The pairs (or repeat) transactions index model is the most widely used method for constructing housing market index and for estimating individual house prices. However, this model is unable to provide price level information on the housing market and encounters several statistical problems when individual houses have more than two transactions. In order to overcome the shortcomings of pairs transactions model, we propose a multi transactions model with the panel data approach to estimate market index and to predict house prices. The empirical results based on 5,000 houses in Howard County Maryland demonstrate that the proposed multi transactions model: (1) can provide price level information on the housing market; (2) produces market index with smaller standard errors; (3) is more robust in terms of difference between the sub-sample estimate and full sample estimate; (4) has smaller index revision volatilities; and (5) performs better in the out-of-sample test on the prediction of individual house prices. Thus, the multi transactions model is recommended for constructing housing market index and assessing individual house prices.