On price formation and quantity adjustment in Swedish housing markets

Abstract: This thesis consists of four essays on price formation and quantity adjustment in Swedish markets for owner-occupied houses.Essay I: As a means of describing the time-series behaviour of single-family house prices, univariate forecasting equations are estimated by seemingly unrelated regression techniques using data for 20 urban areas over the period 1967-92. The predictability of real house price changes is found to be quite significant. Real appreciation rates are positively correlated in the short run and negatively correlated in the long run. This mean reverting pattern applies in all urban areas and is most pronounced in the latter half of the period.Essay II: Based on a dynamic capital asset market model, a restricted error-correction model of real house prices is estimated on Swedish panel data for 1968-94. A cointegrating relationship is identified between house prices, total income, user costs and construction costs, all measured in real terms. The model is used to investigate whether part of the dramatic house price fluctuations in Sweden over the last decade can be referred to speculative behaviour in the housing market. It is shown that house prices are well explained by variations in fundamental demand and supply conditions. Therefore, without any evidence on the reasoning of economic agents, there is no basis for referring these fluctuations to speculative behaviour.Essay III: Informational imperfections in the housing market imply that the adjustment of house price expectations for individual units is likely to be slow. There may also be asymmetries in buyers' and sellers' responses so that the market exhibits some quantity adjustment. Empirical evidence on the contemporaneous relationship between prices and sales suggests that sellers respond more slowly than buyers. From a search theoretic model consistent with this result sales are predicted to respond prior to prices following a shock to demand. To test this empirically, estimates of a VAR-model of the after-tax interest rate, house prices and sales are used. The results support the hypothesis that sales respond prior to prices following a shock to the after-tax interest rate.Essay IV: When the typical seller's horizon is finite, standard search models postulate that sellers' reservation prices decline during the search process. This is tested using micro data on sales of owner-occupied houses in Sweden 1995-96. In regressions of the selling price, time to sell yields a negative coefficient. Unlike other mechanisms that might lead to this result, a tendency in seller's reservation prices to decline over time would also imply that the probability of selling the house, given that it has not yet been sold, increases with time. To test this hypothesis, a Weibull hazard and a semi-parametric hazard model are estimated. The results provide no support for the hypothesised positive duration dependence in the hazard rate.