3.2  A summary measure of skill-intensity

As is evident from Appendix Table A2, it is possible to generate a large number of tabular breakdowns of qualification composition. While this provides an unrestricted account of changes in the distribution of qualifications over time, it is difficult to construct a clear and concise understanding of overall patterns. To this end, we aggregate the qualification distribution into a single index of skill intensity. There are many different ways of calculating such an index – see Laroche and Mérette (2000) for a discussion of different approaches.

In this paper, we use an economics-based approach to measure the skill intensity of a subgroup of the population. This approach entails, first, calculating the average income of each qualification group across the whole population in 1986 and treating this as the “price” of that qualification and, second, weighting each qualification’s price by its subgroup share.[13] Specifically, we calculate the skill-intensity for group-i in year-t as

(1)    

where y_q86 is the average annual income for all people with (highest) qualification-q in 1986, and s_qit is the qualification-q share of the group-i population in year-t – ie, s_qit = N_qit/N_it, where N_qit is the number of people in group-i in year-t with qualification-q, and N_it is the total number of people in group-i in year-t.

This index of skill intensity holds constant the “prices” of different qualification (ie, skill) levels across population subgroups and over time, and normalises the size of the population. Thus, changes in group-level skill intensity are due entirely to changes in the qualifications distribution within the group, and do not reflect either changing employment levels or incomes. The use of average annual income as the “price” of skill tends to confound the unit price (ie, hourly wage) effect with employment intensity (ie, hours of work), which may vary systematically across qualification groups, over time, and/or across population subgroups. A final point to note about this analysis (and also the human capital analysis later in the paper) is that the construction of the skill-intensity index uses qualification-group average incomes based on only those who report incomes but the full qualification-cell population shares (ie, including those who do not report incomes). Implicitly, this assumes that the income distribution (or, at least, the relevant moments) within qualification groups is the same for the non-specified income group as for the group with specified incomes.[14]

The aggregate skill intensity index for 1986 (SI_·86) is simply the population average income in 1986. In general, however, the measure differs from actual income for two reasons. First, qualification average incomes are measured in 1986, so that any changes in qualification-specific incomes over time will not factor into the skill intensity index. Second, qualification average incomes are measured at their national level, so that the average income of any group that pays a wage premium will exceed the skill intensity measure for that group. We return to this issue shortly.

We can examine the extent of changes in relative skill-intensity of the population as a whole, as well as the skill intensity of various sub-groups of the population, over time using this index. Tables 4a, 4b, and 4c summarise the results of such an analysis for each of the employed workers, labour force, and whole populations aged 20-59. In each table, the first two columns show the 1986 and 2001 measures of skill intensity, the third column presents the change in skill intensity for each group between 1986 and 2001, and columns four to six present the change in skill intensity over each of the three intercensal periods of our study: 1986-91, 1991-96 and 1996-2001. The skill intensity in the first row pertains to the full population, and is expressed relative to the 1986 measure. The skill intensity in subsequent rows pertain to different population subgroups, and each is expressed relative to the overall skill intensity in that year.

Consider first the skill intensity for subgroups of the employed population, reported in Table 4a. The first row shows that the overall skill intensity of workers increased by 4% between 1986 and 2001: 2.6% between 1986 and 1991, 0.3% between 1991 and 1996, and 0.6% between 1996 and 2001. At least some of the increase in the skill intensity of workers between 1986 and 1991 may be due to relatively high unemployment among lower-skilled workers in 1991, which had the effect of raising the average skill level of the employed workforce. For example, Table 1 shows that unemployment is more concentrated in low-skills, and also was higher in 1991 than other years.

The second and third rows of Table 4a pertain to skill intensity levels of male and female workers. In 1986 male skill intensity was 1% higher than average, while female skill intensity was 2% lower. However, growth in female skill intensity between 1986 and 2001 was 5.7%, compared to only 2.1% for males. This relatively faster growth resulted in no apparent gender difference in the skill intensity in 2001.

The next set of rows describes the skill intensity of different age groups of employed workers. This shows that younger workers (aged under 40) are relatively more skilled than older workers, with the peak skill-group aged 25-29. In addition, the skill intensity of younger workers also tended to increase more than older workers, from 5.7% for 20-24 year olds to 2.3% for 50-59 year olds, reflecting the rapidly increasing levels of education being acquired during this period.

