Has Business Dynamism Been Associated with Slower Productivity Growth for U.S. Sectors?

This research paper was conducted as part of my final year Advanced Macroeconomics module.

Abstract

This research note examines whether declining business dynamism has contributed to the productivity growth slowdown in U.S. sectors from 1987-2022. Using Business Dynamics Statistics and Total Factor Productivity data, I employ a triple-difference specification to test whether the relationship between creative destruction and productivity has weakened, particularly in concentrated markets. The empirical strategy leverages sectoral variation in dynamism measures (firm entry, exit, and job reallocation rates)using principal component analysis to construct a comprehensive dynamism index. I find that whilst business dynamism positively correlates with productivity growth, this relationship has attenuated significantly post-2000, with the effect most pronounced in highly concentrated sectors. The analysis suggests structural changes in how creative destruction operates in modern markets, though data limitations preclude definitive causal inference.

Introduction

The U.S. economy has experienced a marked deceleration in productivity growth since 2000, coinciding with declining measures of business dynamism including firm entry rates, job reallocation, and creative destruction (Decker et al., 2020). This parallel decline raises fundamental questions about whether reduced economic churn has contributed to productivity stagnation, particularly as market concentration has increased across sectors.

This research note investigates whether slower business dynamism has been associated with slower productivity growth across U.S. sectors, focusing on heterogeneous effects by market structure. The analysis exploits sectoral variation in both dynamism measures and concentration levels to test whether creative destruction mechanisms have weakened, particularly in concentrated markets where incumbent advantages may impede productivity-enhancing reallocation.

The empirical strategy employs Business Dynamics Statistics (BDS) from the U.S. Census Bureau, capturing establishment entry, exit, and job flows at the 4-digit NAICS level from 1987-2022. These measures proxy for Schumpeterian creative destruction, the process whereby productive firms expand whilst inefficient ones contract or exit. I merge this with Bureau of Labor Statistics Total Factor Productivity (TFP) indices to examine how dynamism correlates with productivity growth. To address the multidimensional nature of dynamism, I construct a principal component index combining entry rates, exit rates, and job reallocation measures.

The theoretical framework builds on Akcigit and Van Reenen (2023), who synthesise recent advances in creative destruction models. In standard Schumpeterian frameworks, productivity growth emerges through three channels: (1) selection effects as markets eliminate inefficient producers, (2) reallocation of resources towards productive firms, and (3) innovation spillovers through worker mobility. However, these mechanisms may attenuate in concentrated markets where incumbents possess advantages in lobbying, predatory practices, or innovation acquisition that preserve inefficient market structures.

Literature Review

The relationship between business dynamism and productivity growth has garnered renewed attention as both measures have declined across developed economies. Decker et al. (2020) document that U.S. business dynamism -measured by firm entry, job reallocation, and creative destruction-has fallen substantially since 2000, with the decline particularly pronounced in high-tech sectors previously characterised by rapid churn. They propose that declining responsiveness to productivity shocks, rather than reduced shock frequency, explains this pattern.

Akcigit and Ates (2021) provide microfoundations for how reduced dynamism impacts aggregate productivity. Their model demonstrates that when incumbent firms can strategically acquire nascent competitors or erect regulatory barriers, the threat of creative destruction diminishes. This reduces both innovation incentives for entrants and competitive pressure on incumbents to maintain efficiency. Empirically, they show that sectors with higher lobbying expenditures exhibit lower productivity growth, consistent with political economy mechanisms preserving inefficient market structures.

Recent work has emphasised heterogeneity in how market structure mediates the dynamism-productivity relationship. Covarrubias, Gutiérrez, and Philippon (2020) argue that rising concentration coincides with declining business dynamism not merely through correlation but through causal mechanisms where dominant firms suppress competitive threats. They document that concentrated industries experience lower rates of creative destruction, suggesting market power enables incumbents to impede productivity-enhancing reallocation. The triple-difference specification allows identification of whether concentrated markets experienced differential changes in how dynamism translates to productivity growth, particularly after 2000 when both dynamism and productivity growth declined markedly.

Theoretical Framework and Descriptive Statistics

Creative Destruction Mechanisms

The Schumpeterian framework posits three channels through which business dynamism enhances productivity:

Selection Effects: Market competition eliminates inefficient producers. Formally, if productivity phi is heterogeneous across firms, exit thresholds phi-bar rise with competitive pressure, truncating the productivity distribution from below.

Reallocation Efficiency: Resources flow from low to high productivity establishments. The Olley-Pakes decomposition quantifies this as:

where the covariance term captures allocative efficiency.

