Shift-share analysis
Shift-share analysis is a widely used decomposition technique in regional economics that decomposes observed changes in a regional economic variable—such as employment, output, or income—over a specified period into three additive components: the national growth effect, which captures the impact of overall national economic trends; the industry mix effect, which reflects the influence of the region's industrial composition relative to the national average; and the regional shift (or competitive) effect, which isolates unique local factors driving differential performance.[1] This method provides a structured framework for understanding whether regional economic changes stem from external macroeconomic forces, structural differences in industry shares, or region-specific competitiveness.[2] Originating in the early 1940s, shift-share analysis was first articulated by Daniel Creamer in his examination of locational shifts in U.S. manufacturing industries, building on earlier ideas from reports like the 1940 Barlow Commission.[1] The approach gained formal structure in 1960 through the work of Edgar S. Dunn and collaborators, including Harvey Perloff, who outlined the classic three-part decomposition in their analysis of regional growth patterns across U.S. regions.[1] Over subsequent decades, the technique evolved to address limitations, such as static assumptions about industry shares; notable advancements include the dynamic shift-share model proposed by Barff and Knight in 1988, which incorporates time-varying weights, and the Esteban-Marquillas extension in 1972, which adjusts for non-homothetic industry structures.[1][3] In practice, shift-share analysis is applied retrospectively to evaluate economic performance at various scales, from metropolitan areas to states or countries, often using data on employment by industry from sources like national statistical agencies.[4] The national growth effect is calculated as the regional base times the national growth rate, the industry mix effect as the difference between national industry-specific and overall growth applied to regional shares, and the regional shift as the residual difference between actual and expected changes.[3] A positive regional shift indicates competitive advantages, such as innovation, labor quality, or infrastructure, while negative values suggest disadvantages.[2] Commonly employed in policy analysis, the method informs economic development strategies by highlighting industries with strong local competitiveness for targeted investments or workforce training.[5] Despite its simplicity and interpretability, critics note its descriptive rather than causal nature and sensitivity to base-year choices, leading to integrations with econometric models or input-output frameworks in advanced applications.[1]Introduction
Definition and Purpose
Shift-share analysis is a decomposition technique in regional economics that breaks down changes in regional economic indicators, such as employment or output, into three primary components: a national growth effect reflecting overall economic expansion, an industry composition effect capturing the influence of sectoral mixes, and a regional competitive effect highlighting local advantages or disadvantages.[6] This approach enables the isolation of region-specific factors from nationwide trends and industrial structures, providing a structured way to evaluate why a region's economy may grow faster or slower than expected.[6] The core purpose of shift-share analysis is to identify the drivers of regional economic performance by distinguishing between exogenous influences—like national growth and industry-specific trends—and endogenous ones, such as local competitiveness, thereby supporting targeted economic development strategies and policy formulation.[6] By quantifying these effects, the method helps policymakers and analysts pinpoint thriving sectors for investment or underperforming areas needing intervention, fostering a deeper understanding of structural economic changes at the regional level.[7] Shift-share analysis originated in the post-World War II era of regional economics, emerging as a tool to analyze postwar economic disparities and guide planning efforts amid rapid industrialization and reconstruction.[1] It presupposes basic knowledge of economic indicators, including employment shares by industry and region, to interpret the relative performance of local economies against benchmarks.[6]Historical Development
