قیمت و حجم معامله در بازار مسکن هلند

Housing markets typically exhibit a strong positive correlation between the rate of price increase and the number of houses sold. We document this correlation on high-quality Dutch data for the period 1985–2007, and estimate a VEC-model that allows us to study the mechanism giving rise to the correlation. The data identify the flows of new houses offered for sale as well as the number of houses sold. According to the estimated model, shocks to market fundamentals (the mortgage rate) have an immediate and significant impact on the rate of sale, little impact on the rate of entry of new houses for sale, and a gradual impact on the house prices. This pattern is consistent with an economy where buyers and sellers gradually learn about changes in market conditions.

Between the mid 1980s and today real house prices more or less doubled in most industrialized countries. Even before the recent crisis, this was not a smooth process of continuous growth. All countries experienced cycles where booms with price increases above trend were followed by busts with stagnating or falling prices. But price fluctuations alone do not fully characterize the ups and downs of the housing market. In price booms, the market is typically also more liquid with frequent transactions and houses selling quickly, whereas in busts there are fewer sales and many houses remain on the market for a long time. Since the housing stock is fixed in the short run and most transactions are driven by households moving from one home to another, i.e. being present more or less simultaneously on both sides of the market, this is something of a puzzle. Fig. 1 and Fig. 2 illustrate this pattern for the Dutch market for owner-occupied homes, based on the data being analyzed in this paper. Peaks and troughs of price changes and the number of sales coincide clearly in some periods (e.g. the trough in 1995 and the peak in 1999), but over other periods the correlation is weak. Once we relate sales to the number of houses for sale there is a much stronger correlation. Fig. 2 shows that the rate of sale (sales divided by houses on the market) and price changes follow each other very closely (the correlation coefficient is 0.71).Market developments reflect the decisions of thousands of homeowners to offer their houses for sale and to set reservation prices that they are prepared to accept. At the same time, prospective homebuyers shop around for good deals. Given the search nature of the process, the market does not clear continuously and variations in the time on the market and the rate of sale may accommodate inertia in prices. Previous empirical studies have largely been confined to looking at data on transaction prices and the number of sales. In this paper we are able to give a richer picture of the process. We analyze detailed Dutch data that allow us to distinguish between the rate of entry of new houses offered for sale and the rate at which the houses on the market are being sold. We are also able to distinguish between the list price at which a house is offered for sale and the final transaction price. The data set contains observations on these variables for the entire Dutch housing market over a period of more than 20 years. For nearly two million dwellings sold between 1985 and 2007, we observe the original list price and the date when the dwelling was put on the market as well as the final sales price and the date of sale. Beyond providing descriptive statistics on these variables, we estimate a dynamic model of the housing market where prices and quantities are driven by disturbances to two fundamental demand factors, unemployment and the interest rate. In our estimated vector error-correction model shocks to demand fundamentals have an immediate but temporary impact on the rate of sale, a gradual and permanent impact on prices (both list price and sales price) and little impact on the rate of entry. This dynamic pattern is in line with previous studies by Hort (2000) on Swedish data and Andrew and Meen (2003) on British data, although these studies look at total sales rather than the rate of sale. It stands in some contrast to a study on US data by Clayton et al. (2010), which finds that the effect on both price and sales peak already one quarter after the shock. We discuss the interpretation of our results against the background of different theories of the interaction between housing prices and sales. We conclude that it is consistent with a housing market where agents gradually learn about changed market conditions, e.g. because sellers have a good overview of the houses offered for sale whereas sellers do not have a corresponding overview of potential buyers searching for a new dwelling. The next section of the paper reviews the empirical literature on the price-quantity correlation. Various theoretical explanations for this correlation have been advanced. In Section 3 we briefly discuss the main candidates: liquidity, asymmetric information, the interaction between credit constraints and demand and supply, and loss aversion. Next we present the data and some descriptive statistics in Section 4. The vector-error-correction model is specified in Section 5 and the estimation results, primarily in the form of impulse-response functions, are presented in Section 6. We conclude in Section 7 that the results are consistent with the asymmetric information view but not with the credit-constraint story.

Details of the estimated model are given in Table 3. Let us first look at the equilibrium properties represented by the cointegrating relations and then turn to the dynamics of the model. 6.1. The cointegrating relations Imposing exactly identified coefficient restrictions as discussed in the previous section, we get the following estimates of the cointegrating vectors. equation(2) tp=−0.233u−0.267i+0.002ttp=−0.233u−0.267i+0.002t equation(3) lp−tp=−0.0011u−0.017i−0.0002tlp−tp=−0.0011u−0.017i−0.0002t equation(4) s=0.108u+0.278i+0.002ts=0.108u+0.278i+0.002t The equilibrium elasticities of price with respect to unemployment and interest are − 0.23 and − 0.27, respectively. There is a positive price trend of 0.2% per month reflecting the influence of income, demographics and other factors not present in the model. To account for the time trend, note that the Kranendonk and Verbruggen (2008) and Francke et al. (2009) both report income elasticities on the order of unity. Combined with a period of income rising with a couple of percent per year, this translates into a positive price trend comparable to the one we have estimated. The interest elasticity is higher than the − 0.086 estimated by Francke et al. in their one-equation error-correction model. The second equation gives the equilibrium elasticities of the premium of list price over sales price as a function of interest and unemployment. Both elasticities are negative; the higher is the interest rate (the cost of waiting) and the higher is unemployment the more eager are households to sell quickly and the lower is the list price relative to the sales price. The elasticities may seem quite small, 1.1 and 1.7%, respectively. But remember that these are elasticities of the ratio of list price to sales price (not of the premium). To get a sense of the magnitudes consider the impact of the real interest rate falling from its peak value of 7.5% (recorded in 1990) to the lowest value of 1.2% (in 2005). A change of this magnitude would cause the ratio of list price to sales price to fall by 3.1%. This effect should be related to the fact that the list-to-sales-price ratio has fluctuated between 1.057 and 1.024, i.e. a fall in this ratio of 3.1% is equivalent to a fall in the list price premium from its maximum to its minimum observed value. The third equation shows, consistent with the second, that the rate of sale is related positively to the interest rate and unemployment. The higher the interest and unemployment, the lower is list price and the higher is the rate of sale. The fact that u and i enter with opposite signs in the third compared with the first cointegrating relation, indicates that the levels of price and rate of sale will be negatively correlated across long-run equilibria. This suggests that an explanation for the observed positive correlation between price and sales should be found in the dynamic adjustment process rather than in properties of long-run market equilibria. The residuals from the cointegrating relations are displayed in Fig. 11a–c. In the latter half of the 1990s, when interest and unemployment rates were both falling, the residual of the first equation went from positive to negative. Interpreting the residual as the deviation from the fundamental market price, this indicates that house prices went from over- to under-valuation. As prices continued to increase despite interest rates picking up from 1999, under-valuation quickly turned into over-valuation. More recently, the residual has been quite volatile. Following several years of stagnating house prices after 2000, our model indicates that Dutch house prices were in fact under-valued by 7% in relation to fundamentals in 2007. The residuals from the second equation indicate that the list price premium was below its equilibrium value in the period of increasing sales prices in the late 1990s, perhaps reflecting that prices were increasing faster than expected when list prices were set. Conversely, there was a large premium, i.e. sales prices were low relative to list prices, when the market was stagnating in the early 1990s and after 2000.