An Investigation of the Swedish Consumption Function : An Error-Correction Approach

This thesis examines the Swedish aggregate consumption function using the concept of cointegration, and explores whether consumption, income, financial wealth, and housing wealth share a long-run trend. The goal of the study was to determine the strength of this cointegrating relationship, the relative roles of housing wealth and financial wealth in the consumption function, and a suitable method for forecasting consumption. The strength of the cointegrating relationship, incorporated in a vector error-correction model (VECM), is evaluated by that model’s forecasts. Its forecasting performance is assessed via accuracy tests and comparisons to alternative models. The study also examines the effect of allowing for intercept correction in the VECM’s forecasts. The preferred VECM demonstrates that consumption is the variable that error corrects to restore the long-run relationship after a disequilibrium, with a correction time of less than three quarters. The results indicate the vital role of financial wealth in the consumption function, while revealing that housing wealth has no significant effect on consumption in either the long run or in the short run. When it comes to forecasting, the alternative, data-driven Bayesian vector autoregressive model generated relatively accurate short-run forecasts, while the theory-driven VECM generated relatively accurate long-run predictions. Moreover, the results further indicated the existence of the cointegrating relationship as the VECM outperformed the vector autoregressive model at both long horizon and short horizon forecasts. Additionally, intercept correction slightly improved short-run, but not long-run, forecasts, giving no indication of a misspecified VECM, or a significant structural shift during the forecast period, 2010-2015.