RDP 9215: The Evolution of Employment and Unemployment in Australia 4. Employment, Productivity and Wages by Sector

Thus far we have described, in considerable detail, the differences in the patterns of employment and unemployment for men and women, and for full-time and part-time workers. In this section, we show that the principal cause of these differences has been the relative decline during the 1970s of those sectors where male employment has been largest, principally manufacturing and construction and the quite substantial increases in employment which have occurred in the service sectors; however, new employees in these industries (largely women working part-time) have tended to come to these jobs from outside the labour force. Thus, increases in employment have resulted in only modest corresponding reductions in unemployment.[20]

Figure 18 shows employment levels by sector, over, the period 1966 to 1991, for men and women, respectively. The major points to note are:

Figure 18: Employment by Sector: 1966–1991
Figure 18: Employment by Sector: 1966–1991

Source: The Labour Force : Australia, ABS Cat. No. 6203.0.

  • the decline in manufacturing employment which peaked, in absolute terms, in 1973 for men and 1974 for women. As a result, manufacturing's share of total employment has fallen over this period from about 27 per cent to 18 per cent for men and from 21 per cent to 10 per cent for women;
  • the fall in construction employment for men from 1975 to 1983. Not until 1989 did male employment in this sector recover its 1975 level. Construction employment is also highly cyclical, recording large falls in 1982–83, strong growth from 1984 to 1989, and large falls again recently;
  • the rapid increase in employment in finance[21] and community services. The finance sector employed just over five per cent of men in 1966 but nearly 10 per cent of men in 1991. For women, the proportions were eight per cent and 13 per cent. The proportion of men employed in community services increased over this period from just under six per cent to nearly 11 per cent, while for women the increase was from nearly 20 per cent to nearly 30 per cent; and
  • the declining importance of agricultural employment for men. Nearly 11 per cent of men were employed in agriculture in 1966; this figure fell to seven per cent in 1991.

Figure 18 gives us some insight into the cause of the rise in the equilibrium unemployment rate that began around 1974, and continued until the end of the 1970s. As we have noted, male employment in manufacturing and construction recorded negative growth during this time. A fall in employment in a particular sector (or sectors) should not in itself lead to growing unemployment, provided that other sectors expand at the same time, and there is no skills mismatch which prevents surplus labour in one sector from being employed elsewhere. Indeed, such a change would be expected – and desirable – as a part of any economy's normal path of structural adjustment. Employment in agriculture, for example, fell by over 10 per cent between 1966 and 1974 with no discernible effect on the unemployment rate. However, after 1974 no other sectors expanded sufficiently quickly to absorb the excess male labour; the consequent slow down in employment growth led to an increase in unemployment duration and hence in the equilibrium rate of male unemployment.

Female employment in manufacturing also started to fall in the mid 1970s. Unlike the male labour market, employment in one important sector – community services – grew strongly, but still insufficiently to absorb the excess female labour, which was exacerbated by fast growth in the female labour supply during this time.

Thus, it would appear that the rise in the equilibrium rate of unemployment that was triggered in the mid 1970s was due to the economy's inability to adapt to the shocks of the time. These have been well-documented elsewhere (see e.g. Gregory and Duncan (1979), and comments on their paper). In particular, the very rapid increase in real wages, relative to both labour productivity and the cost of capital, appears to have been the crucial influence in initially increasing the rate of unemployment[22]. More fundamentally, however, it was the inability of real wages to adjust downwards in response to adverse shocks (such as the slowdown in productivity growth which began at this time in Australia, and the rest of the world) which raised the equilibrium unemployment rate.[23]

Apart from the real wage effects, it might be argued that the decline in manufacturing employment was in part caused by the 23 percent fall in effective assistance to that sector which occurred in 1973/74.[24] In Section 4.1 we argue that there is in fact no simple causation running from falls in manufacturing protection to falls in manufacturing employment. Moreover, cuts in protection to one sector should have no lasting influence on aggregate unemployment, provided that the economy makes the necessary structural adjustments. Thus, if sector-specific shocks do have persistent effects on unemployment, the problem lies not in the occurrence of the shocks per se, but in the impediments to the flows of resources from declining to incipiently expanding sectors.

Figures 19 and 20 decompose the sectoral employment figures into full-time and part-time, over the period 1978 to 1991. Note:

Figure 19: Male Employment by Sector: 1966–1991
Figure 19: Male Employment by Sector: 1966–1991

Source: The Labour Force : Australia, ABS Cat. No. 6203.0.

Figure 20: Female Employment by Sector: 1966–1991
Figure 20: Female Employment by Sector: 1966–1991

Source: The Labour Force : Australia, ABS Cat. No. 6203.0.

