Principal investigator: Omar Barroso Khodr
Authors: Philippe Aghion, Antonin Bergeaud and John Van Reenen
Original title: The Impact of Regulation on Innovation
Location of the Intervention: France
Sample Size: 182,347 distinct firms and 1.66 million observations
Primary Variable of Interest: growth rate in
number of priority patent applications – between t − 1 et
Type of Intervention: Proposal for the relaxation of regulatory standards in technological innovations and labor policies.
Methodology: OLS; Marginal Effects on Economic Shocks; Dynamic Econometric Analysis; and Schumpeterian Growth Models.
Summary
Aghion, Bergeaud, and Reenen (2023) present a framework that can be used to assess the equilibrium impact of regulation on endogenous innovation in heterogeneous firms. The authors implement this model using panel data from French firms, where there is a sharp increase in the burden of labor regulations on companies with 50 or more employees. Consistent with the qualitative predictions of the model, a decrease is found in the fraction of innovative firms just to the left of the regulatory threshold. Furthermore, a reduction in the innovative response of firms to demand shocks just below the threshold was found. Regulation reduces aggregate innovation by 5,7%.
- Policy Problem
According to Aghion, Bergeaud, and Reenen (2023), the study identifies a significant, yet often overlooked, policy problem: the negative dynamic impact of company size-based labor regulations on technological innovation and long-term economic growth. While much policy analysis focuses on the static costs of regulation, this research argues that the effects on growth are far more consequential. According to the authors, the specific policy in question is a French labor law that imposes a substantially higher regulatory burden—including mandatory workers' councils and union representation—on companies with 50 or more employees. This creates a strong disincentive for companies to grow beyond that threshold.
In this context, the authors emphasize that the main political implication is that such regulations, while intended to protect workers, create a substantial implicit tax on firm growth. This tax distorts firm behavior, leading to a well-documented “bloat” of firms that stagnate just below the 50-employee mark. Crucially, the study finds that this also creates an “innovation valley,” where firms just below the threshold drastically reduce their investments in innovation to avoid triggering the burdensome regulation. Using a structural model and firm-level data, the authors estimate that this regulation functions as a 2,6% profit tax, reducing aggregate innovation by approximately 5,7% and decreasing welfare by at least 2,2%. Thus, they conclude that static analyses of regulatory cost significantly underestimate the true long-term damage, since the vast majority of welfare loss results from reduced innovation, not just resource misallocation.
Furthermore, the study identifies a subtle distortion in the 'type' of innovation sought. Regulation discourages incremental, growth-oriented Research & Development (R&D), but it can bias innovative companies, encouraging them to invest in more radical, labor-saving technologies, far exceeding the established limit. This suggests that policy may unintentionally stifle the incremental and constant improvements that drive productivity, while also altering the direction of technological change.
To address the identification challenge in this field, the study proposes and demonstrates a robust methodological approach. Instead of relying on broad international regulatory indices, this study explores the sharp, almost experimental, variation created by the specific limit of 50 employees in a single country. It combines this variation with an analysis of how firms of different sizes respond to exogenous shocks in export demand, providing clear causal evidence that firms near the limit are reluctant to innovate, even when market opportunities expand. This methodology offers a model for policymakers to accurately assess the innovation costs of specific and point-in-time regulations.
In summary, the main policy conclusion is that well-intentioned labor protections, when conceived as rigid mandates conditioned on company size, can have serious unintended consequences for innovation and growth. The authors reiterate that policymakers should be aware that the greatest costs of such regulations are dynamic, not static, and should consider more flexible mechanisms or broader-based policies that achieve social objectives without creating strong disincentives to company growth and technological investment.
- Policy Implementation Context
The theoretical context of the study is a formal economic model that provides a precise analytical framework for understanding how labor regulations contingent on firm size affect innovation and growth. It is based on the Schumpeterian growth model established by Klette and Kortum (2004), adapting it to a discrete-time framework in which firms expand by innovating to capture new product lines.
