# Economics 312

## Spring 2010Take-Home Midterm QuestionsDue by electronic submission: Noon, Friday, March 5

### Instructions

1. This part of the exam is open-book; you may use whatever outside materials you wish. However, you may not communicate with anyone except the instructor about the exam.

2. The exam is due electronically at noon on Friday, March 5. There is no time limit, but you should not need to spend more than a few hours on it.

3. You are responsible for making sure that you understand each question clearly. In case of any ambiguity, be sure to consult the instructor.

4. When you discuss the estimated effects of one variable on another, but sure to consider both statistical significance and economic significance.

5. If you use Stata commands/options or statistical methods that have not been discussed in class, you should briefly describe what the command/test/method does and justify why it is appropriate to use. You should supplement your answers with supporting Stata output when it is useful.

### Question #1

Hedonic price analysis is used to model the value of commodities with varying characteristics. Early applications of this analysis included cars and asparagus (!). More recent applications have included computers and, very commonly, houses. Hedonic analyses of house prices are usually done with the individual house as the unit of analysis, but this question asks you to apply the analysis to averages of house values over small areas.

The data set Hprice.dta contains a set of variables for each of 506 neighborhoods. The variables in the data set are defined below:

nnumber = neighborhood number

price = median neighborhood house price (\$)
crime = crimes committed in neighborhood per capita
nox = atmospheric concentration of nitrous oxides (p/100m)
rooms = average number of rooms in neighborhood houses
dist = weighted distance of neighborhood to 5 employment centers
proptax = property tax per \$1000 valuation in neighborhood
stratio = student-teacher ratio in neighborhood schools
lowstat = percentage of neighborhood people of "low status"

a. What determines the median house price in these neighborhoods? Use regression analysis to estimate a model of house prices. Explain and defend your choice of specification and estimation method (functional form, assumptions of the estimator, etc.). You may present multiple alternative models in a table for comparison purposes but be clear about which model you prefer and why.

b. Suppose that you have been hired by the board of directors of neighborhood number 130's homeowners' association. The board believes that their neighborhood's property is undervalued and wants to increase house values. They have asked you to advise them on the following propositions using your statistical analysis. For each, write a short report supported by relevant tables showing your estimates and calculations.

i. Are board members justified in believing that property values for neighborhood 130 are too low, given the neighborhood's characteristics?

ii. What are the major factors affecting values in neighborhood 130 both positively and negatively relative to other neighborhoods?

iii. Colin Commuter, one of the board members, has argued that improving a major road connecting the neighborhood to downtown would increase property values. The road improvement would increase the value of the radial access index from 4 to 8, but the increased traffic would increase nitrous oxide concentrations from 6.24 to 6.50. Does your analysis support Mr. Commuter's argument? How confident are you of your answer?

iv. Board member Sarah Scardicat is worried about neighborhood crime. She has proposed an increase in police service that would require an increase in property taxes from 43.7 to 47.7 and lower crime from 0.881 to 0.6. Would this significantly increase property values?

### Question #2

The effects of campaign expenditures on voting outcomes are of interest to both economists and political scientists. The dataset Vote.dta contains data for 173 two-party races for the U.S. House of Representatives in 1988. The variable voteD contains the percentage of the votes received by the Democratic candidate. Expenditures by the two candidates (in thousands of dollars) are given by expendR and expendD. The variable prtystrD measures the Democratic party's general strength in the district as the percentage of the 1988 Presidential vote in that district that went to the Democratic candidate.

a. Estimate the relationship between voting outcomes and expenditures using an appropriate econometric model. Provide an argument supporting the model and estimator that you use. How effective are campaign expenditures in explaining voting?

b. Is campaign spending an important determinant of vote shares? Explain.

c. Does an equal increase in spending in a campaign (by both candidates) favor Democrats, favor Republicans, or have no effect? How strong is this result statistically and economically? How does this conclusion affect your choice of econometric model?

d. Is an additional \$1000 of campaign spending by Democrats on Congressional races more effective at increasing the Democratic share of the vote in districts that tend to vote Democratic (based on Presidential voting)? Is an additional \$1000 of campaign spending by Republicans on Congressional races more effective at increasing the Republican share of the vote in districts that tend to vote Republican?

e. Using your estimated relationship, find the additional amount that the Democratic candidate in the 6th district of Florida (FL) would have had to spend in order to win that race, given the amount spent by the Republican.

f. Suppose that the Democratic Party had received an additional \$10,000,000 in campaign funds. They would like to use it to maximize the number of Democrats winning Congressional seats. Based on your analysis, what advice would you give them (retrospectively) about which Congressional campaigns should have receive the additional funds?