  • the slow growth in full-time male employment. The only sectors to record any appreciable growth were community services and finance, and to a lesser extent, wholesale and retail trade;
  • the doubling of part-time male employment in wholesale and retail trade and the growth in part-time male employment in recreation (but the absolute numbers are small);
  • the increased number of women employed full-time in community services and finance; and
  • the very large rise in part-time female employment in finance, community services and wholesale and retail trade. By 1991, these last two categories accounted for nearly 60 per cent of part-time female employment. In contrast, the growth in full-time female employment in community services and wholesale and retail trade was much smaller.

An elementary, but nonetheless useful, method for interpreting these facts is to assume that sectoral employment levels are determined by the intersection of a downward sloping demand for labour curve and an upward sloping supply curve. We can then attribute declining employment over time to one of two factors: either inward shifts of the demand curve that may be due to factors such as declining demand for that sector's output; or inward shifts of the supply curve, due to declining numbers of people with the requisite skills or desire to work in that sector, for example. Similarly, increases in employment could be the result of outward shifts of the demand curve for labour, e.g. financial deregulation in the 1980s could plausibly have led to both increases in the demand for financial services and the number of people supplying them; or they could be the result of exogenous increases in labour supply, e.g. the steadily increased labour force participation of married women.

By examining the joint movement of wages and employment, we can ascertain the relative importance of these forces in the determination of sectoral employment. Shifts of the demand curve will result in employment and wage levels moving in the same direction, while shifts in the supply curve will result in them moving in opposite directions. If, for example, employment levels and wages have both tended to increase over time, we can conclude that outward shifts in the labour demand curve have dominated any shifts in the supply curve.

An important measurement issue which arises here is the treatment of the wage. In standard one-sector expositions, the relevant wage is “the” real wage W/P, where W is the nominal wage and P is the price of the good. However, when there is more than one sector, this choice is not clear. The real wage relevant to the demand for labour in sector i is Wi/Pi, where Wi and Pi are, respectively, the nominal wage paid in that sector, and the price of that sector's product. However, for labour supply decisions, the relevant real wage is Wi/P, where P is a general price index i.e., an average of all prices. Another problem is that we want to abstract from outward shifts in the labour demand curve due to productivity improvements (which will tend to increase real wages, however measured) and outward shifts in the supply curve due to population increases (which will tend to reduce real wages).

The second problem can be resolved by deflating the industry real wage by an economy-wide real wage, so that shifts in the demand and supply curves lead to changes in relative real wages. We resolve the first problem by assuming that the relevant real wage is that which determines labour supply, implying that the relative real wage is equal to the relative nominal wage.[25] Under this assumption, the demand curve for labour is shifted by changes in relative product prices. A decrease in the relative price of manufactures, for example, will lead to an inward shift of the labour demand curve in that sector.

Figures 21 and 22 show scatter plots of wages and employment from 1976 to 1991 in selected sectors of the economy, for men and women, respectively. The wage, shown on the vertical axis of each figure, is the relative hourly nominal wage, as discussed above. Employment, shown on the horizontal axis, is constructed as total hours worked in each industry divided by total hours worked in the economy.

Figure 21: Employment vs. Hourly Wage – Males
Figure 21: Employment vs. Hourly Wage – Males
Figure 21: Employment vs. Hourly Wage – Males

Source: The Labour Force : Australia, ABS Cat. No. 6203.0, Weekly Earnings of Employees (Distribution) : Australia, ABS Cat. No. 6310.0.

Figure 22: Employment vs. Hourly Wage – Females
Figure 22: Employment vs. Hourly Wage – Females
Figure 22: Employment vs. Hourly Wage – Females

Source: The Labour Force : Australia, ABS Cat. No. 6203.0, Weekly Earnings of Employees (Distribution) : Australia, ABS Cat. No. 6310.0.

Some interesting patterns are apparent. The first panel of Figure 21 shows that male manufacturing hours, as a proportion of total hours, declined significantly from 1976, with two separate trends occurring in relative wage levels. The first, from 1976 to 1983, saw relative manufacturing wages increase (implying an inward shift in the labour supply curve), while relative manufacturing wages fell quite sharply from 1983 to 1991, implying an inward shift of the demand curve.[26]

In sharp contrast, consider the relative wage and employment levels of males in finance, shown in the second panel. The increase in employment is quite evident, as is the increase in wages since 1983, coinciding, and no doubt caused by, the deregulation of the financial system and property boom of the 1980s. Community services in the third panel also saw a large relative increase in male employment. However, wages in this sector tended to fall, suggesting a shift in labour supply rather than demand. The final panel of Figure 21 shows steadily declining employment in agriculture, and a volatile relative wage, possibly due to inward shifts of both the demand and supply functions.[27]

The first panel of Figure 22 shows a fall in the hours worked by women in manufacturing, but unlike men, an increase in wages over the 1980s. This suggests the dominance of supply factors rather than demand. Why this occurred is not obvious, although the increasing attractiveness of employment in other sectors is a possibility.

