How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey

Size: px
Start display at page:

Download "How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey"

Transcription

1 Australian Perry & Wilson: Journal of The Labour Accord Economics, and Strikes Vol. 7, No. 3, September 2004, pp How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey Jane Harrison Economics, Murdoch University Abstract This paper provides evidence on the degree of gender segregation in Australian workplaces using the Australian Workplace Industrial Relations Survey. The paper finds that the extent of horizontal workplace gender segregation fell significantly over the 1990 to 1995 period. In terms of vertical segregation, the paper examines the degree to which internal labour market structures apply to workplaces and considers the way in which women and men are positioned within workplace hierarchies and their potentially differentiated experience in respect to movements up career ladders, promotions, and the attainment of managerial positions. The paper finds that the internal labour market structure operates in Australia but at a weaker level than previously assumed. It operates much more strongly for men than for women. 1. Introduction Gender segregation studies seek to measure the extent of unevenness in the distribution of economic and social outcomes between women and men. In Australia, as elsewhere, the focus of attention has been on trends in horizontal occupational gender segregation the extent to which women and men are unevenly distributed between occupations. 1 There are two reasons for this emphasis on horizontal occupational gender segregation. The first and most obvious reason is that occupational titles convey significant information on labour market outcomes such as wages, social standing, authority and responsibility. The second reason for the relatively large number of occupational gender segregation studies in Australia is a more practical one; namely, the ready availability over a long period of time of occupational data on a gender basis (censuses, labour market surveys of one kind or another). As compared with the case of occupations, there has been a paucity of analyses related to gender segregation between workplaces (horizontal workplace gender segregation) and the hierarchy of jobs within workplaces Address for Correspondence: Jane Harrison, Economics, Murdoch University, Murdoch WA 6150, Australia, Tel: , Fax: , j_harris@murdoch.edu.au. I would like to thank Gavin Wood, Alison Preston and anonymous referees for their helpful comments on previous drafts of this paper. 1 For examples of Australian studies see, Karmel and Maclachlan (1988), Kidd (1993), Lewis (1982), (1996), Lewis and Shorten (1991), Miller (1994), Miller and Volker (1985), Power (1975), Rimmer (1991), Vella (1993), Watts and Rich (1991, 1992). The Centre for Labour Market Research, 2004

2 330 Australian Journal of Labour Economics, September 2004 (vertical workplace gender segregation). Yet such studies are potentially of significant value if we are to understand the position of women in the labour market. Workplaces differ in terms of the tasks, responsibilities and wages they offer and assign employees from the same occupational background (e.g., clerks in one workplace may be treated differently from clerks in another workplace). Ignoring workplace effects, therefore, results in a loss of explanatory power in understanding the variation of labour market outcomes. Moreover, gender-crowding effects may be more evident at the workplace level than at any other labour market level. Workplaces with a high proportion of women may offer lower wages to their female-dominated workforce than similar workplaces with a more even gender distribution of workers. Finally, it seems clear that the analysis of vertical segregation and hierarchy is inherently a workplace matter. How else are we to determine the degree of authority and power an individual has in the labour market other than in the context of their position in the hierarchy of the workplace in which they work? The well-known glass ceiling effect, which hypothesises that women face barriers in breaking into more senior positions in job hierarchies, can only be understood in the context of the workplace. So far, the absence of large-scale unit record matched workplace-employee datasets has impeded the study of workplace gender segregation in Australia. However, the recent availability of workplace-based data in the Australian Workplace Industrial Relations Survey (AWIRS) has opened up new opportunities for research into gender segregation in the Australian labour market. 2 Drawing largely on the second or 1995 wave of AWIRS (referred to in this paper as AWIRS95) and the matched employee-employer file contained in this dataset, this paper broadens the focus of gender segregation studies through an investigation of the degree and structure of horizontal and vertical workplace gender segregation. The paper is organised as follows. Section 2 reviews the relatively brief workplace segregation literature and discusses the related field of the operation of internal labour markets. Internal labour market theory provides a useful springboard to the examination of vertical workplace segregation. In section 3, we present, for the first time for Australia, estimates of horizontal workplace gender segregation and compare trends (albeit only for the two waves of AWIRS) in horizontal workplace segregation with similar estimates for horizontal occupational gender segregation. Is the degree of horizontal workplace segregation higher or lower than the degree of horizontal occupational segregation. Do trends in the two series move in the same direction in the 1990s? In this section, we also model the gender share of workplace employment to understand why it is that some workplaces have a much higher proportion of female employees than the national average while, in other workplaces, women represent a much smaller share of workplace employment. The main point of focus of the paper is that of vertical workplace segregation, which is examined in section 4 of the paper. Here we consider the way in 2 For an example of the ways in which the AWIRS data can be used to examine the position of women in the labour market see Pocock and Alexander (1999) who utilise the AWIRS data to examine the gender wage gap.

