The paper shows that the average periodic return decreases with the holding period, both for residential real estate in the Netherlands and England and Wales, as well as for commercial real estate in the United States: The longer the holding period is, the lower on average the periodic return is.

The literature provides several reasons why the average periodic returns are higher for shorter holding periods. The first reason is the disposition effect: investors tend to sell more quickly a `winner' and to hold onto a lower-performing property longer. The second reason is that there might be improvements just after purchasing the property. 

The first implication of this finding is that the widely used repeat sales (RS) model is misspecified, because it does not differentiate between holding periods. The second implication is that systematic revisions in RS indices are due to the changing distribution of holding periods over time. This link has so far not been provided in the RS literature on index revision.

This paper proposes an adjustment to the RS model by including dummy variables for each holding period, apart from a baseline holding period to avoid perfect collinearity.  The estimated price index represents the left-out holding period. This model solves the misspecification issue. Moreover, it is shown that (systematic) index revisions are much smaller in RS models including holding period dummy variables compared to a standard RS model.