The Economic Cost of Selection Neglect in Portfolio Choice: Evidence from Australian Fine Wine Auctions

We estimate whether investors incur economic costs when they observe a selected subset of market outcomes. Using data from Langton’s, Australia’s principal wine auction house, we apply a Markov Chain Monte Carlo model that jointly estimates latent prices and the probability of trading, correcting selection bias. Our auction setting offers direct observability of four outcomes – sold, passed-in, withdrawn, and never offered – making the selection mechanism transparent. We demonstrate that investors who ignore such selection effects create suboptimal mean-variance portfolios, forgoing 1.2% per month in certainty-equivalent returns, driven mainly by underestimated expected returns rather than misestimated covariances.