Family provisions at the workplace and their relationship to absenteeism, retention, and productivity of workers: timely evidence from prior data.

Brandon, Peter D. ; Temple, Jeromey B.

Introduction

Changes in family structure and in the number of mothers working full- or part-time have altered employment patterns, the composition of the work force, and the roles that mothers and fathers perform in families. The new 24-7 economy, the mixing of traditional gender roles, and the feminization of the work force have been noticed by employers. Indeed, a growing number of employers know that family and work demands often conflict and that this conflict can affect the morale, productivity, and retention of workers. The concern that difficulties in juggling family and work responsibilities can negatively affect worker performance has led some employers to provide on-site child-care or add family leave to benefit packages. If such initiatives, employers have assumed, reduce friction between family responsibilities and work demands, then worker productivity should increase and unexcused absenteeism and unnecessary turnover should decrease.

The assumption that workers and employers both benefit when employers provide child care and family leave is appealing but heretofore it lacks empirical verification. Few studies have established that worker performance is related to such family provisions or leave policies. Indeed, few studies have possessed a representative sample of workplaces to investigate the question: Do work-based child-care and family-leave provisions improve worker performance? This study, however, uses a representative sample of Australian workplaces from the early 1990s to test the hypothesis that these provisions are associated with the productivity, absenteeism, and turnover of workers at the workplace. Though these data used here are ageing, they remain one of the only sources of workplace data in the world capable of generating findings that have enormous implications for contemporary Australian workplaces and families. The implications of the findings from these data for workplaces and families, and for future data collection efforts on workplaces and families are discussed thoroughly in the conclusions section of this study.

Background

Due to demographic changes in the work force, the welfarist approach to human-resource management has resurfaced. In the past, benefits to individual workers were emphasised; today, family provisions are the centerpiece of the new corporate welfarism. Although employer-supported child care and family leave characterise contemporary corporate welfarism, most research has focused only on employer-supported child care. Friedman (1986) cites figures suggesting that child care is becoming a major employee benefit. In the late 1980s, over 3,500 major companies in the United States offered some form of child-care support to employees (Edgar, 1988); 775 of these companies supported on-site child-care centers. In fact, Mayfield (1985) and Grant, Sai-Chew, and Natarelli (1982) argue that Canadian employers have supported various forms of child care for 25 years.

The limited scope of existing studies makes generalizations about the extent of employer-supported child care across industries difficult. Burud, Aschbacher, and McCroskey (1984) claimed that employer-based child care in the United States has grown, even though their tabulations were generated from choice-based samples taken at different points in time in different parts of the country. They found that in 1978, 71% of employer-based child-care programs were found in hospitals and only 9% in industry; the other 20% were in public agencies and unions. In contrast, by 1982 47% of the programs were found in hospitals and 47% were found in industry; the rest were found in public agencies and trade unions.

Burud et al.'s study typifies most studies of employer-based child care: documenting the prevalence of employer-based child care, sometimes stratified by public and private sectors. Few studies have moved beyond this line of research to identifying workplaces with family leave provisions, to examining how workplaces with child care differ from workplaces without child care, and to gauging how child-care provisions influence key worker outcomes.

Although existing evidence is anecdotal and not generalisable, it is still used to argue that benefits accrue from providing on-site child care. Advocates of employer-based child care claim that it reduces absenteeism and labor turnover. Mann (1984), Perry (1981), Burud et al. (1984), and Alisberg (1984) reported falls in worker absentee and turnover rates after employer-sponsored child care was introduced. Alisberg (1984) cited a survey of 58 companies whose absentee rate fell by 72%. In another study, researchers used a cost-benefit approach to quantify the gains to employers from establishing on-site child-care. The researchers estimated that reductions in absenteeism and turnover amounted to 2% and 15% of the average employee's salary, respectively (Department of the Prime Minister & Cabinet, 1989).

Apart from suggesting reductions in labor-related costs, studies argue that employer-based child care improves employee morale, reduces recruitment costs, enhances corporate image, improves industrial relations, shortens maternity leave, and has tax advantages. Yet, except for research on employee morale, (1) the claims are unconfirmed.