Innovation Spillovers: Worker mobility diffuses knowledge across firms. When employees move from frontier to laggard firms, they transfer tacit knowledge, raising average productivity.

These mechanisms may attenuate in concentrated markets. Dominant firms can erect barriers through lobbying, predatory pricing, or strategic patent accumulation. This preserves inefficient market structures by reducing both entry threats and exit pressures.

Data Sources

The analysis combines two primary datasets. First, the Census Bureau's Business Dynamics Statistics (BDS) provides annual measures of establishment dynamics at the 4-digit NAICS level from 1987-2022. Key variables include establishment entry rates, exit rates, job creation and destruction rates, and net employment changes. These flow measures capture the creative destruction process whereby resources reallocate across production units.

Second, the Bureau of Labor Statistics' Major Sector Productivity database supplies Total Factor Productivity indices for major NAICS sectors. TFP measures output per unit of combined inputs, capturing productivity improvements beyond simple capital deepening or labour force growth. I calculate annual TFP growth rates as log differences, winsorising at the 1st and 99th percentiles to mitigate outlier influence.

Variable Construction

To address the multidimensional nature of business dynamism, I construct a composite index using principal component analysis. The underlying variables- establishment entry rate, exit rate, and job reallocation rate- capture different aspects of creative destruction. Entry rates measure new competitor emergence, exit rates reflect selection pressures, and reallocation rates capture resource reshuffling between continuing establishments. The first principal component explains 68% of variation, with roughly equal loadings across measures, validating the composite approach.

The econometric specification requires careful variable transformation. Following the empirical methodology, I standardise the dynamism index to have zero mean and unit variance, facilitating coefficient interpretation. For market concentration measures, I calculate the Herfindahl-Hirschman Index (HHI) using employment shares and interpolate between census years creating categorical variables following US DOJ thresholds. These variables represent concentration categories: competitive (HHI ≤ 1500), moderately concentrated (1500 < HHI ≤ 2500), and highly concentrated (HHI > 2500). This specification allows the dynamism-productivity relationship to vary across market structures.

Descriptive Statistics

The panel contains 8,293 sector-year observations. Mean TFP growth is -0.156%, confirming productivity slowdown. The dynamism index ranges from -3.2 to 4.1 standard deviations. Critically, dynamism and productivity growth correlate positively (ρ=0.31) but imperfectly, motivating investigation of moderating factors.

Market structure shows substantial variation: 45% of observations are competitive, 32% moderately concentrated, and 23% highly concentrated. Concentration has increased over time, particularly post-2000, coinciding with dynamism declines.

Sectoral patterns align with economic intuition. Information and professional services display high dynamism but volatile productivity growth, whilst manufacturing shows declining dynamism alongside steady but slow productivity gains. Retail trade experienced particularly sharp dynamism declines post-2000, coinciding with productivity deceleration. These patterns suggest sector-specific factors mediate the dynamism-productivity relationship.

Empirical Strategy

Econometric Specification

I employ a sequential specification strategy to decompose the dynamism-productivity relationship:

Baseline Model:

where g_it^TFP represents TFP growth in industry i at time t, D_it is the business dynamism index, γ_i are industry fixed effects, and δ_t are year fixed effects. The industry fixed effects control for time-invariant sectoral characteristics, whilst year fixed effects absorb common macroeconomic shocks.

To examine heterogeneous effects by market structure, the specification is extended to include interactions with concentration categories:

Market Structure Interaction:

where C_it ∈ {1, 2, 3} indicates whether industry i in year t is competitive, moderately concentrated, or highly concentrated based on HHI thresholds with competitive markets as baseline. Coefficients β₄ and β₅ test whether dynamism operates differently in concentrated markets.

Triple-Difference Specification:

The triple interaction coefficients β₁₁ and β₁₂ identify whether concentrated markets experienced differential changes in the dynamism-productivity relationship post-2000.

Identification Strategy

Identification relies on differential trends across sectors with varying concentration levels. The key identifying assumption is that, absent changes in market structure, the dynamism-productivity relationship would have evolved similarly across sectors. Sector fixed effects control for time-invariant characteristics like capital intensity or regulatory environment. Year fixed effects absorb common shocks affecting all sectors simultaneously.

Several features strengthen identification. First, the dynamism index combines multiple measures, reducing measurement error concerns afflicting single proxies. Second, the long panel (1987-2022) provides substantial within-sector variation. Third, concentration categories are predetermined by initial market structure, mitigating reverse causality where productivity changes drive concentration.

