Unobserved apps quality. Apps likely differ in quality levels, which we cannot observe completely. Higher-quality apps are ranked higher, receive higher ratings, and can charge higher prices.Moreover, apps can have different privacy settings (not observed in our data), which affect both their popularity and price (Kummer and Shulte 2016). This leads to a correlation among price, rating, and the error term. Therefore, we need to account for unobserved app-level effects.We do so by including time-invariant fixed effects, vj, in our dynamic panel model: