About WTS Prof. Clair Brown Faculty, Students and Affiliates Research Areas Online Research Reports Working Papers

THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN RESOURCES PROJECT:

Second Interim Report
CSM-32
Clair Brown, Editor

9. The Evolution of Skill Demand and the Nature of the Employment Relationship in a Technology-Intensive Firm
Vincent M. Valvano

9.4 Earnings Determination at NewTech

Promotions are a primary incentive mechanism for employees in large firms and have a major impact on earnings over the length of a career. Promotions are typically defined by the firm in relation to a hierarchy of jobs or levels. A standard description of the hierarchical structure of a large firm is based on job titles. The bottom level of the hierarchy consists of entry level jobs in which the majority of entrants are new hires from outside the firm. The next level in the hierarchy consists of jobs which are primarily fed by entry level jobs, and subsequent levels are mapped in similar fashion. A promotion is defined as a move from a job title situated at a lower level of the hierarchy to one at a higher level.

The hierarchy at NewTech differs in definition from the classic hierarchical structure based on job titles. For managers and professionals, a promotion is defined as a move between job grades rather than a move between job titles. As is evident in Table 9-3, relatively few employees change job titles more than once at NewTech. (However, because Table 9-3 makes no adjustment for length of employment and includes employees with relatively short employment spells, the importance of job title changes is likely to be understated.) Moves between job grades are more common. For example, between 1993 and 1994, 8.7% of continuing managerial and professional employees at NewTech changed job title, while 22% changed job grade.

Table 9-3. Percentage Distribution of Unique Job Spells for Managers and Professionals, 1983-1994.

Panel A. Job Spells-Entering Managers

Number of Job Spells Managerial

Jobs (%)

Professional

Jobs (%)

Total

(%)

0 . 82.4 .
1 54.8 14.6 45.2
2 38.2 2.4 39.6
3 5.0 0.4 11.0
4 1.7 0.1 3.1
5 0.2 . 0.8
6 0.1 . 0.2
7 0.1 . .
8 . . 0.1
N     1,652

Panel B. Job Spells-Entering Professionals

Number of Job Spells Professional

Jobs (%)

Managerial

Jobs (%)

Total

(%)

0 . 91.1 .
1 52.5 7.3 48.1
2 37.5 1.3 38.0
3 8.0 0.2 10.1
4 1.6 0.1 2.7
5 0.3 . 0.9
6 0.0 . 0.2
7 . . 0.0
N     8,034

Note: Derived from a pooled sample of new entrants to the firm during 1983-1994. Includes left-censored employees (those who transferred in from another division in the firm or changed from non-exempt to exempt status) and right-censored employees. (.) denotes empty cell. An employee is initially designated Manager or Professional based on his/her entry occupational classification. A job spell is a continuous period of time in one job title.

Although the hierarchy at NewTech is defined by job grades, it does not necessarily differ in qualitative respects from the standard hierarchy defined above. A hierarchy based on job titles may be largely identical to one based on job grades, if subsets of job titles in the standard hierarchy are closely related in function. For example, at another technology firm that is similar to NewTech in many respects, the concept of job level is explicit in job titles. Electrical engineers at this firm move through five closely related levels. Differences between levels are based on differences in job complexity and responsibility. Engineers ostensibly move through the level structure as they acquire more skill and experience in their job. An outside researcher examining this firm's personnel records will observe engineers moving between job titles, beginning with Electrical Engineer III, then Electrical Engineer II, and so on, with a top level of Staff Electrical Engineer. At NewTech, we observe engineers who remain in the (hypothetical) job title Electrical Engineer, but who move through a series of job grades, A, B, C, etc. There is no reason to believe a priori that the two systems differ in respect to the ranking and promotion of employees.

Professional and managerial employees at NewTech move through a common set of 11 ordered job grades. In addition, six higher grades are populated by top managers. Figure 9-2 presents a distribution of employees by grade for 1988. Entry by professionals is concentrated in the lower grades but is not restricted to those grades, as is evident in Table 9-4. Minimum tenure values of zero in six of the bottom seven grades suggest that new hires are entering in each of those grades. Average tenure increases with grade through the bottom half of the grade structure before leveling out. But there are also high-tenure employees in the lower grades, which suggests that some employees experience few or no promotions during their tenure at NewTech.

