Precarious Work Schedules in Low-Level Jobs:
Implications for Work-Life Interferences and Stress
Julia R. Henly & Susan Lambert
University of Chicago
March 2010
Draft: Please do not quote or cite without permission of authors.
Address correspondence to either author at 969 E. 60th St. Chicago, Illinois 60637, USA
jhenly@uchicago.edu, slambert@uchicago.edu
Note: The authors would like to thank Lauren Gaudino and Ellen Frank for excellent research
assistance. This research would not have been possible without the generous support of the Russell
Sage Foundation, the Ford Foundation, and the Annie E. Casey Foundation, to which the authors
are very grateful.
precarious work schedules
2
precarious work schedules
3
Abstract
Drawing from a survey of employees in a national apparel chain, this study considers the
precariousness of employment in the retail sector by examining two dimensions of work schedules
(schedule unpredictability and limited employee input) that are hypothesized to increase general
work-to-family conflict, interfere with workers’ ability to plan and structure nonwork time, and
heighten levels of perceived stress. We find that net of individual, family, and other work
characteristics, unpredictability and limited input into schedules are related to each of the outcomes.
Unpredictability appears to have a stronger and more significant influence across model
specifications. Overall, the findings for unpredictable schedules are robust to several model
specifications, including the inclusion of enabling factors such as supervisor support that may be
thought to ameliorate the negative impacts of precarious schedules (Voydanoff, 2004). Limitations
of the study and future research directions are discussed.
precarious work schedules
4
The current economic recession is increasing the vulnerability of workers at all levels of the
U.S. labor market. Unemployment increased from 4.9 to 7.2 percent in 20081, continuing its rise to
9.17 percent in February 2010. The destabilizing effects of the economic downturn, however, are
felt not only by those who lose their jobs. Those who remain employed face increasingly
precarious work conditions. For example, involuntary part-time work has reached a 30-year high
(8.8 million) and the length of the average workweek has fallen to a record low of under 34 hours2.
Although the current recession has extended the reach of precarious work to a broader
segment of workers, it did not give rise to these conditions. Indeed, job characteristics and
employer-worker relations have become increasingly precarious over the past three decades, due to
a combination of factors including macro-economic changes in the structure of the labor market
(shifts from manufacturing to service, for example), the decline in unionization, the growth of
business strategies emphasizing cost containment, and the loosening of government labor standards
(Blank, Danziger, & Schoeni, 2006; Kalleberg, 2009; Lambert, 2008). Precariousness has increased
globally, although the focus of this paper is the United States where worker protections are
particularly weak.
Growing precariousness is evidenced by increases in nonstandard employment
arrangements, such as part-time, contingent and temporary jobs, as well as in the use of scheduling
practices that allow employers to make quick adjustments to staffing levels in standard jobs (Henly
& Lambert, 2005). Notably, in good times and bad, employees at the front lines of many of today’s
firms bear the brunt of routine fluctuations in demand for services and products through their work
schedules. Hourly workers have limited control over their work schedules and increasingly
experience variable, unpredictable, and reduced work hours that can compromise their job
1
All statistics in this paragraph taken from Bureau of Labor Statistics, Current Population Survey, accessed March 13,
2010; http://www.bls.gov/news.release/empsit.nr0.htm.
2
This is the lowest average number of work hours for production and nonsupervisory workers in private firms since this
statistic has been tracked, beginning in 1964.
5
precarious work schedules
performance and their ability to earn an adequate living (Lambert, 2008). Indeed, the scheduling
practices commonly employed across key industrial sectors in the U.S. economy contribute to a
growing precariousness of employment that has important implications for individuals, families,
and communities (Kalleberg, 2009; Lambert, 2008; Henly, Shaefer, & Waxman, 2006).
In this paper, we focus on the precariousness of work schedules for hourly non-management
employees in the retail sector. Drawing from a survey of employees in a national retail apparel
chain, we examine two dimensions of work schedules that we hypothesize to influence both
behavioral and stress-related outcomes of workers. In particular, we consider (1) the
unpredictability of employees’ work schedules, that is, the extent to which employees have limited
advance notice of their work schedules and cannot count on or anticipate getting a particular work
schedule (in terms of days, shifts, and the number of hours worked) and (2) the lack of employee
input into the time, days, and hours they work (e.g., lack of schedule control). We view these two
dimensions of work schedules as indicators of employment precariousness in everyday work and
hypothesize that net of other employment characteristics and personal and family circumstances,
both dimensions of precariousness will increase general work-to-family conflict, interfere with
workers’ ability to plan and structure their nonwork time, and heighten levels of perceived stress.
Building from Greenhaus and Beutell’s (1985) now classic distinction between strain-based and
time-based work-to-family conflict, we further hypothesize that input into scheduling may be
particularly important for reducing stress and will compensate to some degree for an unpredictable
schedule in terms of reducing workers’ stress but will not significantly reduce interferences with
nonwork activities. Our survey findings lend support to these hypotheses (with one exception) and
are robust to several model specifications, including the inclusion of enabling factors such as
supervisor support that may be thought to ameliorate the negative impacts of precarious schedules.
The exception concerns our findings for schedule input and stress. We do not find that increased
precarious work schedules
6
schedule input is more highly associated to stress than the other dependent variables under
investigation, nor do we find that it significantly reduces the association between schedule
unpredictability and perceived stress, as hypothesized.
BACKGROUND
The Growing Precariousness of Work
In his 2009 presidential address to the American Sociological Association, Arne Kalleberg
spoke of the growing precariousness of work and increasing insecurity of workers in the United
States and globally. By precariousness, Kalleberg refers to “employment that is uncertain,
unpredictable, and risky from the point of view of the worker” (p.2, 2009). He states:
Precarious work has far-reaching consequences that cut across many areas of concern to
sociologists. Creating insecurity for many people, it has pervasive consequences not only
for the nature of work, workplaces, and people’s work experiences, but also for many
nonwork individual (e.g., stress, education), social (e.g., family, community), and political
(e.g., stability, democratization) outcomes. It is thus important that we understand the new
workplace arrangements that generate precarious work and insecurity.
Thus, work has become more precarious over the past 30 years as employers increasingly
pass risk from the market onto workers and demand for labor flexibility – the ability of employers
to flexibly and quickly adjust the number of employees and their work hours – has increased
(Lambert, 2008). In jobs paid by the hour, maintaining a close link between employees’ work hours
and variations in labor demand is one strategy employers use in their efforts to contain labor costs.
Lambert (2008) demonstrates several ways in which labor flexibility practices operate in
workplaces—for example, through strategic use of work status categories and “just-in-time”
scheduling practices—suggesting that employer-driven variation in work hours, rather than
flexibility that benefits workers, is common practice in many workplaces, especially in lower-level
precarious work schedules
7
jobs. Whatever the reasons for growing employer reliance on labor flexibility (e.g., global
competition, cost over quality strategies, eroding labor standards), the practices used to achieve
labor flexibility shift risk from firms to workers, resulting in increased instability and insecurity in
hours and income (Lambert, 2008). The literature on firm-level labor markets provides strong
evidence that passing instability onto workers leads to problems with performance, such as
heightened absenteeism and turnover (Appelbaum, Bailey, Berg, & Kalleberg, 2000; Appelbaum,
Bernhardt, & Murnane, 2003; Baron & Bielby, 1980; Jacobs, 1994; Osterman, 1999).
Dimensions of Precarious Work in Low-Level Jobs
Nonstandard Employment Status and Work Schedules
Workers employed in hourly, non-management jobs are at particular risk of precarious
employment. Research demonstrates that low-level workers disproportionately hold jobs with
nonstandard status and nonstandard schedules (Golden, 2005; Presser, 2003; Henly & Lambert,
2005). Regarding employment status, the literature demonstrates that employers are relying less on
“regular, full-time” employment status categories (Herzenberg, Alic, & Wial, 1998; Tilly, 1996),
replacing regular status workers with part-time, contingent, and temporary employees and adopting
new work statuses such as “full-time flex” and “reduced compensation professional.” These new
statuses serve to restrict access to benefits and facilitate employers’ ability to apply flexible labor
practices to a wider set of workers beyond the traditional part-time/full-time status distinction
(Lambert & Waxman, 2005; Lambert, 2009). Henly & Lambert (2005) argue that employment
status is often quite ambiguous in low-level jobs, such that “one’s status on paper may not be
indicative of the number of hours worked or even whether an employee is working at all” (p.479).
