To test the hypothesis that characteristics alone explain the difference between the losses of men and of women, we calculate bootstrapped standard errors of the
overall male advantage due to behavior in each period. We use 50 bootstrapped samples from the overall data to calculate the mean overall male advantage, which
is our measure of women’s losses attributable to discrimination. For all quarters, the 95 percent confidence interval for the impact of behavior on proportionate earnings
losses does not include zero. This leads us to reject the hypothesis that the observed disparity in long-term losses is caused only by differences in observed personal,
employer, job, or injury characteristics.
V. The Impact of Nonemployment on Gender Differences in Losses
Our next step consists in looking for alternative labor-market features that may cause the observed gender disparity in long-term losses. The first hypothesis
we examine is that women and men differ in their employment patterns after injury. Our data can tell us whether individuals had positive covered earnings in each ob-
served quarter, allowing us to measure the impact on losses caused by different post- injury employment patterns.
To do this, we estimate the impact of injury on the probability of employment positive earnings in each quarter with logistic regression, using the White-
Huber adjustment White 1980 to account for within-person correlation. These estimates use the same independent variables as the loss estimates Equation 3.
Hardware limitations prevent us from using more than 9,100 observations in the logistic regressions, so we randomly sampled men and women to get sample sizes
under this limit. Selected coefficients from the logistic regressions are displayed in Table 6.
From these estimates, we derive the average impact of being injured on the post- injury employment of women and men. Figure 7 displays average changes in em-
ployment for injured men and women relative to the comparison group. In the three years after the post-injury quarter, women’s nonemployment rates averaged 4.1
points above those of the comparison group and men’s 2.7 points above. These results suggest that nonemployment accounts for a substantial proportion of post-
injury losses. Because the percentage change in earnings is the sum of the percen- tage changes in wages and employment rates, the injury-related drop in wages
after accounting for nonemployment is about 5.1 9.2–4.1 points for women and 3.8 6.5–2.7 points for men, leaving a gender gap of 1.3 points. Because our data
consist of quarterly earnings and not hourly wages or hours worked, we cannot tell how much of this remainder in gender disparity in post-injury losses reflects a
reduction in time worked reduced weekly hours worked, or spells of nonemploy- ment that do not span full calendar quarters or a decline in wages. We have deter-
mined an upper bound point estimate of wage discrimination: 1.3 percentage points. However, this estimate could be reduced if we knew how much post-injury hours
changed.
Table 6 Logistic Estimates of the Probability of Being Employed, Men and Women with
Lost-Time Injuries
Men Women
n ⫽ 9,050 n ⫽ 9,013
Pre-injury trend ⫺
0.03 0.06
0.03 0.04
Injury impact Change in quarter after injury, Quarter 2
⫺ 0.94
⫺ 1.78
0.37 0.39
Additional change in Quarter 3 0.97
0.94 0.30
0.33 Additional change in Quarter 4
⫺ 0.30
0.02 0.22
0.24 Trend after Quarter 4
⫺ 0.07
0.00 0.03
0.03 Individual characteristics
Age 15–24 Change in quarter after injury, Quarter 2
0.27 0.11
0.13 0.13
Additional change in Quarter 3 ⫺
0.20 0.13
0.12 0.10
Additional change in Quarter 4 0.28
0.10 0.11
0.12 Trend after Quarter 4
0.02 ⫺
0.01 0.01
0.01 Age 25–54
base case base case
Age 55⫹ Change in quarter after injury, Quarter 2
⫺ 0.37
⫺ 0.41
0.19 0.18
Additional change in Quarter 3 ⫺
0.02 0.20
0.16 0.15
Additional change in Quarter 4 0.13
⫺ 0.39
0.10 0.12
Trend after Quarter 4 ⫺
0.04 ⫺
0.03 0.01
0.01 Tenure ⬉6 months
Change in quarter after injury Quarter 2 0.93
1.03 0.27
0.29 Additional Change in Quarter 3
⫺ 0.43
⫺ 0.29
0.20 0.20
Additional Change in Quarter 4 0.09
0.18 0.16
0.15 Trend after Quarter 4
0.04 ⫺
0.02 0.03
0.03
Table 6 continued
Men Women
n ⫽ 9,050 n ⫽ 9,013
Tenure 6.1 months to 1 year Change in quarter after injury Quarter 2
⫺ 0.06
⫺ 0.30
0.28 0.30
Additional change in Quarter 3 ⫺
0.39 ⫺
0.38 0.21
0.21 Additional change in Quarter 4
0.06 0.16
0.17 0.15
Trend after Quarter 4 0.12
0.07 0.03
0.03 Tenure 1.1 to 5 years
Change in quarter after injury, Quarter 2 0.06
⫺ 0.39
0.30 0.33
Additional change in Quarter 3 ⫺
0.02 ⫺
0.09 0.16
0.22 Additional change in Quarter 4
0.02 ⫺
0.07 0.17
0.16 Trend after Quarter 4
0.10 0.01
0.03 0.03
Tenure 5.1 to 10 years base case
base case Tenure more than 10 years
Change in quarter after injury 2 0.34
0.34 0.31
0.38 Additional change in Quarter 3
⫺ 0.27
⫺ 0.36
0.23 0.24
Additional change in Quarter 4 0.04
0.25 0.18
0.17 Trend after Quarter 4
0.07 0.04
0.03 0.03
Wald X
2
201 11,091.84
10,824.81
Significant, p ⬍ .05 Note: These results are derived from randomly selected samples of the original data set, with different
sampling rates for men and for women. The omitted categories used imply that the estimates of injury impacts displayed in this table are for the base case of a 25–54 year old unskilled blue-collar worker in
a durable manufacturing industry with five years of tenure in a firm with over 1,000 employees. This worker had constant earnings in the pre-injury period and the injury did not involve low back pain or
carpal tunnel syndrome. Each person is observed for 24 time periods. This results in 217,200 observations for men and 216,312 observations for women. Standard errors are adjusted for within-person correlations.
Observations are weighted to account for the relative undersampling of men relative to women. Standard errors are in the parentheses.
Figure 7 Estimated Injury-Related Nonemployment Workers Injured in Wisconsin, 1989–
1990
VI. Other Possible Explanations for Observed Differences