Table 9-4. Tenure (in Years) by Job Grade for U.S. Managers and Professionals, 1988

Job Grade N Mean

Tenure

S.D. Min. Max.
A 294 2 2 0 30
B 64 3 3 0 12
C 625 5 5 0 33
D 130 7 8 1 33
E 759 12 9 0 37
F 493 14 8 0 36
G 380 17 9 0 46
H 275 17 8 1 37
I 156 20 8 2 38
J 60 21 7 10 37
K 21 21 6 11 32
L 14 20 7 9 36
M 8 21 5 12 29
N 3 25 8 16 30
O 2 16 0 16 17
P 1 20 . . .
S 1 27 . . .
All Grades 3,286 11.4 9.3 0 46

Mean salary increases in job grade as does wage dispersion (Figure 9-3). There is also wage overlap across grades, which indicates that wages are not determined solely by job grade. In an exercise not reported here, I have examined mean salary by grade separately for managers and professionals. In general, the means are close and professionals often have a slight earnings advantage in many of the grades in which they are represented. However, the professional track does not extend to the upper echelons of the grade hierarchy. The proportion of professionals in grades I, J, and K falls rapidly, and there are no professionals in the highest grades, L and above, which account for less than 1% of total managerial and professional employees.

Of central interest is the relationship between the grade hierarchy and earnings variation within the firm. Several interpretations of this relationship are conceivable. We have already seen that earnings are not perfectly defined by grade (Figure 9-3), and so can discount a rigidly administered pay policy. However it is still possible that administrative rules which, for example, limit within-grade salary dispersion, are important for earnings variation.

A second interpretation of the relation between job grades and earnings variation would maintain that grades are a "superficial" explanation of pay and that other variables actually drive earnings variation. The candidate variables may be standard human capital variables, such as education and tenure, or omitted variables, such as differences in individual ability or productivity. If omitted human capital or individual effects are important for earnings variation, we would expect to find observationally equivalent individuals advancing through the grade hierarchy at different rates. The wide range of tenure observed across employees within the same grade (Table 9-4) suggests that this is occurring to some extent at NewTech.

Finally, of course, it is possible that administered rules, human capital, and individual effects each contribute to earnings variation in the firm. The results of an initial analysis of earnings variation among managers and professionals are presented in Tables 9-5 and 9-6. Three models are estimated (each controlling for year effects): a standard human capital earnings function, an equation that includes only organizational variables (job grades), and a combined model that includes human capital and job grade variables (Table 9-6). The regressions use pooled observations from the sub-sample of U.S. managers and professionals during 1976-1994. The combined model was also estimated using cross-section data from 1988. The dependent variable in each equation is the log of real annual pay, including salary and bonus (in constant 1994 dollars).

Table 9-5. Components of the Variance of Pay for Professionals and Managers, 1976-1994

Source of Variation Percent of Sum of Squares Degrees of Freedom
1. Job Grade 40.2 14
2. Human Capital 0.2 7
3. Year 1.3 18
4. Joint Job Grade, H.C., Year 49.6 C
5. Total Between Cells 91.3 39
6. Interactions / Residual 8.7 64,062
Total 100.0 64,101
Total Sum of Squares 5,855

Notes: Each of the main effects in lines 1-3 is conditional on the other two. Line 5 is the R-squared of the combined model in Table 9-6. Line 4 is the difference between line 5 and the sum of lines 1-3. Line 6 is 100 minus line 5. The human capital factor includes age, firm tenure, firm tenure squared, and a dichotomous variable for education level. Pay is the natural log of base salary plus bonus.

Table 9-6. Effects of Human Capital and Job Grade on Annual Pay Levels

    1976-1994 Pooled Regressions  
Independent

Variables

Sample Means,

Characteristics

a. Human Capital b. Job Grades c. Combined 1988 Cross Section
Year Dummies   yes yes yes  
Intercept   10.40

(.0056)

10.54

(.0018)

10.51

(.0025)

10.48

(.0085)

High school diploma or equivalent 7.1% -.164

(.0036)

  -.0002*

(.0015)

-.0045*

(.0065)

Associate degree 1.1% -.164

(.0081)

  -.0048*

(.0034)

-.0061*

(.013)

Master's degree 29.7% .106

(.0019)

  .013

(.0008)

.0072*

(.0034)

Doctorate 5.7% .296

(.0037)

  .036

(.0017)

.022

(.0076)

Age (years) 37.1 .006

(.0002)

  .0009

(.00007)

.001

(.0003)

Tenure (months) 119 .003

(.00003)

  -.00001*

(.00001)

.0001*

(.00007)

Tenure2   -.000006

(.00000008)

