RDP 9505: Labour-Productivity Growth and Relative Wages: 1978-1994 3. Labour-productivity Growth in Specific Industries

Over the business cycle that ran from March 1983 to June 1991, four industries experienced a decline in the measured level of labour productivity. In the finance, property and business services sector this decline can largely be explained by problems with the measurement of labour input (see previous section). This section analyses productivity outcomes in the other three sectors: two service industries – wholesale and retail trade and recreation, personal and other services – and the construction industry.

3.1 The Wholesale and Retail Trade Industry

The wholesale and retail trade industry is the largest industry in Australia. In 1994, it accounted for nearly 18 per cent of total output in the non-farm economy (nearly 20 per cent if the ownership of dwellings, import duties and imported bank service charges are excluded) and for about 22 per cent of total hours worked.

Figure 9 shows output per hour worked in the industry over the period from March 1978 to June 1994. Labour productivity is clearly pro-cyclical; falling in recessions and increasing in booms. Perhaps more importantly, the trend rate of productivity growth appeared to change in the mid 1980s. After output per hour worked increased by around 1 per cent a year from 1978 to 1983, it fell, on average, by around 0.1 per cent a year over the following eight years.[12] More recently, labour-productivity growth has again been positive. Despite the recent rise, labour productivity in June 1994 had still not reached its previous peak in March 1984.

Figure 9: Labour Productivity in the Wholesale and Retail Trade Industry
(March quarter 1978 = 100)
Figure 9: Labour Productivity in the Wholesale and Retail Trade Industry

While comparable data for other countries are scarce, the low productivity growth in the wholesale and retail trade industry appears to be unusual by international standards.[13] The OECD publishes data on total-factor productivity for different sectors for member countries (Meyer Zu Schlochtern and Meyer Zu Schlochtern 1994). Unfortunately, the data for wholesale and retail trade also include the output of the hotels and restaurants sector (which under the ASIC classification system is included in the recreation, personal and other services sector). The data show that between 1970 and 1989, total-factor-productivity growth in this sector in Australia was −0.34 per cent per annum. This performance was worse than that in any other country for which data are published – on average, across the 13 countries, total productivity growth in the sector was 0.8 per cent per annum.[14]

In light of the international experience, and given the perception that there has been substantial technological innovation in the distributive trades over the past decade, the fall in labour productivity is surprising. The use of scanning devices and computer-operated stock control systems have become common place. Stores have become larger and many firms have rationalised their operations. Despite these changes, the average worker in June 1994 was processing slightly fewer goods per working hour than a worker ten years ago. We now examine why this is so.

3.1.1 Measurement issues

Firms in the distribution industry provide three core outputs: (i) distribution of goods; (ii) product information; and (iii) convenience or the minimising of transaction costs. (Some might argue that they also provide entertainment.)[15] In obtaining an estimate of the growth rate of output for the industry, statistical agencies typically only consider the first of these outputs. In Australia, in the wholesale industry, base-year output for six industry groups is extrapolated by constant price estimates of wholesalers' sales in each industry group. In the retail industry, base-year output for 16 industry groups is extrapolated by constant price estimates of industry retail turnover. This methodology assumes that output is a fixed proportion of turnover for each industry group. As such, it only takes account of the ‘distribution’ output of the sector, ignoring the supply of product information and the provision of convenience. Given current measurement practices, changes in the quality of service, opening hours, the extent of product information and the quality of the shopping environment do not lead to changes in output.

There are two principal sources of data on measured output and labour input for the wholesale and retail trade industries. The first is the periodic censuses of the retail industry and surveys of the wholesale industry carried out by the ABS. For retail trade, a census has been conducted every five or so years since 1948.[16] For wholesale trade, the most recent surveys were conducted in 1981/82 and 1991/92. The second source is that used above for the calculation of industry labour-productivity indices – that is, quarterly output data from the national accounts and labour input from the Labour Force Survey. The census/survey data provide comprehensive snap-shots at particular points in time, while the national accounts provide time-series data.

