Table 1: Description of Variables

(Note: This analysis is for men only)

Variable Name	Description
Lnwage	The log of wages. Lnwage=ln(wage)
Wage_re	Wage in $ per hour (not used in this analysis)
Asian, Black, Hispanic	1 if the respondent is in that race/ethnic group, 0 otherwise (Non-Hispanic white (“white” ) is the excluded category)
Years of Education	The respondent’s years of education
(group-name)*Edyrs	The interaction term between the race/ethnic group and education (in Stata, the result of including the term i.re*edyrs, where “re” is the categorical variable for race/ethnicity)

(Note: When discussing differences in log wages, you can call them “log points”. Using log wages is useful because we are estimating proportional differences in wages.

For example ln(10)-ln(5) = ln(2) – ln(1) because 10/5 = 2 = 2/1

ln(10) = 2.302

ln(5) = 1.609

ln(2)=.693

ln(1)=0

If you doubled everyone’s wage, the results for the gender gap using log wages would not change.)

Table 2: Summary Statistics, by Race/Ethnicity

(The mean of lnwage and edyrs for each race/ethnic group)

Ignore the column “weight”

. by re: sum lnwage wage_re edyrs [w=weight]

-------------------------------------------------------------------------------

-> re = white

(analytic weights assumed)

Variable | Obs Weight Mean Std. Dev. Min Max

-------------+-----------------------------------------------------------------

lnwage | 44133 87748824.8 2.847951 .593882 .9618849 5.318708

wage_re | 44133 87748824.8 20.58432 13.11469 2.616624 204.12

edyrs | 44133 87748824.8 13.80586 2.725675 1 21

-------------------------------------------------------------------------------

-> re = black

Variable | Obs Weight Mean Std. Dev. Min Max

-------------+-----------------------------------------------------------------

lnwage | 3383 9538143.66 2.587464 .5316631 1.021423 5.230694

wage_re | 3383 9538143.66 15.43376 9.603858 2.777143 186.9224

edyrs | 3383 9538143.66 13.18191 2.301107 1 21

-------------------------------------------------------------------------------

-> re = asian

Variable | Obs Weight Mean Std. Dev. Min Max

-------------+-----------------------------------------------------------------

lnwage | 1905 4701733.92 2.918734 .6587708 1.175573 5.125326

wage_re | 1905 4701733.92 22.91397 15.92213 3.24 168.2289

edyrs | 1905 4701733.92 15.11951 3.267622 1 21

-------------------------------------------------------------------------------

-> re = hispanic

Variable | Obs Weight Mean Std. Dev. Min Max

-------------+-----------------------------------------------------------------

lnwage | 4637 13299069 2.407758 .4745804 1.013054 4.537548

wage_re | 4637 13299069 12.60776 7.619075 2.754 93.46136

edyrs | 4637 13299069 10.21205 3.737197 1 21

Table 3: Regression Estimates of Race and Ethnic Inequality for Men, 2005 Current Population Survey

	(1)	(2)	(3)
	lnwage	Lnwage	lnwage
Black	-0.260	-0.202	-0.168
	(0.009)**	(0.008)**	(0.046)**
Asian	0.071	-0.053	-0.173
	(0.013)**	(0.011)**	(0.053)**
Hispanic	-0.440	-0.103	0.567
	(0.008)**	(0.008)**	(0.023)**
Years of education		0.094	0.106
		(0.001)**	(0.001)**
(Black)*Edyrs			-0.002
			(0.003)
(Asian)*Edyrs			0.007
			(0.003)*
(Hispanic)*Edyrs			-0.061
			(0.002)**
Constant	2.848	1.552	1.388
	(0.003)**	(0.011)**	(0.013)**
Observations	54058	54058	54058
R-squared	0.07	0.26	0.28

Standard errors in parentheses

* significant at 5%; ** significant at 1%

1. a. How does the coefficient on the variable “Asian” in Model 1 of Table 3 relate to the average values of lnwage by race and ethnicity listed in Table 2? [I.e., if you only had Table 2 you could tell me what the results of Model 1 of Table 3 would be (but not the other models)….why?]

b. What is the 80% confidence interval for the coefficient on Asian in Model 1 in Table 3? [You can use the standard normal distribution rather than the t-distribution because the number of cases is large. If the z-score is between two numbers, choose the closest one.]

2. Test the hypothesis that the coefficient on Asian in Model 1 equals 0.05,

3. Explain why the coefficient on Asian has a positive effect in Model 1 and a negative effect in Model 2. How could you argue that the coefficient on Asian in Model 2 is measuring the effect of discrimination in the labor market? What information from Table 2 is useful in understanding what happens to the coefficient on Asian going from Model 1 to Model 2 of Table 3?

4. a) In Model 3, what is the size of the predicted wage gap (in terms of log wages) between white and Hispanic men with 12 years of education? (I.e., what is the predicted difference in log wages for a white man with 12 years of education and a Hispanic man with 12 years of education?)

b) In Model 3, does the size of the wage gap (in terms of log wages) between white and Hispanic men get bigger or smaller as the number of years of education increases?