In Lambert’s study of low-level jobs across four industrial sectors (financial, transportation,
hospitality, and retail), the primary distinction that is found between part-time and full-time status
hourly employees is the classification’s implication for benefit access and employment and training
precarious work schedules
8
programs. Neither full-time nor part-time status reliably indicated the number of hours worked, with
hour fluctuations proving routine for both full-time and part-time hourly workers. Thus, one’s
employment status may indeed signal greater or lesser job security—especially in terms of benefit
access—but it is an insufficient proxy for work precariousness. Even workers in full-time jobs may
have difficulty getting enough hours and may face nonstandard work schedules into which they
have little input and are given little advance notice.
Thus, in addition to employment status, it is important to consider the ways in which the
scheduling practices that employers use can contribute to the precariousness of employment. Much
of the literature on nonstandard work schedules concerns the timing of work (e.g., nontraditional
hours). The growth of the service industry and increasing globalization have increased demand for
nonstandard hour workers (Presser, 2003). Based on 1997 Current Population Survey (CPS) data,
Harriet Presser (2003, p.1) estimates that “two-fifths of all employed Americans work mostly at
nonstandard times”, and even more employees regularly work at least some of their hours outside of
regular daytime, weekday hours.
Nonstandard schedule work is common across occupations, but service and laborer positions
disproportionately require nonstandard hours (Presser, 2003). These occupations also heavily
employ workers for whom precarious work schedules are likely to be most detrimental; for
example, workers with limited education and skills, limited financial resources, minorities, and
single parents (Presser & Cox, 1997; Presser, 2003).
Unpredictability of Work Schedules and Limited Employee Control Over Schedules
In addition to concerns about the timing of work (e.g., nontraditional hours), other
nonstandard scheduling features may also be important to understanding precarious employment.
In particular, when someone works may be a less relevant indicator of precariousness than the
predictability of one’s work or the control a worker has over his or her work schedule.
precarious work schedules
9
There has been limited research attention to work schedule predictability—that is, the extent
to which workers can count on or anticipate receiving a particular work schedule from week to
week, and the length of advance notice given regarding work hours. Recent studies suggest that
unpredictable schedules in low-level hourly jobs are widespread. Lambert’s previously mentioned
study of low-level jobs across four industrial sectors found that unpredictable scheduling practices
(limited advance notice and frequent last-minute changes) were a typical employer strategy for
managing fluctuations in consumer demand, observed in all 22 workplaces studied (Lambert &
Waxman, 2005). In Henly’s companion study of low-income mothers employed in six of the retail
settings studied by Lambert, the majority of participants reported having less than one week’s
notice regarding the following week’s schedule, and changes to schedules were often made after the
schedule was posted (Henly, Shaefer, & Waxman, 2006). Moreover, two-thirds of participants
reported that posted schedules routinely varied week to week. Workers were regularly called in for
unscheduled, “last-minute” hours, sent home early, or pressed to stay later than a shift’s scheduled
end-time. Thus, schedule unpredictability – distinct from the timing of work hours – may be an
important dimension of precarious work that has implications for workers and their families.
The benefits of a predictable schedule for workers may be attenuated in situations where
schedules are not sufficiently flexible to accommodate nonwork-related responsibilities. Indeed,
schedule predictability may turn into rigidity if workers have limited input into their work schedules
or if posted schedules are not amenable to change. Thus, another key dimension of work schedules
is the control (or input) that employees have over the hours, days, and shifts they work.
Unlike the paucity of work on predictability, control over schedules has received relatively
more attention, especially by work-family scholars concerned with flexibility. It is important to
distinguish between employee-driven flexibility, which is discussed here as indicating employee
input into the number and timing of work hours, and employer-driven variation, which indicates
precarious work schedules
10
employer control over the scheduling process, as in the aforementioned concern over an increasing
reliance on labor flexibility practices by employers (see Lambert & Waxman, 2005; Moss,
Salzman, and Tilly, 2005; Henly, et al., 2006). Although the work-family literature does not always
distinguish these two very different sources of variation in work hours, there is little ambiguity that
it is the first – employee-driven control over schedules – that is assumed to benefit workers.
Flexible work is often heralded as a critical workplace benefit that has the capacity to
deliver workers greater control over how, where, and when they carry out their work, thereby
easing the integration of work with family (and other nonwork) roles (Kossek, Lautsch, & Eaton
2005). For example, employee input into work schedules may make it possible for a parent to
arrange work hours around a child’s school schedule or allow a worker to make a last minute
schedule change to accommodate an unanticipated medical appointment. Employee surveys
consistently demonstrate that workers desire flexible work arrangements, believe flexibility in hours
would improve their quality of life, and would even trade other forms of work opportunity for
flexibility (Golden, 2005). Yet, research based on national samples of workers indicates that
employee-driven flexibility in work hours is rare among low-level workers (Galinsky, Hughes, &
David, 1990; Golden, 2005). For example, Current Population Survey (CPS) data demonstrate that
low-skilled unmarried mothers – who may arguably have the greatest need for flexible work
arrangements – are least likely to have them (Golden, 2005). In fact, flexible work schedules are
distributed according to predictable stratification statuses: men more than women, whites more than
nonwhites, and those with higher education over lesser education benefit disproportionately from
flexible work schedules (Golden, 2005). Smaller, more focused studies of specific job sectors
demonstrate how the process used to set work schedules for low-level, hourly, nonmanagement
workers typically provides only very restricted formal avenues for employees to exercise input over
precarious work schedules
11
their work hours. As a result, employee input into scheduling is negotiated informally, if at all, and
is at the discretion of the manager (Henly et al., 2006).
Implications of Precarious Work on Nonwork Spheres and Employee Stress
Work-life and work-family scholars have long been interested in the various ways in which
work conditions interact with nonwork domains. Much of this literature has focused on
professional, white collar employment (and middle- and upper-class two-parent families), although
there have been notable exceptions (Lambert, 1990;1999), and in recent years there has been
increased attention to work-family linkages across a broader range of individuals and a more
diverse set of jobs (Grzywacz, Almeida, & McDonald, 2002; Perry-Jenkins, 2005; Swanberg,
2005). The growth in precarious employment underscores the importance of examining the
adequacy of our understanding of the work-family interface for today’s workers (Lambert, 1999;
Rayman & Bookman, 1999). Several recent studies of the low-wage labor market that examine the
characteristics of low-level jobs (Acs & Loprest, 2008; Kalleberg, Reskin, & Hudson, 2000), firmlevel practices (Lambert, 2009; Hacker, 2006) and work and family life among low-income and
working class families (e.g., Lein, Benjamin, McManus, & Roy, 2005; Perry-Jenkins, 2005;
Heymann, 2000) suggest that key concepts and findings from the work-family field may benefit
from refinement to better reflect common but understudied job characteristics as they are
experienced by workers with limited economic means.
Toward this end, the current paper investigates whether the two dimensions of precarious
work schedules discussed above are related to both behavioral and stress outcomes in a sample of
low-wage, hourly, non-management retail workers. In particular, we consider whether the
unpredictability of employees’ work schedules and the amount of input employees have into the
time, days, and hours they work are associated with general work-to-family conflict, and whether
precarious work schedules
12
these indicators of precarious work contribute to perceived stress and interfere with workers’ ability
to plan and structure nonwork time.
Our conceptual framework joins the research on firm practices in the low-wage labor market
(as discussed above), poverty research on work-family management strategies, and the work-family
literature on job demands and work-family conflict. By bringing these three literatures together,
our goal is to contribute to a fuller understanding of the consequences of precarious work demands
for women employed in low-level retail jobs.
Work-to-family conflict is conceptualized as “a form of interrole conflict in which the role
pressures from the work and family domains are mutually incompatible in some respect” (p.77,
Greenhaus & Beutell, 1985). According to Greenhaus and Beutell, the sources of work-to-family
conflict can be time-based, in which the time pressures of one role interfere with the time demands
of another role, or strain-based, in which the strain symptoms produced by one role interfere with
one’s ability to carry out another role.3 Importantly, work-family conflict is conceptualized in the
literature as bidirectional (e.g., Frone, Russell, & Cooper, 1997; Frone, Yardley, & Markel, 1997;
Greenhaus & Beutell, 1985; Kelloway, Gottlieb, & Barham, 1999; Voydanoff, 2005). Work roles
can interfere with family (work-to-family conflict) and family roles can interfere with work (familyto-work conflict). Because the current study is specifically interested in the behavioral and stress
implications of precarious work schedules, our focus in unidirectional – from work to family –
although we certainly recognize the importance of better understanding family-to-work conflict as
well for workers in precarious jobs.4
3
Greenhaus & Beutell also identify behavior-based work-family conflict, in which the behaviors expected in one role
(compassion with family) are difficult to adjust to another role (competitiveness on the job). Our models do not address
behavioral flexibility across roles.