  .0000002

(.00000004)

-.0000003*

(.0000002)

Grade B 2.1%   .057

(.0027)

.055

(.0027)

.040

(.0114)

Grade C 19.8%   .149

(.0014)

.141

(.0014)

.153

(.0061)

Grade D 4.1%   .194

(.0021)

.183

(.0021)

.208

(.0091)

Grade E 24.1%   .34

(.0013)

.323

(.0016)

.341

(.0068)

Grade F 12.7%   .49

(.0015)

.464

(.0018)

.492

(.0077)

Grade G 11.9%   .619

(.0015)

.595

(.0019)

.641

(.0083)

Grade H 7%   .754

(.0017)

.724

(.0022)

.772

(.0091)

Grade I 4.4%   .901

(.0020)

.869

(.0024)

.935

(.0103)

Grade J 1.5%   1.03

(.0031)

.995

(.0034)

1.07

(.0133)

Grade K 0.8%   1.13

(.0041)

1.10

(.0043)

1.21

(.0201)

Grade L 0.3%   1.30

(.0067)

1.26

(.0068)

1.36

(.0244)

Grade M 0.2%   1.39

(.0082)

1.36

(.0083)

1.49

(.0315)

Grade N 0.1%   1.53

(.0126)

1.49

(.0125)

1.59

(.0503)

Grade O and above 0.3%   1.65

(.0065)

1.61

(.0067)

1.98

(.0440)

R2   .511 .911 .913 .927
R2 without year dummies   .501 .898 .900  
N   64,102 64,102 64,102 3,513
Dependent Mean   10.90 10.90 10.90 10.94

Note: All regressions are OLS. The dependent variable is the annual observation of the log of real annual salary plus bonus in constant 1994 dollars. Standard errors are in parentheses. Coefficients that are not significant at the one percent level are indicated by an asterisk.

An analysis of variance of earnings (Table 9-5) provides an estimate of the explanatory power of the human capital and job grade variables in the combined (main effects) model in Table 9-6. Following the method of Groshen (1991) and Leonard (1989), OLS regressions are used to measure the separate and joint contributions of each main effect (human capital, job grade, and year) to earnings variance. Controlling for the other main effects (human capital and year), job grades can unambiguously account for 40% of earnings variance. This is a large effect and indicates that job grades importantly structure pay in the firm. In contrast, human capital variables (tenure, tenure squared, age and education) together can unambiguously account for less than 1% of variation, controlling for the other main effects. By themselves, typical measures of human capital (age, tenure and education) contribute little to explaining pay differences. This result is not entirely surprising, given the highly-skilled and highly-educated character of this employee sample. Year effects can explain about 1% of pay variance, suggesting that earnings do not importantly adjust to external economic shocks or variations in firm performance. A significant amount of earnings variance (50%) cannot be unambiguously attributed to any one main effect because of collinearity between the main effects. For example, it has already been noted that tenure increases in job grade (Table 9-4). The scope for interactions between human capital, job grade and year is relatively small, accounting for 9% of pay variance. While there is some variation within job grade-human capital-year class, its importance is overshadowed by the variation across job grades. The size of the interaction effect also represents an upper bound on individual variation after conditioning by the main effects. Individual variation within job grade is perhaps not negligible but it does not go far towards explaining overall earnings variation in the firm. This does not mean, however, that individual performance is unimportant in determining pay. Rather, it implies that to increase pay significantly, employees must change job grades, i.e. be promoted. The ANOVA results thus point to an investigation of the determinants of promotion and a consideration of other variables that may better measure individual performance or proxy for unobserved human capital.

The longitudinal aspect of the NewTech data set allows a number of additional issues to be addressed. One important issue is the extent to which pay adjusts to product or labor market shocks. In Figure 9-4, real mean salary is plotted by grade for the period 1976-1994. There is some evidence of adjustment, although the overall impression is of stability of real earnings over time, especially for the lower grades. Higher grades display more variability in real earnings. This stability is confirmed by a comparison of earnings equations estimated with and without year controls (Table 9-6). The R2 for each equation is presented for each of the three models. Evidently, controlling for year provides little additional explanatory power in each case.