3 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 331 which women and men are positioned within workplace hierarchies and to their potentially differentiated experience in respect to movements up career ladders, promotions, and the attainment of managerial positions in workplaces. A key objective in this section is to extend international studies of segregation and internal labour market models to the Australian context (see Baker, Gibbs and Holmstrom, 1993, 1994a, 1994b; Pudney and Shields, 2000a, 2000b; and Treble, van Gameren, Bridges and Barmby, 2001). Much of the international literature on vertical workplace segregation is built on a case study approach. The importance of the AWIRS dataset is that it allows for the examination of this issue in the context of a large set of workplaces. Section 5 concludes the paper. 2. Workplaces and Internal Labour Markets Workplace Effects The earliest recognisable gender segregation studies, which present results on a workplace basis, are McNulty (1967) and Buckley (1971). Utilising U.S. Bureau of Labor Statistics data for , and 1970 respectively, they found that occupational wage averages were higher for women in workplaces with a mix of women and men than in workplaces employing women only in the relevant occupations (a gender crowding effect). Blau s Equal Pay in the Office study published in 1977 found that men were more likely to be employed in high-pay workplaces and women in low pay workplaces. Furthermore, the more segregated the workplace, the higher the wage paid to men and the lower the wage to women (all other things being equal). Conversely, the more integrated the workplace the higher the wage paid to women and the lower the wage paid to men. In a series of papers in the mid-1980s, Bielby and Baron (e.g. Bielby and Baron, 1986a, and 1986b) emphasised the importance of the workplace to an understanding of women s position in the labour market. They found that men and women in the same occupational category were often assigned different job titles within workplaces with consequent implications for pay and status. Moreover, Bielby and Baron found that some workplaces primarily employed men (of a given occupation) while others primarily employed women (of the same occupation) leading to higher levels of gender segregation than would otherwise be expected. As with Blau (1977), Bielby and Baron found that organisations, which principally employ men, provide high wages and better conditions than similar workplaces that primarily employ women. Carrington and Troske s (1995) study focuses on small employers. Their main point of interest is in the issue of horizontal firm gender segregation and the implications of horizontal firm gender segregation for the gender wage gap. Consistent with Bielby and Baron s work they suggest that the level of gender segregation in small businesses is higher than comparable estimates of the degree of horizontal occupational gender segregation. Their key findings include the fact that male business owners employ fewer women than do similar women business owners, larger businesses tend to employ more women than smaller firms, and low paying firms tended to employ more women than higher paying firms. Across all industries,

4 332 Australian Journal of Labour Economics, September 2004 horizontal firm gender segregation was found to account for 55 per cent of the total gender pay gap. Firms with an over representation of men tend to pay more than firms with an over representation of women. However, a great deal of variation is found among different industries with regard to the explanatory power of the gender pay gap. 3 Millward and Woodland s (1995) study of workplace gender segregation in the United Kingdom, using the Workplace Industrial Relations Survey, indicated horizontal workplace gender segregation was also widespread among British workplaces. Craig, Garnsey and Rubery (1985) drew similar conclusions based on their analysis of U.K. workplaces in four manufacturing and two service industries. They also concluded that there was significant levels of gender segregation within workplaces based on the feminisation of jobs. The assignment of women to particular job titles resulted in these jobs titles being more likely to be specified as being low skilled and consequently low paid. This occurred irrespective of the actual skills content of the work performed. Internal Labour Markets and Hierarchies Much of the focus of this paper is on the structure of hierarchies within Australian workplaces and the position of women within those hierarchies. The long-standing literature on the operation of internal labour markets and a more recent set of studies on hierarchy provide a useful organising framework for our analysis. This is inspite of the fact that much of the original internal labour market literature did not emphasise a gender dimension. An internal labour market exists when long-term relationships prevail between employees and organisations, recruitment occurs largely at lower ports of entry, and there is a corresponding high rate of internal promotion as compared with external recruitment to positions higher up the career ladder (Doeringer and Piore, 1971; and Osterman, 1984). Wachter and Wright (1990) have characterised the internal labour market as reflecting the outcome of joint rational behaviour on the part of organisations and workers in an environment in which long-term, match-specific investments between establishments and employees (such as on-the-job training) generate surpluses for the firm. The internal labour market structure is also more likely to develop when high transaction costs militate against match-specific investments being made between independent contractors and firms. Tournament theory provides an alternative perspective on the existence and nature of internal labour markets within firms. Under tournament theory, the spread in pay between job titles in the workplace provides the opportunity of future prizes to employees and hence imparts positive incentive and productivity effects for current employees (e.g., Bognanno, 2001; Bronars and Famulari, 1997; Drago, Garvey and Turnbull, 1996; and Lazear and Rosen, 1981). The greater the wage spread, the greater are the incentives created to gain a promotion. When firms recruit from outside the workplace, the incentive effects of the tournament for existing employees are muted. The internal labour market structure, which emphasises internal 3 For further relevant U.S. studies see Groshen (1991a, 1991b) and Carrington and Troske (1998).

