9781118041581

Realtors will tell you that buying a house is a no-lose investment. By owning

your dream home, you not only enjoy the housing services you would otherwise

have to pay for in the form of rent, but you also have an asset that is sure to

appreciate in value. After all, housing prices never go down.

With the benefit of hindsight, we know that the last statement is untrue

(housing prices can crash) and that the case for home ownership was way over-

sold. By looking at the past pattern of house prices, could we have recognized

the unusual nature of escalating house prices over the last 20 years? Should we

have been concerned that the housing price bubble might pop and prices

plummet? Figure 4.6 depicts an index of average housing prices for the period

1975 to 2010. The figure shows the level of realhousing prices after netting out

the underlying rate of inflation in the U.S. economy.^8 The message of the fig-

ure is very clear. Between the mid-1970s and the mid-1990s, real housing prices

160 Chapter 4 Estimating and Forecasting Demand

The Housing

Bubble and

Crash

CHECK

STATION 6

A utility that supplies electricity in Wisconsin is attempting to track differences in the

demand for electricity in the winter (October through March) and the summer (April

through September). Using quarterly data from the last five years, it estimates the regres-

sion equation Q 80.5 2.6t 12.4W, where W is a dummy variable (equal to 1 in the

winter quarters, 0 otherwise). Has the utility made a mistake by not including a summer

dummy variable? Now, suppose the utility believes that the rate of increase in demand

differs in the winter and the summer. Think of an equation (using an additional dummy

variable) that incorporates this difference.

The last four variables, W, S, U and F, represent the seasons of the year (U denotes

summer). They are called dummy variables. They take on only the values 0

and 1.For instance, the winter dummy (W) takes on the value 1 if the particular

sales observation occurs in the winter quarter and 0 otherwise. When we perform

an OLS regression to estimate the coefficients b, c, d, e, and f, we obtain:

As we expect, the coefficient for fall is greatest and the coefficient for winter is

lowest. To forecast winter toy sales, we set W 1, S U F 0, generating

the equation Qt1.89t 126.24. Analogously, the predictive equation for fall

toy sales is Qt1.89t 164.38. In essence, we have a different constant term

for each season. To generate a forecast for winter 2005, we use the winter equa-

tion while setting t 41. The computed value for next quarter’s sales is Q 41 

203.73. Contrast this with the prediction of 223.08 based on the simple trend.

Accounting for seasonality via dummy variables provides a much more realis-

tic prediction.

Qt1.89t126.24W139.85S143.26U164.38F.

(^8) So an annual change in the index from 100 to 101.5 means that average house prices rose 1.5 per-
cent faster than the rate of inflation. If inflation averaged 2.5 percent, then nominal house prices
increased by 4 percent.
c04EstimatingandForecastingDemand.qxd 9/20/11 9:14 AM Page 160