In this thesis we investigate the possible ways of generating macroeconomic scenarios that serve as input for an asset and liability management (ALM) study of a pension fund. Firstwe discuss the different type of models that can be used and we conclude that the vector autoregressive (VAR) model is the most suited. We estimate a VAR model ourselves extended with an affine term structure model of interest rates, in this way there are no arbitrage opportunities. The simulations from our model are used in an ALM study for a typical Dutch pension fund. Different time periods of historical data are being considered and we find that these have a great impact on our estimation results and corresponding scenarios. At last the term structure model is extended for longer maturities, but this extension does not lead to a better model fit.