The following panel shows the large increase in employment and wage levels of women in finance, again no doubt due to the effects of financial deregulation and the property boom. It is possible that women moved from manufacturing to finance, attracted by higher wage levels brought about by the outward shift in demand for workers in the finance sector. (It is also possible that women were attracted into finance from outside the labour force, making the wage lower than it would otherwise have been. However, since the net effect of these moves was a clear increase in finance sector wages, demand factors must have been dominant.) The relative decrease in female employment in wholesale and retail trade, shown in the third panel appears to have been due to both demand and supply shifts, reflected in the volatile wage. The same appears to be true of employment in community services.

In Table 4 we show average growth rates of output, employment and hours worked for all sectors, over five periods. These periods are 1966/67 to 1973/74[28], a time of strong economic growth and very low unemployment; 1974 to 1981, a period of weak growth and a steadily rising unemployment rate; the two recessions 1982–1983 and 1990–1991, which saw very rapid increases in unemployment, and 1984–1989, a period of strong output and employment growth, and a steadily but slowly falling unemployment rate. The major points of interest are:[29]

Table 4: Average Annual Percentage Changes in Output (Q), Employment (E) and Hours Worked (H) by Sector
Sector 1966/67–1973/74 1974–1981 1982–1983 1984–1989 1990–1991
Agriculture 1.9 0.0 0.3 3.3 −0.2 −0.9 1.2 0.0 −0.2 2.9 0.4 −0.2 2.9 −0.3 0.3
Mining 18.3 3.4 3.8 2.8 4.0 4.2 3.5 1.8 −1.9 7.0 1.0 2.7 6.0 −5.0 −3.4
Manufacturing 5.8 1.7 1.7 1.6 −1.2 −1.7 −4.4 −4.4 −7.0 4.4 1.3 2.3 −3.5 −4.3 −4.5
EGW 7.4 0.4 −0.1 4.9 3.4 2.6 3.6 4.3 3.0 4.8 −2.8 −2.3 3.0 −6.1 −5.1
Construction 5.4 3.2 3.1 3.8 −0.5 −1.3 −6.3 −7.9 −10.3 5.1 6.6 7.8 −8.0 −5.0 −6.0
WRT 5.2 2.6 2.1 2.3 0.9 0.1 −1.2 −1.2 −1.8 4.4 4.3 4.1 −1.6 −0.1 −0.1
TSC 6.9 2.3 2.3 4.8 1.2 0.4 1.0 2.1 0.5 6.1 1.3 2.1 2.1 −0.5 0.3
Finance 4.8 4.6 4.1 3.4 4.2 4.0 1.0 1.4 0.2 7.4 7.2 8.0 −1.4 1.4 1.5
Public 3.4 4.8 4.6 2.5 3.2 2.8 1.5 2.6 1.1 1.9 1.1 1.7 1.5 4.4 4.6
Comm Serv 6.4 5.2 4.0 5.1 5.0 4.9 4.5 1.6 0.9 4.1 4.2 4.5 3.9 2.9 3.7
Recreation 4.9 3.2 2.3 1.8 1.4 1.0 1.8 0.7 −0.1 3.2 5.6 5.8 0.6 4.3 3.3
Total 5.6 2.6 2.2 2.9 1.3 0.6 −0.4 −0.9 −2.3 4.8 3.5 3.8 −0.3 −0.1 −0.2

Notes: Agriculture is Agriculture, Forestry, Fishing and Hunting; EGW is Electricity, Gas and Water; WRT is Wholesale and Retail Trade; TSC is Transport, Storage and Communication; Finance is Finance, Property and Business Services; Public is Public Administration and Defence; Comm Serv is Community Services; and Recreation is Recreation, Personal and Other Services.

Source : The Labour Force : Australia, ABS Cat. No. 6203.0, Australian National Accounts, Gross Product, Employment and Hours Worked, ABS Cat. No. 5222.0, Australian National Accounts: National Income and Expenditure, ABS Cat. No. 5204.0.

  • the very large growth rates in manufacturing productivity during the two periods of expansion. From 1966/67 to 1973/74, manufacturing output grew on average by 5.8 per cent per year, while employment and hours each grew by only 1.7 per cent per year. From 1984 to 1989, annual output growth averaged 4.4 per cent, while growth in employment and hours averaged only 1.3 per cent and 2.3 per cent, respectively;
  • the large growth in productivity in 1984–1989 in electricity, gas and water, and transport, storage and communication which continued in 1990–1991. In contrast, output per employee in these sectors fell over 1982–1983. This difference probably reflects the commercialisation of many government-owned business enterprises in recent years;
  • large falls in output, employment and hours worked occurred in manufacturing and construction during the two recessions, 1982–1983 and 1990–1991. However, unlike manufacturing, employment in construction grew strongly from 1984 to 1989;
  • agricultural productivity grew strongly in most periods; and
  • productivity in mining grew exceptionally strongly during the two periods of expansion, 1966/67 to 1973/74 and 1984 to 1989. However, this growth was probably driven by capital accumulation, and the mining sector is, in any case, a relatively small part of the economy, especially in employment.

