Reagan Baughman, University of Michigan, Ann Arbor, Michigan, USA
Daniela DiNardi, Federal Trade Commission, Washington, DC, USA
Douglas Holtz-Eakin, Syracuse University, Syracuse, New York, USA
The views expressed here are those of the authors and do not necessarily represent the views of the Federal Trade Commission. The authors would like to thank members of the Success by Six Committee of the United Way of Central New York and Stacy Dickert-Conlin for helpful suggestions. Peggy Austin, Mary Santy, Denise Paul and JoAnna Berger provided invaluable administrative support for this project. Roberta Nasto provided outstanding research assistance.
Family-supportive employment benefits have become increasingly popular in recent years as an employer response to the increasing labor force participation of women, and the consequent need to balance work and family life. Economic theory predicts that these types of fringe benefits could at least partially pay for themselves through a combination of increased productivity and lower wages. A survey of 120 employers in an upstate New York county provides data on benefits packages and outcome measures that are used to test this hypothesis. We find that employers who offer flexible sick leave and child care assistance experience measurable reductions in turnover. Employers who offer benefits like flexible scheduling policies and child care also appear to offset part of the cost of these benefits by paying lower entry-level wages than do their competitors.
Family-friendly organizations; Fringe benefits; Productivity; Wages; Surveys; USA.
International Journal of Manpower
MCB UP Ltd
As the labor force participation of women and the fraction of dual-earner households in Western countries have increased dramatically in recent years, employers have responded by providing employment benefits, ranging from flexible scheduling to daycare to parental leave, which support a balance between work and family life. Several recent research studies document the increased prevalence of these so-called “family-friendly” benefits in the workplace, but none have considered either the costs or potential benefits that employers face in deciding to offer them to employees (Galinsky and Johnson, 1998; Whitehouse and Zetlin, 1999). However, employers’ decisions to offer benefits are shaped not only by the demands of their employees, but also by how they believe the benefits package will affect their bottom line. When one US employer, the Owens Corning company, implemented a family benefits package in late 1994, its managers cited the results of a study conducted by Kwasha Lipton. The study found that 82 percent of the managers believed that “work/life benefits will become increasingly important to ensure productivity and reduce absenteeism and stress” (Rose and Ahrens, 1996). Clearly, many employers not only expect family-supportive fringe benefits to provide personal value to workers, but to improve (or at least not hurt) productivity and profitability.
Despite the fact that, from the point of view of economic theory, one might reasonably expect productivity gains to result from family-supportive fringe benefits, there exist few useful quantitative studies that document the effect. Barton (1992) notes the “absence in the literature of any serious cost benefit analysis associated with the offering of day care for either parents or children”. To provide some evidence on this topic, we conducted an employer survey and an analysis of the specific impacts of family-supportive policy on factors likely to affect productivity and profitability. Using detailed data for a sample of 120 employers, including turnover rates, entry-level wages and components of fringe benefits packages, we seek to answer the following major questions:
- Do employers who offer “family-friendly” benefits enjoy measurable gains in the productivity of their workforce?
- Are these employers able to recoup all or part of the cost of providing benefits by paying lower wages?
Before describing the survey and the results of our analysis, we briefly discuss the predictions that basic economic theory would make about the implementation of family-supportive benefits. The predicted impacts of the benefits for an employer generally can be thought of as taking two forms. The first is in productivity changes, or any effect that implementing benefits has on an employer's ability to find and retain enough high quality workers, and to keep them as productive as possible. The second impact is through financial tradeoffs. Although employers may pay large sums of money to provide certain family-supportive benefits, economic theory predicts that these expenditures will be met with reductions in other forms of employee compensation, namely cash wages. Both productivity gains and financial tradeoffs have the potential to reduce or even negate the cost to employers of providing family-supportive benefits. We discuss each of these potential impacts in turn.
There are four areas in which we might expect family benefits to affect productivity: morale, turnover rates, absenteeism and recruiting effectiveness. Changes in morale may directly change productivity levels by altering the pace and quality of individual work; they may also have intermediate effects on turnover and absenteeism. Unfortunately, changes in levels of morale are extremely hard to measure, and even when they are measured it is very difficult to determine how they affect productivity.
