In this paper we apply the Kalman filter to a state formulation of a multi-factor term structure model allowing for measurement errors in the data. We estimate one and two factor models using panel data allowing the cross sectional and dynamic implications of the yield curve to be taken into account. The panel data approach has the advantage of using all the information in the yield curve across and over time compared to a time series approach only. The models are estimated on data for the Belgian franc, British pound, Danish krone, Dutch guilder, French franc, German mark, Japanese yen, Italian lira and Swiss franc. Our empirical results indicate that the two factor model represents a good description of the yield curves in these markets.