Conduct further regression analysis on the fuel price and consumption data.

Please answer the questions concisely in your report, and include figures and tables of regression results where called for. These can be copied and pasted from Excel or R. Make sure they are sized and formatted to be legible, number them, and refer to them by number in you write-up (e.g., "... as can be seen in Table 2...").

Using the Required Data Set linked on this page, answer the following questions.

1. Fuel demand in OECD countries: This largely replicates what you saw in the lectures.
   • Restricting your data to the OECD subset for now, create two scatter plots showing the relationship between fuelcon and fuelprice and between fuelcon and gdppc. Briefly describe what they imply about the relationships between these variables.
   • Use multiple regression (for the OECD sample) to estimate the demand function for motor fuels (again using fuelcon and fuelprice), with quantity as a function of both the price and income (per capita GDP). Run the regression once in a linear specification and once in terms of the natural logs of the variables. Display the tables of results.
   • Discuss the regression results: interpret the slope coefficients, discuss their statistical significance, and discuss the goodness of fit of the regressions. Are the results consistent with what you would expect for a demand relationship? How do you interpret the slope coefficients in the regressions in logs? Are they of a magnitude you would expect?
2. Fuel demand for non-OECD countries:
   • Repeat all parts of #1 for the sample of non-OECD countries.
   • Compare the results for the OECD and non-OECD samples. Do the intercepts and the effects of price and income appear to be similar across these groups of countries? In comparing across regressions, use the 95% confidence intervals to give a rough idea whether the estimates are "close" to each other or quite different. What do you think might account for any substantial differences?
3. A skeptic claims that price and income are far less important in determining cross-country differences in fuel consumption than other factors, such as whether the country has an abundant supply of fossil fuels, or whether people have to drive to get where they need to go, or how far they have to drive when they do. Using variables available in the spreadsheet (as well as any new variables you might construct from them), assess these claims, with scatter plots and multiple regression(s). Make sure you explain what you have done, and include controls for price and income in all regressions. Does the evidence bear out the skeptic's claims? Does any of these purported factors help account for any major differences between the OECD and non-OECD samples? Explain how you know, using additional data analysis if needed.

Data Excel Sheet 1:

Data on vehicle fuel consumption and prices, c. 2010 Countries with valid data on gasoline and diesel, GDP per capita Variable Definition oecd = 1 if OECD country, 0 otherwise code Country abbreviation area Land area in square km gappc GDP per capita in US$ pop population fuelcon consumption of gasoline plus diesel fuel in gallons per year per capita fuelprice average of gasoline and diesel price in US$ per gallon, weighted by consumption shares oil_rents_gdp Oil rents (net income from petroleum production) as percent of GDP pop_urban Percent of population living in metro areas of more than 1 million population Data from online appendix to Lucas W. Davis The Economic Cost of Global Fuel Subsidies AMERICAN ECONOMIC REVIEW 104(5) May 2014 pp. 581-585. https://www.aeaweb.org/articles?id=10.1257/aer. 104.5.581 except ... Land area, oil rents, and urban pop share from World Development Indicators https://databank.worldbank.org/source/world-development-indicators