4
Although our models are based on traditional work-family frameworks derived from a role conflict perspective (e.g.,
Kahn, Wolfe, Quinn, Snoek, & Rosenthal, 1964), we do not assume the inevitability of conflict across work-family
roles more generally. Role accumulation can have positive benefits on worker health and well-being under particular
circumstances (Barnett & Hyde, 2001; Marks, 1977; Perry-Jenkins, Repetti, & Crouter, 2000; Ruderman, Ohlott,
precarious work schedules
13
Critical to our conceptualization is the notion of job demands as sources of time-based and
strain-based work-to-family conflict. Accordingly, job demands (in this case, demands resulting
from unpredictable schedules and limited input into work schedules) “compete for a person’s time”
(Greenhaus & Beutell, 1985), draining resources and complicating temporal work-family patterns
that would otherwise facilitate the fulfillment of obligations and desired activities in the family
domain (or another domain, in an expanded work-life conflict framework). Importantly, our
conceptualization reflects a modification of conventional notions of work-to-family conflict in that
our concern is about time-based and strain-based conflicts resulting from unpredictability and lack
of input into schedules rather than a concern with excessive job demands that overload roles and
create time pressures – as in the problem of too much work or a preoccupation with work during
nonwork times (see Major, Klein, & Ehrhart, 2002; Pleck, Staines, & Lang, 1980; Kahn et al.,
1964).
Regarding time-based sources of work-to-family conflict, the evidence is strong that
excessive time demands create work-to-family conflict. Whether the work conditions of low-level
hourly workers, who often have trouble getting enough hours (see Lambert, 2008), create timebased work-to-family conflict has not been sufficiently studied. Thus, we consider the behavioral
consequence – in terms of interferences in nonwork domains – of unpredictable and uncontrollable
time–based work demands. Informed by theories of social time (e.g., Almeida & McDonald, 2005;
Presser, 2003), we argue that work demands that create ambiguities around the timing of work
complicate the activities central to other spheres of life. For example, unpredictable work schedules
over which employees have limited control are hypothesized to interfere with the planning of
Panzer, & King, 2002; Moen & Yu, 2000; Greenhaus & Powell, 2006; Grzywacz & Bass, 2003). Moreover, we
appreciate that enabling resources such as job autonomy and boundary-expanding resources such as work supports
increase the likelihood of facilitation rather than conflict between work and family roles (Voydanoff, 2004). Still, we
maintain that a work-family conflict perspective is more appropriate than a role accumulation/enrichment or facilitation
perspective for this study, given what is known about the limited availability of work-family supports in low-level jobs
and the generally precarious job characteristics of hourly work in the retail industry.
precarious work schedules
14
nonwork activities such as doctor’s appointments, social outings, or arranging other family
activities, as planning is facilitated by advance knowledge of and control over one’s available time.
We assume that both unpredictability and limited schedule input will be sources of time-based
conflict, and that even when employees have input into their schedules, schedule unpredictability
may interfere with nonwork activities.
We know of no large scale studies that examine the relationship between schedule
unpredictability and time-based work-to-family conflict per s; however, other aspects of work
schedules, such as nonstandard work-timing and irregular work hours are associated with timebased work-to-family conflict (Major, Klein, & Ehrhart, 2002; Pleck, Staines, & Lang, 1980;
Fenwick & Tausig, 2005). Moreover, there is evidence that greater control over work schedules can
reduce time-based conflict (Fenwick & Tausig, 2005; Kossek, 2005), although other research
suggests that the degree of employee schedule control must be significant for it to matter in any
meaningful way (Greenhaus & Beutell, 1985; Bohen & Viveros-Long, 1981). Relatedly, there has
been a growing concern about whether and how the temporal nature of nonstandard work schedules
influences the organization and conduct of family life (Almeida & McDonald, 2005; Presser, 2003).
Survey findings reveal significant associations between nonstandard work hours and the time
parents spend with children (Almeida & McDonald, 2005) and on specific parent-child activities
(Presser, 2003; Heymann, 2000). A number of qualitative studies also illustrate the challenges that
nonstandard work schedules pose for family routines and child care, and these studies suggest that
timing, unpredictability, and limited control over schedules matter for family functioning (Roy,
Tubbs, & Burton, 2004; Henly & Lambert, 2005; Scott et al., 2005).
A stress model underlies much of the work-family literature (see reviews in Eckenrode &
Gore, 1990; Edwards & Rothbard, 2005), and the transmission of stress between work and family
has been characterized as a manifestation of negative spillover (Pleck, 1995; Voydanoff, 2004).
precarious work schedules
15
Over the past quarter century, a substantial laundry list of work stressors with demonstrated
relationships to an equally impressive list of strain symptoms and stress-related outcomes has
accumulated. For example, different measures of work demands have been associated with
depression (Frone, Russell, & Barnes, 1996; Googins, 1991), fatigue (Pleck et al., 1980; Googins,
1991), irritability (Pleck et al., 1980), somatic complaints (Burke, 1988), general well-being (GrantVallone & Donaldson, 2001), heavy alcohol use (Frone et al., 1996), psychological distress
(Matthews, Conger, & Wickrama, 1996) and problematic marital relationships (Matthews et al.,
1996).
Greenhaus & Beutell (1985) point out that whereas some work stressors that induce strain
are independent of time-based work demands (for example, role ambiguity, role conflict and
boundary spanning activities), “extensive time involvement in a particular role also can produce
strain symptoms,” (p. 81) suggesting that time-based and strain-based conflict can result from the
same work stressors. We extend Greenhaus and Beutell’s argument to our concern with schedule
precariousness, which we hypothesize will contribute to strain symptoms. Staines and Pleck (1983)
provide some preliminary support for this view. They find that atypical work hours and days, as
well as variable hours, relate to work-family role strain, whereas greater schedule flexibility (i.e.,
control) relates to less role strain. Fenwick & Tausig (2005) find that schedule control is related to
work-life conflict measures, as well as spillover and distress-related outcomes; and importantly,
their findings suggest that schedule control operates independently from the timing of work.
Relatedly, Campbell and Moen (1992) find that when considered in the same model, limited control
over work schedules, but not nonstandard work hours, contribute to strain. Given these results, we
expect that unpredictable schedules and lack of input into scheduling will both induce strain, but
that having greater schedule input may compensate for (and therefore, reduce) the strain of
unpredictable schedules.
precarious work schedules
16
Worker Vulnerability to Employment Precariousness
Certain workers may be particularly vulnerable to both precarious work conditions and
stress and work-family conflict (Kalleberg, 2009). Research on labor market inequalities
demonstrates how common markers of social status such as low education, single parenthood,
minority status, and being female increase the likelihood of employment in an occupational or
industrial sector with fewer worker protections and in jobs with lower wages, fewer benefits, and
less standard scheduling practices (Padavic & Reskin, 2002; Grzywacz, Almeida, & McDonald,
2002; Golden, 2005; Kalleberg, et al., 2000; Christensen, 1998; Gottschalk & Danziger, 2005).
Demographic variables such as education, age, race and ethnicity, marital status, and health status
also show important relationships to measures of work-family conflict and stress (Frone, Yardley,
& Markel, 1997; Grzywacz, Almeida, & McDonald, 2002; McLoyd, Toyokawa, & Kaplan, 2008;
Moen & Yu, 1999; Presser, 2003). Moreover, resources available in workers’ households, such as
other sources of income and a partner to share caregiving, may be related to workers’ experiences
of precarious employment as well as stress and work-to-family conflict.
Thus, in this paper, we include several vulnerability measures as controls in the multivariate
analyses in order to assess the extent to which unpredictability and lack of input are associated with
stress- and time-based conflict independent of other personal and household factors that may put
workers at risk for precarious employment and work-to-family conflict.
Study Hypotheses
The primary focus of the current study is on two dimensions of precariousness – schedule
unpredictability and lack of schedule input– and their association with general work-to-family
conflict, time-based conflict (interferences with planning nonwork activities), and perceived stress,
in a sample of women working in hourly retail jobs. Based on the findings reviewed above from
the work-family literature, we advance a behavioral and stress model to explain the hypothesized
precarious work schedules
17
relationships between these variables. Overall, we expect both unpredictability and lack of input to
be related to the outcomes. However, because we argue that input into work hours can reduce the
stress of unpredictable schedules, we expect lack of input to have a stronger relationship to the
stress outcome than the others, and we expect the association between unpredictability and stress to
be attenuated when both dimensions (unpredictability and input) are considered together in the
same model. However, we argue that input into schedules is insufficient to fully compensate for the
interferences in nonwork activities that unpredictability can cause; thus, we expect that schedule
unpredictability will have a robust relationship to the behavioral measures that tap into interferences
with nonwork activities, even after the contribution of schedule input has been considered in the
model.
