Question

In attempting to forecast future quarterly sales of a company, you fail to reject the unit-root hypothesis on the series of ln(sales in millions), so instead run a regression on the first differences in the log of sales. The coefficient estimates are as follows:

	Coefficient	Standard error
Intercept	0.17	0.02
ln(St-1)-ln(St-2)	-0.58	0.14

The most recent quarterly sales figure is 7.27 million, and the one before that is 7.69 million. Based on the regression coefficients, what is your forecast for the next quarter's sales?

Hint: the regression model allows you to easily forecast the difference in ln(sales), specifically ln(St+1)-ln(St), based on the coefficients and the previous difference in sales, ln(St)-ln(St-1) Once you have that forecast, use some algebra to rearrange and solve for St+1.

Answer 1