Beautiful Data

Analysis

For a broad overview of the changes in the housing market, we'll start with the evolution of the average sale price and number of sales. Since the data is reported weekly, that's a natural time unit to use.

Figure 18-1 shows weekly average sale price and number of sales for the 293 weeks in the data. There are some very interesting patterns. The behavior of the average price is striking, with an increasing trend until June 2007 and then a precipitous drop to the present day—a clear illustration of the boom and bust in housing prices.

Sales look quite different. Most years (especially 2004 and 2005) show a marked seasonal effect, with a peak in mid- to late summer and fewer sales in the winter months. (This may be a good place to note that the data only rarely includes the true closing date, so we're using the date when the sale was reported in the newspaper, which may be four to six weeks later than the closing.) Once we look past the seasonal effect, we see something else. From the middle of 2006 until early 2008, sales volume decreases, surely an indicator of the housing bust. However, the sharpness of the drop in early 2008 may also reflect the winter slowdown in sales. And what about the increase starting in early 2008? One possibility is that by this point house prices had dropped enough that buyers were shopping for bargains with the arrival of spring. Another possibility is that some of this increase is due to foreclosure sales. Perhaps the explanation will be clearer ...