DOES INDUSTRIAL DIVERSITY ALWAYS REDUCE UNEMPLOYMENT? EVIDENCE FROM THE GREAT DEPRESSION and AFTER

Abstract

A portfolio model of employment predicts that cities with more diversified employment opportunities should experience lower unemployment rates than less diversired cities. Empirical analysis of the diversity-unemployment relationship using Census data support the portfolio theory for the years 1950, 1960, and 1970. During the early months of the Great Depression, however, industrially more diversired cities experienced higher, rather than lower, rates of unemployment. By combining the portfolio model of employment with the Lucas-Phelps islands model, the anomalous effect of diversity in 1931 is explained as the result of employers' diffculty of distinguishing real from nominal shocks. I. INTRODUCTION The high unemployment of the 1930s continues to engage the imaginations of economists and economic historians. Although a large and impressive literature now covers most aspects of the Great Depression, few studies discuss geographic variations.[1] Variations in unemployment across cities are, nonetheless, of interest. The cross-sectional variation in unemployment makes it possible to study the effect of industrial diversity on unemployment. Industrial diversity is the evenness with which employment is distributed across industries within a city. Suppose that a nation's employment is divided into, say, twenty industries. A city is defined to be perfectly specialized when all employment is concentrated in a single industry, and perfectly diversified when each of the twenty industries employs an equal share of the city's labor force. Recent empirical work by Simon [1988] and Diamond and Simon [1990] shows that during the late 1970s and early 1980s, cities with more diversified employment opportunities experienced lower rates of unemployment. The results from the 1970s and 1980s raised the possibility that some part of the extremely high unemployment of the 1930s could be explained by the low degree of industrial diversity in many American cities. Indeed, in the 1930s, articles by McLaughlin [1930] and Tress [1938] suggested the possibility of an inverse relationship between diversity and unemployment.2 We tested the relationship between unemployment and industrial diversity at the onset of the Great Depression in 1930 and 1931. We found, much to our surprise, that industrially more diverse cities experienced higher rates of unemployment in 1931. The surprising effect of employment diversity on unemployment in 1931 led us to speculate that the effects of diversity might differ under different macroeconomic conditions. Specifically, we hypothesize that industrial diversity has two effects. Industrial diversity normally reduces unemployment through what we call a portfolio effect. We argue that the anomalous effects of diversity on unemployment at the onset of the Depression could have resulted if diversity of employment opportunities slowed essential wage adjustments required as a result of the massive decline in aggregate demand. During normal periods, the portfolio effect dominates and diversity reduces unemployment. The magnitude of the Great Depression could have caused the wageadjustment effect to dwarf the portfolio effect, thus reversing the normally inverse relationship between diversity and unemployment. II. DIVERSITY AND UNEMPLOYMENT The Portfolio Effect The risk-reducing property of diversity is well-known to economists. Conroy [1975] first applied portfolio analysis to local labor markets. Simon [1988] developed a simple heuristic theory of the portfolio effect of industrial diversity. Imagine a multi-industry world where only real shocks affect the demand price of firms. All output is assumed to be sold in the national market. The real shocks are assumed to be equivalent to observations on a random variable with zero mean, independently and identically distributed across industries. The shocks may be temporary or permanent. Although the distribution of shocks is known, the duration of any particular shock is not. …