Turnover, on the other hand, is a more precisely defined concept and its link to a firm's bottom line has been documented. Phillips (1990) suggests that of the cost of employee turnover averages approximately 1.5 times the annual salary of the worker being replaced. This estimate includes not only direct recruiting costs associated with filling a position, but the lost productivity in the workplace as departing employees prepare to leave and then as positions are left vacant until a replacement is found. There are two ways that turnover may be linked to the benefits package an employer offers its workers. First, offering benefits packages that contain flexible scheduling or leave provisions could contribute to lower turnover if workers are no longer forced to leave jobs (voluntarily or involuntarily) when they must take time away from the job to take care of children or elderly parents. This applies to workers with long-term caretaking responsibilities as well as those who may frequently find themselves absent or late when available child care does not match their work schedules. A second possibility is that a family supportive benefits package that is seen as meeting employees’ personal needs might improve their morale and satisfaction with their employer. The predicted impact of family benefits on absenteeism is very similar to that on turnover. Flexible scheduling or child care provisions in a family benefits package also have the possibility to reduce absenteeism if employees are frequently absent because of conflicts between work and child care schedules.
Finally, employers that offer family-supportive benefits may see turnover-related productivity gains because they do better in the recruiting process. Workers decide to apply for jobs not only because of total compensation levels, but based upon how well the mix of cash wages and offered benefits meets their needs. If many workers demand benefits like flexible scheduling, paid leave and child care, then companies that offer these benefits will have the largest pool of well-qualified applicants from which to hire. If this scenario is correct (and family benefits do help with recruiting), then productivity should rise for two reasons. First, higher quality newly-hired workers will improve productivity directly. Second, there will be a cost-savings effect of reducing the time needed to fill vacant positions.
Employers that provide family benefits incur costs; however, not only might employers expect to recoup costs through productivity gains, there are also likely to be associated cost savings in other forms of employee compensation. Two likely suspects are the sick and personal leave that most employers already offer employees and cash wages.
A recent study of workplace absenteeism in the USA by CCH Incorporated found that 21 percent of workers using sick time at work are in fact not ill, but are taking time away from the job to care for children or elderly relatives (Keller, 2000). This raises the possibility that when employers provide either leave or flexible scheduling that lets employees better manage work and family, they may find that employees use less of the sick or personal time offered to them. The extent to which there is a noticeable change in sick and personal time usage depends upon how much of these types of leave employers offer to begin with and how flexible their time banking policies are.
The second major area in which one would expect to see a financial tradeoff associated with implementing a family benefits package is in wage compensation. In the most basic economic model of the labor market, employers pay workers in proportion to their individual marginal productivity. It is critical to note that the total value of a compensation package is chosen this way, but not its breakdown. Compensation encompasses not only cash wages, but also benefits ranging from health and life insurance to retirement savings accounts to family benefits. In the long run, American employers generally have the flexibility to decide how compensation packages are broken down and to adjust cash wages. The important exception to note is that employers of the lowest-wage workers are constrained by having to pay the federal or state minimum wages.
The implication is that when the mix of compensation between wages and benefits is altered, the value of the total package does not change, at least not in the long run. Therefore, when family benefits packages of some value are offered to workers, one would expect equal reductions in cash wages or other benefits to eventually be made by employers. In practice, however, employers are often reluctant to provide additional fringe benefits because they feel they bear the full cost. This wage/fringe benefit argument is not unique to family benefits, but applies to health insurance, pensions, and all other benefits. Pauly (1999) addresses the issue in the context of employer-provided health insurance in the USA and notes that many employers do not believe that employees bear some of the cost of insurance through reduced wages, despite considerable empirical evidence to support the presence of wage/benefit tradeoffs.
The data we use to analyze the relationship between family benefits, productivity and wages comes from a survey of a randomly selected group of employers in Onondaga County, New York. The sample was drawn from two sources. The first was the Dun & Bradstreet 1999 Survey of Business listing of employers in Onondaga County and the second was the local Chamber of Commerce member list. The Chamber of Commerce was used as a secondary data source because it was more likely than the nationally-published Dun & Bradstreet lists to contain smaller employers and employers who had only recently opened businesses. From both the Dun & Bradstreet and Chamber of Commerce lists we first eliminated all state and federal government offices and all employers with less than five employees, and then selected two random samples that totaled approximately 2,000 employers when duplicates were removed. We were eventually able to set up and conduct interviews with 120 of them between October 2000 and January 2001. The interviews were conducted over the phone, but advance copies of the questionnaire were sent to participants because some information we were gathering (such as numbers of employees hired in the past year) had to be looked up in human resources records.
