The first six questions on this exam are inspired by the paper:

Michael F. Lovenheim and Joel Slemrod, 2010, "The Fatal Toll of Driving to Drink: The Effect of Minimum Legal Drinking Age Evasion on Traffic Fatalities," Journal of Health Economics 29,p. 62-77.

In the 1970s, 39 states of the United States lowered their minimum legal drinking age (MLDA) from 21 years old to 18, 19 or 20 years old. These changes were followed by increases in the numbers of fatal traffic accidents involving teenage drivers, raising concern that allowing teens to drink was putting them (and others) at greater risk of fatal traffic accidents. By the late 1970s, some states started raising their MLDAs back to higher levels. In 1984 the National Minimum Drinking Age Act mandated that all states raise their MLDAs to 21 years old or forfeit federal highway funds. By 1987, the minimum legal drinking age was 21 in all states. Thus, during the 1970s and 1980s, the minimum legal drinking age varied across states and over time.

Many researchers have used data from the 1970s and 1980s to study the effect on "fatal traffic accidents involving 18-year-olds" (a left-hand-side variable) of state government choices regarding whether to prohibit or allow drinking by 18-year-olds. When a state's minimum legal drinking age is 19 or older, the state prohibitsthe drinking of alcohol by 18-year-olds. When the state's minimum legal drinking age is 18 years old, the state allows the drinking of alcohol by 18-year-olds. Many studies confirm that prohibiting drinking by 18-year-olds reduces the total number of fatal traffic accidents involving 18-year-old drivers.

The authors of the paper cited above point out that prohibiting drinking by 18-year-olds might have a more complicated effect on fatal traffic accidents than is sometimes allowed for in econometric studies, because some states that prohibited drinking by 18-year-olds had neighboring states where drinking by 18-year-olds was still allowed. They suggest that when a state prohibits the drinking of alcohol by 18-year-olds, the effect on fatal traffic accidents involving 18-year-olds might differ across locations within the state, depending on whether a location is near to or far from the border of a state where 18-year-olds can still drink legally. For 18-year-olds who live farfrom the border of a state where they may still drink legally, the prohibition of legal drinking by 18-year-olds in their home state should tend to reduce their drinking and to reduce their risk of fatal traffic accidents. For 18-year-olds who live near to the border of a state where they may still drink legally, however, the prohibition of legal drinking in their home state might increasetheir risk of being involved in a fatal traffic accident, by creating an incentive for them to drive across the border to drink legally, thereby encouraging them to drive longer distances when drinking.

Table 1 reports the results of two regressions from the paper (with some details changed for the purposes of this exam).The regressions employ a dataset in which each observation describes a particular U.S. county in a particular year. A county is a small geographic and administrative sub-unit of a state in the United States. Each state is made up of many counties. In any one state, some counties are near to borders with other states and some are not. The data pertain to 3108 counties, each observed in most years between 1977 and 2002. The variables employed in the regressions are defined in Table 2.

Table 1

OLS Regression Estimates

Dependent Variable: Traffic Fatalities Involving 18-year-old Drivers

(Standard Errors in Parentheses)

	Regression 1	Regression 2
Intercept Prohibited Prohibited*(Near to legal drinking) Ln(VMT per capita) Seatbelt law Ln (Real beer tax)	0.970 (0.020) -0.092 (0.028) -0.637 (0.050) -0.037 (0.022) -0.480 (0.070)	0.896 (0.022) -0.122 (0.023) 0.179 (0.035) -.519 (0.047) -0.010 (0.023) -0.463 (0.054)
Number of observations	70,997	70,997

Table 2

Variable Definitions and Descriptive Statistics

Variable	Definition
Traffic Fatalities Involving 18-year-old Drivers	Number of fatal traffic accidents in which a driver was 18 years old, in the county and year relevant to the observation (The average across counties and years in this sample is 0.69. The standard deviation is 1.72.)
Prohibited	Dummy variable equal to 1 if 18-year-olds were legally prohibited from drinking alcohol in this county and year. That is, this variable equals 1 for counties and years in which the minimum legal drinking age was 19 or older. It is equal to zero for counties and years in which the minimum legal drinking age was 18, meaning that 18-year-olds were legally allowed to drink alcohol.
Near to legal drinking	Dummy equal to 1 if the county is "near" to a place where 18-year-olds may drink legally, in the sense that the average person in the county lives 25 miles or less from the border of a state in which 18-year-olds may drink legally (0 otherwise)
Ln(VMT per capita)	Natural logarithm of the average per capita vehicle-miles traveled in the county and year (mean=1.112, std. dev.=9.033) [This variable is meant to pick up differences across counties and time periods in the intensity of driving and traffic.]
Seatbelt law	Dummy equal to 1 if a law requiring drivers to wear seatbelts is in force in that county and year (0 otherwise)
Ln(Real beer tax)	Natural logarithm of average beer tax in county and year, where the average beer tax is measured in real 2005 cents per gallon (mean=0.302, std. dev.=4.256)

1. (4 points) State as completely and clearly as possible (i.e. using relevant "units" and numbers) the meaning of the estimated regression coefficient on log(Real beer tax) in the first column of Table 1 (-0.480). One well-constructed sentence should suffice.

2. (6 points) Assume that Regression 1 in Table 1 satisfies all the assumptions required for the standard theory underlying confidence interval estimation (as discussed in class). Calculate as accurately as possible a 90 percent confidence interval for the estimated coefficient on Seatbelt Law in Regression 1 of Table 1.Please show:

a. the formula you use,

b. all the numbers you plug in, and

c. any information you use to identify the correct critical value.

3. (10 points) Consider the confidence interval you estimated for Question 2. Please answer the following questions completely and carefully, making explicit reference to any information that influences your conclusions.

a. Based on this confidence interval estimate,what conclusions (if any) can you draw regarding the size and social science importance of the effect of Seatbelt Law on Traffic Fatalities Involving 18-year-olds? Explain.

b. Do you consider this estimate precise? Why or why not?