6.2.1Shift-Share Analysis

It was only in the late 1950s that economists started to study the differences in economic performance between U.S. regions. Edgard Dunn (1960) introduced a new technique called shift-share analysis, which distinguishes between growth factors operating uniformly at a national level, and specific growth factors. The first step of the technique is to compute the expected employment level in each region in a target year, say 1990, based on the percentage increase in total employment in the nation between a base year, say 1980 and the target year. The gap between the expected and the actual target year total employment level in each region is called the net shift, and is expressed as a percentage of the base year total employment. Dunn then divides the net shift into differential and proportional components. The differential shift is the sum of the shifts by industry sectors for the region, where each industry’s share is defined as the difference between regional and national growth in the industry. The proportional shift is merely the difference between the total and differential shifts. While Stilwell (1970) and others noted that the technique of shift-share analysis has little basis in the theory of regional growth, the technique remains a useful descriptive tool as when studying historical data on population or employment for instance.

Norcliffe (1977) built an interesting model for disaggregating regional productivity performance into average, mix, scale and residual effects. This model is similar in form to shift-share analysis but with the addition of two structural effects. The structural effects are incorporated in the productivity component associated with the particular mix of industries in a region on one hand, and the size distribution of establishments within each industry, on the other hand. Thus, in Norcliffe’s model, regional productivities differ from each other in four ways: averages (the difference with global productivity), mix and scale effects (the structural effects), and residuals.