Modeling and estimating returns to seller reputation with unobserved heterogeneity in online auctions

•We convert the online auction data as an unbalanced panel dataset.•This study considers seller's unobserved effect under sample selection framework.•The empirical finding confirms the existence of unobserved heterogeneity of sellers.•The returns to reputation is overestimated when a pooled dataset is used.•Seller's reputation remains an important factor for buyers when bidding on items.

This study provides an alternative approach that takes account of the unobserved effects of each seller under a sample selection framework. Empirical findings, after converting the Yahoo! Japan Auctions dataset into unbalanced panel data, show that seller reputation remains an important factor for buyers when bidding on items, while the sample data confirm the unobserved heterogeneity of sellers and the sample selection problem.