5 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 333 promotion up job ladders, is thought to ensure the greatest realisation of tournament incentive benefits. In subsequent sections of the paper we shall examine the role and nature of the operation of internal labour markets in Australia and the position of women within that structure but we first turn to an analysis of the extent of horizontal workplace gender segregation in Australia. 3. Horizontal Workplace Gender Segregation Segregation indices are commonly employed to study trends in the level of horizontal occupational gender segregation in the labour market. We now employ these same indices to estimate the degree of workplace gender segregation in Australia using the AWIRS data. Table 1 Workplace Segregation Measures, AWIRS90, AWIRS95 AWIRS 90 AWIRS 95 AWIRS 90 AWIRS 95 Main Main Main Main Survey Survey Survey Survey Duncan and Duncan Index Karmel and McLaughlin Index Marginal Matching Index Gini Index Atkinson Index: delta= delta= delta= Table 2 Occupational Segregation Measures, 1991 and 1996 Censuses, and AWIRS95 AWIRS Main Employee Census Census Survey Survey Major Group Level Duncan and Duncan Index Karmel and McLaughlin Index Marginal Matching Index Gini Index Atkinson Index: delta= delta= delta= Unit Group Level Duncan and Duncan Index na Karmel and McLaughlin Index na Marginal Matching Index na Gini Index na Atkinson Index: delta= na delta= na delta= na 0.666

6 334 Australian Journal of Labour Economics, September 2004 Two waves of the AWIRS have been conducted. The first wave was based on interviews conducted in workplaces between October 1989 and May 1990 and is referred to throughout as AWIRS90. The second wave of AWIRS was conducted in 1995 and is referred to as AWIRS95. Both AWIRS90 and AWIRS95 collect data at the workplace unit of analysis. It is important to emphasise the fact that AWIRS is a workplace-based dataset, and not an organisational-based dataset, as the results obtained at the workplace level may not translate strictly to the organisaltional level. For the purposes of AWIRS, a workplace is defined as a single physical area from which the operations of an enterprise or organisation are conducted. Enterprises or organisations can, of course, include more than one workplace. If they do, the workplace is said to be part of a multi-workplace enterprise. Threequarters of all workplaces in AWIRS95 are multi-workplace organisations while one-quarter are single-workplace organisations. The AWIRS dataset contain a number of survey components. What is termed the Main Survey incorporates data drawn from a set of questionnaires put to general and employee relations managers and union delegates at workplaces. Workplaces included in the Main Survey had 20 or more employees at the time of sample selection. In addition to the Main Survey, AWIRS95 is made up of two additional components; an Employee Survey, and a Panel Survey. The Employee Survey is based on a questionnaire provided to a sample of employees at the same workplaces surveyed in the AWIRS95 Main Survey. The results from this survey can be matched to the workplace-based survey to provide matched workplace-employee data. Respondent workplaces to the Panel Survey are those workplaces that were sampled as part of AWIRS90. Tables 1 and 2 provide estimates of trends in horizontal workplace and occupational gender segregation in Australia over the 1990s based on both the AWIRS data and Census data (for one set of the occupational comparisons) across a standard set of indices used in the segregation literature. The estimates of horizontal gender segregation are the first to be presented for Australia. The horizontal gender segregation estimates based on AWIRS90 and AWIRS95 contained in tables 1 and 2 reveal a number of things. First, they show a relatively high degree of horizontal workplace gender segregation. For example, the Karmel and McLauglin (1988) Index value for workplaces in the AWIRS95 Main Survey was (0.227 for workplaces in the 1995 Panel Survey). This estimate significantly exceeds comparable estimates for horizontal occupational segregation at the major group occupational level for both the 1996 Census (0.185) and AWIRS95 Main Survey (0.175). The horizontal workplace gender segregation estimates are, however, somewhat lower than estimates of occupational segregation at the unit group level of occupations (0.267 for the 1996 Census data and for the 1995 AWIRS Employee Survey). Workplaces combine (unit group) occupational classifications with different gender compositions together. It may have been expected that the extent of horizontal workplace segregation would, therefore, be closer to the major group occupational segregation estimates than the minor group level estimates, but the opposite is the case. Workplaces appear more gender segregated than one might have expected.