Overall, although the argument that employers gain more than they lose if on-site child care is provided and flexible family leave are offered is compelling and directly relevant to recently-passed workplace reform legislation, the argument still lacks empirical verification. And to complicate matters, there are no current national-level workplace data available that can address the strengths and weaknesses of the argument, despite its clear importance to workplace reforms taking place in Australia today. Fortunately, however, data does exist, albeit less-recent than preferred, that can squarely address the veracity of the argument that family-friendly workplace provisions benefit employers. Those data come from the 1991 "Australian Workplace Industrial Relations Survey" (AWIRS). The AWIRS and its early British equivalent, the "Workplace Industrial Relations Survey" (Daniel, 1987; Millward et al., 1992) are the only data existing that permit comparing worker-related outcomes across workplaces with and without some sorts of measures of family-friendly workplace provisions.

Theory

Firms create employment conditions that maximise employee performance and minimise unnecessary losses of workers. On-site child care or flexible family leave would therefore become part of benefit packages when these provisions meet the latter criteria. On the other hand, employees work where they can receive the best compensation packages. Workers with family responsibilities take jobs that either compensate them for being unable to meet family duties or permit them to jointly meet work and family demands.

Given that some workers demand family services and that some firms willingly supply them, mutually beneficial employment contracts that include family services should occur between workers and firms. Moreover, a theory of compensating differentials predicts that the value of on-site child care or family leave will be capitalized into workers' wage rates (Rosen, 1986). That is, workers with family constraints accept lower wages in return for family amenities. Hence, in models of worker performance, family provisions, such as on-site child care and family leave, should decrease turnover and absenteeism and increase worker productivity.

Other factors should affect provision of child-care facilities or family leave, such as the age, experience, and education of workers at a specific firm. The gender composition of employees at a workplace should have an impact, too, since working mothers, rather than fathers, make the child care arrangements for children (Brandon, 1999). Higher representations of women at workplaces should increase the likelihood that child-care services are provided.

Organisational features of workplaces are also expected to influence productivity outcomes. This study tests whether the presence of traditional mechanisms that firms and workers have used to assess and address each others' demands-namely, human-resource managers and unions-influence turnover, absenteeism, and productivity. Finding that union presence is important may mean they remained pivotal to worker performance in the 1990s.

Moreover, the size of a firm's work force, how the firm organizes its production process, and whether it operates for profit should also play key roles in determining worker performance. Workplaces that are independent entities, not parts of larger entities that are commercial, private enterprises seeking profits, should be more aware of the costs of poor worker performance. Though these data cannot address this conjecture, possibly independent workplaces or profit-driven entrepreneurs are less likely to train workers so they are easily replaced.

Many workplaces with rotating shifts could affect worker performance as well. Research suggests that shift work influences working parents' child-care choices (Presser, 1986). Also, many managers believe that recruiting and retaining shift workers is easier if child-care services are offered. Alternatively, shift-work may identify workplaces, such as, hospitals or schools that are predisposed to offer employer-supported family services.

Methods

Data Description

Analyses use data from the Australian Workplace Industrial Relations Survey (AWIRS). The AWIRS collected data so that the patterns in industrial relations could be summarized and differing equity and efficiency outcomes at the workplace level could be analyzed. The survey was comprised of a main survey of 2,004 workplaces with 20 or more employees covering all industries with the exception of defense and agriculture, across all states and territories and a smaller survey of managers at 349 workplaces with between five and 19 employees. The main survey is based on a multistage, stratified probability sample of workplaces across Australia. The response rate to the survey was 87.1%.

The main survey consisted of four different questionnaires, three of which were administered to the on-site managers who had the day-to-day responsibility for workplace industrial relations (see AWIRS [1991]). Data used in these analyses come from 3 of the 4 survey modules: the Employee Relations Management Questionnaire (ERMQ), the General Management Questionnaire, and the Employee Profile Questionnaire.

The AWIRS is ideal for analyzing determinants of workplace family provisions and for assessing the effects of the provisions on worker outcomes. The comprehensiveness of the information allows for an investigation of the determinants of family provisions at the workplace. Heretofore, no nationally representative sample of workplaces with micro-level data has been available to examine the provision of family services with other workplace characteristics such as demographic composition, earnings, organizational structure, and industry sector.