Unfortunately, the two sources present quite different pictures. The quarterly data show that in the second half of the 1980s, labour productivity fell in both the wholesale and retail industries, with the fall being more pronounced in the wholesale industry (see Figure 10). This is at odds with the results from the Wholesale Industry Surveys and the Retail Censuses. Data from the survey of wholesalers show that real turnover increased by 19 per cent between 1981/82 and 1991/92, while employment fell by 4 per cent; implying an increase in turnover per person of some 23 per cent. Over the same time period, the national accounts data show an increase in output in the wholesale sector of around 8 per cent, and the Labour Force Survey shows an increase in employment of 19 per cent (and an increase in hours worked of 23 per cent). This implies a fall in output per person of 15 per cent between 1981/82 and 1991/92.[17] In retailing the contrast is less marked. The census data appear to show slightly slower productivity growth than do the national accounts/labour force survey data, although a complete assessment is made difficult due to changes in industry definitions brought about by the switch from ASIC to ANZSIC classifications.

Figure 10: Labour Productivity Indices for Retail Trade and Wholesale Trade
(March quarter 1978 = 100)
Figure 10: Labour Productivity Indices for Retail Trade and Wholesale Trade

3.1.2 The wholesale industry

If the survey data are correct, then labour-productivity growth in wholesaling was considerably higher than that suggested by Figure 10. Table 2 presents employment and turnover data from the Wholesale Industry Survey for various sub-categories of wholesale trade. It shows that in 1991/92, almost two-thirds of total employment in wholesaling was in sub-sectors that supply investment goods (the first four categories). This explains much of the cyclicality of the industry. In all sub-sectors, with the exception of machinery and equipment, employment actually fell over the ten years to June 1992. Perhaps more importantly, the data show that in all but one sub-sector, real turnover per person increased between 1981/82 and 1991/92. The increase was particularly large in minerals, metals and chemicals, where labour productivity increased by 70 per cent over the ten year period. Given the widespread gains in labour productivity throughout the industry, it seems unlikely that labour productivity in the industry as a whole could have declined.

Table 2: The Wholesale Industry (1991/92)
Type of wholesaler
Turnover $m
% change in employment 1981/82–1991/92 % change in turnover per person 1981/82–1991/92
Builders hardware dealers 44,155 6,408 −19.7 20.8
Machinery and equipment 102,677 17,616 4.9 23.5
Minerals, metals and chemicals 31,012 24,438 −20.7 70.7
Farm properties and produce 19,565 7,954 −14.9 −6.1
Food, drink and tobacco 43,466 14,516 −14.8 3.1
Textile and clothing 14,875 3,076 −9.3 6.7
Household goods 12,977 3,235 −26.3 28.6
Other specialists 39,524 7,874 −0.1 34.2

Notes: (a) Nominal turnover data have been deflated by the implicit price deflator for the wholesale industry.
(b) ASIC and ANZSIC data have been used to construct the percentage changes. The differences in the definitions are relatively minor with a couple of exceptions. In the builders hardware dealers category, building completion services are not included in the data for 1991/92 (in 1991/92 this sector accounted for around 7 per cent of sub-sector employment). The machinery and equipmentcategory excludes motor vehicles wholesalers in both periods.

Source: ABS Cat. No. 8638.0, 1981/82 and 1991/92.

3.1.3 The retail industry

Stores selling food account for the largest share of employment in the retail industry (see Table 3). Food stores have also been the source of the fastest jobs growth within retailing, with the number of jobs increasing by 22 per cent between 1986 and 1992. Employment in stores selling bread and cakes grew particularly strongly, increasing nearly 140 per cent over the six year period, while the number of establishments increased by nearly 90 per cent.[18] Employment growth was also very rapid in takeaway food retailing (up 41 per cent). Outside the food industry, employment grew quickly in stores selling domestic hardware and housewares (up 43 per cent) and in stores selling newspapers, books and stationary (up 28 per cent) (see Appendix A).

Table 3: The Retail Industry
Square metres
per location
per square metre
per employee
Type of store
Supermarket and grocery 179,619 6.6 560 23.3 4,918 5.4 145,269 8.9
Specialised food 238,028 37.0 105 11.2 3,294 −5.0 63,581 −18.7
Total food 417,647 22.0 186 10.9 4,164 1.0 98,713 −6.2
Department stores 86,576 −9.7 9,084 3.1 2,508 −10.5 113,172 3.5
Clothing and soft goods 91,653 8.9 152 12.0 2,597 −5.0 94,891 −1.3
Furniture, housewares appliances 72,503 23.7 401 20.2 2,072 −8.8 163,869 0.2
Recreational goods 56,297 17.8 155 4.6 3,323 1.6 113,168 −6.0
Other personal and household goods 111,026 39.4 169 8.8 1,955 1.5 82,909 4.1
Total personal and household goods 427,688 14.2 248 8.4 2,331 −7.1 108,626 −0.1
Motor vehicle retailing and services 215,198 n.a. n.a. n.a. n.a. n.a. 203,040 n.a.