The authors explain that the core of the theory models the regulatory boundary as a tax on profit implicit that it applies only to companies whose number of employees exceeds a specific level (denoted by In this model, the size of a company is directly proportional to its number of product lines – represented by nThe regulation is incorporated assuming that companies with n ≥ n̄ (The equivalent in product lines to the limit of 50 employees) face an additional marginal tax rate.
regarding their profits, while smaller companies do not.
Within this framework, the model derives clear and testable predictions about the behavior of firms. The main theoretical results of this experiment are that firms just below the threshold of n They have a lower intensity of innovation. Thus, according to the author, the owner faces a disincentive; a successful innovation would lead the company to exceed the threshold, triggering the profit tax and reducing the net gain from growth. Additionally, the authors found that the relationship between total innovation (patents) and company size shows different slopes on each side of the threshold. The intensity of innovation is higher for small companies far from the threshold and moderately high for large companies already subject to the tax.
Finally, the model demonstrates that the ratio between these inclinations towards large and small businesses is directly related to the implicit regulatory tax. This provides a crucial theoretical link that allows the authors to use empirically observed patterns in the relationship between innovation and firm size to quantitatively estimate the magnitude of the regulatory burden.
In summary, the theoretical framework translates the intuitive political problem—that a regulatory cliff discourages growth—into a rigorous economic model. It formally demonstrates how the threshold distorts R&D investment decisions, provides specific geometric predictions for the data (the “innovation valley” and slope shifts), and establishes a structural basis for transitioning from qualitative evidence to a quantitative assessment of the welfare of dynamic policy costs.
- Evaluation Details
The primary data source for this study is the administrative balance sheet information collected annually from 1994 to 2007 for all French companies by the French Federal Revenue Service (FICUS). The sample is restricted to non-governmental companies. Patent data is incorporated by merging this dataset with the PATSTAT database, using an algorithmic matching based on the similarity between the name and address of the patent assignee and the registered name and address of the company. Due to the lower accuracy of the matching for very small companies, the analysis focuses on companies with more than 10 employees. Given the research interest in a regulation affecting companies with 50 or more employees, the final main sample is further restricted to companies that had between 10 and 100 employees in 1994 or in their first year of observation.
In this way, the authors had to use additional data sources, including BACI trade data and French customs records, to construct measures of demand shock. The resulting dataset comprises 1,66 million observations across 182.347 distinct firms. Descriptive statistics reveal that innovation, measured by the number of patents filed, is rare and highly asymmetrical; the average firm files 0,009 patents per year, while the average innovative firm files 0,28. The study documents a strong, almost linear relationship between firm size and the likelihood of being an innovative firm, consistent with the economic literature. However, a key empirical finding is a sudden drop in the share of innovative firms just below the 50-employee threshold, creating an “innovation valley,” which is consistent with the article’s theoretical model. It is also observed that the slope of the relationship between innovation and size is flatter for firms above the threshold compared to those below it. The authors note that this variation in slopes, along with the distribution of firm sizes, provides empirical patterns that will later be used to assist in estimating model parameters, such as the tax rate implied by regulation.
- Method
According to the authors, the central objective of the model is not to reproduce every detail of the real world, but to capture the main economic trade-offs generated by regulation within a clear and tractable framework. Thus, they seek to understand how companies close to the regulatory boundary can behave differently and, from this, derive testable predictions for empirical analysis.
The model conceives of companies as sets of product lines. The size of a firm is determined by the number of lines in which it operates, which is directly related to its number of employees. Regulation is represented as an additional tax on profits that comes into effect when the company exceeds a certain size limit. The main behavioral intuition is that the owner—who manages the company over two periods—decides how much to invest in innovation (R&D) to create new product lines. This decision involves weighing the future profits from growth, the immediate cost of innovation, and the risk of losing existing lines to competitors.
For a company situated just below the threshold, expanding through even a single new product line triggers the regulatory tax. This creates a strong disincentive: the owner significantly reduces investment in innovation to avoid this costly “tax jump,” even when there is a positive demand opportunity.