An early precursor to shift-share analysis appeared in the 1940 Barlow Report by the Royal Commission on the Location of the Industrial Population in the United Kingdom, which used a rudimentary form of the technique to assess interregional employment changes and inform post-war planning.[1] The method originated in the early 1940s with the work of economist Daniel Creamer, who developed it as a method to examine shifts in regional manufacturing employment relative to national trends, initially applied to U.S. data on labor distribution across industries.[8] Creamer's approach laid the groundwork for decomposing regional economic changes into national growth, industrial mix, and regional share components, though it remained largely descriptive at this stage. The method gained formal structure in 1960 through the contributions of Edgar S. Dunn Jr. and collaborators, including Harvey Perloff, who refined it into a systematic analytical tool. Dunn's paper "A Statistical and Analytical Technique for Regional Analysis" introduced the "Dunn cross-classification" to better isolate and interpret the interactive effects between national and regional factors.[9] This formalization was influenced by pioneers in regional science, such as Walter Isard, whose 1960 book Methods of Regional Analysis: An Introduction to Regional Science integrated shift-share into broader frameworks for studying spatial economic dynamics and emphasized its role in understanding regional development as a subset of national growth. The collaborative work, including Perloff et al.'s Regions, Resources and Economic Growth, elevated shift-share from an ad hoc measurement to a standardized technique, widely adopted in academic and policy analyses during the 1960s to evaluate employment trends and inform redevelopment strategies in economically distressed areas.[1] Extensions to the core model began in the 1970s, addressing limitations in handling structural changes over time. In 1972, J.M. Esteban-Marquillas proposed a reinterpretation that incorporated homothetic employment assumptions to separate allocation effects more precisely, enhancing the model's ability to account for differential industry distributions.[10] This was further developed in 1984 by Francisco J. Arcelus, who extended the framework to include additional interaction terms, improving the decomposition of competitive advantages.[11] Meanwhile, efforts to incorporate temporal dynamics traced back to earlier critiques but culminated in the 1988 dynamic model by Richard A. Barff and Prentice L. Knight III, which used continuous growth rates to mitigate biases in static comparisons over extended periods.[12] These advancements solidified shift-share as a cornerstone of regional economic analysis by the late 20th century.Core Methodology
Components of the Traditional Model
The traditional shift-share model decomposes changes in a regional economy, such as employment or output, into three distinct components that isolate the influences of broader economic trends, sectoral composition, and local factors. This decomposition allows analysts to attribute regional performance to national dynamics, structural characteristics, and competitive elements without implying causation. Developed as a descriptive tool, the model applies to various geographic scales, including states, metropolitan areas, and subnational regions, providing insights into how local economies align with or diverge from national patterns. The national growth effect, also known as the proportional or national share effect, represents the portion of regional change attributable to overall national economic expansion. It assumes the region grows at the same rate as the nation as a whole, reflecting what the region's economy would experience if its industry shares remained constant relative to the national total. This component captures the baseline influence of macroeconomic conditions, such as aggregate demand or policy shifts, on the region. For instance, during periods of national recovery, a positive national growth effect would indicate that the region's expansion mirrors broader growth, independent of its unique industrial makeup.[13] The industry mix effect, sometimes termed the structural or differential industry effect, measures the impact of the region's sectoral composition compared to the national average. It accounts for whether the region specializes in industries that expand or contract faster than the national economy overall. A favorable industry mix occurs when a region has a higher concentration in high-growth sectors, contributing positively to its performance beyond national trends; conversely, over-reliance on declining industries yields a negative effect. This component highlights structural advantages or vulnerabilities in the regional economy's diversification. Terminology variations include "industry share" or "proportionality effect," emphasizing the role of sectoral specialization.[14] The regional share effect, referred to alternatively as the competitive, differential, or regional shift effect, isolates the residual change after accounting for national growth and industry mix influences. It reflects local factors—such as labor quality, infrastructure, innovation, or policy environments—that drive performance above or below national and structural expectations, indicating a region's competitive edge or disadvantage. A positive regional share suggests "regional advantage," where local conditions enable outperformance in specific industries, while a negative value points to inefficiencies or external challenges unique to the area. This effect is central to identifying policy levers for enhancing competitiveness.