7 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 335 Second, the segregation estimates suggest that the degree of horizontal workplace gender segregation has fallen over the 1990 to 1995 period. Each of the segregation indices registers a decline in horizontal workplace gender segregation and this decline is evident for both the AWIRS Main Survey workplaces and the Panel Survey set of workplaces. This trend of declining gender segregation in the early to mid 1990s is also evident from Census data for occupations at the more disaggregated unit group occupational level. In the case of the major occupational categories, the majority of segregation indices actually reveal a slight rise in the degree of gender segregation. Modelling the Female Share of Workplace Employment What determines the gender composition of workplaces? If only occupation matters then the gender composition of workplaces will be explained by the underlying occupational structure of workplaces. However, if a broad range of determinants matter, then a statistical analysis of the female share of employment will reveal a large range of other significant effects. Our starting hypothesis is exactly that; in addition to occupational structure a large range of factors explain the gender composition of workplaces. If this is confirmed, it underlines the need to move beyond a focus on occupations in terms of the examination of gender segregation. In this section, we estimate a model of the female share of workplaces based on the AWIRS95 data. One obvious additional factor explaining the gender segregation of workplaces is the industry location of the workplace. The industry location of the workplace can be important in explaining female employment shares for a number of reasons. First, historical patterns of development have assigned particular industries as either male or female dominated. In sociological explanations of gender segregation, historical patterns are important in affecting socialisation processes and the choices made as to preferred industry location. Mining and construction, for example, are seen as being male dominated industries. Irrespective of occupational categories within these industries, women may have less of a preference to work in these industries as compared to those which have been seen as more female dominated (e.g., health and community services, cultural and recreation services, and personal and other services). Second, the degree of security of job tenure, choice over working hours, the average level and dispersion of wages, the wage discount for skill atrophy, and the environmental conditions of work may all differ by industry. The existence of systematic gender differences on these indicators would lead to differences in the workplace female shares across industries irrespective of the occupational pattern of employment in those workplaces. The third set of independent variables relate to the type of market the workplace operates in and the ownership structure of the workplace. Two dimensions of market structure are specified: Openness (whether the workplace is wholly domestically focussed or has greater exposure to the world); and self-assessed degree of competition faced by the workplace. A fourth set of independent variables relate to production characteristics of the workplace. We incorporate production characteristics to pick up on both

8 336 Australian Journal of Labour Economics, September 2004 supply side and demand side forces. On the supply side women may have a greater aversion to workplaces with round the clock operations as such hours of operation will interfere with greater childcare and home commitments. We, therefore, include a set of variables for hours of operation. On the demand side, the relative costs of employee turnover and training are potentially significant variables. Turnover costs refer to the time and money it takes to turn an outsider (a non-employee) into a fully-fledged insider. These costs include training costs together with personnel costs associated with the recruitment process. Workplaces that exhibit high turnover costs are, all other things being equal, less likely to employ workers who have a higher propensity for quits, as the costs of training will not be fully recouped. Workplaces that have a perception of higher voluntary turnover among women than men may tend to reduce their employment of women irrespective of the (imperfectly known) likely turnover rates for individual women (and men). Measures of turnover time are, therefore, included in our estimated model. These variables reflect management s perceived time in converting an outsider into an established insider. We also incorporate a variable which specifies whether the workplace undertakes direct formal training programs for its employees. Other workplace characteristics recorded in the AWIRS dataset, which are included in the estimated model, are self-assessed autonomy of the workplace (from Head Office), organisational change (whether the workplace has experienced recent organisational change), net employment growth, and self-assessed structured management (whether a workplace has a highly structured management system). Structured management reflects evidence of a range of characteristics from employee monitoring, to disciplinary and grievance procedures and health and safety structures. Workplaces which record ordered structures across a majority of structure characteristics are given a value of one for this variable while other workplaces are given a value of zero. Personnel practices, wage structures, and the employee composition of workplace represent the fifth group of explanatory influences. A two-way process is possible in this area. Personnel practices may affect the decision of women and men as to whether they wish to take a position in a given workplace. At the same time, the female composition of a workplace may in turn influence the degree to which these practices are employed. Personnel practices incorporated in the model include maternity leave available, childcare assistance provided, workplace childcare facilities, financial assistance for childcare, equal employment or affirmative action, and employment targets for women. Table 3 presents the tobit Maximum Likelihood results of the AWIRS95 Main Survey models of female share of employment. The tobit model accounts for the censoring of female share from above (100 per cent) and from below (0 per cent). As is evident from the results, the occupational structure of the workplace plays an important part in determining the share of workplace employment taken by women. Not surprisingly, the female share of employment rises with a greater proportion of clerks and falls in workplaces

