4.2  Upskilling and employment change

We now turn to the relationship between upskilling and changes in employment over the period. Upskilling may occur either within types of jobs (ie, as a result of a shift from lower to higher qualifications of workers in particular jobs), or between different types of jobs (ie, as a result of a shift in employment from lower to higher skilled jobs). Examining the pattern of qualification changes across different job types will help us to identify the factors behind qualification upgrading, and also the relative importance of different patterns of upskilling. For example, upskilling may be due to changes in production technology, biased towards higher skills, and reflected in increased demand for more qualified workers. Such an explanation is often referred to in the literature as “skill-biased technological change” although, as Card and DiNardo (2002) note, without any explicit consideration of technology measures this labelling is tautological. Alternatively, upskilling may be the result of changes in product market demand biased towards industries with higher skill levels, and/or work organisational changes biased towards occupations with higher skill levels.

Furthermore, many different patterns of upskilling could generate an equivalent increase in measured skill intensity. For instance, improvements may occur at the lower end of the qualifications spectrum (eg, a shift in workers with no qualifications to school qualifications), or at the higher end (eg, a shift in workers with post-school qualifications to university qualifications). As is evident from the tables already presented, the reality is a more complex configuration of changes reflecting a general increase in qualifications across the population over the period.

This section examines patterns of employment growth over the period, and the relative importance of industry, occupation, and qualification-related factors in shaping employment growth. It should be borne in mind that this analysis is descriptive, and cannot identify what the specific growth factors are; only whether they operated more strongly on workers in particular industries, occupations, or with particular types of qualification. Three analyses are presented: first, a regression analysis of employment growth, which focuses on the proportion of the variation in employment growth accounted for by industry, occupation, and qualification factors; second, an analysis of the covariation of employment growth between different qualification groups; and third, a graphical summary of the relationship between employment growth, initial skill intensity, and changes in skill intensity.

4.2.1  Regression analysis of job-cell employment changes

We first focus on the extent to which the change in job cell employment is associated with upskilling (qualification effects), changes in product markets (industry effects) and/or work organisation (occupation effects). Specifically, we examine regression models of the form:

(2)     

where is a measure of the change in employment of workers in industry-i, occupation-o, with qualification-q, between 1986 and 2001; is the “explained” component of employment growth, which may vary by industry, occupation, and/or qualification; and is the regression residual representing the unexplained component of growth. The measure of employment growth that we use is:

(3)     

where the time subscript “t-1” denotes the initial year (eg, 1986), and “t” denotes the final year (eg, 2001). This employment growth measure differs from the commonly used employment growth rate in that the change in employment is divided by average employment rather than by initial employment. This employment growth index has the properties that it is bounded between –2 and 2, is monotonically related to the usual measure of growth, and treats employment increases and decreases symmetrically.

Table 5 presents results for various specifications of equation (2), which differ in the specification of the intercepts. All regressions are weighted by the initial 1986 employment level (Emp_ioqt-1). For each regression, we report the coefficients on the qualification variables (when included), an indication of whether industry, occupation or their interaction effects have been included, and the regression R-squared and marginal R-squared associated with the various sets of controls.[18] The results in column (1) pertain to a regression that includes just indicator variables for each of the qualification groups. That each of the qualification coefficients are positive and statistically significant indicates there was employment growth for all groups relative to the (omitted) “no qualifications” group. The R-squared from this regression is 0.33, implying that over one-third of the variation in employment growth across cells defined by industry, occupation, and qualification is associated with qualification differences in growth.

The regressions reported in columns (2) and (3) include, respectively, just industry and occupation dummy variables. The R² from these regressions is 0.37 and 0.24 respectively. When we include both industry and occupation dummy variables additively (column (4)), the R² rises to 0.46. This implies industry and occupation marginal R²s of 0.26 and 0.09, which are substantially lower than the R²s reported in columns (2) and (3), and imply the industry and occupation effects on employment changes are quite strongly correlated. Column (5) includes the full set of industry and occupation interactions, and the R² rises to 0.68.

Columns (6) to (9) in Table 5 repeat the specifications in columns (2) to (5) with the addition of the set of qualification dummy variables. The qualification coefficients in these columns are remarkably stable and are very close to the coefficients in column (1), with the exception of the coefficients on “Higher degree” and “Bachelor’s degree” which fall by 10-15% from column (1), and the coefficients on “Higher school” which falls by about 5%. Furthermore, the qualification marginal R²s from these regressions range from 0.23 to 0.25. Compared to the (full) R² for qualifications of 0.33 in column (1), these results suggest that the sizeable contribution of qualification effects to employment changes across jobs is largely unrelated to industry and/or occupation factors. This implies that the relatively high growth in employment for workers with high qualifications is quite broad and not merely a reflection of their being in fast growing industries and occupations.

The stability of the marginal R² for qualifications is in contrast to those of industry and occupation, where the proportion accounted for by each of these factors reduces markedly if the other factor is controlled for. There is clearly a strong relationship between industry and occupation effects.

The patterns of stable coefficients on qualification effects and marginal R²s also holds over the three intercensal sub-periods as well (we report the results in Appendix Tables A4a to A4c). The primary difference across the three periods is that qualifications are much less important in understanding the changes between 1986 and 1991, associated with only about 13% of the variation in employment changes, than in the subsequent two periods where qualification effects account for about 40% of the variation in employment change in each period. Thus, while a large fraction of the variation in employment change over the 1986-91 period appears to be related to job-cell specific effects (industry-occupation interactions account for over three quarters of the variation in this period), qualifications play a much greater role in the subsequent two periods.

The key inference to be drawn from the regression analyses presented here is that, except possibly during the first intercensal period, factors operating at the industry and occupation level cannot account for the more rapid growth in employment experienced by more highly qualified workers. It would appear that upskilling reflects changes in factor markets rather than in product markets or in occupational mix alone. Furthermore, the same positive relationship between qualifications and employment growth is apparent both within and between job groups. That is, upskilling has occurred both as a result of a shift in employment to higher skilled jobs, and as a result of an increase in skill levels of workers in particular jobs.