Furthermore, data on the utilization of workers across workplaces and in the distribution of award payments and conditions of employment allow contrasts between workplaces with family provisions and those without such provisions. Such comparisons are of interest because no previous studies have examined whether workplaces with family provisions differ from other workplaces on award agreements, worker utilisation, and fringe benefits. Other attractive features include workplace level information about unionisation, human-resource specialists, and other work sites within the organization, and measures of workplace efficiency. Particularly important for the empirical work are the measures of labor turnover, absenteeism, and labor productivity, which are needed to test propositions that child care and family leave increase worker productivity and worker retention. Workplace weights (see AWIRS [1991]) adjust sample statistics and parameter estimates for workplace size and industry, thereby allowing valid inferences about Australian workplaces.

The survey has some deficits, however. One problem lies with the definition of the workplace. The workplace is equivalent to the Australian Bureau of Statistics (ABS) definition of a work location. The ABS defines a work location as "a single, unbroken physical area, occupied by an enterprise, which ... is engaged in productive activity on a relatively permanent basis ..." (ABS, 1983). The difficulty lies with multilocation organisations where the AWIRS-sampled workplace is centrally controlled. Hence, estimates are upwardly biased because effects of organizational structure are incorrectly added to the estimated effects of variables measured at the workplace level. Some of this bias is eliminated by use of questionnaire items that identify single, independent workplaces and that pinpoint the level that makes managerial decisions.

Added to problems of workplace definition is a lack of information on workers' ages, educational levels, and job experiences. This dearth of information on workers means that sorting models cannot be exploited and that a compensating differentials theory cannot fully explain findings. Though data on workers' educational levels and ages are not included in the models, data on occupational groupings at workplaces allow testing whether differences across occupational types affect worker performance. Of interest is testing whether workplaces with larger managerial and professional staffs (those who are usually more costly to replace and better educated) are more likely to perform better. Another difficulty is endogeneity of workplace sites. Firms do not randomly select geographic locations for workplaces. They systematically pick work sites to minimize costs and that may mean locating where there are location-specific tax incentives and specific types of labor. If site selection is related to spatial factors, then estimates are biased (Dye & Antle, 1984).

Hence, the statistical portrait of employer-based family provisions and analyses of how these services affect worker outcomes are only partial. The nature of these data cannot refute rival hypotheses that could account for the systematic patterns observed in these cross-sectional data. Nevertheless, the richness of the sample provides a unique opportunity to study this understudied, yet important, aspect of modern-day working life for Australian families.

Statistical Modeling

The three dependent variables for this study are worker turnover, unapproved absenteeism, and a worker's productivity relative to others. Because worker turnover is measured as the percentage of permanent employees who voluntarily resigned over the past 12 months, a weighted reduced-form multivariate linear regression model is used to estimate the effect of the presence of child care and family leave provisions on variation in worker turnover across workplaces. Because the other two dependent variables, worker productivity relative to others and unapproved absenteeism, were measured as ordinal scales, weighted reduced-form multivariate nonlinear models are used. Before estimation, the latter two dependent variables were converted to binary variables with each possessing mutually exclusive and exhaustive categories. For instance, a workplace either has high unapproved absenteeism or it does not. Logistic regression models are used and the coefficients yielded from the models are interpreted as the partial derivatives of the log of the odds ratio of the two alternatives (Aldrich & Nelson, 1986). Thus, the influence of any independent variable on the odds of having high absenteeism is relative to not having high absenteeism. Table 1 defines these dependent variables and also defines the independent variables that are contained in the models. Table 1 also provides descriptive statistics for all variables. (2)

Findings

Table 2 shows the differences across workplaces by the provision of on-site child-care and family-leave. The descriptive statistics are for the sample of 1,536 workplaces that had either a child-care facility (N=31) or a family-leave policy (N=119) in 1989. (Since Table 2 indicates that the incidence of on-site child care is rare compared with the incidence of family-leave, doubt is cast on the argument that on-site child care occurs commonly.)