9 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 337 with a greater proportion of tradespersons. However, there is a clear role for non-occupational categories in explaining female share of employment. Critical here is the role played by industry variables. There exists a highly segregated workforce based on industry categories even after controlling for the occupational structure of these industries. Relative to the default (manufacturing) the female share of employment rises appreciably in health and community services and to a lesser extent in education and finance and insurance. Significant negative effects are found in mining, electricity, construction, wholesale trade, and transport and storage. Table 3 Women s Share of Workplace Employment: 1995 AWIRS Main Survey, Maximum Likelihood Censored Regression Model Marginal Coefficient Effects T-ratio Mean Constant Occupation Professionals Para-professionals Tradespersons Clerks Sales and personal workers Plant and machine operators Labourers Industry Mining Electricity Construction Wholesale trade Retail trade Accommodation-restaurants Transport and storage Communication services Finance and insurance Property and business services Government administration and defence Education Health and community services Cultural and recreation services Personal and other services Market structure/workplace characteristics Log of employment size Domestic market Domestic some export Primarily export Intense competition Strong competition Moderate competition Some competition Predominantly Australian owned Equal Australian foreign owned Predominantly foreign owned Wholly foreign owned Government business hours of operation hours of operation hours of operation hours a day Employee turnover time - 1 week to a month Employee turnover time months Employee turnover time months Employee turnover time - 6 months to a year

10 338 Australian Journal of Labour Economics, September 2004 Table 3 (continued) Women s Share of Workplace Employment: 1995 AWIRS Main Survey, Maximum Likelihood Censored Regression Model Marginal Coefficient Effects T-ratio Mean Employee turnover time - greater than a year Net employment growth Organisational change Formal training in the workplace Structured management Per cent part-time Per cent non-core workers Influence of largest occupation Childcare assistance provided Maternity leave available Workplace childcare facilities Financial assistance for childcare Equal employment or affirmative action Employment targets for women (managerial) Employment targets for women (non-managerial) Sigma Sample Size 1608 Log Likelihood Even controlling for industry and occupation mix, female share is affected by market and workplace production characteristics, personnel characteristics and the industrial relations environment. However, the influence from the remaining set of variables appears weaker than for industry and occupation suggesting that industry and occupation combined account for the major explained variation of female share. Workplaces that operate primarily, but not wholly, in the domestic sector and face limited competition have a higher female share than other workplaces. The female share of the workplace employment is also strongly positively related to part-time employment levels. As anticipated women s share of workplace employment is negatively related to the hours of operation of the workplace (those workplaces operating on a continuous or close to continuous basis having a lower female share). The role of turnover time and training appears weak. 4. Vertical Workplace Segregation: Internal Labour Markets and Hierarchies An analysis shows the extent of vertical workplace gender segregation in Australian workplaces. The AWIRS dataset (the focus is on the richer AWIRS95 data) in this section is particularly well suited to examine vertical segregation in workplaces as it overcomes a key deficiency of many empirical studies examining the role of hierarchies. That deficiency relates to the reliance, in previous studies, on personnel data drawn from a single individual organisation rather than on a range of organisations. The question of whether certain types of workplaces are more likely to exhibit internal labour market features as compared with other workplaces can only be answered with databases that contain a broad range of workplaces.