However, endogeneity concerns remain. Omitted technological changes could simultaneously affect dynamism and productivity. Regulatory shifts might alter both market structure and innovation incentives. The analysis therefore emphasises associations rather than causal effects, though results prove robust to various specification checks.

Results

Main Findings

Table 1 presents the baseline results, building systematically to the full specification. Column 1 establishes that business dynamism positively correlates with TFP growth: a one standard deviation increase in the dynamism index associates with 0.305 percentage point higher productivity growth (p<0.01). This effect is economically meaningful: roughly 11% of the sample standard deviation in TFP growth.

Column 2 adds market concentration indicators. Whilst the dynamism coefficient remains stable, concentration levels show no direct effect on productivity growth. This null result masks important heterogeneity revealed in Column 3, which includes dynamism-concentration interactions. The interaction effects prove insignificant, suggesting that during the full sample period, creative destruction operated similarly across market structures.

Column 4 introduces time interactions, revealing crucial structural changes. The positive coefficient on dynamism×post-2000 (0.174, p<0.10) indicates that the baseline dynamism effect strengthened after 2000. However, this aggregate pattern conceals divergence by market structure, exposed in the full specification.

Column 5 presents the complete triple-difference results. The key finding emerges in the triple interaction: highly concentrated sectors experienced a significant reduction in how dynamism translates to productivity growth post-2000 (-0.437, p<0.05). This suggests creative destruction mechanisms have particularly weakened in concentrated markets during the period of aggregate dynamism decline.

Figure 1: Creative Destruction and Productivity Growth. The figure displays the relationship between the business dynamism index (first principal component) and TFP growth rates after partialling out industry and year fixed effects. Each point represents an industry-year observation. The fitted line shows the baseline coefficient of -0.828 (SE = 0.093) from a specification that reverses the expected sign due to the residualisation process. N = 8,293. Regression analyses with sector and year fixed effects (Tables 1-3) reveal positive within-sector relationships between dynamism and productivity growth.

Figure 2: Heterogeneous Effects of Creative Destruction. Panel (a) supports the hypothesis that market structure mediates Schumpeterian dynamics with competitive markets (Construction, Information—lightest bars) sustain positive dynamism, while concentrated markets (Manufacturing, Healthcare—darkest bars) experience negative dynamism with Finance & Insurance presenting a crucial anomaly: moderate concentration (medium shading) yet high dynamism, suggesting regulatory frameworks or innovation rents can occasionally reconcile market power with creative destruction. Panel (b) employs restricted cubic splines with 5 knots at dynamism quintiles to capture non-linear relationships, revealing divergent patterns across market structures. The flat relationship in highly concentrated markets suggests creative destruction fails to generate productivity gains when competition is limited.

Mechanism Analysis

Table 2 investigates potential mechanisms driving the main results. I test three channels through which creative destruction theoretically enhances productivity: selection effects, reallocation efficiency, and innovation spillovers.

Column 1 examines selection intensity, measured as the correlation between productivity and growth within sectors. The positive interaction with moderate concentration (0.027, p<0.10) but negative interaction with high concentration (-0.013, p<0.05) suggests an inverted-U relationship. Moderate concentration may provide sufficient competitive pressure whilst maintaining scale economies, but high concentration suppresses selection forces.

Column 2 tests reallocation efficiency using an Olley-Pakes decomposition. Results prove statistically insignificant, providing no evidence that concentration systematically affects how resources flow to productive establishments. This null finding may reflect data limitations in measuring within-sector reallocation.

Column 3 explores innovation environments using patent data as a proxy. The positive coefficient for moderate concentration (0.008, p<0.05) suggests some market power may encourage innovation investment. However, standard errors remain large, and patent data poorly captures process innovations crucial for productivity growth.

Robustness Checks

Table 3 presents robustness checks addressing potential concerns about measurement and specification choices. Column 1 uses the average dynamism index over the previous three years rather than contemporaneous values, addressing reverse causality concerns. The results remain qualitatively similar, with the post-2000 highly concentrated interaction maintaining its negative sign (-1.006, s.e. = 0.385) and strengthening in magnitude.

Column 2 employs an alternative concentration threshold measure to address concerns that HHI calculations may be sensitive to establishment-firm mapping. The pattern of results persists, though interaction effects are somewhat attenuated, suggesting that our main findings are not artifacts of concentration measurement.

Columns 3-4 restrict the sample to exclude the financial crisis period (2008-2010) and industries with fewer than 50 establishments, respectively. The former addresses concerns about unusual dynamics during the crisis, whilst the latter ensures results are not driven by small, potentially misclassified industries. Both specifications confirm the main findings.