Based on this structure, the model generates a central prediction about the intensity of innovation. It shows that innovation is highest among small companies, well below the threshold (which can grow freely); it is the second highest among large companies, already above the threshold (which have already internalized the cost of the tax); and it is lowest among medium-sized companies located immediately below the threshold. This "valley" in innovative effort is the critical distortion produced by regulation.
The model is then extended to the aggregate level, characterizing the steady-state distribution of firm size. It describes how the flows of inflow, outflow, expansion, and contraction balance over time. Regulation, by restricting the growth of firms near the boundary, shifts the size distribution to the left—resulting in more small firms and fewer large firms than would exist in the absence of the rule. Because smaller firms innovate less, this shift in distribution further reduces the economy's aggregate innovation.
Finally, the model incorporates an exogenous demand shock to generate the specific hypothesis tested empirically. It analyzes how a firm's innovative response to a positive shock depends on its proximity to the regulatory threshold. The theory predicts that, while all firms tend to increase innovation in the face of higher demand, this response will be weaker for those just below the threshold. The fear of exceeding it and incurring the tax outweighs the incentive to explore the new opportunity. This produces a clear and falsifiable prediction: in the data, the interaction between a demand shock and proximity to the regulatory threshold should have a negative effect on innovation growth. The entire empirical strategy is built to test this central prediction of the theoretical framework.
- Main results
The study investigates how company innovation, measured by patent growth, responds to positive shocks in export demand and how this response is affected by French labor regulations that impose additional costs on companies with more than 50 employees. The main conclusion is that, while a positive shock in market size generally increases innovative activity, this response is significantly attenuated for companies just below the regulatory threshold of 50 employees.
Specifically, according to the authors, a 10% increase in market size increases patents by approximately 1,1% on average, with larger companies tending to respond more intensely. However, the interaction between the demand shock and a variable dummy For companies with 45 to 49 employees, the trend is negative and significant. The authors point out that this indicates that companies near the limit are much less likely to increase innovation in response to new demand opportunities. The authors interpret this as evidence of a “growth rate”; companies avoid investing in innovation to minimize the risk of exceeding the limit and incurring substantial regulatory costs. This pattern holds even after controlling for other non-linearities related to company size.
In this way, quantitatively, the study estimates that this regulatory distortion leads to an aggregate loss of innovation of approximately 5,7% compared to a scenario without regulation. This translates into a significant reduction in the steady-state economic growth rate. The loss is driven primarily (about four-fifths) by reduced innovation rates among established firms, with the remainder due to a leftward shift in the distribution of firm size (towards smaller, less innovative firms) and lower innovation by new entrants. The regulatory rate parameter It is the most critical factor in these losses.
Finally, further analyses show that the negative effect on the threshold is particularly pronounced for high-quality patents (in the top citation decile), while the pattern is not evident for lower-value patents. The authors conclude that the dynamic losses resulting from distorted innovation incentives are substantial, approximately twice the conventional static welfare losses due to misallocation.
- Lessons in Public Policy
The study's results show that rigid regulatory thresholds, such as the cost increase when exceeding 50 employees, generate strong distortions in company behavior. Firms close to the limit avoid growth and, above all, avoid innovation, even in the face of positive demand shocks. Since innovation is the main driver of long-term growth, these dynamic effects become much more relevant than the static costs traditionally considered in regulatory assessments.
Additionally, the study also reveals that these distortions do not only affect companies directly at the threshold: they alter the size distribution of firms, shifting the economy towards an equilibrium with more small companies and therefore, with less aggregate innovation. Furthermore, the losses fall disproportionately on high-quality innovations, amplifying the negative impact on productivity.
The main implication the authors seek is clear: policies based on cutoffs should be replaced by gradual mechanisms that avoid abrupt cost jumps and reduce incentives for companies to "freeze" their growth. Reforms that soften or eliminate these thresholds can generate substantial productivity and growth gains by restoring incentives for innovation, especially among companies with greater expansion potential.
References
Klette, Tor Jakob, and Samuel Kortum. 2004. “Innovating Firms and Aggregate Innovation.” Journal of Political Economy 112 (5): 986–1018.
