📈 economics

Threshold-based liability regulation and the timing of radical innovation: Evidence from Korea's Serious Accident Punishment Act

Using a difference-in-differences approach on Korean manufacturing data, this study finds that the 2022 Serious Accident Punishment Act, which imposes criminal liability on officers of firms with 50+ employees, caused a transitory dip in radical innovation among covered firms during the enactment window without evidence of a permanent level shift or subsequent recovery beyond their pre-existing trend.

Original authors: Jihwan Woo, Nari Kim

Published 2026-08-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jihwan Woo, Nari Kim

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of business and law, rules are often written with a specific size in mind. Imagine a regulation that applies only to companies with fifty or more workers, leaving smaller businesses completely free from its requirements. This kind of size-based cutoff is a common tool for governments, but it creates a strange divide. It splits the business world into two groups: those who must follow the new, strict rules and those who do not. Economists have long wondered how these rules change the way companies behave. Do they simply stop growing to stay small? Or do they change the very nature of the new things they create? The most exciting kind of business growth comes from radical innovation—creating products that have never existed before in the market. These are the risky, high-reward ideas that often come from smaller, agile companies. If a new law makes it dangerous to launch such a product, it could stifle the engine of progress.

This question became a real-life experiment in South Korea with the passage of a law called the Serious Accident Punishment Act. Before this law, if a workplace accident resulted in a fatality, the company as a whole might face fines, but the people running the company were rarely held personally responsible. The new law changed that. It made corporate officers personally liable, facing up to a year in prison and unlimited fines, if a worker died due to a failure in the company's safety system. However, the law included a specific cutoff: it only applied to companies with fifty or more employees. Smaller companies were exempt for a while. This setup created a perfect opportunity for researchers to see what happens when a group of businesses suddenly faces a heavy new risk, while a similar group of businesses right next to them does not.

Researchers from Amazon Web Services and the Korea Institute of Procurement set out to measure exactly how this law affected the creation of new products. They looked at data from a massive survey of Korean manufacturing companies conducted every two years between 2018 and 2024. This survey asked companies about the new products they had launched in the previous three years. The researchers focused on two specific types of innovation. The first was "market-first" innovation, which means a product that was new to the entire domestic market. This is the most radical kind of invention. The second was "firm-first" innovation, which is a product that is new only to that specific company but might already exist elsewhere. By comparing the companies that were forced to follow the new law with those that were not, the team could see if the threat of personal liability changed what these companies decided to build.

The story the data told was not one of permanent damage, but of a temporary pause. Before the law was passed, the larger companies were already slightly ahead of the smaller ones in creating these radical, market-first products. When the law was announced in early 2021 and set to take effect in early 2022, something interesting happened. During the year between the announcement and the actual start date, the larger companies suddenly slowed down. Their rate of launching new-to-market products dropped significantly compared to their previous trend and compared to the smaller companies. The data showed a dip of nearly eight percentage points in their innovation rate during this specific window. It was as if the companies hit the brakes, waiting to see how the new rules would actually be enforced. The uncertainty was at its highest during this time; companies knew the law existed, but they did not know how often prosecutors would bring charges or what exactly would count as a safety failure.

However, this slowdown did not last forever. Once the law had been in effect for a couple of years and companies had gained experience with how it worked, the larger firms returned to their original path. By 2024, their rate of radical innovation had bounced back to where it would have been if the law had never been passed. There was no permanent loss of creativity. The companies had not given up on big ideas; they had simply delayed them until the fog of uncertainty cleared. The researchers found that during the pause, these companies did not switch to making safer, less risky products in large numbers. Instead, they seemed to wait. This behavior matches a theory that when the future is unclear, businesses prefer to wait before making irreversible commitments. Once the rules became clearer, they resumed their normal pace of innovation.

The study also looked at whether this law caused companies to stop growing so they could stay under the fifty-employee limit. While the data showed that the number of companies in the fifty-to-ninety-nine employee range did decrease, the researchers could not definitively prove that companies were shrinking on purpose to avoid the law, because the survey did not track individual companies over time. What they could say with certainty was that for the companies that remained large, the law did not kill their ability to invent. It only delayed their most ambitious projects for a short time. The findings suggest that when governments create laws with sharp size cutoffs, the immediate reaction is often a hesitation born of fear and uncertainty, rather than a permanent change in strategy. For policymakers, this means that the true cost of such regulations might be a temporary slowdown, not a permanent loss of innovation, provided that the rules become clear and predictable over time.

Technical Summary: Threshold-based Liability Regulation and the Timing of Radical Innovation

Problem and Research Question
Modern regulation frequently employs size thresholds (e.g., 50 employees) to delineate regulatory regimes, creating a "covered" class and an "exempt" small-firm class. While existing literature establishes that such thresholds distort firm-size distributions and aggregate productivity, it remains unclear how they affect the timing and composition of innovation, particularly radical (new-to-market) innovation.

