Revisiting structural change and market integration in late 19th century American capital markets

Abstract

This article draws on a variety of time series tools to more deeply explore issues surrounding the emergence of a national capital market in the late 19th century. Our focus is on the timing of the emergence of a national capital market. Rather than relying on the absolute narrowing of regional interest rate differentials, which is a common approach in this literature, we use Gregory and Hansen cointegration tests, which allow us not only to test for cointegrating relations in the interest rate series but to identify unknown structural change dates as a byproduct. We also use dynamic conditional correlations to determine the dates at which regional interest rate correlations began increasing. Our results suggest that structural changes are centred around the year 1900, which is consistent with Sylla's argument that the lowering of capital requirements by the Gold Standard Act of 1900 increased bank entry and competition and facilitated regional capital market convergence.

Acknowledgements

The authors thank Thomas Weiss and Howard Bodenhorn for valuable comments on an early draft of this article. We also appreciate the use of Bodenhorn's interest rate data. K. Choi acknowledges helpful suggestions made by Chulho Jung and B. Dupont would like to acknowledge the Department of Economics at Wellesley College, where he was a visiting assistant professor during some of this research.

Notes

¹ See Wright, Old South, New South and Ransom and Sutch, One Kind of Freedom for some explanations of why the Southern capital markets were so radically different from the rest of the country.

² Previous work on cointegration and market integration includes Ravallion (1986), Kim (1990), Goodwin (1992) and Alexander and Wyeth (1994).

³ A time series variable is defined as (weakly) stationary if the mean and auto-covariances of the series do not depend on time. If any series violates the stationary properties, it is said to be a nonstationary time series variable.

⁴ In other words, when two variables y _{1 t} and y _{2 t} are cointegrated, they have a common stochastic trend.

⁵ It is easy to infer that there are six possible combinations in our model.

⁶ ϵ _t should be an I(0) process.

⁷ The following equation generates the ADF test: where y_t is a variable, ϵ _t is the error term and Δ is a difference operator.

⁸ All of the unit root tests for the first differenced series are rejected at the 5% level. Testing results of the first differenced series are available upon request.

⁹ Bodenhorn (1992) and LaCroix and Grandy (1992) found that financial markets were integrated in antebellum America but Davis (1965) found poor integration in the immediate aftermath of the Civil War so this is perhaps properly thought of as re-integration of capital markets.

¹⁰ The Federal Reserve Act was approved and signed into law in December 1913.

¹¹ Dynamic conditional correlation model required the stationary variable and we found that the level of four different interest rates has unit root. Therefore in this analysis, we use the first differenced of interest rates.