Table 2 shows that at workplaces with family leave, workers are less often paid above the award rate than are workers in work sites without family leave. However, at workplaces with child-care facilities, workers are paid above award rates about as often as workers at work sites without child care. Differences also exist in use of non-core workers (i.e., non-permanent labor). Work sites with child-care facilities are more likely than other sites to use non-core workers, but work sites that have family leave are less likely than others to use non-core labor. Furthermore, there is a higher likelihood that females are represented among the pool of non-core workers when workplaces have child-care facilities. A greater percent of workplaces that have family leave and of workplaces that do not have child care are unionised. Finally, settings with either child care or family leave have higher wage bills than do settings without such provisions.

Table 3 shows that the provision of on-site child care lowers the likelihood of high absentee rates and raises the likelihood that managers judge workers as relatively more productive. Although on-site child care lowers turnover rates, the effect is statistically insignificant. These findings support those of previous studies concerning the effects of on-site child care. The difference, though, is that the present results have been generated from multivariate models that use a representative sample of workplaces and control for other factors correlated with worker retention and performance.

This study also differs from others in arguing that two forces drive the significant estimate of on-site child care's effect on absenteeism. First, when on-site child care is provided, workers have less need to take unapproved leave to meet family duties-child care's direct effect. But also, when on-site child care is provided, managers can verify the legitimacy of workers' time-off requests and account for the time workers spend away from work. (3) This indirect effect that on-site child care has on the monitoring of workers has not been considered in other studies. (4) Other variables significantly affect absenteeism, turnover, and relative productivity as well. Using non-core workers lowers the probability that workplaces have high absentee rates but increases worker turnover; increases in the weekly cost of labor lower the probability that workplaces have high absentee rates and increase workers' relative productivity; unionised workplaces are more likely to have higher absenteeism but lower turnover rates; and workplaces that directly monitor worker performance are more productive.

Table 4 shows that although the estimated effects of family leave on worker turnover and relative productivity are statistically insignificant, family leave also lowers absentee rates. And again, workplaces that directly monitor worker performance have more productive workers. In addition, workplaces with larger managerial staffs are more likely to have higher worker productivity; increases in weekly costs of labor lower absenteeism and turnover rates and increase the probability that employees are considered relatively more productive; unionised workplaces are more likely to have higher absenteeism but lower turnover rates and lower reported worker productivity; and workplaces that have long-term employees have lower turnover rates and higher reported worker productivity.

There are several explanations for the findings on family leave in Table 4. Family leave is more flexible than, say, maternity or paternity leave, which are strictly tied to births. If family leave was only approved for major events, such as births, its effects on worker outcomes (e.g., worker performance) would be meager, because such events occur so infrequently. (5) The effect of family leave is large precisely because family leave can be taken for a number of circumstances, major and minor. Moreover, even if parental leave could be used at any time, female working parents, not male working parents, historically have been the parents to take unauthorized time from work to care for young children. Furthermore, because family leave is part of workers' benefit packages, its effect, if any, should be more closely associated with worker retention outcomes, not necessarily with worker performance outcomes. In these data, the direction of the estimated coefficient for family leave suggests lower turnover rates, though the magnitude of the effect is statistically insignificant.

The tables not only provide evidence about the effects of family provisions on worker outcomes but they implicitly argue for more empirical work is needed to further understand how businesses and workers come to agree upon family-leave provisions. Each type of leave is different and, as apparent from the analyses, the impact of each policy on worker outcomes is different. Better measures of productivity, larger samples, and more data on workers and their employment contracts and wage rates would have helped decipher whether any fixed effects for family leave existed.

Conclusions

Firms have cautiously approach the creation of family provisions because hitherto no studies have shown that child-care provisions and family leave significantly lower absenteeism and turnover rates or increase labor productivity. The findings reported here indicate that family provisions enhance firms' abilities to retain workers and increase worker performance but that these effects operate in conjunction with other workplace features. That is, the constellation of economic, demographic, and organizational features need to be ascertained and controlled before the effects of child care and family-leave on worker retention and performance can be detected.