11 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 339 There is, however, one caveat to the use of the AWIRS95 dataset in examining hierarchies and the operation of internal labour markets. That caveat relates to the fact, noted previously, that the AWIRS95 dataset, as a workplace-unit database, contains workplace-level variables. The only exception occurs when questions are specifically addressed to the organisational level. One reason for making clear this distinction between the workplace and the organisation is that hierarchical rank-orderings of positions in a workplace may not necessarily equate to rank orderings for the organisation in which the workplace is located. Moreover, as discussed previously, models of internal labour markets are conditioned on the role of tenure, the internal labour market model positing that employees move up the internal job ladder over successive years in the one organisation. However, the crucial tenure variable in AWIRS95 is a workplace tenure variable and not an organisation tenure variable. In other words, respondents are asked about their tenure in a workplace and not an organisation. Employees who move up an organisation s job ladder by obtaining a promotion in another workplace within the organisation (e.g., at the organisation s head office) will record zero workplace tenure at the time of the promotion (as will, of course, new entrants to the organisation) despite perhaps having spent a considerable period of time in the organisation. (This assumes, of course, that respondents strictly and consistently answer questions on tenure using a workplace rather than organisational frame and we can t discount the possibility that respondents indeed blur workplaces and organisations in their responses.) The consequence of this is that the proportion of new entrants who are managers will be artificially inflated in multi-workplace organisations, but not in single-workplace organisations, leading to a possible erroneous rejection of the internal labour market model. In the case of single workplace organisations, workplace tenure is organisational tenure and there is no room for possible ambiguity. 4 Because of the importance of the tenure variable to an analysis of the internal labour market model, much of the analysis in this section of the paper will be presented at the single workplace level. However, the multi-workplace results are not without some interest. This is because a comparison between single-workplace organisational and multi-workplace organisational results potentially highlight an interesting but little researched feature of internal labour markets; namely, the role played by intra-organisational mobility for progression through the workplace. If tenure profiles differ between single and multi-workplace organisations, having controlled for other differences between the two types of workplaces, then that difference can validly be attributed to the impact of intra-organisational mobility the role played by movement from one hierarchical position in the organisation to another higher position in another workplace within the organisation. 4 As emphasised above, it is, of course, possible that many respondents to the survey interpreted the question as an organisational tenure question or that the degree of internal organisational mobility within multi-workplaces is in fact relatively low. Single-workplace organisations are somewhat smaller than multi-workplace organisations. Single-workplace organisations have on average around 100 employees in the AWIRS95 dataset as compared to around 150 employees in multiworkplace organisations. There are 14,474 employees in multi-workplace organisations as compared to 2,056 employees in single workplace organisations.

12 340 Australian Journal of Labour Economics, September 2004 Determining the Hierarchy of Positions within Workplaces In a large matched workplace-employee dataset such as AWIRS95, hierarchical structures must be built by the researcher using the available variables in AWIRS95 they are not available from the organisational charts that are often used in studies based on a single organisation. Potential sources of information in AWIRS95 include data from workplace managers on the occupational structure of the workplace, occupational data from the Employee Survey, self-assessed levels of responsibility assigned to the employee, promotion and seniority-based wage rises experienced by employees and, of course, the wage paid to the employee. Our approach is to use a combination of measures of heirachy comprising (1) location in managerial occupational positions ( General Managers, Specialist Managers and Managing Supervisors are the AWIRS95 categories), (2) estimates of within-workplace responsibility and (3) wage rankings of individual employees. Managerial location, responsibility and wage rankorderings can then be evaluated against workplace tenure to assess the applicability of the internal labour market model. The responsibility scale adopted in our analysis is an employee self-assessed scale. The scale is constructed using the following set of questions from the AWIRS95 Employee Survey. In general, how much influence or input do you have about the following? 1. The type of work you do 2. How you do your work 3. When you start and finish work 4. The pace at which you do your job 5. The way the workplace is managed or organised 6. Decisions which affect you at this workplace The respondent options for each of the above questions are: A lot; Some; A little; None. These options are rescaled so that the highest degree of responsibility ( a lot ) is given the highest value of 4 and the lowest ( none ) is given a value of one. The response from each question is then normalised to lie between 0 and 100 using the transformation [[(Actual raw score - lowest possible raw score)/possible raw score range] X 100]. The mean of the transformed values from the six questions is determined so that a unique responsibility value is assigned to each employee in the Employee Survey. The responsibility value is constructed so that the higher the value, the greater the degree of employee control, responsibility or autonomy. To provide a within-workplace assessment of the position of the employee in the workplace, each employee s responsibility value is given a ranking in that workplace. These rankings are then placed in five approximately equal categories (ties in rankings prevent a perfectly proportioned quintile division). The same ranking process is applied to the hourly wages of employees in the Employee Survey to derive hourly wage rankings. This ensures that a

