Skip to content
Back to search results

Shift-Share Analysis

Introduced by
Edgar S. Dunn (Daniel Creamer credited with early use)
Year
1960
Type
Descriptive decomposition of regional growth
Family
Regional economic analysis
Font size

What does it do?

Shift-share analysis is a descriptive technique that decomposes the change in a regional variable — most often sectoral employment — into three additive components: the part attributable to overall national growth, the part attributable to the region's industry mix, and the part attributable to the region's own competitive performance.

Formalized by Edgar Dunn in 1960, it answers whether a region grew because the national economy grew, because it specializes in fast-growing industries, or because its industries outperformed (or underperformed) their national counterparts.

The intuition behind the method

Suppose a region's employment grew 8% over a decade. Was that impressive? Shift-share asks three sequential questions. First, the whole nation grew 5%, so 5 points are just the rising tide lifting all boats — the national share. Second, this region happens to specialize in technology, a sector that grew faster than average nationally, adding a favorable industry-mix bonus. Third, whatever growth remains beyond what national trends and the region's industrial composition predict must reflect something local — better policy, lower costs, agglomeration — captured as the competitive or regional share. The three pieces add up exactly to the observed change.

History

Shift-share analysis emerged in U.S. regional economics in the 1940s and 1950s — Daniel Creamer is often credited with early applications — and was given its rigorous additive formulation by Edgar S. Dunn in his 1960 paper in the Papers of the Regional Science Association. Jose Manuel Esteban-Marquillas reinterpreted the technique in 1972, introducing a homothetic employment benchmark and an allocation effect to disentangle the industry-mix and competitive components. The method has remained a staple of regional and urban economics for decades and is the conceptual ancestor of the shift-share (Bartik) instrument now central to applied causal labor economics.

When to use it

Use shift-share analysis as a first descriptive pass when comparing the growth of a region (or any sub-unit) against a reference area and you want to separate macroeconomic, structural, and local-competitiveness contributions. It requires only sectoral counts at two points in time for the region and the benchmark, making it cheap and widely applicable in regional economics, labor-market studies, and industry analysis. It is a descriptive accounting decomposition, not a causal model: the competitive-share residual absorbs everything not explained by national trends and industry mix, so it should not be read as a clean estimate of policy effects. For causal identification of local labor-demand shocks, the related shift-share (Bartik) instrumental-variables approach is used instead.

Limitations and considerations

  • Purely descriptive: the competitive-share component is a residual that mixes all unexplained factors and is not a causal estimate.
  • Results depend on the chosen base year, sectoral aggregation, and the start-versus-end period weighting (which classic component carries known weighting biases).
  • The classic three-way split confounds the industry-mix and competitive effects when a region is specialized; the Esteban-Marquillas reformulation addresses this.
  • It is a static comparison of two endpoints and ignores the path and timing of change between them.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Exact additive identity with three economically meaningful components and no residual, making results easy to communicate.
  • Minimal data requirements — sectoral counts at two time points for a region and a benchmark — so it is inexpensive and broadly applicable.
  • Separates the influence of national trends and industrial composition from genuinely local performance.
  • Serves as an intuitive descriptive benchmark and a stepping stone to more formal regional models and causal designs.

Applications

  • Explaining why a region's employment grew faster or slower than the nation by separating national, structural, and competitive forces.
  • Comparing the economic performance of cities, states, or countries against a common benchmark.
  • Analyzing the sources of change in exports, output, productivity, or emissions by industry over time.
  • Diagnosing whether regional decline reflects an unfavorable industry mix or a loss of local competitiveness, informing development strategy.
  • Providing descriptive structure that motivates the construction of shift-share (Bartik) instruments for causal analysis.

How it works

How it works

1

The starting point is the observed change Δe_i in the regional variable for industry i between a base year (0) and an end year (t). The decomposition explains this single number as the sum of three economically interpretable forces.

Formula
\Delta e_{i} = e_{i}^{t} - e_{i}^{0}
2

The national-share component attributes to industry i the growth it would have experienced had it simply grown at the overall national rate g. It is the base-year level e_i^0 scaled by g, representing the macroeconomic backdrop common to every region and sector.

Formula
NS_i = e_{i}^{0}\, g
3

The industry-mix (or proportional-shift) component captures the effect of the region's sectoral composition. It multiplies base-year employment by the difference between the industry's national growth rate g_i and the overall national rate g, so a region specializing in nationally fast-growing industries earns a positive mix effect.

Formula
IM_i = e_{i}^{0}\,(g_i - g)
4

The competitive-share (or differential/regional-shift) component is the residual that reflects local competitiveness. It multiplies base-year employment by the gap between the industry's regional growth rate r_i and its national growth rate g_i, isolating how much the region's own industries outperformed or lagged their national peers.

Formula
RS_i = e_{i}^{0}\,(r_i - g_i)
5

By construction the three components sum exactly to the total observed change, Δe_i = NS_i + IM_i + RS_i, with no residual. Summing across all industries gives the region's total decomposition, making the technique a transparent accounting identity rather than a statistical estimate.

Formula
\Delta e_i = NS_i + IM_i + RS_i

Common pitfalls

  • Interpreting a positive competitive-share as proof of effective local policy, when it actually absorbs every unmodeled influence including measurement error.
  • Comparing decompositions computed with different sectoral classifications or benchmark areas, which are not comparable.
  • Ignoring the allocation effect: classic shift-share cannot tell whether a strong competitive share stems from the region specializing in industries where it is competitive (the Esteban-Marquillas decomposition separates this).
  • Using highly aggregated sectors, which washes out the industry-mix signal the method is designed to reveal.

Frequently asked questions

What is the difference between shift-share analysis and the shift-share (Bartik) instrument?

Classic shift-share analysis is a descriptive accounting decomposition that splits observed regional growth into national, industry-mix, and competitive components. The shift-share or Bartik instrument reuses the same national-growth-times-local-industry-shares construction, but as an instrumental variable to identify the causal effect of local labor-demand shocks in a regression, leveraging variation in industry composition as exogenous. One describes; the other identifies a causal parameter.

Why does the competitive-share component cause interpretation problems?

The competitive (differential) share is computed as a residual — everything in regional growth not explained by national trends and industry mix. It therefore absorbs genuine local competitiveness together with measurement error, data classification artifacts, and any omitted structural factor. Esteban-Marquillas (1972) showed it is also entangled with specialization, and his reformulation adds an allocation effect to separate 'competitive because specialized in the right industries' from 'competitive across the board.'

Does it matter which year's weights are used?

Yes. The classic decomposition applies base-year employment as weights, which makes the components sensitive to the choice of start year and can introduce weighting bias when sectoral structure shifts substantially over the interval. Dynamic and average-weighted variants recompute the decomposition over sub-periods or use symmetric weights to reduce this dependence, analogous to the choice of reference structure in other decomposition methods.

Related methods

Sources

  1. Dunn, E. S. (1960). A statistical and analytical technique for regional analysis. Papers of the Regional Science Association, 6(1), 97–112.

    Open DOI reference
    Copied.
  2. Esteban-Marquillas, J. M. (1972). A reinterpretation of shift-share analysis. Regional and Urban Economics, 2(3), 249–255.

    Open DOI reference
    Copied.