Moreover, past studies on this topic have not emphasized enough how managers match workers to workplaces so that workers will perform at their best and will stay at their job. Family services in these models, which control for other factors, appear part of that process.

And, past studies have not sufficiently considered how the provision of collective amenities, (i.e., public goods)--like child care--are conditioned on how many workers demand similar types of amenities. Nor have these studies considered that workers have preferences over reimbursement packages; some workers may prefer wage supplements to benefits like family-leave which tend to lower wage rates. History has chronicled how the welfarist approach has waxed and waned with workers' desires to protect their jobs and wages (Dunford, 1992).

In addition, it remains unclear whether work organisations are best equipped to provide child-care services. State agencies may be better able to provide child-care services for low-income workers who can afford neither expensive child care nor wage cuts, while the market may be better able to serve the needs of higher-income parents who are willing to pay for quality child care. Indeed, it may be imprudent for today's public policymakers to saddle employers with new duties that they are ill-equipped to handle and for which there are uncertain returns. Child-care policies that encourage employer-supported family initiatives may be preferable to policies mandating employer-based family initiatives. There are subtle differences between the two approaches, and each has important implications for the provision of child-care services, the allocation of firm resources, and the structure of workplaces.

Since the collection of the AWIRS data in 1991, the industrial relations landscape in Australia has been transformed. Following the 2004 election, the Federal government has pressed ahead with workplace reforms, including the removal of the No Disadvantage Test and unfair dismissal laws, and introducing the Work Choices legislation. The Federal government

argues that the new Australian Workplace Agreements (AWA), which is an integral part of its Work Choices package, will empower employees to better negotiate with employers for benefits like more family-friendly workplace provisions. In contrast, critics argue that by further eroding the power of unions, the proliferation of AWA's will lead to a decline in paid maternity leave, parental leave and other family-friendly benefits (ALP, 2006). Clearly, this new industrial relations environment calls out for new workplace data that can examine the effect of current and future reforms on the wellbeing of employees and their families, and to further understand the interconnections between family demands and absenteeism, retention and the productivity among Australian workers. We argue that a promising path for future workplace data is to collect information on workplaces through the existing Household Income and Labour Dynamics in Australia (HILDA) survey instrument (Watson and Wooden, 2002). By linking information on the workplaces of employed persons surveyed in HILDA, the potential exists for a new, representative sample of Australian workplaces. A key advantage of nesting workplace data within HILDA's panel structure is that researchers can observe the effects of the current and future industrial relations reforms on Australian families over time. New data collection would also enable the operationalisation of more reliable measures of workplace characteristics, including uniform measures of geographic and spatial variables.

Overall, these results using a nationally representative sample of workplaces from an earlier period in Australia's industrial relations system support findings from case studies and select samples that were conducted around the same historical time period. In summary: on-site child care is negatively associated with absenteeism; managers are more likely to judge workers as more productive at workplaces with on-site child care; and, family-leave provisions are negatively associated with absenteeism. More findings are needed, however, from other probability-based samples to buttress the findings here.

Acknowledgements

First author note: Brandon would like to thank Matt Kahn, Robert Kaufman, Andrew Cognard-Black, and Gerald Garvey for helpful suggestions on earlier drafts of this manuscript. Comments from researchers at the Australian Graduate School of Management as well as those at the National Opinion Research Center in Chicago, Illinois, U.S.A are also appreciated.

References

ABS. (1983). Australian Standard Industrial Classification: Volume 1. The Classification ABS Cat. No. 1201.0. ABS Canberra.

Aldrich, J.H., & E D. Nelson. (1986). "Linear Probability, Logit, and Probit Models." Sage University Papers, No. 45. Beverly Hills, Calif.

Alisberg, H. (1984). Public/Private Partnership: A Cost Effective Model for Child Care Services. United Way of Connecticut, New Haven, CT.

ALP. (2006). WorkChoices: A Race to the Bottom: The adverse effects of the Government's extreme industrial relations changes. Labor Parliamentary Taskforce on Industrial Relations. Available from: www.aph.gov.au, accessed July 2006.