Thus, we assess the following hypotheses:
Hypothesis 1: Controlling for indicators of vulnerability, the unpredictability of work
schedules is positively related to general work-to-family conflict (1a), work interferences in
nonwork activities (1b), and perceived stress (1c).
Hypothesis 2: Controlling for indicators of vulnerability, lack of input into work schedules
is positively related to general work-to-family conflict (2a), work interferences in nonwork
activities (2b), and perceived stress (2c).
Hypothesis 3: Controlling for indicators of vulnerability, the positive association between
unpredictable work schedules and work interferences (3) will remain when input is also included in
the model (because we assume that input cannot compensate for the time-based conflict that
unpredictability in scheduling can create).
Hypothesis 4: Controlling for indicators of vulnerability, the positive association between
unpredictable work schedules and perceived stress (4) will be attenuated when input is added to the
precarious work schedules
18
model (because we assume that input into schedules reduces the strain-based conflict that
unpredictability in scheduling can create).
We do not advance an a priori hypothesis regarding whether the coefficients for
unpredictability and control will remain significantly related to the general work-to-family conflict
measure when they are both included together in the model because the work-to-family conflict
measure is a general measure that includes dimensions of both time-based and strain-based work-tofamily conflict.
METHOD
Data
The analyses are based on two waves of data from the Retail Employee Scheduling Survey,
a telephone survey conducted in 2008 of hourly retail workers in the Midwest. Sample eligibility is
based on employment within one of 21 stores from a national retail chain participating in a
workplace-based intervention. Overall, 136 respondents participated in Wave 1 and 156
respondents participated in Wave 2 of the study. Eligibility was determined at Wave 1 based on
employment at the participating store during the recruitment phase. Eligibility was determined at
Wave 2 based on eligibility at Wave 1 and/or employment at the participating store during
recruitment at Wave 2 (as described further below). For the purposes of this study, we exclude all
managers from the sample and all respondents who were not working at the store at the time of the
Wave 2 survey. This results in an analysis sample of 123 for Wave 1 and 120 for Wave 2. Further,
for the analyses that rely on both waves of data (Table 6), we only include non-manager
respondents that were employed at the store at both waves 1 and 2 of the survey and for whom we
have information at both time periods. This restriction limits the sample to 92 for all analyses that
rely on both waves of data.
Sampling Strategy
precarious work schedules
19
The research team was provided the name of all employees but not their contact information
(telephone numbers and/or addresses) prior to recruitment. Because contact information for
employees in the 21 participating stores was not initially available, sample recruitment took place
through a two stage process at both Waves. In the first stage, a letter for every employee was
provided by the research team to the store manager, who distributed them to store employees with
their weekly paycheck. The manager’s role was simply to distribute the letter, not to encourage
employees to participate in the study. (Managers were also eligible for participation, but were not
included in the present analyses given they are responsible for the scheduling process and would
therefore have missing data on our independent variables.) The letter explained the purpose of the
survey and requested that employees provide the researchers with contact information by returning
a self-addressed, stamped, contact card to the researchers. Employees were not asked at this stage to
commit to participation, but rather for their willingness to provide the research team with contact
information so that they could be contacted regarding the survey. After the letter was distributed, a
follow up letter was provided. In addition to a follow-up letter, research team members visited the
21 stores in person (sometimes multiple times) in an effort to collect additional cards. Employees
were provided with a $10 gift card for supplying contact information. The first stage of recruitment
for Wave 1 yielded a response rate of 68.0% -- that is, 172 of 253 eligible store employees provided
contact information during the Wave 1 recruitment.
Of the 253 employees eligible for participation in Wave 1, 215 were eligible in Wave 2.
Individuals were only deemed ineligible for Wave 2 if we had no way to contact them because they
did not return contact information at Wave 1 and they no longer worked in one of the 21
participating stores. Individuals who had left the firm but had provided contact information at
Wave 1 were eligible for Wave 2. In addition to the 215 eligible Wave 1 employees, an additional
precarious work schedules
20
41 individuals became eligible in Wave 2 because they had been hired into one of the 21 stores
since recruitment took place for Wave 1. This resulted in 256 eligible Wave 2 participants.
For Wave 2, we followed a similar process to Wave 1 to collect updated contact information
(or new contact information for employees hired since Wave 1 recruitment and employees who did
not return contact information during the first recruitment period). For respondents who were no
longer employed at one of the 21 stores and for whom we had contact information, we sent a letter
requesting updated telephone contact information to their home address. All other letters were
distributed in the stores in a manner consistent with that of Wave 1 recruitment. Overall, we
received contact information from 190 of 256 eligible participants at Wave 2. Thus, we attempted to
interview 74.2% of eligible Wave 2 participants.
The University of Wisconsin Survey Center was responsible for the second stage of the
recruitment for both Waves 1 and 2, which involved calling employees who had provided contact
information, conducting informed consent, and administering the survey. All survey respondents
received a $25 check for their participation at Wave 1 and a $50 check for participation at Wave 2.
At Wave 1, the second stage of recruitment yielded a response rate of 79.1%; that is, 136 of the 172
employees who returned response cards participated in the survey. Thus, the overall response rate
for Wave 1 is 53.8% (136 respondents out of 253 eligible store employees). At Wave 2, the second
stage of recruitment yielded a response rate of 82.1% (156 respondents out of 190 who returned
response cards), resulting in an overall response rate for Wave 2 is 60.9% (156 out of 256). Eightyeight percent of Wave 1 respondents were successfully reinterviewed in Wave 2 (120 of 136).
For the purposes of this paper, we exclude all managers (13 in Wave 1 and 15 in Wave 2)
and we exclude any Wave 2 respondents who were no longer working at the retail establishment
(20, 1 manager and 19 non-managers). This results in a sample of 123 Wave 1 respondents and 122
Wave 2 respondents. Of the 123 Wave 1 respondents in the restricted sample, 92, or 74.8%, were
precarious work schedules
21
reinterviewed at Wave 2; the remaining 21 Wave 2 employees in the restricted sample had joined
the company after Wave 1.
Measures
The majority of variables were constructed from responses to self-report survey questions
(see Appendix A for specific items). Three control variables (race, age, and employment status)
come from the personnel records of the corporation that were obtained as part of the broader
workplace intervention.
Independent Variables
Two indices created for the purposes of this study are used to capture precariousness of
employment. Schedule unpredictability is a measure that assesses the amount of advance schedule
notice and the extent to which respondents can predict their schedules based on a three-item index
(alpha=.65) that captures the extent to which respondents agreed or disagreed that they can
anticipate and/or count on the number and timing of hours that they will be working week-to-week.
Based on these two measures (advance schedule notice and hours/timing predictability), we
constructed a measure of schedule unpredictability that has three possible values including Most
Predictable (more than one week notice of work schedule and a score in the top half of the median
split of the 3-item index of hours/timing predictability, coded as 1), Somewhat Predictable (one
week or less notice of work schedule and a score in the top half of the median split of the 3-item
index of hours/timing predictability, coded as 2) and Unpredictable (one week or less notice of
work schedule and in the bottom half of the median split of the 3-item index of hours/timing
predictability, coded as 3). We treated Schedule Unpredictability as a continuous variable in the
multivariate analyses reported here; although the results are unchanged when Schedule
Unpredictability is considered categorically, with Most Predictable serving as the excluded
category.
precarious work schedules
22
Lack of Input is a 4-item index (alpha=.82) that assesses the amount of input into the
number of hours respondents report working each week, the days they work each week, the days
they have off, and their starting and ending times5.
Supervisor Support is a 4-item emotional-support subscale developed by Hammer, Kossek,
Yragui, Bodner, & Hanson (in press) that assesses the extent to which workers agree that their
immediate supervisor is willing to talk about and is interested in their work and nonwork conflicts.
(Higher scores indicate moresupport.) This variable is used in Model 5 and Model 5b to test the
robustness of our hypotheses to alternative model specifications.