Note: Turnover is in average 1989/90 prices.

Source: ABS Cat. No. 8623.1.

Value added per employee tends to be highest in those sub-sectors selling high-value goods such as furniture and motor vehicles. According to data published from the 1985/86 Retail Census, value added per employee in these sub-sectors is about three times that in the sub-sector with the lowest value added per employee (milk bars and take away stores). While large differences exist in the level of labour productivity between sectors, international evidence suggests that large differences also exist within sub-sectors.

In general, one of the most significant determinants of the level of labour productivity for a given class of store is the size of the store. In a comprehensive comparison of retailing productivity in Britain, the United States and Germany, Smith and Hitchens (1985) conclude that, in all three countries, labour productivity increases with shop size, where size is measured by turnover. They attribute much of the superiority of retail productivity in the US to the larger size of shops in that country. A similar conclusion is reached by Baily (1993) who attributes the large size of shops in the United States to relatively weak zoning regulation, few restrictions on shopping hours and a well-developed private transport system. An OECD study into the distribution systems of a variety of countries (OECD 1992) concludes that average sales per employee tend to be 50 to 80 per cent higher in the largest class of stores than the smallest, with value added per employee being as much as 100 per cent higher. The study also reports work by Noyelle (1990) which concludes that in France, the shift to larger stores between 1980 and 1986 contributed over 1 per cent per annum to labour-productivity growth in retailing (out of a total labour-productivity growth in the sector of 2.4 per cent per annum).

As Table 3 shows, this shift to larger stores has also occurred in Australia. The increases have been particularly pronounced for supermarkets and stores selling furniture, housewares and appliances. In both cases, the average sizes of stores increased by over 20 per cent between 1986 and 1992. In most cases, real turnover has not increased as quickly, so that turnover per square metre has fallen.

Not only has the physical size of stores increased, but so too has the number of employees per store. In 1980, on average, 5.5 people worked at each retail location. By 1992, this had increased to 6.2 people. In part, this increase is due to the growth of part-time employment, with the number of full-time employees per location actually falling. The increase in the size of businesses can also be seen in the number of ‘small’ and ‘large’ businesses (small businesses are those that employ fewer than 20 people);between 1980 and 1992 the number of small businesses increased by 18 per cent, while the number of large businesses rose by 32 per cent (Kiel and Haberkern 1994).

Despite the increase in store size and the adoption of new technologies, measured labour productivity in the retail industry has fallen.[19] As is the case for the economy as a whole, changes in labour productivity within a particular sector can, theoretically, be decomposed into changes due to compositional shifts and changes due to increases in labour productivity within sub-sectors. Unfortunately, detailed data on output and hours worked by sub-sector of retailing are not published. In their place, data from the 1985/86 and 1991/92 Retail Census can be used to obtain a rough estimate of the effect of compositional shifts. These data suggest that had the employment structure that existed at the time of the 1991/92 Census existed in 1985/86, value added per employee in 1985/86 would have been up to 2 percentage points lower.

The single most important contributor to this negative compositional effect is the expansion of the fast food industry (milk bars and take-away stores); over the six years between the Retail Censuses, the share of total hours (in food and personal and household good retailing) worked in this sub-sector increased by about 5 percentage points to around 17 per cent. With value added per hour worked in this sub-sector (in 1985/86) being roughly 60 per cent of the average, this change of 5 percentage points in employment shares contributed almost 2 percentage points to the decline in retail industry productivity. An additional 1 percentage point was contributed by the contraction of hours worked in department stores, as these stores tend to have higher than average levels of labour productivity. The rapid expansion of relatively low-productivity bread stores also contributed to the slowdown.