The survey instrument that we used first asked for information about the employer itself, including number of employees, industry, for-profit status, percent of workforce that is female and percent that is unionized, as well as a list of fringe benefits that we did not label as explicitly family-supportive (health and dental insurance, retirement plans, life insurance, tuition assistance, employee assistance plans and stock ownership). We collected a more detailed set of information about the benefits we labeled as family-supportive, including:
- if they were offered at all;
- when they had been implemented;
- whether they were available to the entire workforce; and
- whether the policy was formal (defined as being written in an employee handbook or posted on a company website).
Following is the full set of family-supportive benefits that we asked about:
- paid family leave;
- unpaid family leave;
- flexible sick leave policies;
- compressed work weeks;
- other flexible scheduling arrangements;
- job sharing;
- sick child day care;
- on-site day care;
- subsidies for off-site day care;
- reserved slots (unsubsidized) in off-site day care;
- child care resource and referral service; and
- dependent care flexible savings accounts (FSAs).
Defining measures of these benefits is an issue that deserves some discussion. American employers of more than 50 workers are required by the Family and Medical Leave Act (FMLA) of 1993 to provide 12 weeks of unpaid family leave, a provision that covers 60 percent of workers (Cantor et al., 2001). Therefore, our measure of unpaid leave refers to any unpaid leave offered in addition to FMLA requirements (or any leave offered by employers of less than 50 workers). Flexible sick leave policies allow employees to use their own allotted sick time to care for sick family members. We defined compressed work weeks as policies that shortened work weeks (usually to 4 days), but increased workday length to compensate, so that employees still worked standard 35-40 hour weeks. Other flexible scheduling arrangements referred to employee flexibility in choosing their daily work hours in a five-day workweek. Job-sharing arrangements allow two employees to work part time to fill what an employer considers to be one full-time position. We split child care benefits into as many categories as possible to capture the specific nature of an employer's policy. Dependent care flexible savings accounts are tax-preferred accounts set up by employers, similar in many ways to 401(k)s, which allow employees to save up to $5,000 before taxes every year to pay for child or elder care.
Finally, we asked for information on a number of outcome measures, including turnover, absenteeism, recruiting and wages. In this analysis we focus on the two for which we gathered the most complete and reliable data: turnover and wages. First, the questionnaire asked for the number of employees that had been hired, fired and retired in the past year, as well as the total number of full time employees, so that a ratio could be used to be used to measure turnover. We also recorded starting wages and minimum education levels for up to three entry-level jobs for each employer. We chose entry-level jobs because they allowed us to study the relationship between benefits and wages without having to account for the impact of job experience on wages, and then collected data on minimum educational requirements for those jobs because they have such a strong impact on wage levels. We asked about as many as three entry-level jobs, recognizing that for many employers there are different entry-level jobs at different education levels. For example, entry-level janitorial positions may not even require a high school diploma, while entry-level computer programming jobs may require graduate degrees. It is not immediately clear which is the best measure of entry-level pay, so we collect as much information as possible and try different measures in our analysis. The difference between the different wage measures will be discussed in the results section.
Table I describes the employers in our sample. We use appropriate sampling weights when tabulating descriptive statistics to account for the fact that we drew a stratified random sample by employer size and industry. More detail about the sampling strategy and weights is available from the authors upon request. This sample of 120 employers, who employ 19,596 Onondaga County workers (or approximately 9 percent of total county employment in 2000), is made up of predominantly for-profit firms and larger employers. The survey respondents are spread across eight major industry groups, but come primarily from services, manufacturing and public administration. Compared to overall employment in the county, this sample tends to over-represent employers in the service and public administration sectors, and under-represent retail trade employers.
Nearly all of the employers in the sample offer some type of health insurance plan, and more than 85 percent offer dental insurance. Most also seem to offer some sort of retirement plan, in the form of a 401(k), other defined contribution plan or a defined benefit pension plan. Tuition assistance, stock ownership and Employee Assistance plans are relatively less common. The percentage of employers that offer basic non-family benefits, such as health insurance, suggest that our sample is over-representative of companies that provide generous benefits packages. For example, nationally, only about two-thirds of American employers offer health insurance coverage (US Department of Commerce, 2000). The fact that our sample contains employers with who are particularly like to offer all types of benefits may limit the generalizability of our results.
Among the family-supportive benefits, we find that unpaid leave, flexible sick leave and other flexible scheduling policies (which we have defined as within-workday schedule flexibility) are the most prevalent. And, in general, employers are much more likely to be offering policies that give time flexibility rather than financial support. Many more employers offer unpaid (77.7 percent) than paid (42.5 percent) leave. Direct company involvement in day care provision is also relatively infrequent. Only 4.2 percent of the companies in our sample provide on-site day care, and only 1.1 percent subsidize it at other sites. The most common form of financial support for child care is flexible savings accounts, which are offered by 27.2 percent of these employers.