The sectoral shares of output, employment and hours worked, in each period, are shown in Table 5. The declining importance of manufacturing and agriculture is quite apparent, as is the increasing importance of the service sectors.

Table 5: Sectoral Shares of Output (Q), Employment (E) and Hours Worked (H)
Sector 1966/67–1973/74 1974–1981 1982–1983 1984–1989 1990–1991
Agriculture 8.6 8.1 10.0 4.8 6.6 8.4 4.3 6.5 8.3 4.3 5.9 7.4 4.1 5.5 6.8
Mining 3.2 1.4 1.4 6.7 1.3 1.4 6.4 1.5 1.6 7.5 1.4 1.5 8.3 1.2 1.4
Manufacturing 28.6 24.8 25.2 21.4 21.0 21.7 19.9 18.6 19.0 18.8 16.6 17.4 18.1 15.0 15.9
EGW 3.6 1.9 1.9 3.3 1.9 1.8 3.8 2.2 2.1 3.8 1.9 1.8 4.0 1.3 1.3
Construction 7.5 8.4 8.7 9.0 8.1 8.4 8.6 6.9 7.0 8.2 7.1 7.4 7.3 7.3 7.5
WRT 16.6 20.4 20.4 18.1 20.1 20.0 17.7 19.8 19.9 17.0 20.2 20.0 16.5 20.8 20.1
TSC 8.1 7.6 7.6 6.6 7.6 7.6 7.2 7.9 8.0 7.7 7.4 7.6 8.1 6.9 7.3
Finance 8.8 6.8 6.4 10.0 7.9 7.7 10.8 9.1 8.9 11.8 10.4 10.4 12.6 11.5 11.7
Public 3.8 3.7 3.4 4.5 4.6 4.3 4.4 4.7 4.4 4.2 4.7 4.4 3.9 4.6 4.4
Comm Serv 7.3 10.9 9.5 11.1 14.8 13.2 12.4 16.5 15.0 12.5 17.5 15.8 13.0 18.2 16.7
Recreation 3.9 6.1 5.5 4.5 6.2 5.5 4.5 6.4 5.9 4.3 6.9 6.3 4.2 7.6 6.9

Notes: See Table 4 for sectoral abbreviations.

Source : The Labour Force : Australia, ABS Cat. No. 6203.0, Australian National Accounts, Gross Product, Employment and Hours Worked, ABS Cat. No. 5222.0, Australian National Accounts : National Income and Expenditure, ABS Cat. No. 5204.0.

4.1 Manufacturing

We now turn to a more detailed analysis of manufacturing, the only sector, apart from agriculture, where employment has been falling, in absolute terms, over the past 20 years. Table 6 shows that, within the manufacturing sector, the growth rates of output, employment and productivity have been far from uniform. For example, between 1984 and 1989, productivity growth in the transport equipment industry was very small. On the other hand, output growth in basic metal products averaged 4.3 per cent per year, while employment fell by 1.8 per cent per year. Very large productivity gains also occurred in paper, non-metallic mineral products and fabricated metal products. This period of high productivity growth followed a savage reduction in employment during the recession of 1982–83, especially in the metals industries.

Table 6: Average Annual Growth Rates of Output (Q) and Employment (E) in Manufacturing
Sector 1977–1981 1982–1983 1984–1989 1990–1991
Food 1.0 −2.5 −0.2 3.6 3.1 0.6 3.0 −3.9
Textiles 1.4 −2.9 −5.4 −6.9 5.2 1.5 0.3 −6.5
C&F 0.7 −0.2 −2.0 −2.7 1.7 0.0 −10.1 −6.5
Wood 1.0 2.8 −5.1 −4.9 5.0 4.0 −6.7 −6.7
Paper 3.9 1.3 −1.8 2.2 6.5 1.5 −2.0 −1.6
CP&C 3.2 −0.6 −0.5 −4.9 3.1 0.9 0.8 −5.1
NMP 0.9 −1.4 −7.3 −3.3 5.6 1.0 −14.0 0.7
BMP 3.4 2.9 −5.1 −8.5 4.3 −1.8 −3.9 −5.3
FMP 4.5 0.3 −10.0 −6.8 6.8 2.5 −1.5 −3.2
Transport 0.3 −2.1 −2.9 −9.6 2.6 2.0 −14.0 −8.3
Other M 2.1 −0.3 −10.7 −10.9 5.2 1.2 −4.4 −2.7
Misc 3.5 −0.6 −4.7 0.0 5.6 2.8 −5.6 −3.3
Total 2.1 −0.4 −4.4 −4.4 4.4 1.3 −3.5 −4.3

Food is Food, Beverages and Tobacco; C&F is Clothing and Footwear; Wood is Wood, Wood Products and Furniture; Paper is Paper, etc, Printing and Publishing; CP&C is Chemical, Petroleum and Coal Products; NMP is Non-Metallic Mineral Products; BMP is Basic Metal Products; FMP is Fabricated Metal Products; Transport is Transport Equipment; Other M is Other Machinery; Misc is Miscellaneous.