[13] These components are interlinked and additive, summing to the total observed regional change, which enables a clear isolation of the regional share as a measure of localized dynamics. The national growth effect provides the foundational trend, the industry mix adjusts for compositional differences, and the regional share captures idiosyncratic performance, together offering a layered understanding of economic shifts across regions like U.S. states or European metropolitan areas. Variations in terminology, such as "regional drift" for the competitive effect, arise from contextual adaptations but preserve the core decomposition logic.[14]Formula and Calculation
The traditional shift-share model mathematically decomposes the change in regional employment for a specific industry into three additive components: the national growth effect, which captures overall national economic expansion; the industry mix effect, which accounts for the differential growth rates of industries at the national level; and the regional share effect, which isolates the region's competitive advantage or disadvantage relative to the national industry average.[15] This formulation, originally developed by Edgar M. Dunn in 1960, provides a structured way to attribute regional economic changes to broader and local factors.[16] Key notation includes: E_{rj}^0, the base-period employment in region r for industry j; g_n, the national growth rate across all industries over the period; g_{nj}, the national growth rate specific to industry j; and g_{rj}, the growth rate of industry j in region r.[17] These variables are typically derived from employment data spanning two time periods, such as census years, with growth rates calculated as (E^t - E^0)/E^0, where superscript t denotes the end period.[15] The core equation for the change in regional industry employment, \Delta E_{rj}, is expressed as: \Delta E_{rj} = g_n E_{rj}^0 + (g_{nj} - g_n) E_{rj}^0 + (g_{rj} - g_{nj}) E_{rj}^0 Here, the first term represents the national effect, the second the industry mix effect, and the third the regional share effect.[17] To compute these for a single industry, one first calculates the growth rates g_n, g_{nj}, and g_{rj} from available employment data, then multiplies each differential by the base employment E_{rj}^0.[15] For the total regional economy, the effects are aggregated across all industries j: the total national effect is \sum_j g_n E_{rj}^0 = g_n E_r^0, where E_r^0 = \sum_j E_{rj}^0; the total industry mix effect is \sum_j (g_{nj} - g_n) E_{rj}^0; and the total regional share effect is \sum_j (g_{rj} - g_{nj}) E_{rj}^0.[17] This summation verifies that the overall regional employment change \Delta E_r = \sum_j \Delta E_{rj} equals the sum of the three total effects, ensuring the decomposition is exhaustive.[15] The model assumes a static analysis between two discrete periods, treating growth rates as constant within each period and applying uniform national growth to all regions in the absence of industry-specific or regional factors.[16] It further presumes that national trends provide a neutral benchmark, with deviations attributable solely to industrial composition and local competitiveness.[17]Example and Interpretation
To illustrate the traditional shift-share model, consider a hypothetical example of manufacturing employment changes in a U.S. state, such as California, from 1950 to 1960, drawing on the style of historical economic data analyses from that era.[6] Suppose the state's manufacturing sector employed 1,000,000 workers in 1950, while the national economy grew by 5% over the decade, the national manufacturing sector grew by 3%, and the state's manufacturing sector actually grew by 6%, reaching 1,060,000 workers by 1960. This setup allows decomposition of the total 60,000-job increase into the three components. The calculation begins with input data: base-year (1950) employment for the state’s manufacturing industry (E_{ir0} = 1,000,000), national growth rate (r_n = 0.05), national industry growth rate for manufacturing (r_{jn} = 0.03), and regional industry growth rate (r_{ir} = 0.06).| Component | Calculation | Jobs Gained/Lost | Percentage Share |
|---|---|---|---|
| National Growth | $1,000,000 \times 0.05 | +50,000 | +5% |
| Industrial Mix | $1,000,000 \times (0.03 - 0.05) | -20,000 | -2% |
| Regional Share | $1,000,000 \times (0.06 - 0.03) | +30,000 | +3% |
| Total Change | Sum of components | +60,000 | +6% |
Advanced Models
Dynamic Model