The survey also asked employers about the implementation of their family benefits. Most of the employers in the sample first offered “family-friendly” benefits packages between 1989 and 1993, the period immediately preceding the passage of the FMLA of 1993. The figures presented above referred to the percentage of employers who offer either formal or informal policies; in fact only about half of leave policies and one-third of flexible schedule policies are formal. About two-thirds of leave policies and one-half of flexible schedule policies are offered to all employees within an organization. The policies most likely to be formal and available to all employees are on-site day care and dependent care FSAs.
In order to use the survey data to determine the causal relationship between family benefits and turnover and entry-level wages, we estimate a series of multivariate ordinary least squares (OLS) regression models. We use regressions, as opposed to simply comparing outcome measures among employers that offer different benefits packages so that we can control for differences in characteristics like industry and number of employees. For example, if it were true that larger employers are likely both to offer family-supportive benefits and to have higher turnover rates, then we could not determine the direct impact of benefits on turnover without some means of controlling for size differences. Therefore, we estimate baseline regression models that control for: for-profit status, percent of workforce that is female, percent of workforce that is unionized, industry and number of full-time employees. We also include in our models controls for the other benefits (such as health insurance) that employers offer, recognizing that employers that offer generous non-family benefits are also likely to offer generous family benefits and both types of benefits affect the same outcomes.
We note, however, that what we are measuring in these basic models are correlations and not necessarily causal impacts. In running a simple regression, we have not taken into account that, for example, some employers in our sample may have very recently instituted flexible scheduling programs because they knew that child care problems were causing turnover among their employees. Therefore, we develop a technique for separating the correlation between current outcomes and recent policy changes from the long-term effects that are more likely to be causal. The way that we do this is to make a distinction between “new” and “old” benefits. “Old” benefits are defined as those that have been offered for five or more years as of the survey date. (The results that we present in the next section are not sensitive to the precise definition of “old” benefit; however, three and seven year definitions provide similar results.) We use the impact of these benefits to measure the causal impact of family policy on productivity. The rationale is that benefits that have been in place for several years have had the time to affect all of the outcomes we are studying, and the time lag removes correlations between our family benefits indicators and pre-benefit turnover rates and wages. For both turnover and entry-level wages we present both the basic regression and the “old” benefits model and discuss differences in the results of the two models.
We first estimate a basic OLS regression model of the family benefits on turnover rates. The measure of turnover that we use is voluntary turnover, so the variable is the number of quits in the past year as a fraction of total full-time employees in an organization. The day care variable that we use only refers to on-site day care, because we found that too few employers provided the other day care benefits (sick child day care, off-site subsidization and/or placement) for us to measure impacts of those benefits. For the remainder of this section, when we code an employer as offering a benefit, we are referring to the provision of either a formal or an informal benefit.
The first column of Table II presents results of the basic model. Two statistically significant results emerge from the regression. First, there is a strong positive correlation between child care resource and referral services and turnover, and a strong negative correlation for dependent care flexible savings accounts. Based upon the predictions that we discussed earlier, the latter result seems to make more sense than the former. As was discussed in the previous section, we do not interpret the results of this model as causal impacts. It seems more likely that we are finding significantly higher turnover for employers with child care referral programs because the family benefits packages were implemented in response to high turnover.
The second column of Table II presents the OLS regression model that measures the impacts of “old” benefits. The results of this model are quite different. We now find that providing flexible sick leave and child care referral services are both associated with significant decreases in turnover. We feel that these results are more likely to reflect causal impacts, rather than simple correlations. The differences between the two sets of results illustrate the role that outcomes like turnover has in employers’ decisions to offer benefits packages, and the corresponding difficulty that develops for researchers trying to determine the true impacts of family-supportive benefits.
We now turn our attention to financial tradeoffs. We have hypothesized that providing family-supportive benefits could reduce employers’ costs by decreasing both sick and personal time used and wages. In Table III we estimate a basic model of the relationship between the highest entry-level wage an employer reported offering and its family benefits package. Offering compressed work weeks seems to be associated with higher entry-level wages, while there is a significant negative correlation between job sharing policies and wages.
