This paper examines the relationship between innovation and firm growth in high-tech sectors using a quantile regression approach. The authors, Alex Coad and Rekha Rao, focus on four high-tech sectors: industrial and commercial machinery, electronic and electrical equipment, transportation equipment, and measuring instruments. They argue that while firms generally experience modest growth, the growth rates are heavily skewed, making standard regression techniques less suitable for analyzing the impact of innovation. By using quantile regression, they find that innovation is crucial for a small number of "superstar" firms with rapid growth. The study also discusses the policy implications of these findings, suggesting that broad-based innovation policies that support multiple firms in various directions of search may be more effective than targeted policies focusing on a few firms. The authors conclude that while innovation is important, it is highly uncertain and lacks persistence, and firm growth is best modeled as a random walk.This paper examines the relationship between innovation and firm growth in high-tech sectors using a quantile regression approach. The authors, Alex Coad and Rekha Rao, focus on four high-tech sectors: industrial and commercial machinery, electronic and electrical equipment, transportation equipment, and measuring instruments. They argue that while firms generally experience modest growth, the growth rates are heavily skewed, making standard regression techniques less suitable for analyzing the impact of innovation. By using quantile regression, they find that innovation is crucial for a small number of "superstar" firms with rapid growth. The study also discusses the policy implications of these findings, suggesting that broad-based innovation policies that support multiple firms in various directions of search may be more effective than targeted policies focusing on a few firms. The authors conclude that while innovation is important, it is highly uncertain and lacks persistence, and firm growth is best modeled as a random walk.