While compositional effects may explain part of the poor productivity outcomes, they can provide no more than a partial explanation, as labour productivity appears to have fallen in a wide range of different types of stores. A comparison of the 1985/86 and 1991/92 Retail Censuses, shows that in 15 out of the 25 sub-sectors for which comparative data are available, turnover per employee fell; the largest falls being recorded in retail stores selling marine equipment, takeaway food, bread and cakes, liquor and floor coverings (see Appendix 1). A similar picture emerges when disaggregated quarterly retail trade turnover data are examined.

Why did turnover per employee fall in such a wide range of retail stores in the second half of the 1980s? One possible answer is the spread of part-time employment – workers who work fewer hours process fewer goods through the check-out. Certainly, the share of part-time employment did increase more quickly in the second half of the 1980s than it did over the first half of the decade. Over the 1980s as a whole, the ratio of employment to hours worked in the industry increased by about 4 per cent. This suggests that even if an adjustment is made to take account of a reduction in average hours, labour productivity in a variety of stores would still have fallen.

The strongest candidate for explaining the slow labour-productivity growth in retailing is the deregulation of trading hours. While deregulation has occurred at different rates in different States, stores in all states are now open for longer hours than was the case in 1980. For the 30 years following World War II, shopping hours were heavily regulated with most shops opening for around 48 hours per week. By the end of the 1970s, all States had introduced late night shopping on one or two nights per week. This had the effect of increasing average opening hours a little, to just over 50 hours per week in 1980. Nevertheless, opening hours still remained heavily regulated.

In New South Wales, the regulations began to be eroded in late 1984 when shops were permitted to trade on Friday nights and Saturday afternoons. The success of these longer hours saw increasing pressure for further deregulation, which finally came in 1989 when unrestricted trading hours on Monday to Saturday were introduced. While regulations concerning Sunday trading remain, they are in large part ineffective. Deregulation has also occurred in other states, but in most cases the process has been slower. For example, in Victoria and South Australia, the extension of Saturday trading to 5:00 p.m. took place in 1987.

Kiel and Haberkern (1994) estimate that deregulation has led to an increase in average opening hours in Australia from 52 hours per week in the early 1980s, to 56 hours in 1986 and to 61 in 1992. In New South Wales and the ACT, where deregulation has been more extensive, they estimate that shops were open for an average of 66 hours in 1992 – 15 hours a week more than in 1980. These changes represent an increase of almost 30 per cent in the hours that the average retail store is open in NSW, and an increase of almost 20 per cent Australia wide.

While unpublished data from the ABS show a slightly smaller increase in average shopping hours, the increase is nevertheless significant; the data show that in 1992, the average retail store was open 57 hours. On average, supermarkets were open 75 hours per week; a 12 per cent increase since 1986. Since 1992, hours of operation of many supermarkets have been extended further with a number of supermarkets now open 24 hours a day. With the exception of household appliance stores, all categories of stores recorded longer shopping hours in the 1991/92 Retail Census than in the 1985/86 Retail Census.

Longer shopping hours have increased shopper convenience and thus have led to an increase in the broad concept of output of the retail sector. However, it is unlikely that they have had any substantial effect on the standard measure of output, as opening stores for longer hours is unlikely to change savings-consumption decisions. On the other hand, it would be surprising if an increase of nearly 20 per cent in average opening hours did not require an increase in the number of hours worked. As a result, longer shopping hours imply a reduction in the average level of labour productivity. Longer hours may have also contributed to the negative compositional effects. A good example is bread shops where deregulation of baking hours has led to a proliferation of specialist bread shops which tend to have low levels of labour productivity. Working in the opposite direction is the idea that deregulation of shopping hours has encouraged the move to larger stores with higher levels of labour productivity.[20]

The ‘shopping-hours’ explanation is supported by state-based data on turnover per person employed in food stores. The data suggest that labour-productivity growth was slower in those states that undertook the most extensive deregulation of shopping hours. The shopping hours explanation is further supported by the fact that the largest declines in labour productivity in the retail sector took place at the same time that employment was expanding rapidly. Further support is suggested by the work of Baily and Gordon (1988) who provide a ‘back-of-the-envelope calculation’ of the effect of regulation of shopping hours on labour productivity. They estimate that if German shopping-hours regulations were applied to the United States, total GDP might be 5 per cent higher (assuming that the workers ‘released’ from the retail sector are employed elsewhere in the economy). While this estimate is almost surely on the high side, Baily and Gordon argue that, at least conceptually, making a reasonable adjustment for the output of ‘convenience’ might offset all of the productivity slowdown in US retailing. In reality, however, the extension of shopping hours in the US considerably predates the productivity slowdown. This is not the case in Australia.