The ways in which employers might decide to provide family-supportive benefits based upon the outcomes we are measuring is slightly different for wages than for the productivity outcome measures. But for a different reason, estimating a model using only benefits in place for five or more years is even more important here. Very few employers enjoy immediate flexibility in adjusting wage levels. Contracts and group bargaining agreements cover many workers, and often wage adjustments can only be made when new employees are hired or contracts are re-negotiated. Therefore, we estimate an “old benefits” model here because we could only hope to see causal impacts of benefits packages on wages a few years after the benefits were first offered. The results of this model are presented in the second column of Table III. The results here are consistent with the hypothesis of wage/fringe benefit tradeoff that was developed earlier. Several of the family benefits coefficients are negative and statistically significant. Flexible sick leave, flexible scheduling policy and provision of on-site day care are all associated with significantly lower entry-level salaries.
One further point about the wage regressions deserves discussion. The results that are presented in Table III measure the effect of benefits on the highest entry-level wage an employer offers. Many employers in our sample also reported one or two lower entry-level salaries for less-educated entry-level workers. In models for which we do not report results, we have estimated the wage regressions using the lower entry-level wages and have found no significant effects. This is in line with our predictions, because the lowest entry-level wages that firms in our sample offer are at or close to the US state and federal minimum wages, and employers have little or no flexibility in decreasing them.
The goal of this analysis has been to expand a previously very small body of knowledge about the productivity and profitability effects of family supportive benefits policies. Using data from a survey administered to 120 randomly-selected employers in Onondaga County, New York, we have attempted to gauge the impact of employment benefits like family leave, job sharing and child care on employers’ bottom lines. Economic theory predicts that these types of fringe benefits could pay for themselves through a combination of increased productivity and lower wages. The empirical results of our analysis provide weak support for this theory.
We do not find evidence the majority of family supportive benefits increase productivity. The exception is that employers offering flexible sick leave and child care assistance have significantly lower turnover rates. Unfortunately, because of the small size of our sample and requirements of the appropriate statistical models, we are unable to say anything conclusive about the impact of family benefits on absenteeism, recruiting effectiveness or use of sick and personal time. The results of our wage tradeoff models provide evidence more strongly in line with the predictions of economic theory. While we do not find statistically significant decreases in wages associated with all of the family benefits we consider, we do find that employers offering child care, flexible sick leave and flexible scheduling all pay significantly lower entry-level wages. This result suggests that employers offering family supportive benefits recoup at least part of the cost of these benefits by paying lower wages.
A final interesting theme to emerge from the analysis is the striking difference in the productivity and wage impacts we estimated depending upon whether employers were “new” or “old” adopters of benefits policies. The pattern of, for example, positive correlations between turnover and family benefits overall, but negative correlations among organizations with benefits in place several years, suggests to us that employers did not select benefits policies randomly. It seems likely that many managers chose to implement family benefits packages specifically to address problems with turnover, recruiting and absenteeism. This last finding should be taken into consideration by future researchers wishing to measure causal impacts of “family-friendly” benefits policies.
Table ICharacteristics of employers in sample
Table IIResults for turnover regressions
Table IIIResults for wage regressions
Barton, L. (1992), "Corporate sponsored child care: a benefit with high satisfaction, questionable future", International Journal of Manpower, Vol. 13 No.1, pp.12-24.
Cantor, D., Waldfogel, J., Kerwin, J., McKinley Wright, M., Levin, K., Rauch, J., Hagerty, T., Stapleton Kudela, M. (2001), Balancing the Needs of Families and Employers: The Family and Medical Leave Surveys, 2000 Update, United States Department of Labor, Washington, DC, .
Galinsky, E., Johnson, A. (1998), Reframing the Business Case for Work-Life Initiatives, Families and Work Institute, New York, NY, .
Keller, L. (2000), "Playing hooky", available at: www.cnn.com, No.15 November, .
Pauly, M. (1999), Health Benefits at Work: An Economic and Political Analysis of Employment-Based Health Insurance, University of Michigan Press, Ann Arbor, MI, .
Phillips, J.D. (1990), "The price tag on turnover", Personnel Journal, No.December, .
Rose, K., Ahrens, A. (1996), "The changing nature of work/life benefits: Owens Corning's experience", Benefits Quarterly, Vol. 12 No.4, pp.8-11.
US Department of Commerce (2000), Statistical Abstract of the United States, US Government Printing Office, Washington, DC, .
Whitehouse, G., Zetlin, D. (1999), "‘Family friendly’ policies: distribution and implementation in Australian workplaces", Economic and Labor Relations Review, Vol. 10 No.2, pp.221-39.