This paper investigates whether a threshold-based liability shock produces a permanent reduction in the radical innovation of covered firms or a transitory timing distortion. The study focuses on the 2022 Korean Serious Accident Punishment Act (SAPA), which imposes personal criminal liability on officers of firms with 50+ employees for workplace fatalities, while exempting smaller firms until 2024. The core question is whether the liability rule permanently compresses the radical innovation margin or induces a temporary delay as firms navigate enforcement uncertainty.

Data and Setting
The analysis utilizes anonymized repeated cross-sections from the Korean Innovation Survey (KIS) manufacturing waves (2018–2024), comprising 16,095 observations across 24 KSIC industries.

Treatment: Firms with $\ge$ 50 employees (covered class).
Control: Firms with 10–49 employees (exempt class).
Timeline: The 2020 wave captures the pre-Act period. The 2022 wave captures the enactment and preparation window (post-pass, pre-effective). The 2024 wave captures the first two years of the effective regime.
Outcome Variables: Primary outcomes include market-first (new-to-market) product innovation and the share of sales from such products. Secondary outcomes include firm-first (new-to-firm) innovation.

Methodology
The study employs a Difference-in-Differences (DID) framework adapted for repeated cross-sections to estimate population-level changes in mean innovation outcomes.

Baseline Impact: A standard DID compares the change in innovation outcomes between the covered and exempt classes from 2020 to 2022.
Trend Adjustment: Crucially, the authors identify a pre-existing positive trend in market-first innovation for the covered class relative to the exempt class (+4.6 percentage points per wave). Consequently, the primary specification estimates event-study coefficients against a covered-class-specific linear trend derived from the 2018–2020 pre-treatment period, rather than a flat counterfactual.
Robustness Checks:
- Reweighting: Estimates are reweighted to the 2020 industry-by-size-class composition to account for sample drift.
- Functional Form: Logit models (Average Marginal Effects) and log-linear specifications are used to test sensitivity to functional form assumptions.
- Threshold Specificity: "Donut" specifications (excluding the immediate 50–99 employee band) and narrow-band specifications are used to address potential bunching or sorting.
- Inference: Standard errors are clustered at the industry level, with wild-cluster bootstrap inference used to address the small-cluster problem (24 industries).

Key Results

Transitory Dip in Radical Innovation:
- Unadjusted: The covered class experienced a raw decline in market-first innovation relative to the exempt class between 2020 and 2022.
- Trend-Adjusted: When accounting for the pre-existing upward trend of the covered class, the dip in market-first innovation during the enactment window (2022) deepens significantly to −7.8 percentage points ( $p=0.022$ ). This represents a deviation of roughly two-thirds of the 2020 base rate.
- Sales Share: The independent measure of new-to-market sales share shows a similar deviation of −3.3 percentage points ( $p=0.052$ ).
Recovery to Trend:
- The positive coefficient observed in the 2024 wave (+8.1 percentage points unadjusted) is fully explained by the pre-existing trend. The deviation from the trend in 2024 is statistically indistinguishable from zero (−1.0 percentage point, $p=0.87$ ).
- There is no evidence of a permanent level effect; the covered class returned to its prior trajectory within two years.
Substitution Effects:
- Firm-first (incremental) innovation showed a positive shift (+10.4 percentage points) in the unadjusted contrast, suggesting a potential substitution toward lower-liability innovation paths. However, due to the steep pre-existing trend in this margin, the authors characterize this substitution finding as suggestive rather than definitive.
Mechanism:
- The observed pattern (dip followed by a return to trend) aligns with a real-options mechanism driven by liability uncertainty. Firms delayed radical projects during the period of maximum uncertainty (between enactment and the accumulation of enforcement experience) and resumed them as the regime became clearer. This contradicts the prediction of a permanent compliance tax, which would imply a persistent negative deviation.

Contributions and Significance
The paper makes three primary contributions to the literature on size-dependent regulation and innovation:

Distinguishing Level vs. Timing Effects: It demonstrates that the innovation cost of a liability threshold can be a transitory timing distortion rather than a permanent level effect. This challenges the assumption that regulatory shocks necessarily cause permanent reductions in innovation volume.
Methodological Rigor in Trend Analysis: The study highlights the necessity of accounting for pre-existing differential trends when evaluating size-threshold regulations. Failing to adjust for the covered class's prior upward trajectory would have led to an underestimation of the initial shock's magnitude and a misinterpretation of the recovery as a "bounce back" rather than a return to trend.
Liability Rules and Radical Innovation: It provides empirical evidence that liability rules, which price the unknown safety properties of new products, specifically compress the radical (market-first) margin. This supports the theoretical view that uncertainty regarding liability costs disproportionately affects high-risk, high-reward innovation projects.

Conclusion
The authors conclude that the SAPA threshold shock produced a temporary delay in radical innovation for covered firms, concentrated in the enactment window, followed by a return to the pre-existing trend. The distinction is critical for policy evaluation: assessments of size-threshold regulations must track multiple post-reform periods and measure innovation against the covered firms' own prior trends. Otherwise, temporary delays risk being misread as permanent innovation costs, obscuring the dynamic adaptation of firms to liability uncertainty.

Technical Summary: Threshold-based Liability Regulation and the Timing of Radical Innovation

More like this