This paper makes an attempt to construct a property price index using the transaction price data for apartments in Tokyo. Property price indices are used to indicate economic vitality in urban areas in many cases; therefore, they are shown in US and European countries. In Japan, property price indices are constructed using the hedonic method, but it is criticized that the indices tend to be higher than the indices of land or detached houses due to omitted variable bias. The repeated sales method where only properties with multiple sales are analyzed creates the problems of sample selection bias and the problem of demonstrating the time-dependent effects of property attributes and neighborhood environment. I adopted the two-step Heckman correction to solve the problem of sample selection bias and analyzed the time-dependent influence of the environment and attributes on the apartment prices in order to create the property indices using the repeated sale prices, which are similar to the price indices of land for detached houses.