Please rotate your device to landscape mode to view the charts.

The Use and Misuse of Patent Data: Issues for Finance and Beyond

Journal: Review of Financial Studies

Date: 20220601

Author: Lerner, Josh; Seru, Amit

Abstract:
Patents and citations are powerful tools increasingly used in financial economics (and management research more broadly) to understand innovation. Biases may result, however, from the interactions between the truncation of patents and citations and the changing composition of inventors. When aggregated at the firm level, these patent and citation biases can survive popular adjustment methods and are correlated with firm characteristics. These issues can lead to problematic inferences. We provide an actionable checklist to avoid biased inferences and also suggest machine learning as a potential new way to address these problems.

Link: Google Scholar

Key Findings

Patent Data Bias Issues

Significant biases exist in patent and citation counts when aggregated at firm level, particularly affecting newer patents and citations. These biases are systematically related to firm characteristics.

Technology & Regional Variation

Patent and citation biases vary substantially across technology classes and geographical regions, with computer/electronics patents and states like California showing the largest disparities.

Machine Learning Solutions

Machine learning approaches using firm-level information perform significantly better than traditional adjustment methods in addressing patent and citation biases.

Firm-Level Patent and Citation Bias Correlations

  • Larger firms show greater patent bias (0.0548 coefficient) and citation bias (0.121 coefficient)
  • Higher market-to-book ratios correlate with increased patent bias (0.0464) and citation bias (0.0920)
  • R&D intensity shows positive correlation with both patent (0.0235) and citation (0.0417) biases

Machine Learning Model Performance Comparison

  • Machine learning models achieve higher R² values (0.74-0.81) compared to traditional benchmarks (0.43-0.63)
  • Linear SVR performs best with R² of 0.81 and lowest RMSE of 66.22
  • All ML models outperform conventional adjustment methods

Regional Patent Bias Distribution

  • California and Massachusetts show 2.5x increase in patenting between 1990-2000
  • Delaware showed minimal increase in patenting activity
  • Regional differences persist even after traditional adjustments

Contribution and Implications

  • Demonstrates systematic biases in patent data that affect research inferences in corporate finance
  • Provides an actionable checklist for researchers using patent data
  • Introduces machine learning as a promising solution for addressing patent data biases

Data Sources

  • Firm-level bias correlations based on Table 1 regression coefficients
  • Machine learning performance metrics derived from Table 4 model comparisons
  • Regional patent distribution based on Figure 3 patent application trends