186 E.T. Hill Economics of Education Review 20 2001 181–191
training and educational level. But because the data used in this study measure fewer women at lower educational
levels see Appendix A, the data report less remedial education.
Further evidence of employers’ needs to observe a labor market commitment is provided by the finding that
a strong early labor force attachment is associated with on-the-job training later in life but not with education or
other training. Women with more work experience had strong incentives as well as increased opportunities for
on-the-job training compared to those in the labor force fewer years. Although this analysis uses work experience
rather than tenure with the firm, the positive statistically significant association of work experience with on-the-
job training may also be due to the tendency of firms to train workers after they have been with the company for
some time noted by other studies Loewenstein Spletzer, 1997; Barron et al., 1993; Lillard Tan, 1992.
Apparently certain women tend toward training. Women who acquired education and training after the
usual schooling age but before 1967 more likely trained during the survey period.
7
Those who obtained education and other training prior to 1967 acquired more training
by all training methods from 1967 to 1984. Women who trained on-the-job prior to 1967 subsequently acquired
more of both on-the-job and other training. That training in an earlier time period raises the probability of training
later on is similar to the finding of Loewenstein and Spletzer 1994.
Table 3 shows that during the survey years, the methods of training appear to be complements rather
than substitutes: women who received training by each method were more likely to obtain it by the other two
methods. Some employees may signal employers that firms can benefit from training them. Alternatively or in
addition, certain workers may be more interested in acquiring education and training.
5. Training and wages
If education and training are human capital invest- ments made to enhance productivity and productivity is
reflected in wages, analyzing training and wages will show which training method enhances wages most.
Table 4 shows the determinants of the respondents’ wages at older ages. The dependent variable, the log of
the 1984 wage level, is presented with and without the 1984 or most recent occupation. It includes the 1967
educational level because education acquired after 1967 is a training variable. The Heckman 1979 selectivity
7
Measuring the effect of training reported in 1967 is difficult because the training could have occurred in 1967 or more than
25 years before that.
model controls for the fact that not all women worked and had a wage in 1984.
8
Younger women, those with more work experience, and women who lived outside the south experienced
higher wages than other women. Wages differed by occupation: women in professionaltechnical occupations
and in sales experienced higher wages compared to women in clerical occupations while several of the other
occupations showed significantly lower wage levels. All types of training obtained at early ages before 1967
were associated with higher 1984 wage levels.
Concerning education and training during the survey period, when occupations were included, only those who
received on-the-job training from 1977 to 1984 showed higher wages. However, when occupations were omitted,
wage level advantages at older ages occurred from both education and on-the-job training. The benefit from adult
education seems to occur by placing women into better- paying occupations.
But older women need to find the path to greater wage growth. Table 5 analyzes wage increases during the sur-
vey period using the change in the log of the hourly wage rate as the dependent variable. Using the variables from
Table 4 which could change as well as 1967 training variables, the first wage specification in Table 5 indicates
that working more years was associated with wage increases. This could be due to increased work experi-
ence or to informal on-the-job training. Of the three training methods, on-the-job training programs showed
8
Identifying selection corrections is a problem because often the same variables affect wages and the probability of working.
Variables for assets, the local unemployment rate, health prob- lems, and marital status were added to obtain the probability of
working even though these may affect wages also. Results of the probits used for Table 4 are as follows:
Independent variables Coef.
t value Age
20.05 7.13
Race white 0.06
0.76 Educational level 1967
0.01 1.03
Years employed 1984 0.04
12.89 Lived in south 1984
20.05 0.69
Unemployment rate 1984 20.05
3.34 Health problems 1984
20.80 11.66
Assets 1982 20.001
2.75 Never married
0.19 1.01
Marriage ever ended 1.00
1.50 Adult education pre’67
0.07 0.75
On job training pre’67 0.04
0.34 Other training pre’67
0.02 0.31
Education 1967–1984 0.14
1.35 On-the-job training 1977– 0.98
10.25 1984
Other training 1977–1984 0.31 3.68
Constant 1.82
6.12 Significance level:
0.01 level
187 E.T. Hill Economics of Education Review 20 2001 181–191
Table 5 Determinants of wage changes 1967–1984
Independent variables Change in ln wage rates 1967–1984
Coefficient
a
Age 1967 20.01
2.31 Race white
20.11 3.34
Educational level 1967 20.0002
0.05 Change lived in south 1967–1984
20.02 0.26
20.03 0.42
Years employed 1967–1984
b
0.05 5.60
0.05 5.60
Years employed squared 1967– 20.0001
2.63 20.0001
1.93 1984
Adult education pre ’67 20.05
1.27 20.04
1.08 On-the-job training pre ’67
0.02 0.43
0.03 0.55
Other training pre ’67 20.03
0.76 20.02
0.58 Education 1967–1984
20.03 0.79
20.04 0.84
On-the-job training 1977–1984 0.09
2.61 0.09
2.58 Other training 1977–1984
0.06 1.63
0.06 1.70
Constant 0.90
12.90 1.28
8.24 R
2
0.05 0.07
Number of cases 868
868
a
Absolute value of t statistics in parentheses, significance level: 0.01, 0.05, 0.10.
b
At least 6 months and 35 h.
the strongest association with wage growth. The acqui- sition of later education was not associated with
increased wages while off-the-job training was mar- ginally significant.
9
The second wage equation in Table 5 includes several additional variables which might affect wage changes.
Age and race might affect wage growth if age and race discrimination exist. Age, which might also include atti-
tude or other factors affecting productivity, was negative and significant.
Non-white women had more wage growth than white women. However controlling for occupation, they had
similar wage levels by 1984 Table 4 so that older non- white women were apparently ‘catching up’ during this
period. Educational level in 1967 was not significant, probably because better-educated women were paid
higher starting wages.
Additional analyses using wage observations available for most of the survey years between 1967 and 1984
including education and training reported in 1967 showed no pattern and few significant training variables.
The training identified as occurring between 1977 and 1984 displayed no clear pattern with regard to wage
changes for each wave of the survey during that time although there is some indication that on-the-job training
9
Splitting education into that acquired before 1977 all edu- cation levels and college the 1977–1984 measure provided
no advantage nor did using each wave of the education and training responses separately for wage changes between 1977
and 1984.
affected wages immediately while college enhanced wages with a delay.
10
Whether those who receive training are those who would benefit most is difficult to tell since the selection
process into training is unknown. Work experience may include the effect of informal on-the-job training which
was not surveyed directly.
6. Training and labor force participation