Cohort analysis is also a useful way to gain insight into wage determination inside the firm. In Figure 9-5, the real earnings paths for the 1976-1991 cohorts are plotted. A cohort is comprised of all managers and professionals who entered the NewTech business unit during a given year. Some variation in mean starting salary is evident across cohorts, as the classic model of the internal labor market predicts. But the shape of earnings profiles across cohorts also varies significantly over time. This is contrary to the prediction of the classic ILM model, which holds that insiders are insulated from external market pressures on wages. Two effects are evident. First, beginning in 1986, the slope of many of the cohort salary paths became significantly more flat. This effect is consistent with a pay policy change by the firm. For example, perhaps the firm shifted some compensation from base salary to a form of variable compensation such as profit sharing, which is not accounted for here. A second effect is that later cohort wage paths don't achieve the steepness of earlier cohort paths. As consequence, simple returns to tenure vary across cohorts. For example, after ten years, the mean salary (in real terms) of the 1976 cohort had increased by 70%, while the 1980 cohort experienced a 53% gain after the same amount of tenure, and the 1984 cohort experienced a 33% gain. Although the levels of return to tenure may be understated if other forms of compensation besides salary are important, there is no reason to expect the cohort effect to disappear in this case. These cohort effects are currently being analyzed in more detail.


9.5 Career Length and Turnover Patterns

Analysis of turnover in the firm can provide insight on the nature of human capital in the firm and help discern among theories of wage determination. For example, if human capital is predominantly firm-specific, we would expect turnover to decline as employees' investment in firm-specific capital increases. Firms may also use steep earnings profiles to reduce turnover, although such profiles may also reflect shared investments by the firm and its employees in specific training. Existing empirical research on turnover generally finds a strong monotonic decline in the separation rate as tenure in the job increases (see Anderson and Meyer, 1994, who measure separation rates at quarterly tenure intervals). Large firms tend to have lower permanent separation rates than small firms and high-wage firms have lower separation rates than low-wage firms.

An initial review of the data suggests that employment separation patterns at NewTech differ from what might be expected, based on our knowledge of aggregate turnover trends. Figure 9-6 presents yearly hazard rates for employment separation rates for managerial and professional employees in five cohorts. The hazard rate for a given cohort in a given year represents the proportion of employees in the cohort who will leave the firm in the coming year having "survived" up to that year. For example, the hazard of .25 in year 3 for 1976 cohort indicates that of the cohort members who have accumulated three years of tenure, one-quarter will leave between year 3 and year 4. Two notable aspects of turnover at NewTech are evident in Figure 9-6. Separation rates do not monotonically decline as tenure increases, unlike results that have been generally reported. In fact, these rates don't peak until year three of tenure. In addition, separation rates remain high (above 10%) for most cohorts up to the seventh or eighth year of tenure. Finally, differences in separation rates across cohorts are also evident. These patterns suggest that distinctive firm or industry characteristics have an important effect on the nature of the employment relationship and deserve to be explored in additional detail.


9.6 Summary

1. The occupational distribution of NewTech employment, both worldwide and in the United States, shifted toward higher-skill professionals between 1985 and 1994. The share of professional employees in the U.S. grew by 70% during the period, while manager and technician shares both declined and the operative share of employment remained stable.

2. NewTech employment in the United States is relatively more skill intensive than worldwide employment. The employment shares of professionals and technicians are larger in the U.S. than overall at NewTech, while the share of operatives based in the U.S. is lower.

3. The movement of employees between professional and managerial job titles is sizeable and the direction of the flow is two-way. For example, 17% of employees who entered NewTech in a managerial job during 1983-1994 experienced at least one spell in a professional job. Professionals at NewTech are the most important source of entry into managerial titles, accounting for over half of entry into such jobs.

4. Professionals and managers move through a common set of ordered job grades, which importantly structure pay at NewTech. In addition, six higher grades are populated by top managers. By themselves, typical measures of human capital (tenure and education) contribute little to explaining pay differences but omitted measures of human capital or individual performance are still likely to be important determinants of pay.

5. Real earnings over time at NewTech are relatively stable. There is little evidence of adjustment of real mean earnings by grade to labor or product market shocks. Some variation in mean starting salary is evident across cohorts, and the shape of earnings profiles across cohorts also varies significantly over time.

6. Employment separation rates for managers and professionals continue to increase until after the third year of tenure and remain above 10% for most cohorts until the seventh or eighth year of tenure.

End of Chapter 9

Go to Chapter 10
Go to References for this Chapter
Go to Table of Contents for this Chapter
Go to Table of Contents for the CSM-HR Interim Report



 IRLE HOME
 LIBRARY
 PROGRAMS
 
© 2005 Institute for Research on Labor and Employment. 
2521 Channing Way # 5555 
Berkeley, CA 94720-5555