Shadows on Main Street: Mapping America's Financial Architecture from the Yellow Pages of the 20th Century

Abstract

We construct a novel and comprehensive database of the branches of all financial services providers, including banks and nonbank financial institutions, at the county level over 1920-2000 from Yellow Pages published in Indiana, Michigan, and Ohio. We find that about one-third of these branches are shadow banks, i.e., non-regulated entities, and that there is strong county-specific persistence in both the number of branches and the share of shadow banks over time. We examine the statistical and economic implications of these findings for extant empirical literature on the effect of finance on real outcomes, which mostly focuses on easily-measured regulated institutions. Ignoring shadow banks in these analyses creates a non-classical measurement error since shadow banks evolve as a complement to regulated institutions. Using banks alone as a proxy for financial development can understate the true effect by as much as 50% or overstate the true effect by as much as 80%, depending on the empirical application. Routine empirical choices such as whether to use the number of branches or their log-transformed values in the regression analysis create large differences in the direction and extent of bias. We provide suggestions for correcting for these biases in future empirical studies.

Keywords: Banking, Shadow Banks, Financial Intermediation, Measurement Error

JEL Classification: G21, G23, L1