Revisiting Stock Market Signals as a Lens for Patent Valuation

Estimating the private value of patents is important, yet challenging. By developing a method based on stock market returns to produce estimates of individual patent values, Papanikolaou, Seru, and Stoffman (2017) (KPSS) opened venues for new research. We characterize the measurement error in KPSS -- the difference between the latent true patent value and the corresponding KPSS estimate -- and show it is negatively correlated with the latent true patent value. We then investigate the use of KPSS estimates in two different applications. First, we show that using KPSS values to gauge differences in value between different patent groups is internally inconsistent and introduces attenuation bias. We offer two solutions: extending the original KPSS method to allow for patents to be drawn from two distinct value distributions, and using abnormal stock market returns. We compare both to the original KPSS estimates in several contexts relevant to the organizational scholars, such as patents by large and small teams, scientific and non-scientific patents, and offshored and domestically invented patents. Second, we show that KPSS yield unbiased estimates when used as explanatory variables. These analyses allow us to characterize the main trade-offs associated with each approach, and offer practical guidance to researchers.