13 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 341 second dimension to the career ladder within the workplace is identified, so ensuring sensitivity to our results on internal labour markets within workplaces. Promotions and External Recruitment A key proposition of the internal labour market model is that the propensity to use external recruitment to fill positions is largely restricted to bottom rungs in the job hierarchy (i.e., to ports of entry). External entry to higherlevel positions in the workplace is limited as an increasingly important role in the assignment of employees to positions is played by internal promotion procedures. A strong positive relationship between tenure and workplace hierarchical location, therefore, is posited in the internal labour market model. The proportion of new entrants who are managers should be well below the proportion of long-career employees who are managers. New entrants should similarly be much more likely to be assigned lower responsibility levels and a lower wage within the workplace than longer career employees. Baker, Gibbs and Holmstrom (1993,1994a, and 1994b), and Treble, van Gameren, Bridges and Barmby (2001), using personnel data, find higher levels of external recruitment to grades in the career ladder above ports of entry than they suggest would be expected under the internal labour market model. Baker, Gibbs and Holmstrom (1993,1994a,1994b) report external hiring rates for a large US company (i.e., the proportion of new hires hired from outside the firm) of between 18 per cent and 31 per cent. External hires at the top level of the career hierarchy ran at 20 per cent. Mean firm tenure at higher managerial grades in the career ladder was estimated at around 11 years; twice the mean tenure figure for the company as a whole and five times the average of mean tenure at the lowest level in the hierarchy. Treble, van Gameren, Bridges and Barmby (2001) find that external recruitment rates above ports of entry for the large British firm examined were somewhat lower than for the Baker, Gibbs and Holmstrom study and hence are more supportive of the internal labour market model. However, Treble, van Gameren, Bridges and Barmby find a U-shaped external profile with higher rates of recruitment at both lower and higher points in the management hierarchy than in the middle. In terms of the position of women in career ladders, much of the research evidence supports the view that women face difficulties at middle and top ranks in the promotion chain (see, for example, Cotter, et al., 2001; Garcia- Crespo, 2001; Jones and Makepeace, 1996; Ward, 2001). McDowell, Singell and Ziliak (2001) suggest, however, that at least in respect to the economics profession such effects were restricted to the pre-1990s period (see also, Pudney and Shields, 2000a and 2000b), while Groot and Maassen van den Brink (1996) emphasise the fact that women are more likely to find themselves in jobs that offer promotion less frequently rather than have less opportunities to gain promotion for equivalent jobs. (e.g., Garcia- Crespo, 2001; Gibbons and Waldman, 1999; Groot and Maassen van den Brink, 1996).

14 342 Australian Journal of Labour Economics, September 2004 There is no separate indication in the AWIRS95 of success in achieving a promotion within the workplace. Employees are, however, asked whether they have received a pay rise because of age, length of service or promotion and we use this as our proxy for promotions within the workplace. AWIRS95 Evidence on Women and Men in Australian Workplace Hierarchies We now turn to our descriptive analyses of the structure of hierarchies and internal labour markets in Australia using AWIRS95 data. Table 4 The Female Managerial Share in Workplaces, AWIRS95 Single Workplace Multi-workplace Proportion of Category who are Women Organisations Organisations Managers Managing supervisors Specialist managers General managers All managers Non-managerial employees Table 5 Workplace Responsibility and Wage Rank Orderings, AWIRS95 Single Workplace Multi-workplace Proportion of Rank-order Category who are Women Organisations Organisations Workplace responsibility Lowest responsibility grade Second lowest Middle Second highest Highest responsibility grade Wages Lowest wage grade Second lowest Middle Second highest Highest wage grade Table 4 provides estimates of the proportion of women who hold down managerial positions in Australian workplaces. Similar estimates for withinworkplace responsibility and wage rank orderings are presented in table 5. Our results are cross-classified according to the single and multi-workplace structure of organisations. As previously emphasised, AWIRS95 is a workplace-based survey and the tenure variable in AWIRS95 is a workplace tenure variable and not an organisation tenure variable. For single workplaces, the two are the same, allowing for a more direct attack on the question of the applicability of the internal labour market model. However, large differences between the single and multi-workplace organisation in terms of the relationship between workplace tenure and the structure of career ladders may highlight the role played by intra-organisational mobility in the career ladder.

15 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 343 The results for table 4 highlight the glass-ceiling effect. Women s share of managerial positions is only half that of its share of non-managerial positions with the greatest disparity evident for the highest managerial classification; namely, the general managerial position. While women comprise close to 50 per cent of non-managerial positions they hold down less than 20 per cent of general managerial jobs. A similar pattern is evident with respect to rank-orderings of responsibility and wages. As indicated in table 5, women are less likely to be located in the top rank-order categories than are men. Women comprise around 40 per cent of employees in the highest responsibility and wage grades (even lower in the case of multi-workplace organisations and wages) but 50 per cent in the lowest responsibility and wage grades. Table 6 Managers, External Hires and Workplace Tenure, AWIRS95 Single Workplace Multi-workplace Organisations Organisations Women Men Women Men Proportion of managers who are new workplace hires Managing supervisors Specialist managers General managers All managers Proportion of non-managerial employees who are new workplace hires Mean workplace tenure Managing supervisors Specialist managers General managers All managers Non-managerial employees Proportion of all employees who are managers in different workplace tenure categories Less than one year One year Two years Three years Four years Five years to 10 years to 15 years to 20 years years All careers In tables 6 and 7 we cross-classify managerial and responsibility/wage rankorderings by tenure and hiring patterns to explore further the question of whether internal labour markets operate in Australia and the nature of women s position within the internal labour market. The internal labour market model posits that employees move up the internal labour market so that managers will be selected from employees with longer tenures and the rate of external hiring at the higher managerial and responsibility levels will be relatively low.