Australian Workplace Industrial Relations Survey (AWIRS). 1989-1990: Technical Report and Data Release. (1991). Social Science Data Archives. Canberra.

Brandon, P.D. (1999). "Income-Pooling Arrangements, Economic Constraints, and Married Mothers' Child Care Choices." Journal of Family Issues, Vol. 20, No. 3. Pp. 350-370.

Burud, S. L., P. Aschbacher, J. McCroskey. (1984). Employer Supported Child Care: Investing in Human Resources. Boston: Auburn House Press.

Callus, R., A. Morehead, M. Cully, & J. Buchanan. (1991). Industrial Relations At Work: The Australian Workplace Industrial Relations Survey. Commonwealth Department of Industrial Relations. Canberra: Australian Government Publishing Service.

Daniel, W., and N. Millward. (1983). Workplace Industrial Relations in Britain: The DE/ PSI/ESRC Survey. London: Heinemann Education Books. (Reprinted by Gower, Adershot).

Department of the Prime Minister & Cabinet. (1989). "Child Care in the Workplace: A Cost-Benefit Study of Employer Provided Child Care." Canberra, Australia: Australian Government Printing Office.

Dunford, R. W. (1992). Organizational Behavior: An Organizational Analysis Perspective. Sydney: Addison-Wesley Publishing Company.

Dye, Ronald, & R. Antle. (1984). "Self-Selection via Fringe Benefits." Journal of Labor Economics 2(3): 388-411.

Edgar, Don. (1988). "Corporate Responses to Workers with Families." Family Matters. The Australian Institute of Family Studies, Melbourne. Summer.

Friedman, D. (1986). "Child Care for Employees' Kids." Harvard Business Review 64(2): 28-34.

Grant, L., P. Sai-Chew, & E Natarelli. (1982). Children at Work: An Inventory of Work-Related Day Care in Canada. Social Planning Council of Metropolitan Toronto, Toronto, Ontario.

Mann, J. (1984). "Child Care Public Policy Report." Young Children 40(1)..

Mayfield, M. I. (1985). "Employer-Supported Child Care in Canada." Mimeo, Faculty of Education, University of Victoria, Victoria, British Columbia.

Millward, N., M. Stevens, D. Smart, and W. Hawes. (1992). Workplace Industrial Relations in Transition: The The DE/PSI/ESRC Surveys. Adershot: Dartmouth Publishing.

Panckhurst, E (1984). "Workplace Child Care and Migrant Parents." National Women's Advisory Council. Canberra: Australian Government Publishing Service.

Perry, K. S. (1981). Employers and Child Care: Establishing Services through the Workplace. Women's Bureau, U.S. Department of Labor, Washington, D.C.

Presser, H. P. (1986). "Shift Work among American Women and Child Care." Journal of Marriage and the Family 48: 551-563.

Rosen, S. (1986). "The Theory of Equalizing Differences." In O. Ashenfelter & R. Layard (eds.)., Handbook of Labor Economics. Amsterdam: North Holland.

Watson, N. and M. Wooden (2002) 'The Household, Income, and Labour Dynamics In Australia (HILDA) Survey: Wave 1 Survey Methodology,' HILDA Project Technical Paper Series No. 1/02, Melbourne Institute of Applied Economic and Social Research, University of Melbourne.

Zigler, E. F., & Frank, M. (1990). The Parental Leave Crisis: Toward a National Policy. New Haven: Yale University Press.

(1) Friedman (1986) examined employee morale and found that the stress of balancing work and family increased depression among employees.

(2) The response variable "Family leave" is conditioned on the joint distribution of sampled workplaces having established paternity-leave and family-leave policies. Survey items that ask about both policies are in the ERMQ, Section F (AWIRS, 199l).

(3) For example, if a worker at a workplace with on-site child care was excused to check on a child, then that worker's manager could call the child-care facility to verify that the worker was there.

(4) An extension of this argument is that on-site child care may increase the occurrences of approved absences from the workplace.

(5) Separate analyses use paternity leave as a predictor. This variable affected worker outcomes in the expected directions, but as expected, effects were st atistically insignificant.