Source: The Labour Force : Australia, ABS Cat. No. 6203.0, Australian National Accounts, Gross Product, Employment and Hours Worked, ABS Cat. No. 5222.0.

We noted above that the reduction in male manufacturing employment since 1983 was probably due to a reduction in demand for manufacturing workers.[30] One explanation for this decline, quite common in popular discussions, is that technological innovation has reduced the demand for manufacturing workers. At first glance this seems an unlikely explanation, because the way in which technology is usually modelled results in increased output for all combinations of inputs to production and therefore leads to increased, not decreased, employment in the technologically improved sector.

However, in Appendix 2 we describe a two-sector general equilibrium model in which labour-saving technological progress in the capital-intensive sector, e.g. manufacturing, can lead to a reduction in employment in that sector, provided the elasticity of substitution in production between capital and labour in each sector is sufficiently small. We know of no convincing evidence one way or the other on this question.

Another explanation, currently popular in some circles, is that reductions in tariffs and other trade barriers on imported manufactures caused contractions in the output and employment in some manufacturing industries. In Table 7 we show changes in employment and effective protection in the manufacturing sector. Two points are immediately apparent. The first is that most of the job losses in manufacturing, over the period 1976 to 1991, occurred during the recession years 1982–1983 and 1990–1991. In some industries, in fact, over 100 per cent of net job losses over the extended period occurred during those four years. The second point is that there was no significant reduction in effective protection during this time; indeed, effective assistance actually increased in some cases. We therefore find it difficult to conclude that the reduction in manufacturing employment was simply caused by declining industry assistance.

Table 7: Employment and Protection in Manufacturing 1976 – 1991
Industry Change in
1976 to 1991
(per cent)
Change in
1981 to 1983
(per cent)
Change in
1989 to 1991
(per cent)
('000)   ('000)   ('000)  
  1976 90/91   1981 1983   1989 90/91
Food −20.3 18 3 12.8 10 6 −14.7 3 3
Textiles −13.3 51 68 −5.0 55 68 −4.3 72 68
C&F −16.3 118 176 −4.7 172 210 −10.3 172 76
Wood 11.1 18 13 −9.4 15 18 −14.6 16 13
Paper 17.4 30 7 4.8 25 16 −4.2 11 7
CP&C −12.8 24 10 −6.4 15 12 −6.3 12 10
NMP −5.4 9 3 −3.5 4 4 0.7 3 3
BMP −21.5 16 8 −17.4 10 10 −8.3 9 8
FMP −8.5 36 17 −16.5 31 27 −8.1 19 17
Transport −49.6 61 33 −25.8 67 62 −20.9 37 33
Other M −38.5 25 15 −37.8 21 21 −8.2 18 15
Misc 3.2 25 20 −0.3 28 25 −5.5 23 20
Total −154.3 28 15 −108.9 24 21 −104.5 17 15

Notes: See Table 6 for industry abbreviations. The average rate of effective protection calculated as (VA-VA')/VA,' where VA is value added with protection, VA' is value added without protection. Estimates of effective protection for calendar years are averages of adjoining financial years.

Source: Industries Commission and Industries Assistance Commission Annual Reports, various issues.

The decline in manufacturing employment might, however, be associated with the fall in industry assistance in another sense. Table 7 shows that most of the falls in manufacturing employment have occurred during recessions, and that these falls have been largely permanent.[31] That is, we think that in addition to being periods of temporarily slow aggregate economic activity, recessions are also periods of accelerated structural change. Viewed in this light, the slow fall in unemployment following a recession is readily explainable. While the purely cyclical increase in unemployment is eventually reversed, the recession-induced structural adjustment is not.

The data suggest that this structural adjustment has involved a permanent change in the composition of aggregate employment, away from manufacturing and towards services. However, the new jobs have not been filled by manufacturing workers who have lost their jobs in the recession but by new entrants to the workforce. There is no obvious theoretical reason for this to be the case, but it is possible that these new jobs have characteristics (e.g. are part-time) and require skills that are more suited to these new entrants (such as women re-entering the workforce). Declines in industry assistance facilitate this structural change.

We are unaware of any formal theory in the business cycle literature that models this structural adjustment process exactly; however some interesting recent papers by Caballero and Hammour (1991) and Hall (1991a, 1991b) highlight some of the important mechanisms that could be at work. These papers view recessions as periods of economic reorganisation and renewal, an idea first articulated by Joseph Schumpeter (1939)[32].

