You are now working at the Department of Commerce and your boss wants to know how changes in the price of gasoline affect demand for gas in the United States controlling for the country's per capita income, and car usage. You have data on gasoline consumption and prices for 19 years.
YEAR = year, 1960-1978
LGASPCAR = log of consumption per car
LINCOMEP = log of per capita income
LRPMG = log of real price of gasoline
LCARPCAP = log of per capita number of cars
Note: For how these data was originally analyzed see Baltagi, B. and Griffin, J., "Gasoline Demand in the OECD: An Application of Pooling and Testing Procedures," European Economic Review, 22, 1983, pp. 117-137.
You estimate a regression and obtain the following STATA output:
a) Interpret the coefficient on lrpmg. Is it statistically significant? How can you tell? Construct a 95% confidence interval around the same coefficient.
b) Is the coefficient on lcarpcap statistically significant? How can you tell?
c) What do the Wand adjusted- 11, tell you in general? In this example, compare the two values, what do you conclude about the model based on this information?
d) Since gasoline prices might be related from one year to the next, you are concerned that there might be autocorrelation. Assume pure autocorrelation is your only concern. Explain specifically what problems it might cause. Be specific in addressing how it might affect both coefficients and standard errors. What does this mean for hypothesis testing?
e) Conduct a two-sided test using the Durbin-Watson d statistic from the STATA output. Is there evidence of autocorrelation? How can you correct the problem caused by autocorrelation?