Table 1: Definitions of Variables and Their Descriptive Statistics

Variable Definition

Employee number Number of employees
Female Percentage of female workers
Managerial Percentage of workers classified as managers
Wage bill Mean weekly wage bill (1990A$)
Unionized 1 if unionized workplace, 0 otherwise
Paid over-award Percentage of weekly pay which is
 over-award rate
HR manager 1 if human-resource manager present, 0 otherwise
Worker attachment Percentage of workers at site more than 10 years
Sole workplace 1 if only workplace in organization, 0 otherwise
Restructured work 1 if major workplace restructuring occurred, 0
 otherwise
Shift work 1 if workers work shifts, 0 otherwise
Non-core Percentage of non-core workers
Non-core female Percentage of female non-core workers
Private commercial 1 if commercial, private workplace; 0 otherwise
Monitored 1 if workers watched by supervisor, 0 otherwise
Workplace age 1 if workplace over 10 years old, 0 otherwise
Child care 1 if workplace has child-care facility, 0
 otherwise
Family leave 1 if workplace has family leave, 0 otherwise
High absenteeism 1 if the percentage of workers absent without
 approval is above the 67th percentile, 0
 otherwise
Turnover Percentage of permanent employees resigned as
 of September 1989
Rely-prod. 1 if management considers labor productivity a
 lot higher than other comparable workplaces, 0
 otherwise

Variable Mean Standard
 Deviation

Employee number 237.20 460.64
Female 0.389 0.283
Managerial 0.184 0.387
Wage bill 490.85 144.50
Unionized 0.880 0.324
Paid over-award 0.07 0.119
HR manager 0.458 0.498
Worker attachment 0.095 0.294
Sole workplace 0.130 0.337
Restructured work 0.134 0.340
Shift work 0.391 0.488
Non-core 0.173 0.228
Non-core female 0.086 0.147
Private commercial 0.714 0.451
Monitored 0.134 0.341
Workplace age 0.171 0.371
Child care 0.020 0.140
Family leave 0.073 0.270
High absenteeism 0.157 0.364
Turnover 0.264 0.403
Rely-prod. 0.103 0.304

Note: Unless otherwise specified in the definition, all variables
are measured at the level of the workplace.

Table 2: Selected Characteristics of Workplaces by Presence of
Child-Care Facility or Established Family-Leave Policy

Variable With Child-Care No Child-Care
 Facility Facility

Family leave 0.091 0.086
Child care n/a n/a
Employee number 373.36 105.23
Female 0.495 0.416
Managerial 0.557 0.207
Wage bill 632.29 477.29
Private commercial 0.297 0.710
Unionized 0.690 0.833
Paid over-award 0.080 0.070
HR manager 0.491 0.338
Worker attachment 0.127 0.070
Sole workplace 0.160 0.187
Restructured work 0.178 0.132
Shift work 0.651 0.294
Non-core 0.301 0.187
Non-core female 0.134 0.096
Turnover 0.192 0.286
High absenteeism 0.056 0.130
Workplace age 0.219 0.178
N 31 1,505

Variable Family Leave No Family Leave

Family leave n/a n/a
Child care 0.016 0.12
Employee number 144.69 105.95
Female 0.469 0.413
Managerial 0.381 0.198
Wage bill 514.33 476.59
Private commercial 0.387 0.730
Unionized 0.994 0.818
Paid over-award 0.036 0.073
HR manager 0.383 0.337
Worker attachment 0.033 0.073
Sole workplace 0.091 0.194
Restructured work 0.172 0.109
Shift work 0.333 0.296
Non-core 0.147 0.192
Non-core female 0.085 0.098
Turnover 0.150 0.296
High absenteeism 0.129 0.130
Workplace age 0.105 0.185
N 119 1,417

Source: AWIRS (1991). Notes: weighted estimates; n/a = not
applicable.