If this interpretation of events is correct, then measured productivity in the sector would be increased by re-regulation of shopping hours. This would, however, be an absurd reaction. Once shopping hours were deregulated, many retailers moved quickly to extend their hours of operation. Such a move was in response to a clear public demand for greater flexibility in the timing of shopping.[21] This raises an important issue. Normally slower productivity growth is associated with a slower rate of increase in conventionally measured living standards. However in this case, the link between growth and welfare has been weakened. While an evaluation of the benefits that households received from longer shopping hours is beyond the scope of this paper, revealed preference says that these benefits may be quite significant. By not making an adjustment to the quality of the output of the retail sector, the statistician may have underestimated output and overestimated the price of that output. On the positive side, given that hours deregulation has already occurred on a widespread scale, the magnitude of this measurement problem should not grow any further.

3.2 The Recreation, Personal and Other Services Industry

The recreation, personal and other services sector plays only a small role in explaining changes in aggregate labour-productivity growth, but the large fall in the level of labour productivity makes it an interesting case for study. In 1992, the level of labour productivity in the industry was only about 80 per cent of the level in 1978. As can be seen from Figure 11, the bulk of this fall occurred between 1984 and 1991. This was also a period of very rapid employment growth. Over this seven year period, total hours worked in the sector increased by 40 per cent.

Figure 11: Labour Productivity in Recreation, Personal and Other Services
(output per hour worked, March quarter 1978 = 100)
Figure 11: Labour Productivity in Recreation, Personal and Other Services

Using the ASIC data, the industry has three main sub-sectors:

  • entertainment and recreation services;
  • restaurants, hotels and accommodation services (the ‘hospitality’ sub-sector); and
  • personal and other services.

In 1993, the hospitality industry accounted for 54.8 per cent of total industry employment (up 3 percentage points over the past ten years) while entertainment and recreation services and personal and other services accounted for 23.4 and 21.9 per cent respectively (up 0.2 and down 3.1 percentage points respectively). Separate output data are not published for all three sub-sectors, but output and hours worked data are available for similar ANZSIC categories since the mid 1980s. Figure 12 shows indices of labour productivity for the relevant categories. In both cultural and recreational services and accommodation, cafes and restaurants, the level of labour productivity fell significantly through the second half of the 1980s.

Figure 12: Labour Productivity in the Recreation Sector
(March quarter 1985 = 100)
Figure 12: Labour Productivity in the Recreation Sector

In assessing the industry's overall performance, developments in the hospitality sector are particularly important, as this sector accounts for over half of the industry's employment. The hospitality sector (using the ASIC classification) has four principal components: accommodation; pubs, bars and taverns; cafes and restaurants; and licensed clubs. An indication of the importance of these various sub-sectors and the relative levels of labour productivity can be obtained from the ABS's 1991/92 Survey of Hospitality Industries. Similar surveys were also conducted in 1979/80 and 1986/87.

Table 4 shows the share of part-time and full-time hospitality employment accounted for by each of these sub-sectors in 1986/87 and 1991/92. It also shows nominal gross industry product per full-time equivalent worker. Unfortunately, deflators for the different sub-sectors are not published.

Table 4: The Hospitality Industry
  Share of full-time
Share of part-time
Productivity level
($ per full-time
equivalent worker)
1986/87 1991/92 1986/87 1991/92 1986/87 1991/92
Accommodation 25.6 30.8 15.8 16.2 24,882 28,412
Pubs, bars and taverns 26.0 18.7 32.0 25.2 22,543 29,146
Cafes and restaurants 30.7 32.7 31.6 39.8 15,616 20,108
Licensed clubs 17.7 17.9 20.7 18.8 29,284 35,081
Total 100.0 100.0 100.0 100.0 22,440 26,830

Notes: (a) Full-time equivalent employment is calculated as full-time employment plus half of part-time employment.
(b) The measure of output is Gross Industry Product.

Source: ABS Cat. No. 8674.0.