Dependent variables
General Work-to-Family Conflict is a 5-item index adapted from Netemeyer, Boles, and
McMurrian (1996) that combines both time- and strain-based items (alpha=.87). For example,
workers rated how often the demands of their work interfere with their personal or family time,
create strain that makes it difficult to fulfill personal or family responsibilities, and cause them to
adjust their personal plans. Work Interferences in Nonwork Activities is a 4-item index of questions
that asked workers whether they have “more than enough time, just enough time, or not enough
time” to schedule doctor’s appointments, plan activities with friends or family, and arrange to cook
a meal at home (alpha=.84).6 Perceived Stress is an 8-item scale adopted from the 14-item scaled
developed by Cohen, Kamarck, and Mermelstein (1983). The items asked workers to report how
5
We also constructed a measure similar to our lack of input indicator that is based on three items originally intended to
be an instrumental support subscale of supervisor support. These items are "I can depend on my manager to help with
scheduling conflicts", "my manager lets me come in late or leave early to accommodate my family", "when I have to
miss work, my manager helps me make up the hours". In the paper, we do not report results for this measure, however
the multivariate results are similar regardless of whether we use this measure or the lack of input measure as an
indicator of employee input over schedules. Our preference is for the lack of input measure because it assesses
schedule input directly (e.g., simply asking about whether respondents have input over weeks, days, and start and end
times) rather than invoking the role of the manager in providing that input. Thus, we view the lack of input measure as a
cleaner and more independent measure of input than is the manager instrumental support measure.
6
We developed this scale by “unpacking” a single question that Swanberg et al. (2008) used to assess the extent to
which workers in a drug-store chain reported sufficient time to “plan personal, family, or other responsibilities.”
precarious work schedules
23
often in the prior month they experienced feelings of anxiety, personal control, and confidence in
their coping skills (alpha=.82).
Control variables
We include several measures of vulnerability as control variables in the multivariate
regressions. Personnel records provided information to construct Age, a continuous variable
calculated from the workers’ birthdate, Race, a dichotomous variable that differentiates workers
who are recorded as White (coded 1) in the company’s personnel records from other workers
(coded 0), and Part-time, a dichotomous variable that identifies workers who hold part-time (coded
1) rather than full-time (coded 0) positions at the retail firm. A series of additional dichotomous
variables are constructed from workers’ survey responses: High School Education identifies
workers who reported having no more than a high school degree (coded 1) from workers who
reported additional education (coded 0); Child under 18 in the home distinguishes workers with at
least one child under 18 who lives with them (coded 1) from workers without children younger than
18 in their home (coded 0); Respondent has Other Job indicates whether respondent reports a job in
addition to the target retail job (coded 1) or does not (coded 0). To capture both marital status and
employment status of partner, we constructed a trichotomous measure that includes Does Not Have
Partner (coded as 1), Has Partner Who Does Not Work or Works Part-Time (coded as 2), and Has
Partner Who Works Full-Time (coded as 3). Does Not Have a Partner is treated as the excluded
variable in the analyses. Subjective Health is a one-item question asking respondents to rate their
health as excellent, very good, good, fair, or poor. Given the distribution on the measure, it was
dichotomized to indicate excellent or very good health (coded as 0) or good, fair, or poor (coded as
1). We also include a control for whether the respondent worked for an experimental or control
store in all analyses, although we do not evaluate the intervention itself in this study. A variable that
precarious work schedules
24
measures the length of time between Waves 1 and 2 surveys is included in all regressions that use
both waves of data. The average time between surveys was 6 months.
Analytic Plan
First, we present descriptive statistics of the control, independent, and dependent variables
for both Waves 1 and 2. Second, using Wave 2 cross-sectional data, we examine five multiple
regression models for each of the three dependent variables. The models assess the relationship
between each dependent variable and the set of vulnerability variables, which we treat as controls
(Model 1), the unpredictable schedule variable net of control variables (Model 2, Hypothesis 1), the
lack of input variable net of control variables (Model 3, Hypothesis 2), and our two independent
variables entered simultaneously (unpredictable schedules and lack of input over schedules) with
the set of controls (Model 4). Model 4 assesses hypotheses 3 and 4 which are concerned with the
independent associations of each dimension of precarious work schedules net of the other, and
whether input into schedules compensates for the stress induced by unpredictable schedules (HO 4)
but not for the time-related conflicts (nonwork interferences) (HO 3). The fifth model includes
supervisor emotional support in the full model as a robustness check to determine if the associations
observed with the independent variables remain once level of supervisor emotional support is
considered (Model 5).7
The five models are presented together for each dependent variable on Table 3 (General
Work-to-Family Conflict), Table 4 (Interferences with Nonwork Activities), and Table 5 (Perceived
Stress). In interpreting the tables, note that the independent and dependent variables are scaled such
that positive coefficients mean that more unpredictable schedules and more limited input over
schedules are related to higher work-to-family conflict, more interferences with nonwork activities,
7
A similar robustness test was conducted with job autonomy and job challenge in the model instead of (and together
with) supervisor support. Controlling for job autonomy and challenge does not appreciably change the coefficients on
the independent variables of interest.
precarious work schedules
25
and more perceived stress. Because the respondents were drawn from 21 stores, and hence the data
are not independent, the standard errors are adjusted in all models to account for the clustering of
employees by store to obtain a robust variance estimate (Williams, 2000; Rogers, 1993).
The Wave 2 cross-sectional analyses attempt to deal with the problem of correlated error
between the independent variable and the error term by including explicit multivariate controls
measured at Wave 2 (Berk, 2004). Because unmeasured characteristics that may be correlated with
both scheduling practices and work-life outcomes are not included in the models, these multivariate
specifications are still likely to produce biased estimates of the relationships of interest (e.g., the
concern of omitted variable bias, Greene, 2000). In order to reduce this bias, we rerun the models
reported in Tables 3 through 5, taking into account prior measures of the dependent variable of
interest using Wave 1 data (see Table 6). This does not eliminate the problem of correlated error
nor does it exclude the possibility of reverse causality explaining our results. But by removing the
association of prior measures of the dependent variables on the observed Wave 2 relationships,
these regressions reduce the bias and offer a more conservative test of our hypotheses.
More specifically, in the second stage of the analyses, we rerun the five models for each
dependent variable but include in the regression the Wave 1 value of the dependent variable as a
control along with the prior control variables. We also include a control for the time that elapsed
between the Wave 1 and Wave 2 interviews, and we again adjust the standard errors to account for
nonindependence (i.e., clustering). We label these models, 1b, 2b, 3b, 4b, and 5b to differentiate
them from the core Wave 2 analyses and present the results on Table 6. To conserve space, Model
1b (the model with only control variables) is not presented on Table 6 nor are the coefficients for
the control variables for Models 2b through 5b. Table 6 presents the coefficients for schedule
unpredictability and lack of input on the Wave 2 outcomes of interest, and also shows the
precarious work schedules
26
relationship between the Wave 1 and Wave 2 outcomes. All models and their full set of coefficients
are available upon request of author.
RESULTS
Descriptives
Descriptive statistics for the sample of retail employees at each Wave are provided on
Tables 1 and 2. The sample is exclusively female, over three-fourths are non-hispanic white, and
only a minority are parenting a child under 18 years of age (20.3% in Wave 1; 22% in Wave 2). The
sample is relatively diverse by age (mean age is 47.7 years in Wave 1, 48.1 years in Wave 2), with
approximately one-fifth under 35 years of age and more than one-third over 55. Just over 40% in
both waves have more than a high school degree (approximately one-fifth have graduated from
college, not shown on Table 1). Sixty four percent of the sample report that their health was
excellent or very good (as opposed to good, fair, or poor) in Wave 1, whereas 56% reported
excellent or very good health at Wave 2. At both waves, about 40% of respondents are neither
married nor living with a partner, about 42% are married or living with a partner who works full
time, and the remainder are married or living with a partner who is either unemployed or working
part time. Wages for respondents are above the legal minimum, but arguably low at $9.36 per hour
at Wave 1 and $9.20 at Wave 2. The average wages of full-time employees are approximately four
dollars higher than part-time employees. Well over one-third of respondents (38.2% in Wave 1,
40% in Wave 2) hold a job at another place of employment in addition to the position they have at
the retail firm.
Table 2 reports the means and standard deviations of the two independent variables
(unpredictable work schedules and lack of input), supervisor emotional support, and the three
dependent variables for both Waves. As Table 2 indicates, the scores are quite similar across the
two Waves.
precarious work schedules
27
Multivariate Analyses
Table 3 reports the results of the five models for general work-to-family conflict. Model 1
indicates that respondents with more than one job report significantly more work-to-family conflict
than those with only one job and part-time employees compared to fulltime employees report
significantly less work-to-family conflict. These results remain significant across models.