The dynamic shift-share model extends the traditional static framework by addressing its key limitation: the assumption of constant national and industry-specific growth rates throughout the analysis period. Instead, it employs period-specific growth rates to compute the national, industrial mix, and regional (competitive) components annually, then sums these effects across multiple periods for a cumulative assessment. This approach enables more accurate decomposition of employment or output changes over extended time horizons, where growth patterns may fluctuate due to economic cycles or structural shifts. Developed by Barff and Knight in 1988, the model was motivated by the need to mitigate biases in static analyses, particularly when regional growth deviates significantly from national trends over time. The core formula for the total change in regional industry employment, \Delta E_{rj}, under the dynamic model is: \Delta E_{rj} = \sum_t \left[ N_t \cdot E_{rj}^{t-1} \right] + \sum_t \left[ (g_j^t - g_n^t) \cdot E_{rj}^{t-1} \right] + \sum_t \left[ (g_{rj}^t - g_j^t) \cdot E_{rj}^{t-1} \right] Here, the summation occurs over all periods t; E_{rj}^{t-1} denotes employment in regional industry j at the start of period t; N_t (or equivalently g_n^t) is the national growth rate in period t; g_j^t is the national growth rate for industry j in period t; and g_{rj}^t is the regional growth rate for industry j in period t. The first term captures the national growth effect, the second the industrial mix effect, and the third the regional competitive effect, each recalculated using the updated employment base from the prior period. This iterative process ensures that subsequent periods build on prior outcomes, reflecting compounding dynamics. A primary advantage of the dynamic model lies in its ability to capture temporal variations in growth rates, providing a nuanced view of how regional performance evolves relative to national benchmarks. For example, if a region's industry experiences rapid expansion in early periods but stagnation later, the model attributes effects accordingly, avoiding the averaging distortions of the static method. Barff and Knight demonstrated this in their analysis of New England employment growth from 1939 to 1984, where dynamic calculations revealed shifting competitive advantages in high-technology sectors amid national economic recoveries. Overall, it reduces estimation bias in multi-decade studies by incorporating continuous updates to the employment base and growth rates, yielding more reliable insights into long-term regional competitiveness. In contrast to the traditional model, which applies a single set of growth rates derived from endpoint data, the dynamic variant accumulates effects through period-by-period computations, better approximating actual growth trajectories and minimizing under- or overestimation of components in volatile economies.Esteban-Marquillas Model
The Esteban-Marquillas model, proposed by J. M. Esteban-Marquillas in 1972, extends the traditional shift-share framework by further decomposing the regional share effect to distinguish between structural influences and pure competitive advantages at the industry level. This approach addresses a key limitation in the classical model, where the regional component conflates a region's inherent competitiveness with biases stemming from its unique industrial composition relative to the national average. By isolating these elements, the model enables analysts to assess whether observed growth deviations arise from structural factors, such as over- or under-specialization in certain industries, or from region-specific efficiencies and market dynamics.[18] Central to the model is the introduction of homothetic employment, defined as the hypothetical employment in industry j for region r at the base period if the region mirrored the national industrial structure:H_{rj}^0 = E_r^0 \cdot \frac{E_{nj}^0}{E_n^0},
where E_r^0 and E_n^0 are the base-period total employment in the region and nation, respectively, and E_{nj}^0 is the national base-period employment in industry j. This contrasts with the actual base-period regional employment E_{rj}^0. The location quotient L_{rj} = \frac{E_{rj}^0 / E_r^0}{E_{nj}^0 / E_n^0} quantifies the structural deviation, with H_{rj}^0 = E_{rj}^0 / L_{rj}.[18] The regional share effect from the traditional model, \sum_j E_{rj}^0 (g_{rj} - g_{nj}), where g_{rj}, g_{nj}, and g_n denote the growth rates for region r-industry j, national industry j, and national overall, respectively, is reformulated as the sum of two components:
- Competitive effect: \sum_j H_{rj}^0 (g_{rj} - g_{nj}), which applies the regional growth differential to the homothetic base, capturing performance independent of structural biases.
- Allocative (structural) effect: \sum_j (E_{rj}^0 - H_{rj}^0) (g_{rj} - g_{nj}) = \sum_j H_{rj}^0 (L_{rj} - 1) (g_{rj} - g_{nj}), which measures the interaction between compositional differences (via L_{rj} - 1) and the regional growth differential, highlighting how specialization amplifies or dampens competitiveness.
This decomposition ensures additivity to the original regional effect while providing granular insights into industry-specific drivers.[18] For instance, a positive allocative effect indicates that a region's over-specialization ( L_{rj} > 1 ) in fast-growing industries relative to the nation enhances overall performance.[18]