Table 3: Effect of On-Site Child Care on Three Worker Outcomes
(Parameter Estimates and Standard Errors)

 Worker Outcomes

 High Turnover Relative
Variable Absenteeism Productivity

Employee number 0.44 * -0.12 -0.28
 (0.26) (0.54) (0.50)
Female 0.23 0.24 0.65
 (0.23) (0.49) (0.42)
Managerial -0.28 -0.05 0.63 **
 (0.19) (0.03) (0.31)
Wage bill -0.001 ** -0.004 *** 0.13 *
 (0.00) (0.00) (0.07)
Unionized 0.304 ** -0.08 * -0.43 *
 (0.18) (0.03) (0.27)
Paid over-award -0.24 0.10 -0.18
 (0.58) (0.11) (0.91)
HR manager 0.05 -0.007 0.10
 (0.13) (0.02) (0.21)
Worker attachment 0.36 *** -0.10 ** 0.50
 (0.22) (0.04) (0.36)
Sole workplace 0.10 -0.07 ** 0.10
 (0.16) (0.03) (0.28)
Restructured work 0.35 ** -0.02 0.43 *
 (0.17) (0.03) (0.26)
Shift work -0.07 0.05 ** 0.08
 (0.13) (0.02) (0.23)
Non-core -0.99 *** 0.46 *** -0.29
 (0.30) (0.05) (0.48)
Monitored -0.11 0.02 0.61 **
 (0.17) (0.03) (0.26)
Workplace age 0.26 ** 0.05 * 0.28
 (0.15) (0.03) (0.25)
Child care -1.62 ** -0.14 1.75 ***
 (0.94) (0.11) (0.59)
Private commercial 0.003 0.09 *** 0.34
 (0.17) (0.03) (0.31)
Intercept -0.66 * 0.39 *** -3.39 ***
 (0.36) (0.07) (0.61)
Log likelihood -847.19 n/a -343.92
Chi-square 44.90 ** n/a 41.78**
Adj R-square n/a 0.13 n/a
F(16,1393) n/a 13.82 ** n/a
N 1,436.00 1,410.00 1,056.00

Source: Author's calculations based on AWIRS (1991). * p < .10.
** p < .05. *** p < .01. n/a = not applicable.

Table 4: Effect of Established Family Leave on Three Worker Outcomes
(Parameter Estimates and Standard Errors)

 Worker Outcomes

 High Turnover Relative
Variable Absenteeism Productivity

Employee number 0.35 -0.2 -0.13
 (0.25) (0.54) (0.47)
Female 0.24 0.27 0.59
 (0.23) (0.49) (0.42)
Managerial -0.31 -0.05 0.66 **
 (0.19) (0.03) (0.31)
Wage bill -0.001 ** -0.004 *** 0.14 *
 (0.00) (0.00) (0.07)
Unionized 0.34 ** -0.08 ** -0.53 *
 (0.18) (0.03) (0.27)
Paid over-award -0.26 0.10 -0.001
 (0.58) (0.11) (0.90)
HR manager 0.05 -0.008 0.13
 (0.13) (0.03) (0.21)
Worker attachment 0.33 -0.10 ** 0.58 *
 (0.22) (0.04) (0.35)
Sole workplace 0.10 -0.07 ** 0.09
 (0.16) (0.03) (0.28)
Restructured work 0.35 ** -0.02 0.43
 (0.17) (0.03) (0.26)
Shift work -0.08 0.05 ** 0.14
 (0.13) (0.02) (0.23)
Non-core -1.03 *** 0.46 *** -0.08
 (0.30) (0.05) (0.47)
Monitored -0.15 0.02 0.61 **
 (0.17) (0.03) (0.26)
Workplace age 0.26 ** 0.05 * 0.33
 (0.15) (0.03) (0.25)
Child care -0.45 * -0.05 -0.21
 (0.24) (0.04) (0.30)
Private commercial -0.02 0.09 *** 0.21
 (0.17) (0.03) (0.31)
Intercept -0.63 * 0.39 *** -3.28 ***
 (0.36) (0.07) (0.61)
Log likelihood -847.61 n/a -347.81
Chi-square 44.06 ** n/a 34.01 **
Adj R-square n/a 0.13 n/a
F(16,1393) n/a 13.77 ** n/a
N 1,436.00 1,410.00 1,056.00

Source: Author's calculations based on AWIRS (1991). * p < .10.
** p < .0. *** p < .01.