A Composite Digital Financial Development Index via Principal Component Analysis: Construction, Weighting Robustness, and Multi-Indicator Comparison

Abstract

This paper builds an index of digital financial development for 59 countries on the basis of four real indicators: account ownership, domestic credit to the private sector, GDP per capita and internet penetration, by means of the use of principal component analysis (PCA). The first principal component accounts for 63.8% of the total variance across the four standardized indicators, with the loadings evenly distributed across all four indicators (0.44 to 0.54) based on both Kaiser Eigenvalue and scree testing. Three empirical analyses were added as a result of being requested by double-blind review for this revision. First, the rankings of the countries derived from a PCA are virtually equal to those obtained from a simple equal-weighting approach (Spearman ρ = 0.9992; maximum difference in ranks across 59 countries: 2 positions), directly showing that the results reported in this paper are not specific to the chosen weighting strategy. Second, the composite index is individually as strongly correlated with GDP per capita (ρ = 0.86) as with account ownership (ρ = 0.87) but less with domestic credit (ρ = 0.72) and internet penetration (ρ = 0.78), explaining that overall variation in the index is equally well explained by variation in income and account ownership, while the variation in domestic credit and internet penetration is more of a deviation from a simple development story. Third, the shift in percentile of the composite index from a sample median to an observed extreme is calculated directly for a controlled hypothetical example in this case: the movement of domestic credit from its sample median to its observed extreme, which corresponds to a shift from the 49th percentile to the 88th percentile of the composite index. The composite index remains highly correlated with a simple GDP per capita rank (Spearman ρ = 0.86), but countries like Thailand, China and Malaysia are performing much better than their incomes would suggest, largely due to unusually developed domestic credit markets, while Colombia and Albania are performing much worse, largely because of comparatively low levels of account ownership relative to income.