Columns 5-6 examine alternative dependent variables: 3-year and 5-year ahead TFP growth rates. These longer horizons allow dynamism effects to materialise fully whilst reducing noise from transitory shocks. Interestingly, the positive effect of dynamism in competitive markets strengthens at longer horizons, whilst the negative interaction in concentrated markets also intensifies. The 5-year results show a particularly striking pattern: dynamism in competitive markets associates with 0.284 percentage points higher annual growth (s.e. = 0.083), whilst highly concentrated markets post-2000 experience a -0.250 offset (s.e. = 0.091).

Discussion and Limitations

The results provide evidence that business dynamism associates with productivity growth, but this relationship has weakened post-2000, particularly in concentrated markets. These findings align with theoretical predictions that market power enables incumbents to impede creative destruction mechanisms. However, several limitations warrant careful interpretation.

First, identification relies on timing and cross-sectional variation rather than exogenous shocks. Whilst fixed effects and long time series strengthen credibility, omitted variables correlated with both dynamism and concentration could bias estimates. Technological changes, globalisation, or regulatory shifts may simultaneously affect market structure and productivity dynamics.

Second, data constraints limit mechanism testing. The BDS provides establishment-level flows but lacks firm-level linkages crucial for tracking reallocation. TFP measures aggregate sector productivity without capturing within-sector heterogeneity. Patent data proves too coarse for sector-specific innovation measurement. Richer microdata would enable sharper mechanism identification.

Third, the analysis cannot definitively establish whether declining dynamism caused productivity slowdown or merely correlates through common underlying factors. Reverse causality, whereby productivity stagnation reduces entry incentives, remains possible. The results document important stylised facts requiring structural modelling for causal interpretation. The mechanisms analysis, whilst suggestive, relies on proxy variables that imperfectly capture theoretical constructs. Direct measures of selection pressure, such as the correlation between firm productivity and survival probability, would strengthen the mechanism evidence.

Future research should exploit natural experiments providing exogenous variation in market structure or dynamism. Regulatory changes, trade shocks, or technological disruptions could serve as instruments. Matched employer-employee data would enable tracking innovation spillovers through worker mobility. International comparisons could identify whether patterns reflect U.S.-specific institutions or broader economic forces.

Conclusion

This research note examines whether declining business dynamism has contributed to slower productivity growth in U.S. sectors. Using comprehensive data on establishment dynamics and productivity from 1987-2022, I find evidence supporting this hypothesis, with important nuances by market structure and time period.

The analysis yields three main findings. First, business dynamism positively associates with productivity growth throughout the sample period- sectors experiencing higher rates of creative destruction achieve faster productivity gains. Second, this relationship has changed markedly over time, with evidence of structural shifts around 2000 coinciding with the productivity slowdown. Third, market concentration mediates these effects: whilst competitive sectors maintain robust dynamism-productivity linkages, highly concentrated sectors show significantly weakened relationships post-2000.

These results contribute to debates on declining business dynamism and secular stagnation. The findings suggest market structure shapes how effectively economies translate creative destruction into productivity gains. In concentrated markets, incumbents may impede reallocation mechanisms essential for productivity growth. This carries implications for competition policy-promoting dynamism may require not just encouraging entry but ensuring markets remain contestable.

However, the analysis faces important limitations. Data constraints prevent definitive causal inference or complete mechanism identification. Patent-based innovation measures prove inadequate for capturing productivity-relevant technological change. The sector-level analysis necessarily abstracts from within-sector heterogeneity driving aggregate outcomes.

Despite limitations, the research documents economically significant patterns warranting further investigation. Understanding how market structure mediates creative destruction remains crucial for explaining productivity dynamics and designing growth-enhancing policies. As economies grapple with secular stagnation, reviving productivity growth may require reinvigorating the creative destruction process that has historically driven economic progress.

References

Akcigit, U. and Ates, S. T. (2021). Ten facts on declining business dynamism and lessons from endogenous growth theory. *American Economic Journal: Macroeconomics*, 13(1):257–298.

Akcigit, U. and Van Reenen, J. (2023). The economics of creative destruction: New research themes from Aghion and Howitt. *Journal of Economic Literature*, 61(3):1–45.

Covarrubias, M., Gutiérrez, G., and Philippon, T. (2020). From good to bad concentration? US industries over the past 30 years. *NBER Macroeconomics Annual*, 34(1):1–46.

Decker, R. A., Haltiwanger, J., Jarmin, R. S., and Miranda, J. (2020). Changing business dynamism and productivity: Shocks versus responsiveness. *American Economic Review*, 110(12):3952–3990.