1. Paper: (10 Points) In this homework you will start working on the “conceptual model” part of your airport paper. The idea is to develop an argument that helps explain the change in airport size over time. As illustrated in the first lecture of the course, think of Demand and Supply. Your quantity might be thought of as the solution to solving demand and supply so that passengers is a function of demand and supply shifters. Things like capacity expansions (terminal or runway) as easily thought of in terms of supply shifts, while thing like an airline making the airport a hub is a demand shift as might be 911, seasonal dummies, etc. As you move down the road, you might identify factors (from the intro/background section) by looking at the graph of passengers against trend, and look for any major changes? When did they happen? What happened? It may also be helpful to search more about your airport and find out if there is anything unique about the airport over time or about the place that it is located. Try to bring what you know about your airport together with the changes in passenger traffic to come up with a conceptual model. You can do a scholar.google.com search for economics articles on firm size, airport size, growth, etc. to get an idea on how to do it. In the end, tell your stories in terms of the model of demand and supply. 2. Data construction. A. (5) Turn in a do file and graphs and output as described below. Call the HW1 data for your airport (use if origin==”XXX” using hw1.dta) merge it with the quar_inc_pop data, see STATA MANUAL ADDITION 2 and then with GDPPD. This latter comes from the St. Louis Federal Reserve bank data (FRED). 1 B. (5) First, regress passengers on real incomes, population for different distances and in all models include a dummy for quarter 2, 3, 4 and a dummy for 911 (use quarter 4), a trend (t=1 for quarter 1 of 1993), and a dummy interaction between 911 and trend and save the output using outreg2. Find the airport with the highest R2. Note: the passenger data is the result of a 10 percent sample. So multiply it by 10 for an estimated number of origins from any airport. Now, graphically portray the R2 against different distances. See STATA MANUAL ADDITION 3 for help. 3. Heteroskedasticity (20) Call the airport data for quarter 4 of 2009 (all airports). Run a regression of passengers on income(20) and population(20). a.(4) Use a Goldfeld Quandt test to check for heteroscedasticity (5%). Assume the H is due to population, and only population appears on the RHS. There are 100 airports, you might recall in class, sort the data by population. Define group=., then replace group=1 if _n<38 and replace group=2 if group>63 (this was shown in class). Now run a regression for each group (reg pass population if group==1, you could gen rsmall=e(rss) and do the same for group=2. Note: Stata calculates lots of stuff and stores in memory. Type help reg in the command line scroll to the end you will see that a variety of items are stored as e(statistic). For the last regression, it has in memory e.g., e(rss) and other things which you can use. This is only for the last regression. These values are replaced for the next regression unless you define a variable. So after each regression, you might pull rss’s with the command “gen rss1=e(rss), and the same for group 2 after the second regression. Then calculate the statistics. b.(4) Run a regression of passengers on income and population. Use a Whitetest to check for heteroskedasity (5%). (see book and class notes) c.(4) Use outreg2 to present the OLS with and without the heteroskedasity correction. Which do you prefer and why? (See book and class notes) d.(4) Suppose that you have a model passengers as a function of population, and you know that the variance is proportional to population. Please run weighted least squares. Add to the outreg2 file, and compare (see class notes, book, and also stata manual addition 1) e.(4) Now do feasible or estimated GLS with income, population and quarterly dummies as explanatory variables. Assume that the variance is a linear function of each of these variables. Add to the outreg2 table in c. (see class notes).