinations of arithmetic and geometric averages (see e.g. UNDP 2011).

This is similar to the “method of displaced ideal” of Zeleny (1974) in the context of multi-objective optimization programming. In the method of displaced ideal, only the displacement from the ideal point is considered. However, we consider displacement from both the ideal and worst points to compute our IFI, and this makes it somewhat different from the “method of displaced ideal”. The IFI presented in Sarma (2008, 2010) and Sarma and Pais (2011) was based on the distance from the ideal only. This version, presented in Sarma (2012, 2015) incorporates the distance from both the ideal and worst points; thus the present IFI is an improvement over the earlier one and replaces the earlier versions.

The FAS database is an outcome of the initiatives of ‘United Nations (UN) Advisors Group on Inclusive Financial Sectors’, established by the UN in 2006, which decided, in 2008, to involve the IMF and the World Bank in collecting data on access to finance in order to support policy formulation and research. The initial funding for collection of the data was provided by the Government of the Netherlands. In June 2010, the IMF came out with annual data on several indicators of access to finance for the years 2004–2009 on its website. The data used in this chapter was extracted from the website http://​data.​imf.​org/​?​sk=​E5DCAB7E_​A5CA_​4892_​A6EA_​598B5463A34C, last accessed in April 2015.

There may be persons having more than one bank account co-existing with others who may have none. Therefore, number of accounts per capita, is likely to actually provide an overestimation of the proportion of the “banked” population. For example, in 2010, number of bank accounts per 1000 adult people is 2276 in Malaysia, 1324 in Romania, and 1066 in India; this is despite the fact that a significant proportion of population is without bank accounts in these countries.

In this context, it may be noted that Honohan (2008) found a positive and significant association between proportion of banked adults/households and number of bank accounts per 100 adults.

The choice of these weights is motivated by an empirical observation of our data set. In our data set covering 2004–2010, the average ratio of ATM-to-branch per 100,000 adults is found to be 2.13. Thus, on average, there are two ATMs per bank branch, implying that a bank branch, on an average, is equivalent to two ATMs. Thus, the branch index gets a weight of two-thirds and the ATM index gets a weight of one-third in the availability index.

In the literature on the role of finance in economic development, the credit to GDP and deposit to GDP ratios indicate what is known as “financial depth”. In this literature, indicators of financial depth provide a measure of the contribution of the financial system in economic activities. Here, however, we are using these ratios to indicate the volume of credit and deposit generated by the banking system as a measure of the extent of the usage of the banking system due to lack of data on more appropriate measure on this.

For example, the UNDP uses the empirically observed highest observed value as the maximum while computing dimension indexes for the Human Development Index (HDI) (UNDP 2011).

Ardic et al. (2011) estimated that on average, an individual has three deposit accounts in the world. Our choice of Mp is informed both by our data set and the estimates from Ardic et al. (2011). In our dataset, this represents the 90th percentile on the distribution for this dimension.

In our data set this again represents about the 90th percentile on the distribution for this dimension.

In our dataset for 2004 to 2010, we find that the average number of ATMs per bank branch is about 2.13. Our choice of a maximum for ATM (120) being twice the maximum for bank branches (60) is motivated by the above empirical observation. This is about the 92nd percentile observed in the distribution for the ATM dimension.

This represents about 90th percentile observed in the distribution for the usage dimension.

These weights, though they seem a bit arbitrary, are decided based on extensive discussion with banking sector experts and academicians. When appropriate data on all dimensions are available, the weights can be revised accordingly.