In terms of employment, the accommodation and restaurant sub-sectors have clearly been the fastest growing. Cafes and restaurants have increased their share of part-time employment in the hospitality industry from 31.7 per cent to 39.8 per cent and their share of full-time employment by two percentage points to 32.7 per cent. Between 1980 and 1992, the number of businesses operating cafes and restaurants increased by 73 per cent to 8,741 while employment rose by 202 per cent. The accommodation sector also enjoyed strong employment growth, with the number of jobs increasing 92 per cent over this 12-year period. Table 4 suggests that many of these jobs were full-time jobs, with accommodation's share of total full-time employment in the hospitality sector increasing from 25.6 to 30.8 per cent over the period from June 1987 to June 1992.

There are quite large differences in the level of productivity between the various parts of the hospitality sub-sector. Licensed clubs have the highest level of labour productivity, with cafes and restaurants having the lowest level. Using the data for 1991/92, the level of labour productivity in restaurants and cafes is equal to three-quarters of the level of labour productivity for the hospitality industry and 57 per cent of the level of labour productivity in licensed clubs. As a whole, the level of labour productivity in the hospitality sector is lower than the average for the entire recreation, personal and other services sector.

At this level of aggregation, differences in productivity levels, together with changes in the structure of employment appear to have led to relatively large compositional effects on the sector's productivity performance. Within the hospitality industry, the sub-sector with the fastest employment growth has been the lowest labour productivity sector. If the employment shares that existed in 1991/92 had applied in 1986/87, then the level of labour productivity in the hospitality industry in 1986/87 would have been around 2½ per cent lower than was actually the case. While this calculation only provides a rough estimate of the impact of compositional effects, it does suggest that, at least in this component of the industry, these effects may be important.

While the growth of relatively low-labour-productivity industries may have contributed to low productivity growth for the industry as a whole, it is again unlikely that compositional effects provide the full explanation. In a number of the industry's sub-sectors, the available data suggest that there has been relatively limited productivity growth. The data in Table 4 suggest that nominal labour productivity in cafes and restaurants increased by 28.8 per cent over the period 1986/87 to 1991/92. Over this same time period the ‘meals-out’ component of the CPI increased by 30.2 per cent. Similarly, nominal productivity in pubs, bars and taverns increased by 29.3 per cent, while the prices of alcoholic beverages increased by 34.1 per cent.

Low productivity growth of these sectors is also suggested by a comparison of the results from the 1986/87 Services Industry Survey and the 1979/80 Census of Retail and Selected Services Industry. These surveys provide estimates of real turnover (in 1986/87 prices) per employee. These estimates are shown in Table 5 as a percentage of turnover-per-employee in the cafes and restaurant sub-sector in 1979/80.

Table 5: Turnover Per Employee
  1979/80 1986/87
Accommodation 116.2 123.8
Hotels (drinking places) 190.8 178.4
Cafes and restaurants 100.0 91.8
Licensed clubs 163.1 154.3
Hairdressers and beauty salons 60.6 61.0
Laundries and dry cleaners 91.8 97.9
Motion picture theatres 148.8 175.7

Note: Real turnover (in 1986/87 prices) per employee is given as a percentage of real turnover per employee in cafes and restaurants in 1979/80.

Source: ABS Cat. No. 8650.0.

In cafes and restaurants, real turnover per employee fell by 8.2 per cent over the seven years to 1987. This fall is partly due to an increase in part-time employment, but turnover per full-time equivalent worker has also declined. A similar picture emerges for hotels and licensed clubs. Productivity improvements in hairdressing also appear to be extremely small, with real turnover per employee barely increasing over the seven years. Laundries and drycleaners, and especially movie theatres, appear to have done considerably better.

The evidence suggests that many of the sub-sectors of the industry have not experienced significant improvements in productivity. In large part, this reflects the highly labour-intensive and service-oriented nature of these sub-sectors – it is difficult to make hairdressing a more capital-intensive activity and it is difficult to substitute capital equipment for waiters in a restaurant. Perhaps, the greatest scope for ‘technical progress’ in many of these industries is the provision of better quality service with the same labour input. As in the case of retailing, such improvements are difficult to capture in the measure of output used in the national accounts.

While the above data do not provide a complete picture, they do suggest that the declining level of labour productivity in the recreation, personal and other services sector is the result of both compositional effects and stagnant or declining average productivity levels in a number of important sub-sectors. This has occurred at the same time that employment in the sector has grown rapidly – in fact, the periods of most rapid decline in labour productivity have coincided with the periods of most rapid employment growth. One explanation for this combination of low productivity growth and strong employment growth is a decline in the sector's real product wage; this allowed firms to employ more workers even though the new workers were reducing the average level of labour productivity in the industry. We return to this issue in Section 4.