Respondents who report poorer health (or health that is something other than “excellent” or “very
good”) have marginally greater work-to-family conflict in Model 1; the relationship holds when
schedule unpredictability is included in the model (Model 2), but does not remain significant across
the other models. Although age is not significant in Model 1, it becomes so when schedule
unpredictability is taken into account together with the other variables (Models 2, 4, and 5) such
that younger respondents report more work-to-family conflict. Somewhat surprisingly, respondents
with children under 18 in the household report less work-to-family conflict, although the
relationship is only significant in Models 1 and 2.
Controlling for these measures of vulnerability, we find that unpredictable schedules (Model
2) and lack of input into schedules (Model 3) are both postively related to general work-to-family
conflict. Both variables remain highly significant in Model 4 (when schedule unpredictability and
input are considered together). These findings lend support to the first two study hypotheses. Model
5 considers the possibility that supervisor emotional support acts as a resource that can reduce
general work-to-family conflict. The results lend support to this hypothesis as respondents who
report less agreement with the supervisor support items report marginally significantly more workto-family conflict. Moreover, taking supervisor support into account diminishes (but not to
nonsignificance) the relationships between work-to-family conflict and both schedule
unpredictability and input . Thus, these results suggest that for our general work-to-family conflict
measure, which includes aspects of both time-based and strain-based conflict, having supervisor
precarious work schedules
28
support somewhat ameliorates the challenges posed by unpredictable schedules and limited
schedule input, but it does not fully compensate for them.
Table 4 reports the results of the five models for interferences with nonwork activities. We
hypothesized that the time-based conflict induced by unpredictable schedules (HO1) and limited
employee control (HO2) would be positively associated with interferences with nonwork activities,
and that greater input into schedules would be insufficient to eliminate the time-based conflicts
posed by unpredictability (HO3). These hypotheses were supported.
Model 1 indicates that, with the exception of health, none of the vulnerability variables are
significantly related to the interference measure. This lack of association is constant across the
models. As with general work-to-family conflict, we find that unpredictable schedules (Model 2)
and lack of input into schedules (Model 3) are both significantly related to interferences with
nonwork activities. The coefficient for both variables dissipates somewhat in Model 4, when input
into schedules is considered together with unpredictability, although both unpredictability and input
remain significant. Thus, consistent with hypothesis three, the results suggest that although input
into schedules can reduce nonwork interferences, it cannot completely compensate for the
challenges created by unpredictable schedules that contribute to nonwork interferences. In Model
5, when supervisor emotional support is considered together with unpredictability and input, the
coefficients of both input and unpredictability are somewhat reduced. However, schedule
unpredictability remains significantly related to interferences with nonwork activities, input into
schedules is marginally significant, and the supervisor support variable itself is not significant.
Thus, these results do not provide strong evidence that supervisor support mediates the relationship
between precarious schedules and interferences with nonwork activities.
Table 5 reports the results of the five models for perceived stress. We hypothesized that
unpredictable schedules (HO1) and limited input into schedules (HO2) would pose strain-based
precarious work schedules
29
challenges that contribute to perceived stress. We further hypothesized that unlike time-based
conflicts, the strain-based conflict resulting from unpredictable work schedules could be
ameliorated for workers with significant input over their schedules (HO4). As elaborated below,
our first two hypotheses were confirmed. However, we do not find support for our hypothesis that
having input into schedules reduces the association between time-based conflict and perceived
stress. In fact, for this sample, schedule unpredictability is a more powerful predictor of stress than
having input into the scheduling process.
As reported on Table 5, respondents with poorer health report marginally significantly more
perceived stress, and respondents with less education report significantly lower levels of perceived
stress (Model 1). These results are consistent across the models. Although having another job is not
significant in Model 1, it becomes marginally significant in the other models such that respondents
with more than one job report more perceived stress. Respondents who are not married or
cohabiting report marginally less stress than those who are married or cohabiting to a full-time
employed partner, in all models except Model 1.
As hypothesized, having an unpredictable work schedule (Model 2) and having limited input
into work schedules (Model 3) are both related to significantly higher levels of perceived stress.
However, contrary to our hypothesis, we do not find evidence that having input into one’s schedule
reduces the association between unpredictable schedules and perceived stress. To the contrary,
although the schedule unpredictability and lack of input coefficients are both slightly reduced in
Model 4, the coefficient for schedule unpredictability remains significant, whereas the coefficient
for input is reduced to nonsignificance. Adding supervisor emotional support to the model together
with schedule unpredictability and schedule input (Model 5) does not change the size or
significance level of the coefficient for either scheduling variable. Moreover, and somewhat
surprisingly, there is no observable relationship between supervisor support and perceived stress.
precarious work schedules
30
Table 6 summarizes the results of the stage two analyses that replicate Models 1 – 5 (here
referred to as Models 1b – 5b), with the addition of a control variable that takes into account Wave
1 levels of the dependent variable. To conserve space on the table, we do not report the coefficients
for the vulnerability measures or other control variables. The results are presented on Table 6 as
three horizontal panels, one for each dependent variable. As with Tables 3-5, the models are
represented as columns in Table 6, although we have excluded Model 1b (the vulnerability model
without scheduling measures).
As the first row of each panel indicates, and as would be expected, the Wave 1 indicator of
the outcome is highly related to the Wave 2 indicator of the outcome such that Wave 1 general
work-to-family conflict is significantly and strongly related to Wave 2 general work-to-family
conflict, Wave 1 interferences with nonwork activities is significantly and strongly related to Wave
2 interferences with nonwork activities, and so on. Taking those relationships into account, we find
that schedule unpredictability continues to be significantly associated with each of the dependent
measures (Model 2b), and the significant association remains in each of the models even when lack
of input and supervisor support are considered together with unpredictability (Models 4b & 5b).
The scheduling input measure proves less robust. Although in the earlier models lack of input into
schedules was found to be positively related to general work-to-family conflict, interferences with
nonwork activities, and perceived stress, this relationship only holds for perceived stress when the
Wave 1 measure of these dependent variables is taken into account (top and middle panel, Models
3b, 4b, & 5b). That is, when we control for Wave 1 indicators of perceived stress, lack of input at
Wave 2 is significantly related to Wave 2 perceived stress in Model 3b. This relationship
diminishes in Model 4b and 5b, although it is still significant in 5b (when emotional support is
considered). Overall, schedule unpredictability but not schedule input proves to be quite robust
precarious work schedules
31
across the different model specifications and across different outcomes, suggesting its salience as a
contributor to both time-based and strain-based work-family challenges.
DISCUSSION
This study explored a behavioral and stress model in an examination of work schedule
precariousness (unpredictability and limited input into schedules) and its implications for three
outcomes related to work-family well-being. The three measures were chosen to represent a general
measure of general work-family conflict that includes both stress and time-based conflict, as well as
a behavioral measure addressing time-based conflict (interferences with nonwork activities) and a
perceived stress measure, addressing strain-based conflict. The study findings are based on a sample
of retail workers drawn from 21 Midwestern stores of a U.S. retail apparel firm. The exclusively
female sample represented diverse individual and family circumstances, although their employment
was exclusively low-wage and primarily part-time. Focusing on employees retail jobs helps
expands the work-life field by providing insight into the work-life challenges and experiences of
workers in a growth sector of the economy that is characterized by low wages and precarious
scheduling practices (Presser, 2003).
Our analyses lend support to the first three study hypotheses, but not the fourth. In
particular, we find that net of individual, family, and other work characteristics, unpredictability and
limited schedule input are related to each of the outcomes, suggesting that precarious work
schedules are an important source of time-based and strain-based work-to-family conflict. We
further find that, as hypothesized, scheduling input does not significantly reduce the association
between unpredictable schedules and interferences with nonwork activities (such as planning social
activities, a home-cooked meal with family, or making doctor’s appointments). Contrary to our
hypothesis, however, we also do not find that input into scheduling compensates for an
unpredictable schedule in terms of reducing workers’ perceived stress.
precarious work schedules
32
Our central findings are robust to several model specifications. In particular, the general
pattern of results holds, net of our control measures which take into account personal and family
vulnerabilities among the sample, and net of supervisor support which has been shown to affect a
host of work-family related outcomes in previous research (Frye & Breaugh, 2004; Warren &
Johnson, 1995). The results for schedule unpredictability are robust to an even more conservative
test of the hypotheses, that is, when the Wave 1 measure of the dependent variable is controlled for
in the models. Our measure of schedule input proved less robust in these two-wave models,
maintaining its significant relationship to only one of the dependent variables, perceived stress. In
results not shown, we also find that the results are replicated when other work-related factors such
as degree of job control and job autonomy are included in the regressions. Moreover, although
somewhat less stable across model specifications, we find the same pattern of results when a lagged
model is assessed; that is, when Wave 1 measures of schedule unpredictability and lack of input
(together with Wave 1 control variables) are used to predict Wave 2 outcome measures (also not
shown, but available upon request of author). The robustness of our results is especially noteworthy
given the study’s limited power due to a relatively small sample and the large number of control
variables included in the models.