The third dimension along which we describe skill intensity of workers, is their region of residence. In 1986, the regional differences in skill ranged from a high of 4% above the average in Wellington to a low of 3% below the average in Gisborne, Southland and areas outside regional councils.[15] Although all regions experienced an increase in skill levels between 1986 and 2001, the changes in skill intensity over the period tended to favour the urban and more skilled regions, thus increasing regional disparities. By 2001 the skill levels ranged from 5% above average in Wellington to 7% below average in the “Area Outside”, and 5% below average in the West Coast and Southland.

We next consider the skill levels of different industries and occupations. The skill levels of industries in 1986 range from 18% above average in Education and 10% above average in Business Services to 8% below average in Clothing and 7% below average in Agricultural Services. There was substantial variation in the changes in the skill intensity across industries between 1986 and 2001, ranging from an increase of 7.1% in the Electricity, Gas and Water, and 6.2% in Finance and Insurance, industries to a decrease of –1.9% in Construction Services and –0.2% in the Primary Goods industries.

Perhaps not surprisingly, the most marked differences in skill intensity occur between occupations. For example, Science Professionals had a skill-intensity 26% above the average in 1986, while the skill levels of Labourers and Drivers was 10% below the average. Also, and in contrast to changes in skill along other dimensions, the skill intensity of several occupations declined between 1986 and 2001. The changes in skill levels ranged from increases of 5.2% for Managers and 4.5% for Sales workers to decreases of 3.9% for Science Professionals and 3.0% for Administrators, Building Trades and Services workers.

The skill changes for particular subgroups across the three intercensal periods tend to reflect the average skill changes of employed workers in the first row. That is, most of the growth in measured skills tended to occur between 1986 and 1991, during which time average skill increased 2.6%, compared to 0.3% and 0.6% in the subsequent intercensal periods (between 1991 and 1996, and between 1996 and 2001 respectively). In fact, the measured skill intensity of many subgroups, especially defined by industry and occupation, declined in the latter two intercensal periods. For example, the skill intensity of Science Professionals declined by 1%, 2% and 1% respectively in the three subperiods, and the skill intensity of administrators dropped 6% between 1996 and 2001.

As mentioned earlier, some of the fluctuations in intercensal skill changes may be attributable to cyclical factors, which affect the skill-based composition of employment. In order to control for such effects, we repeat the analysis of skill intensity using the broader target populations pertaining to, first, the labour force and, second, the full population aged 20-59. The results are summarised in Tables 4b and 4c. The results are broadly in line with those for employed workers in Table 4a and confirm that, in 1986, employed workers were more skilled than unemployed workers who, in turn, were more skilled than those out of the labour force.

3.3  The relationship between income and skill intensity of jobs

How good a measure of economic skill is the qualifications-based skill intensity index constructed above? To consider this issue we examine the relationship between the average income and the measure of skill intensity, measured across industry and occupation specific cells (described here and in the next section as “jobs”). For this purpose we use the population of employed workers. If the skill intensity index accurately measures the level of economic skill, as measured by average annual income, we would expect to see that average incomes closely track skill intensity.

Figure 1 graphs the relationship between the ratio of average income to skill intensity against the index of skill intensity for each industry-occupation cell in 1986, together with the least-squares fitted regression line. The size of the plotting symbols indicates the relative size of the industry-occupation cells. Figure 1 shows there is a wide dispersion in the relationship between average income and skill intensity (regression R²=0.11), although the fitted line is positively sloped. For example, the slope coefficient is 0.003, which implies that a $1,000 increase in skill-intensity is associated with a 3 percentage point increase in the income to skill ratio. Starting from a typical job with income equal to skill intensity (ie, the point in Figure 1 where the predicted ratio is equal to 100: SI=$29,429), a $1,000 increase in skill intensity (to $30,429) would be associated with an increase in income above this level (to $31,291). This suggests that highly-qualified individuals are over-represented in industries and occupations where incomes are higher for everyone. One interpretation for this finding is that there are spillover or peer-effects operating in jobs dominated by high skills that tend to raise incomes of all workers above the average given their qualifications (see Dillingham 2002). An alternative interpretation for the differences between the average income of a group and its skill intensity is that they reflect the effects of non-qualification based components of skill (eg, experience), which are not priced in the skill intensity measure used here.