3.3 The Construction Industry

Between March 1983 and June 1991, output per hour worked in the construction industry declined at a rate of almost 1½ per cent per year. This deterioration followed a long period of relatively solid productivity growth and occurred at the same time that employment and output were growing particularly rapidly (see Figure 13). While the decline in productivity is apparent for both private and public construction, it is more pronounced in public construction; between March 1983 and June 1991, output per person in private sector construction fell by 5.4 per cent, while in the public sector, labour productivity fell by 17.7 per cent.

Figure 13: The Construction Industry: Output, Employment and Labour Productivity
(March quarter 1978 = 100)
Figure 13: The Construction Industry: Output, Employment and Labour Productivity

The fall in productivity in public sector construction can be explained, in part, by compositional effects. After 1984/85 there was a sharp decline in public sector gross fixed capital expenditure, with the fall in State government expenditure in the electricity sector being particularly pronounced (see Australian Treasury (1994)). While we know of no data that provide estimates of value-added for various sub-sectors of government construction, it is reasonable to assume that value-added in electricity construction was relatively high, so that the decline in this type of construction made a direct negative contribution to productivity growth (this is, of course, not a bad thing). While compositional effects are important for government construction, they are not the complete answer to the declining productivity in the construction sector, as the level of productivity in the private sector also fell.

The extent of the deterioration in private sector construction is surprising and is at odds with data obtained from the Construction Industry Surveys. A comparison of results from the 1984/85 and 1988/89 surveys shows that over this period, value added per employee in the private sector increased by around 16 per cent, with increases experienced in most sub-sectors (see Table 6). The survey data also suggest that compositional effects within the private sector play only a small role in explaining the outcomes. While the share of the non-building construction sector in total employment fell, the share in another high productivity sub-sector – nonresidential building – rose. The biggest fall was in the employment share of residential building construction which has a relatively low level of productivity. In summary, if the 1984/85 employment weights are applied to the 1988/89 productivity data, the level of labour productivity in 1988/89 would have been higher by slightly more than 1 per cent.

Table 6: Private Sector Construction
Type of Construction Share of value of added Share of Employment Value added/employment
  $'000   % change
  1984/85 1988/89 1984/85 1988/89 1984/85 1988/89  
Residential 15.8 13.8 16.1 12.9 29.3 30.0 2.4
Non-residential 15.3 19.2 9.8 12.0 37.9 45.2 19.3
Total building 31.1 33.0 25.9 24.9 29.2 37.3 27.9
Road and bridge 6.5 5.8 3.7 3.0 42.8 53.7 25.5
Other non-building 10.1 7.5 7.0 6.3 35.2 33.5 −4.8
Total non-building 16.7 13.3 10.7 9.4 37.9 40.1 5.8
Trades 52.2 53.7 63.4 65.9 20.0 23.0 15.0
Total 100.0 100.0 100.0 100.0 24.3 28.2 16.0

Note: Value added is at average 1984/85 prices.

Source: ABS Cat. No. 8771.0.

In contrast to the survey data, the national accounts shows a decline in private sector labour productivity of 2.3 per cent between 1984/85 and 1988/89. This difference stems largely from different rates of increase in output, rather than labour input. To obtain quarterly estimates of private-sector construction output, the statistician extrapolates base-year output by various components of gross fixed capital expenditure on construction by the private sector. This approach assumes that input prices increase at the same rate as output prices. This approach also means that any biases in the deflators for gross fixed capital expenditure on construction will also affect the measurement of output of the construction sector. It is difficult to determine to what extent biases exist. Certainly studies in the United States suggest that the deflators for residential construction are biased upwards due to the inadequate measurement of improvement of building quality. Baily and Gordon (1988) estimate that in the US, the residential construction deflator has been upward biased by at least 1 per cent per year for the past 30 years. This is reflected in a pronounced trend decline in productivity in the construction sector. However, it is unlikely that this provides an explanation for the Australian data, since it would require that the systematic mismeasurement of quality became worse in the mid 1980s. There is no obvious reason to believe that this was the case.