These analyses represent an initial attempt to explore relationships between precarious
employment schedules and individual and family life. We are currently subjecting these models to
further scrutiny, considering other factors that may explain the relationships observed among the
cross-sectional survey data, and considering conditions that may moderate the relationships. For
example, we are examining further the mostly null effects for household composition and spousal
employment, and considering alternative methods of specifying relationship status and household
circumstances, and revising our models to reflect the possibility that particular household
arrangements may interact with schedule predictability and input. The survey includes measures on
precarious work schedules
33
the work schedules of partners and other adults in the household as well as including significant
information about the work schedules of respondents’ second job, for those holding more than one
job. These variables may prove important to identifying distinct household-employment
vulnerabilities that interact with precarious schedules in meaningful ways.
In addition, we are developing additional measures of employment precariousness from the
multiple data sources available to us through the broader workplace intervention on which the
current study is based. One source of data is a separate manager telephone survey that collects data
on management hiring and scheduling practices for each of the 21 stores.8 A second source of data
is actual employee schedules and payroll hour reports from each of the 21 stores. Because we have
more than nine months of weekly store schedules for each employee at each store, as well as
information indicating the actual dates that schedules were posted, we are able to calculate an
unpredictability measure that reflects a true measure of advanced scheduling notice (rather than
relying on self-report data). Moreover, because we also have company payroll reports that indicate
the exact days and hours employees worked over this same period, we are able to construct
objective measures of nonstandard schedules (e.g., nonstandard timing), variable schedules
(variation in the days and shifts worked week-to-week), and fluctuating weekly hours. These
objective measures of precarious schedules will allow us to construct analytic models that are less
subject to the endogeneity concerns of our current models that rely on the survey data.
We have attempted to explore the relationships of interest within the constraints imposed by
limited power. Although several model specifications were considered in an effort to limit the
biases inherent in correlational research designs, we are unable to control for omitted variables or
satisfactorily address reverse causality concerns. Thus, we remain cautious regarding our
8
The manager survey was administered to approximately 150 store managers employed by this retail firm. Because the
employee survey on which the current study is based was restricted to the 21 Midwestern stores, further analyses with
these employee survey data may benefit from linking relevant information on labor flexibility practices from the 21
corresponding manager surveys to employee survey data.
precarious work schedules
34
interpretation of the observed relationships, and it is important to consider our results in light of
several endogeneity concerns. For example, although our conceptual model is concerned with
pathways emanating from work to other domains, our Wave 2 cross-sectional results do not rule out
the possibility that our “outcomes” are themselves contributing to precarious work schedules. For
example, respondents who are under significant stress and whose households operate with limited
predictability in routines may experience less predictable schedules because their nonwork lives
preclude them from holding a more regular schedule. Moreover, higher levels of perceived stress
(wherever it originates) may influence the level of input into scheduling that managers offer
employees. Indeed, to the extent that managers consider employee input into schedules a worker
reward to be earned (Henly et al. 2006), employees who are under stress or otherwise not
performing up to manager standards may be given fewer opportunities to provide input into their
schedules. We respond to these alternative interpretation issues by including Wave 1 indicators of
each of the dependent variables in the second set of analyses, but of course these controls do not
eliminate the possibility that Wave 2 outcomes operate on Wave 2 predictors. In future work, we
will further exploit the longitudinal data to examine the hypothesized relationships, constructing
measures of change between Waves 1 and 2 and incorporating store-level data on changes in work
schedules over the course of the intervening period between the two survey waves. Although
longitudinal correlational data still preclude true causal effects to be observed, it is possible to
further examine lagged associations and make use of repeated measures to more carefully examine
correlated changes over time.
Overall, the results reported in this paper suggest that unpredictable schedules and schedules
that offer employees limited input into their work hours have negative consequences for individual
and family life. Precarious scheduling practices are not isolated within a few organizations but
rather reflect growing national and international trends that have profound implications for
precarious work schedules
35
individual and family security (Kalleberg, 2009). Thus, our models for understanding work-life and
work-family linkages must take into account the increasing precariousness of employment
conditions and the changing nature of employer-worker relationships if they are to be useful
explanatory tools.
precarious work schedules
36
Table 1 Sample Descriptives
Wave 1
(n=123)
Wave 2
(n=120)
Race
Black (%)
White (%)
Latina or Hispanic (%)
12.30
77.05
9.02
12.50
80.00
6.67
Child under 18 in home (%)
20.3
22.00
47.72 (14.67)
21.95
37.40
48.05 (13.84)
19.17
35.00
More than high school education (%)
43.09
42.50
Rates health as excellent or very good (%)
Family Structure
Not married or cohabitating (%)
64.23
55.83
40.65
41.67
17.07
42.28
15.83
42.50
$9.36 (2.22)
$9.20 (2.18)
38.21
40.00
Age: mean (sd)
% under 35
% over 55
Married or cohabiting, partner does not have a full-time
job (either not working or working part-time) (%)
Married or cohabiting, partner has a full-time job (%)
Hourly wage: mean (sd)
Additional job (%)
precarious work schedules
Table 2: Descriptives of Independent and Dependent Variables
Wave 1
Wave 2
(n=123)
(n=120)
Unpredictable work schedule
2.33
2.31
mean (sd)
(0.64)
(0.64)
Input over schedule
2.11
2.22
mean (sd)
(0.84)
(0.80)
Emotional support from supervisor
1.66
1.94
mean (sd)
(0.51)
(0.65)
Work-family conflict
2.25
2.27
mean (sd)
(0.79)
(0.82)
Interference with nonwork activities
1.73
1.70
mean (sd)
(0.64)
(0.63)
Perceived stress scale
2.23
2.26
mean (sd)
(0.62)
(0.66)
Independent variables:
Dependent variables:
37
precarious work schedules
Table 3: Dependent Variable: General Work-to-Family Conflict, Models 1 – 5, Wave 2
Model 1
Model 2
Model 3
Model 4
Model 5
1+2+3+
supervisor
vulnerability
1+
1 + lack of
emotional
VARIABLES
(controls)
unpredictability
input
1+2+3
support
Health is good, fair,
or poor
Age
Race (white vs
others)
Education [more