In the model of Caballero and Hammour, product units embodying the latest techniques are continuously being created, while obsolescent units are being destroyed. Demand falls during a recession, leading to an increase in the rate of destruction and a decrease in the rate of creation; the extent to which demand fluctuations are accommodated along either margin depends on the costs of creation. This model is applied to U.S. data on gross labour market flows in manufacturing with the result that the rate of job destruction is found to be much more responsive to changes in activity than the rate of job creation, and that output fluctuations are asymmetric – contractions in output are sharper and more short-lived than expansions. However, these asymmetries are smoothed by the rate of job creation, which is more responsive to expansions in activity than contractions. The rate of job creation, therefore, is roughly symmetric around its mean. Job destruction, on the other hand, amplifies the asymmetries of the output cycle, and so the net effect of the business cycle on employment is asymmetric: unemployment rises sharply during a recession but falls only slowly during the expansionary phase of the cycle.

In Hall's model of business cycles, recessions provide a favourable time to reorganise and undertake productivity-improving activities because the opportunity costs of doing so are temporarily low. Central to this model is the concept of agglomeration externalities, the idea that, because there are complementarities associated with both production and reorganisation, economic activity takes place more efficiently in concentrated periods of time and space.

These models lead to the view that recessions are times of “cleaning up”, when outmoded techniques and products are purged from the economic system. They do not directly address the issue of recessions as periods of accelerated structural change, but it is not difficult to see how the models could be so extended. For example, the model of Caballero and Hammour could be extended to two sectors, with variations in demand leading to adjustment along four margins – the rates of job creation and destruction in each of the two sectors.[33]

These types of the models address the propagation mechanism of recessions, but not the issue of the shocks which initiate the cycle. This is one of the most hotly debated issues in modern macroeconomics, with a consensus view emerging that both nominal and real shocks are empirically important, though their relative significance is still a matter of debate (King et al, 1991). The shock which precipitated the 1982–83 recession is often thought to be the large wage increases which occurred in the metals industries in previous two years, with these increases (especially the wage package of December 1981) acting as a catalyst for wage increases in the rest of the economy (Treasury, 1982).[34] In Table 8, we show some of the increases in real wages in the metals' industries that occurred between 1980 and 1982. It is perhaps not unreasonable to conclude that the massive fall in employment in these industries in the ensuing recession was associated with these increases in real wages, which did not appear to be accompanied by any corresponding increases in productivity. However, if our view of recessions as periods of accelerated structural change is correct, these job losses probably would have occurred eventually anyway.[35] In this sense, the fundamental cause of the decline in employment in the metals' industries in the early 1980s was ongoing structural change in the economy; the prior increase in real wages was the proximate cause which largely determined the pace of that change.[36]

Table 8: National Average Wages for Metal Industry/Metal Trades Award Members 1980–1982
Real Increase
(per cent)
Toolmaker 232.85 299.50 6.3
Fitter 228.40 313.20 13.3
Boilermaker 225.30 308.90 13.3
Welder – 1st Class 222.90 306.90 13.8
Machinist – 1st Class 218.60 288.20 9.0

Notes: These figures do not include the June 1982 Metals Industries Award increase; the estimated real increases are based on the 21 per cent increase in the CPI between March 1980 and March 1982.

Source: Amalgamated Metal Workers and Shipwrights' Union, National Wages and Conditions Survey, June 1982.

4.2 Persistence in Unemployment: Sectoral Evidence

To complete this section, we present data on sectoral unemployment. These data, together with those on employment by sector, duration and participation presented earlier in the paper, allow us to deduce the principal structural causes of persistence in Australian unemployment. Unemployed people are defined as belonging to the sector in which they were last employed full-time, provided that employment was in the previous two years.[37]

Figure 23 shows the very long term unemployed (those unemployed for greater than 104 weeks), the major sectors, and the new entrants to the full-time labour force, as components of the aggregate male and aggregate female unemployment rates.

Figure 23: Decomposition of Unemployment Rates: 1978–1991
Figure 23: Decomposition of Unemployment Rates: 1978–1991 Note: NEVER EMP is never employed full-time, 104+ is unemployed for over two years, CMS is Community Services, WRT is Wholesale and Retail Trade, MFG is Manufacturing, CNS is Construction and OTHER is all other unemployed people classified by industries which are not given separately.

Source: The Labour Force: Australia, ABS Cat. No. 6203.0.

Consider the unemployment rate of men. During the recession of 1982–83 it peaked at just under 10 per cent in 1983; by 1987 it had fallen to about eight per cent, still three percentage points above the rate in 1981. Interestingly, the three sectors that account for over 70 per cent of male unemployment – manufacturing, construction, and wholesale and retail trade, appear to have contributed, between them, only about 20 per cent of persistence.

The sectors labelled “other” (principally agriculture, finance and recreation) contribute about one third of persistence, despite their much smaller contribution to total unemployment. (The sectoral contributions to total unemployment of those who have worked full-time in the last two years can be seen in Table 9.) The persistence of agricultural unemployment can be easily seen in Figure 24, which plots the number of people unemployed in each sector. Unemployment in agriculture in the 1980s hardly fell from its 1983 peak; since agricultural employment was also flat, the agricultural unemployment rate was very persistent. The source of the increased contributions to total persistence from finance and recreation is their increased relative importance in the economy as a whole.

