In the mid-eighties econometric forecasts and ex post simulations of private consumption in Norway began to show clear signs of "structural breakdown" .This evidence lends itself to two interpretations, distinct in their implications for econometric modelling of aggregate consumption. On the one hand consumer behaviour may have been fundamentally changed by the recent deregulation of the financial and housing markets in Norway. Another interpretation stresses the possibility of inherent misspecification of empirical consumption functions. In the latter case one would aim at reconstructing an empirical consumption function, using data which does not include the breakdown period, but which is capable of accounting for the development of private consumption in that period. In this paper we challenge the interpretation that forecast failures provide evidence of "structural shift" by reconstructing the consumption function, using data for the whole sample period. Moreover, the paper puts emphasis on the temporal properties of the data according to tests proposed by Hylleberg, Engle, Granger og Yeo (1988). The empirical evidence implies that consumption and income are not cointegrated at the zero frequency. On the other hand, cointegration tests - including Johansen's Full information procedure - indicate that consumption, income and wealth are cointegrated. The paper shows that an error correction model in consumption, income and wealth is invariant of the changes in the credit and in the housing markets. This shows that the predictive failure of existing consumption functions does not constitute evidence against a conditional modelling approach to private households consumption.