Religion, social capital and business bankruptcy in the United States, 1921–1932

Abstract

We consider the value of social capital that derives from membership in a church. American states with larger churchgoing populations had lower business bankruptcy rates from 1921 to 1932, and states in which the churchgoing population was concentrated in few churches had business bankruptcy rates that were lower still. Both voluntary and involuntary bankruptcy were lower in states with higher church membership. The evidence suggests that church membership acted on bankruptcy through a safety net mechanism and not solely through indicating a preference for honouring commitment.

Keywords:

• business bankruptcy
• church membership
• social capital

Acknowledgements

We thank Jeremy Atack, Robert D. Putnam and David Skeel for their comments on earlier drafts. Pamela Walker Laird made invaluable editorial contributions. Remaining errors, of course, belong to us.

Notes

1. Tolles (1963) discusses how Quakers dealt with the insolvency of their members.

2. For a discussion of changes in the relative significance of business and wage earner (personal) bankruptcy, see Hansen and Hansen (2007).

3. For a summary of the empirical literature on personal bankruptcy rates, see Hansen and Hansen (2007).

4. The Herfindahl index (also called the Herfindahl–Hirshman index) was developed in the literature on anti-trust economics. There the index is used to measure the size of individual firms relative to the size of the industry, and it used as an indicator of the amount of competition among firms in an industry.

5. Because the bankruptcy rate can be thought of as the number of filings out of the population of people who might file, we also considered a binomial regression specification. The binomial specification does not fit our data well.

6. The full model is r_st  = α + C_st β₁ + X_st β₂ + L_s γ + u_s  + ε _st , where r_st is the business bankruptcy rate in each state s and year t. The vector C_st includes our measures of church membership and β ₁ is the estimate of primary interest. The vector X_st includes economic variables, and L_st describes the laws of each state. Year effects are included. The augmented procedure requires that we assume that the state-specific effect u_s is correlated with at least one of the variables in C or X and one of the variables in L.

7. $200 is about equal to one standard deviation in our observations of state income, expressed in 1920 dollars. $200 in 1920 dollars is equivalent to about $2000 in 2007, with adjustments for inflation made using the Consumer Price Index.