Table 9: Distribution of Unemployment by Sector Percentage of Total
Sector 1978–81 1982–83 1984–89 1990–91
Manufacturing 28.9 30.2 25.9 25.7
Construction 18.0 18.7 15.0 19.0
WRT 20.4 19.0 19.8 19.6
Agriculture 7.9 6.9 7.4 6.8
Recreation 6.3 6.4 7.9 6.7
Finance 3.3 3.8 4.6 6.4
Other 15.2 15.0 19.4 15.8
Total 100.0 100.0 100.0 100.0
Sector 1978–81 1982–83 1984–89 1990–91
Manufacturing 20.7 20.3 16.3 17.3
Comm Serv 15.9 15.5 17.5 14.7
WRT 30.4 30.8 29.1 28.4
Recreation 14.4 14.4 15.3 14.6
Finance 7.2 7.8 8.3 11.2
Other 11.4 21.2 13.5 13.8
Total 100.0 100.0 100.0 100.0

Notes: The numbers in the table are the proportions of those who are unemployed and have worked full-time in the last two years, accounted for by each sectoral component. Therefore, “other” includes all those who were employed full-time in the past two years in a sector which has not been included separately. See Table 4 for sectoral abbreviations.

Source: The Labour Force : Australia, ABS Cat. No. 6230.0.

Figure 24: Unemployment by Sector: 1978–1991
Figure 24: Unemployment by Sector: 1978–1991

Source: The Labour Force : Australia, ABS Cat. No. 6203.0..

Slightly less than half the persistence in male unemployment has come from two categories which are not sector-specific. These are the very long term unemployed, and those unemployed people who have never held a full-time job or have been out of the full-time labour force for at least two years. The latter is partly due to the increased difficulties school leavers have in finding employment: in 1987, 18.0 per cent of 15–19 year old males were unemployed, compared with 11.2 per cent in 1981.[38]

Given the negligible growth in manufacturing employment it is puzzling that persistence in manufacturing is not measured to be much larger than it is. One possible explanation is that former manufacturing workers tend to leave the labour force in large numbers, given the rather poor job prospects in their sector. Another is that they account for a large proportion of the very long term unemployed. We have no way of knowing this for sure, since the very long term unemployed are not classified by sector.

However, in Table 10 we show the proportion of those unemployed for between one and two years accounted for by each sector. We can see that former manufacturing workers are disproportionately represented in this group[39]. Assuming that this pattern is repeated among the very long term unemployed, we have some evidence, albeit weak, that the persistence of male unemployment is partly due to the inability of displaced manufacturing workers to find employment elsewhere in the economy. Also disproportionately represented are displaced workers from wholesale and retail trade. This is also puzzling, since employment growth in this sector was strong after the 1982–83 recession. However, most of this growth was in part-time employment; it would appear that these jobs were largely filled by entrants to this sector rather than former full-time employees.

Table 10: Proportion of Unemployed with Durations between One and Two Years by Sector: 1978–1991
Agriculture 3.3 Finance 2.6
Manufacturing 17.3 Public 2.7
Construction 7.0 Comm Serv 4.1
WRT 13.9 Recreation 5.5
TSC 3.0 Other Sectors 1.8

Notes: See Table 4 for sectoral abbreviations. Note also that only 60% of those unemployed between one and two years can be classified by sector, the remainder being those who have not worked full time in the last two years.

Source: The Labour Force: Australia, ABS Cat. No. 6203.0.

The bottom panel of Figure 23 decomposes female unemployment. It is immediately apparent that persistence is less of a problem for women; by 1987, 60 per cent of the increase in female unemployment (from its trough in 1981) had wound back, compared with 40 per cent for men. It is also apparent that the bulk of female unemployment comes from women who are entering the workforce or have not been working full-time. This category also contributes most to the persistence of female unemployment. There appears to be little or no sectoral persistence; this observation is confirmed by the lower panel of Figure 24.

The data therefore suggest that persistence in female unemployment is not driven by the sectoral mismatch which seems to characterise, to a large extent, the persistence in male unemployment. This conclusion is supported by the fact that high median unemployment durations, which decline slowly, do not appear to be as much of a problem for women as for men. The sectoral data are consistent with our conclusions based on the gross flow data: female full-time unemployment fell only slowly over the recovery period 1984 to 1989 because the increases in female employment were largely offset by increases in labour supply.[40] Further evidence for this hypothesis is the slight increase in female unemployment in community services during this period, despite the very large increases in employment in that sector.


