Aveneu Park, Starling, Australia

The points in time (2011 and 2014)

The indicators included the providers’ and the users’
sides. On the providers’ side, the index of FIA introduced in Sahay et. al.
(2015a) covered the number of commercial bank branches and ATMs per one hundred
thousand adults. On the users’ side, a number of indicators were investigated:
share of businesses and investment financed by bank credit, share of the
population with account at a formal financial institution by gender and income
groups, share of firms citing finance as a major obstacle, share of adults
using accounts to receive transfers and wages, share of bank borrowers in the
population and finally, the use of insurance products.

 

The main challenge in building a relationship between
long-run growth and financial inclusion was the absence of long enough time
series of financial inclusion (FI) data. For instance, the index of Financial
Institution Access (FIA) assembled by Sahay and others (2015a) had time series
– number of ATMs and bank accounts – from the IMF’s Financial Access Survey
(FAS) starting in 2004 at the earliest. Since the sample period was between
1980 and 2010, which was combined with a five-year average for all variables
(used in order to smooth out cyclical variations) did unfortunately not provide
robust and usable results in a standard GMM growth regression. Within this
framework, FIA only provided two usable time observations (averages 2000–04 and
2005–10). For this reason, GMM regressions of this type cannot test for the
impact of FIA—or other financial inclusion indicators, for that matter— as the
regressions would not pass the standard diagnostic tests. This paper used OLS
estimation for the growth and inequality regressions.

 

In comparison to the FAS data, the Global Findex data are
certainly more comprehensive and would potentially allow for a more robust
analysis. However, the Global Findex data measure FI at only two points in time
(2011 and 2014) with an assumption that relative financial inclusion did not
vary significantly over time. Hence, the Global Findex data could be interpreted
as a ranking rather than an absolute level

 

An ordinary least squares (OLS) estimation was conducted
taking into account a number of countries, relating an FI measure at one point
in time (or averaged over a period) with growth over a period. Ideally, one
would have initial FI related to subsequent growth (as per the early King and
Levine study) to address reverse causality:

 

in which i denotes country and X denotes
controls.  Additionally, one can also include
a financial depth/development variable (FIN) which could either be (i) privy
(private credit-to-GDP), (ii) FID (index of financial institution depth), or (iii)
FD (the broad financial development index).