Mapping Africa Financial Inclusion- Historical Efforts and MIX's New Approach

 In years past, there have been several attempted efforts to develop supply-side estimates of global financial access. As the methodology behind the Africa landscape data both builds on and differs from these efforts in some respects, we will review several of their components.

In a 2004 review of financial institutions with a “double bottom line,” CGAP identified “alternative financial institutions” that “focus to some degree on extending financial services downward from the economic level of the traditional clients of commercial banks.” They note that, while these institutions reach a large number of clients globally, we lack demographic information on the poverty level of those clients, and thus cannot directly determine the level of supply of financial services to the poor. However, broadening the scope of coverage beyond “traditional” microfinance institutions, in the manner of this study, can still yield insight about the likely distribution of financial services. Following this model, MIX adopted a similar approach in our data collection. [Fn1]

In a 2005 paper, Patrick Honohan reviews existing data on access to finance and recommends four channels for data collection, including data collection from “providers of financial services (whether directly or, probably better, through national regulators)” for supply figures, with a broad scope to include “formal commercial banks, national postal savings or agricultural banks, cooperative credit unions and other entities which do not consider themselves to be specialized microfinance institutions.” Data for these providers may come from stand-alone surveys, national statistical agencies or regulators and from the providers themselves (who have the most detailed information on products and costs). He notes limitations to this method, though, as mainstream financial institutions and “regulatory agencies such as central banks generally have little interest in collecting” client information and may focus primarily on monetary aggregates.

A 2010 review on “Financial inclusion measurement for regulators,” published by the Alliance for Financial Inclusion (AFI), further investigates existing measures of financial access and notes that a first step should be to “take an inventory of existing data collection efforts,” followed by a consideration of use of the data by stakeholders. Surveys involve costs and trade-offs and new efforts must take these into account. Supply-side, rather than demand-side, efforts “may be the easiest and least expensive way to collect financial access data,” but are contingent on regulators or central bank already collecting this data. As we see later, in sub-Saharan Africa, this condition is not met in many markets or for many providers.

Most recently, CGAP and the IMF have launched financial access surveys with online datasets accessible to practitioners, centered on a small set of comparable indicators. Both data sets are global in scope and cover multiple years, although with a primary focus on regulated financial institutions and without segmenting ‘double bottom line’ institutions. (The IMF data focuses exclusively on regulated institutions, while CGAP incorporates data on some unregulated institutions, including microfinance institutions and cooperatives.) Later in this article, we compare the datasets for CGAP, IMF and MIX in more detail.

The focus of this effort is on products and services that are specifically targeted towards or likely to be used by the poor. However, the body of research and the data resources described above yield the following guidelines:

• Focus on a broad range of financial services providers, including unregulated or informal providers.
• Consolidate data from existing national and international resources, stand-alone surveys, and directly from providers to leverage the most data and minimize duplication.
• Collect a small number of easily-comparable indicators.
• Develop methods to mitigate data gaps, in particular a lack of client information from many sources.
• Create a detailed database that links a user to deeper data on products, costs and other factors whenever possible and facilitates auditing and updates for the data.

In addition, we wish to emphasize the process of building and using this data set, as much as the results. Financial services continue to evolve and practitioners and policy-makers need up-to-date, actionable information. Consequently, we provide detailed data behind the high-level numbers, to allow users to investigate sources and identify gaps. A sustainable process for measuring financial access for the poor requires strengthening local data collection efforts and we wish to recognize the impact of those efforts in each market in sub-Saharan Africa.

Comparing existing financial access surveys in Africa

With the availability of multiple resources tracking access to finance around the world, we have a good opportunity to compare these resources and what they tell us about sub-Saharan Africa. We will focus on the IMF and CGAP databases, comparing these to the existing MIX Market data (focused on specialized microfinance institutions (MFIs)) and this landscape data set. Although these measure slightly different things, the comparison can still be informative.

We begin with a comparison of the data on a single country, Mauritius. While it is a small and somewhat non-standard sector, it is an object lesson in the different approaches. Considering these types of outliers is also important if we wish to carry out cross-country statistical or econometric research using the data. Table 1 shows the key metrics for Mauritius from the IMF, CGAP and MIX surveys. (MIX Market has no data on MFIs in Mauritius.)

IMF data from: http://fas.imf.org/ for 2010; CGAP data from: http://www.cgap.org/p/site/c/template.rc/1.11.142569/.

We can see that both the IMF and CGAP surveys cover only commercial banks, which are disjoint samples from the MIX survey, which covers only the cooperatives sector in Mauritius. While the IMF and CGAP surveys cover nominally the same samples, the metrics do not agree across the two studies. In addition, we note the very high levels for credit and deposit metrics from both - by the IMF figures, the average adult in Mauritius has 2.2 loans outstanding! Although we cannot easily track the data back to the original sources, it seems likely that this high outreach is due to the large offshore banking sector active in Mauritius. The MIX data set explicitly excludes this type of activity, which is not targeted at low-income populations. In its place, we have included publicly-available data on the cooperatives sector, via WOCCU, which is missing from the other surveys. (Data and sources for the MIX survey can be found directly in this document.)

Taking these differences into account - the inclusion of commercial financial services providers in the IMF and CGAP surveys and exclusion from the MIX survey - we can still make some basic comparisons of coverage and key metrics. The maps in the following three figures show the number of providers, and the metrics on loans and deposits per 1000 adults from each survey. We also include data from the MIX Market site for 2009, which is focused primarily on specialized microfinance institutions, as a further point of reference. [Fn2]

Fig. 1: Number of providers, by survey

Fig. 2: Loans per 1000 adults, by survey

Fig. 3: Deposits per 1000 adults by survey

We can see that coverage is broadest from the MIX landscape survey, reaching 45 countries in sub-Saharan Africa. The IMF survey has particularly sparse coverage in Africa, reaching only five countries for credit metrics (including Mauritius and the Comoros). The CGAP survey has broader coverage, but with gaps in Central and parts of East Africa. As is to be expected, the MIX landscape data yields lower estimates in countries with stronger commercial banking sectors, particularly in Southern Africa - South Africa, Botswana and Namibia.

We can also check the correlations between the different efforts - if the results are highly correlated, it confirms that the methodology of each supports similar conclusions. Unfortunately, at least for sub-Saharan Africa, cross-survey correlations are low in most cases. The tables below display correlation matrices for three key metrics: number of providers, loans per 1000 adults and deposits per 1000 adults. [Fn3]

Correlation matrices for cross-country financial access surveys

Not surprisingly, MIX landscape data is highly correlated with MIX Market data, since this data is included in the landscape survey. Correlations are otherwise highest with the IMF on credit metrics, although this is also where the IMF sample spans the fewest institutions (5), including the outlier for Mauritius. The CGAP and MIX surveys have moderately high correlations on deposit metrics, likely due to the inclusion of cooperatives, MFIs and state-linked institutions in both surveys. Overall, it is clear that, while major strides have been made within the past few years to have these data sets publicly available, much work needs to be done to make the methodology more robust across surveys and to truly capture access for the full population of developing markets. 

Related Articles:

• Overview
• Data Methodology
• Results Review and Next Steps

Fn1:  One important detail from the 2004 review, is that, in most cases, savings banks were excluded from the results due to a lack of information on outreach or accounts. While this data gap persists, we have estimated outreach using available benchmarks as outlined in a later section.

Fn2 We have excluded the monetary metrics since we expect the MIX data to have substantial gaps from the other surveys due to the inclusion of large commercial banks.

Fn3:  Correlations may not be significant for low-sample size comparisons. We do not provide significance tests in these tables.