There is nevertheless, a question mark over the residential construction deflator, especially in the second half of the 1980s. Historically, this deflator has increased at roughly the same rate as the price index for materials used in housing construction; over the 16 years to September 1986, the dwelling deflator increased at an average rate of 9.6 per cent per year, while the input price index increased at an average rate of 9.5 per cent. In contrast, over the three years to September 1989, the output deflator increased more rapidly than the input price index; 11.3 per cent compared with 8.7 per cent. The rise in the dwelling deflator also exceeded the rise in the non-residential building deflator and the engineering construction deflator (7.5 per cent and 6.1 per cent per annum respectively) and the rise in the household repairs and maintenance component of the CPI (6.0 per cent per annum).

Why did the dwelling deflator increase so quickly? One clue lies in the close correspondence between the dwelling deflator and the Australia-wide price series for new project homes. This series is available from June 1986, and over that period it tracks movements in the deflator quite closely. The project home series is derived from surveys of project home builders, with the ABS asking respondents to exclude land costs. Despite this, there is some circumstantial evidence that the price series may be capturing something other than construction costs. The State-based data show unusually large differences in price increases across States; over the five years following June 1986, the price of a project home rose by more than 60 per cent in Sydney and Brisbane, but only by a little more than 20 per cent in Adelaide and by just over 30 per cent in Perth and Hobart. Such differences are striking, especially when the inputs that go into a house are traded in an essentially national market, and their prices increased at very similar rates in all states. In Sydney – which experienced the country's fastest increases in project home prices in the late 1980s – the pattern of prices changes corresponds quite closely to the general increase in established house prices.

While the builders of project homes may have substantial pricing power in individual markets, the size of the increase in prices, and the variation across States, at least raises the possibility that the dwelling deflator may have been overestimated in the late 1980s. A more comprehensive analysis, however, awaits further research. In any case, even if the dwelling deflator had increased at the same rate as input prices, this would still not fully explain the decline in labour productivity in private-sector construction. A further possible explanation for the slowdown is that the rapid employment and output growth in the construction industry was associated with a reduction in the average skill levels of workers in the industry and to less efficient construction practices. This explanation, however, still has to confront the evidence from the Construction Industry Survey which shows solid productivity growth.


Between 1966/67 and 1978/79 output per person employed in the industry increased at around 1.6 per cent per year. [12]

One source is OECD (1992). This report examines the change in the structure of the distribution industry in seven OECD countries (Australia is not included). [13]

A series of reports published by the OECD suggest that, in a range of countries, labour-productivity growth in the distributive trades was between 1 and 2 per cent in the 1980s (see Betancourt (1993), Dawson (1993), Lachner, Tager and Weitzel (1993), Maruyama (1993), Messerlin (1993), Pellegrini and Cardani (1993) and Wibe(1993)). For the United States, both the Bureau of Labour Studies and the Bureau of Economic Activity estimate that output per hour worked in retailing increased by more than 1 per cent per year between 1977 and 1986. However, over this period labour productivity in food stores fell by 7 per cent (see Baily and Gordon (1988)). [14]

Oi (1992) argues that retail firms also engage in repackaging and supply ancillary services such as delivery and credit. [15]

The 1991/92 census, which is known as the Retail and Services Census, also included certain service industries. In the present paper, this census is simply referred to as the Retail Census. [16]

In part, this difference may reflect the fact that the 1991/92 Wholesale Industry Survey sought data on ‘management units’, while the earlier survey sought data on ‘establishments’. In the latter survey, businesses whose primary activity was not wholesaling were excluded. [17]

In part, this growth stems from the deregulation of baking hours in the mid 1980s. [18]

The fall is less pronounced when data from the 1984/85 base-year national accounts are used. The ABS attributes the difference in the 1984/85 and 1989/90 base-year accounts to a change in the method used to calculate value added of truck retailers (see ABS Cat. No. 5243.0). [19]

See Morrison and Newman (1983) for Canadian evidence that shopping hours regulation favours small stores. [20]

Not everybody perceived this public demand. In 1983, in one of many reports into the regulation of shopping hours, Justice Macken wrote: ‘It is blind fantasy to assume that in the Sydney Metropolitan area there are a significant number of people eager to shop on Saturday afternoons or on Sundays with respect to goods available to them through the week’ (Macken 1983, p. 53 as cited in Kiel and Haberkern (1994)). [21]