16 344 Australian Journal of Labour Economics, September 2004 Table 7 Workplace Responsibility and Wage Rank Orderings, External Hires and Workplace Tenure, AWIRS95 Single Workplace Multi-workplace Organisations Organisations Women Men Women Men Responsibility rank-order categories Proportion of category who are new workplace hires: Lowest responsibility grade Second lowest Middle Second highest Highest responsibility grade Mean workplace tenure: Lowest responsibility grade Second lowest Middle Second highest Highest responsibility grade Wage rank-order categories Proportion of category who are new workplace hires: Lowest wage grade Second lowest Middle Second highest Highest wage grade Mean workplace tenure: Lowest wage grade Second lowest Middle Second highest Highest wage grade Table 6 presents estimates of the proportion of managers and of nonmanagerial employees who are new workplace hires, the mean tenures of managerial employees, and the workplace tenure profile of managers. In Baker, Gibbs and Holmstrom (1993, 1994a, 1994b), and Treble, van Gameren, Bridges and Barmby (2001), a longitudinal perspective is afforded on the role of external hiring. Hence, they can determine the grade at which employees (both past and present) were recruited to the organisation from outside. 5 This is not possible with AWIRS95 as we only have current tenures for existing employees. We can, however, estimate a number of important relevant measures. These include (a) the proportion of currently employed managers (and nonmanagerial employees) who were hired from outside the workplace in the last year (known from the fact that their workplace tenure is less than one year) and (b) the workplace tenure profile of existing managerial and nonmanagerial employees. The former measure we refer to as the marginal external hiring rate. Clearly, the marginal external hiring rate lies below the average external hiring rate (i.e., the proportion of all those in managerial positions who were, at any time in the past, hired from external sources). 5 Baker, Gibbs and Holmstrom (1993, 1994a, 1994b) do, however, refer to left censorship problems in the determination of workplace tenure for employees.

17 Harrison: How Segregated are Australian Workplaces? Evidence from the Australian Industrial Workplace Relations Survey 345 Despite this, the marginal external hiring rate into managerial posts is very high for women in Australian workplaces as evident from the results presented in table 6. Among women managers in single workplace organisations, some 19.5 per cent were new recruits to the workplace (and the organisation). This represents a particularly high figure and indeed equal to the average external hiring rate (for women and men combined) cited by Baker, Gibbs and Holmstrom (1993, 1994a, 1994b) for hires into the top grade of their firm s career hierarchy. For men, the proportion of managers in single workplace organisations who are new recruits to the managerial ranks is much lower at 5.1 per cent. Compare, however, this estimate for men in single workplace organisations with the comparable estimate for multi-workplace organisations. Here the marginal external hiring rate for men rises to some 14.7 per cent. Perhaps the most obvious conclusion that could be drawn from the dissimilar estimates of the managerial external hiring rate in single versus multiworkplace organisations is that the difference reflects the importance for men (but not for women) of intra-organisational moves into and between managerial positions. In other words, intra-organisational mobility is an important means of advancement up career ladders. This is a movement consistent with internal labour market theory but whose effect has not been documented in previous studies. The movement up career ladders over time within the organisation is muted in the case of women. For women, there is no significant difference, in single workplace organisations, between the external hiring rate for managers (19.5 per cent) as compared with the corresponding rate for non-managerial employees (21.1 per cent). This result is sharply at odds with what we would expect under internal labour market theory (and tournament theory). Under internal labour market theory, new recruits should be much more evident in the non-managerial workforce than in the managerial workforce. For men, this is the case as 22.0 per cent of non-managerial employees are new recruits (in itself a large figure) compared to 5.1 per cent of managers. A similar perspective on the position of women relative to men in internal labour markets and hierarchies is evident from estimates at the bottom of table 6 referring to the proportion of employees in a given workplace tenure category who are managers. For women in single workplace organisations, the proportion who are managers barely rises as we move into the higher tenure grades. This compares with a much steeper profile for men where the proportion of men who are managers in very long tenure categories is over five times the proportion in new entrant categories. Mean workplace tenure estimates for managers in single workplace organisations are above those for non-managerial employees for both women and men. However, the ratio of tenures for these two groups is considerably less than the 2:1 benchmark cited in the Baker, Gibbs and Holmstrom (1993, 1994a, 1994b) studies. Mean workplace tenure estimates for managers are also lower for both men and women in the present study than in the Baker, Gibbs and Holmstrom studies suggesting weaker internal