Statistical offices use the matched-model method to compile consumer price indexes to measure inflation. In markets where models turn over rapidly, the matched sample may, by the end of a year, be quite unrepresentative of what is and what was bought. An analytical model is derived to show how bias in a matched-model index arises and extended to shed further light on the influential Aizcorbe, Corrado, and Doms methodology. The empirical work on the bias is undertaken using monthly scanner data for five products: washing machines, dishwashers, television sets, cameras, and vacuum cleaners. Different strategies are explored that might ameliorate the bias, including using ...