Gregory (1991) estimates that, over the period 1966–1988, 58 per cent of new male jobs were filled from the male unemployment pool, while only 23 per cent of new female jobs were filled by unemployed women. [20]

“Finance” is actually “finance, property and business services” which, among others, includes real estate agents, property valuers and business services such as accountancy, advertising etc.. [21]

Real unit labour costs increased by four per cent in each of the financial years 1973/74 and 1974/75. (Foster and Stewart (1991), Table 4.17). [22]

For econometric evidence on this issue, see Layard et al. (1991), Chapter 9. [23]

Effective rates of assistance are published in the annual reports of the Industries Assistance Commission, later the Industries Commission. [24]

The relative real wage is (Wi/P)/(W/P), which is equal to Wi/W, where W is the average nominal wage. We construct this as a fixed weight average of all sectoral wages, with the weights determined by relative employment levels in 1976. [25]

It is interesting to note that this decline in relative manufacturing wages coincided precisely with the operation of the Prices and Incomes Accord, one of the undesirable consequences of which is supposed to be its inability to deliver the relative wage movements necessary for the efficient working of the labour market. As Figures 21 and 22 show, changes to relative wages can occur under a regime of centralised wage fixation. Of course, it is always possible to argue that even greater relative wage movements would have occurred in the absence of the Accord, but this is a difficult proposition to test. [26]

These facts are also consistent with a model in which increases in agricultural productivity, together with falls in the relative price of agricultural goods, produce a fall in agricultural employment. Such a model is outlined in the Appendix 1 and is broadly consistent with two stylised facts: the large increases in agricultural productivity (see Table 4) and a decline in Australia's terms of trade (Foster and Stewart 1991, p29). [27]

We use financial rather than calendar years for this period because the data are available only on a financial year basis until August 1974. [28]

When calculating industry output, the Australian Bureau of Statistics mainly uses employment data from the Survey of Employment and Earnings; by construction, labour productivity in community services, finance, and public administration and defence is zero. Because all our employment data come from the Labour Force Survey, our productivity measures in these three sectors can differ from zero. [29]

We also noted that the relative reduction in female manufacturing employment was probably due to an inward shift of the supply curve. The male/demand effect dominates, however, due to the much larger size of the male manufacturing labour force. Between 1973 and 1991 total manufacturing employment fell by 254,000, comprising a fall of 208,000 men and 46,000 women. [30]

This observation is based on only one observation, the period 1984 to 1989, and so needs to be treated with some caution. However, we argue below that there are good reasons to believe that the loss of manufacturing employment in the current recession will not be recovered at a later stage. [31]

A related literature deals with the interaction between aggregate disturbances and sectoral productivity shocks. Davis and Haltiwanger (1990) investigate the connection between the heterogeneity of establishment-level employment changes and aggregate employment fluctuations over the cycle. They construct a theoretical model which suggests how both aggregate and allocative disturbances can drive fluctuations in job creation, job destruction, productivity, output and unemployment. Their empirical analysis indicates that allocative disturbances were very important in the determination of these variables in U.S. manufacturing over the period 1972 to 1986. See also Davis and Haltiwanger (1992), Aghion and Saint-Paul (1991) and Saint-Paul (1992). [32]

Another class of models which can explain the asymmetric behaviour of U.S. manufacturing employment has been recently developed by Cabellero and Engle (1991, 1992a, 1992b). In these models economic agents at the microeconomic level make infrequent, but large, adjustments in response to random shocks. (Formally, Cabellero and Engle characterise behaviour in terms of an adjustment hazard function; the probability of adjustment increases the larger is the deviation of a variable from its targeted value.) These models are then used to generate aggregate dynamics with properties that are consistent with the data, such as gross employment flows in US manufacturing. Conceivably, models of this type could also be used to explain how large negative shocks, such as recessions, can lead to changing structural behaviour in firms, and the consequent implications for sectoral and aggregate employment. [33]

Other factors, such as the world-wide recession and the rural drought, were also important in determining the length and severity of this recession. [34]

BHP Steel, for example, used the recession of 1982–83 as a catalyst for modernising its plant and equipment, and especially for labour shedding; it now employs about one-third the number of workers than it did before the recession. It seems reasonable to speculate that these changes would probably have occurred anyway over the 1980s, but, in the absence of the wage rises of 1980–82 and subsequent fall in economic activity, would not have occurred in such a concentrated period of time. [35]

We are not, however, asserting that the recession of 1982–83 was, apart from the international influences and the drought, in any sense unavoidable, much less desirable. A good deal of the lost output and employment in this recession could probably have been avoided in the absence of the generalised wage increases of the previous two years. [36]

The very long term unemployed are therefore excluded from the sectoral unemployment data, as are new entrants to the labour force, such as school leavers, and unemployed part-time workers. [37]

Foster and Stewart (1991), Table 4.16. [38]

Over the period 1978–1991, former manufacturing workers accounted for 17.3 per cent of those unemployed between one and two years, and nearly 25 per cent of previously employed full-time workers. However, over the same period manufacturing workers accounted for only 20 per cent of the fully employed work force. [39]

Female employment grew by 27 per cent over the six years to 1989, almost double the growth rate of male employment. Over the same period, however, the female participation rate increased from 44 per cent to 51 per cent, while the male particpation rate fell slightly. [40]