The simulation studies by Lux and Marchesi (1999),(2000) of the noise trader infection model invented by Lux (1998) were among the first to demonstrate detailed agreement of the simulated time series of an artificial market with the main stylized facts of empirical asset returns, e.g. heteroscedastic returns with a realistic tail index and long memory in clustered return volatility despite uncorrelated returns. I extend their univariate model into a multivariate setup containing a second risky asset and a riskless bond, which was not available in their original model. In order to add some further realism, the investment process will be split up into asset allocation and security selection, as is common practice in financial institutions.