than HS (0) vs no
more than HS (1)]
Child under 18
Not Married or
Cohabiting (vs M/C
and spouse is fulltime employed)
Married or
Cohabiting and
spouse is
unemployed or parttime employed
Respondent has
other job
Part-time at retail job
Schedule
unpredictability
Lack of input
Supervisor emotional
support
0.2883
(0.1786)
-0.0059
(0.0039)
0.2827+
(0.1600)
-0.0088*
(0.0035)
0.1995
(0.1770)
-0.0048
(0.0034)
0.2186
(0.1631)
-0.0074*
(0.0032)
0.2221
(0.1558)
-0.0095**
(0.0033)
0.3332
(0.1951)
0.2483
(0.1888)
0.2365
(0.1888)
0.1907
(0.1881)
0.2405
(0.1937)
-0.1958
(0.1515)
-0.4664**
(0.1612)
-0.1760
(0.1393)
-0.3819*
(0.1549)
-0.2334
(0.1467)
-0.2982
(0.1903)
-0.2062
(0.1337)
-0.2653
(0.1746)
-0.1574
(0.1244)
-0.2897+
(0.1546)
-0.0239
(0.1384)
-0.0921
(0.1475)
-0.0700
(0.1439)
-0.1156
(0.1466)
-0.1266
(0.1523)
-0.0216
(0.1826)
-0.0501
(0.1781)
0.0394
(0.1737)
-0.0052
(0.1761)
-0.0383
(0.1608)
0.5668**
(0.1585)
0.6186**
(0.1640)
0.7092**
(0.1539)
0.7199**
(0.1572)
0.6994**
(0.1633)
-0.4670*
(0.1707)
-0.4476*
(0.1717)
-0.5001*
(0.1794)
-0.4710*
(0.1783)
-0.4594*
(0.1774)
0.3211**
(0.1051)
0.3022**
(0.0965)
0.2490*
(0.0967)
0.2263*
(0.0988)
0.1834+
(0.0972)
0.3702**
(0.0969)
0.2378+
(0.1160)
Observations
120
119
120
119
119
R-squared
0.273
0.362
0.349
0.405
0.430
Beta coefficients are unstandardized; Robust standard errors in parentheses. **p<0.01,
*p<0.05, +p<0.1. Note: All models adjusted for intervention condition
38
precarious work schedules
Table 4: Dependent Variable: Interferences with Nonwork Activities, Models 1 – 5, Wave 2
Model 1
Model 2
Model 3
Model 4
Model 5
1+2+3+
supervisor
vulnerability
1+
1 + lack of
emotional
VARIABLES
(controls)
unpredictability
input
1+2+3
support
Health is good, fair, or
poor
Age
Race (white vs others)
Education [more than
HS (0) vs no more than
HS (1)]
Child under 18
Not Married or
Cohabiting (vs M/C and
spouse is full-time
employed)
Married or Cohabiting
and spouse is
unemployed or part-time
employed
Respondent has other
job
Part-time at retail job
0.2735*
(0.1183)
0.0009
(0.0038)
0.2206*
(0.0927)
-0.0010
(0.0033)
0.1951+
(0.1119)
0.0020
(0.0034)
0.1638+
(0.0934)
0.0002
(0.0032)
0.1661+
(0.0957)
-0.0011
(0.0032)
0.0627
(0.1701)
-0.0365
(0.1772)
-0.0225
(0.1610)
-0.0876
(0.1700)
-0.0557
(0.1759)
-0.0714
(0.0941)
-0.2747
(0.1630)
-0.0852
(0.0877)
-0.2180
(0.1538)
-0.1045
(0.0992)
-0.1263
(0.1522)
-0.1120
(0.0877)
-0.1146
(0.1474)
-0.0806
(0.0920)
-0.1302
(0.1433)
0.0894
(0.1072)
0.0298
(0.1181)
0.0487
(0.1051)
0.0090
(0.1115)
0.0019
(0.1114)
-0.0187
(0.1448)
0.0519
(0.1430)
0.0351
(0.1416)
0.0917
(0.1415)
0.0704
(0.1333)
-0.1627
(0.1315)
-0.1191
(0.1353)
-0.0371
(0.1176)
-0.0294
(0.1244)
-0.0425
(0.1249)
0.0153
(0.1575)
-0.0412
(0.1395)
-0.0140
(0.1463)
-0.0619
(0.1359)
-0.0545
(0.1407)
0.2833*
(0.1056)
0.2817**
(0.0579)
0.2208*
(0.0966)
0.2330**
(0.0812)
0.1787+
(0.1019)
Schedule
unpredictability
0.3420**
(0.0639)
Lack of input
Supervisor emotional
support
Observations
R-squared
0.1527
(0.1080)
120
0.126
119
0.255
120
0.226
119
0.312
Beta coefficients are unstandardized; Robust standard errors in parentheses. **p<0.01,
*p<0.05, +p<0.1. Note: All models adjusted for intervention condition
119
0.329
39
precarious work schedules
Table 5: Dependent Variable: Perceived Stress, Models 1 – 5, Wave 2
Model 1
Model 2
Model 3
VARIABLES
Health is good, fair, or
poor
Age
Race (white vs others)
Education [more than
HS (0) vs no more than
HS (1)]
Child under 18
Not Married or
Cohabiting (vs M/C and
spouse is full-time
employed)
Married or Cohabiting
and spouse is
unemployed or part-time
employed
Respondent has other
job
Part-time at retail job
vulnerability
(controls)
1+
unpredictability
1 + lack of
input
1+2+3
Model 5
1+2+3+
supervisor
emotional
support
0.3486+
(0.1707)
-0.0007
(0.0033)
0.0882
(0.1755)
0.3297*
(0.1519)
-0.0025
(0.0032)
0.0209
(0.1792)
0.2930+
(0.1622)
0.0001
(0.0029)
0.0277
(0.1593)
0.2914+
(0.1502)
-0.0017
(0.0029)
-0.0136
(0.1688)
0.2914+
(0.1518)
-0.0017
(0.0029)
-0.0126
(0.1724)
-0.2467*
(0.1166)
-0.0822
(0.1169)
-0.2426*
(0.1122)
-0.0284
(0.1110)
-0.2702*
(0.1056)
0.0231
(0.1309)
-0.2606*
(0.1011)
0.0414
(0.1245)
-0.2597*
(0.1016)
0.0409
(0.1236)
-0.1821
(0.1201)
-0.2299+
(0.1249)
-0.2110+
(0.1170)
-0.2440+
(0.1213)
-0.2442+
(0.1212)
-0.0833
(0.1380)
-0.0734
(0.1452)
-0.0451
(0.1549)
-0.0466
(0.1620)
-0.0472
(0.1642)
0.2150
(0.1373)
-0.1202
(0.1402)
0.2508+
(0.1414)
-0.1296
(0.1402)
0.3041+
(0.1513)
-0.1410
(0.1399)
0.3113+
(0.1516)
-0.1436
(0.1413)
0.3109+
(0.1531)
-0.1433
(0.1427)
0.2010*
(0.0899)
0.2242**
(0.0771)
0.1490
(0.0938)
0.2227*
(0.0882)
0.1477
(0.0970)
Schedule
unpredictability
0.2649**
(0.0810)
Lack of input
Model 4
Supervisor emotional
support
Observations
R-squared
0.0047
(0.1085)
120
0.165
119
0.232
120
0.211
119
0.256
Beta coefficients are unstandardized; Robust standard errors in parentheses. **p<0.01,
*p<0.05, +p<0.1. Note: All models adjusted for intervention condition
119
0.256
40
precarious work schedules
Table 6: Wave 2 Regressions Controlling for Wave 1 Outcome Levels, All Dependent Variables
General Work-to-Family Conflict
Wave 1 General work-to-family
conflict
Schedule unpredictability
Model 2b
Model 3b
Model 4b
Model 5b
0.3949**
(0.0948)
0.2950**
(0.0809)
0.4429**
(0.0956)
0.1263
(0.1162)
0.3884**
(0.0946)
0.2803**
(0.0823)
0.0472
(0.0947)
0.3752**
(0.0941)
0.2118*
(0.0779)
0.0104
(0.1092)
0.1799+
(0.0955)
92
0.583
92
0.534
92
0.585
92
0.602
Model 2b
Model 3b
Model 4b
Model 5b
0.4143**
(0.0895)
0.2757**
(0.0718)
0.4833**
(0.0971)
0.0319
(0.1244)
0.4323**
(0.0929)
0.2855**
(0.0730)
-0.0411
(0.1050)
0.4243**
(0.0980)
0.2558*
(0.0939)
-0.0540
(0.1123)
0.0788
(0.1189)
92
0.431
92
0.359
92
0.433
92
0.437
Model 2b
0.6769**
(0.0826)
0.2083**
(0.0720)
Model 3b
0.7067**
(0.0994)
Model 4b
0.6664**
(0.0933)
0.1753*
(0.0670)
0.1023
(0.0655)
Model 5b
0.6887**
(0.0930)
0.2090**
(0.0732)
0.1212*
(0.0580)
-0.0965
(0.0937)
Lack of input
Supervisor emotional support
Observations
R-squared
Interferences with Nonwork
Activities
Wave 1 Interferences with
Nonwork Activities
Schedule unpredictability
Lack of input
Supervisor emotional support
Observations
R-squared
Perceived Stress
Wave 1 Perceived stress
Schedule unpredictability
Lack of input
Supervisor emotional support
0.1531*
(0.0643)
Observations
92
92
92
92
R-squared
0.611
0.595
0.621
0.627
Beta coefficients are unstandardized; Robust standard errors in parentheses
**p<0.01, *p<0.05, +p<0.1
Note: All regressions adjusted for experimental condition, time between survey waves, health
status, age, race, education, kids, partner job, respondent other job, and part-time retail job.
Sample includes only those employees who were working at the store at both waves of the survey.
41
precarious work schedules
42
References
Acs, G. & Loprest, P. (2008). Understanding the demand side of the low-wage labor market.
Research Report. Washington DC: Urban Institute.
http://www.urban.org/url.cfm?ID=411680
Almeida, D.M. & McDonald, D. (2005). The time Americans spend working for pay, caring for
families, and contributing to communities. In J. Heymann (Ed.) Unfinished Work: Balancing
Equality and Democracy in an Era of Working Families, (pp.180-203). New York: The
New Press.
Appelbaum, E., Bailey, T., Berg, P., & Kalleberg, A. (2000). Manufacturing Advantage: Why High
Performance Work Systems Pay Off. Ithaca, New York: Cornell University Press.
Appelbaum, E., Bernhardt, A., & Murnane (Eds.) (2003). Low-wage America: How Employers are
Reshaping Opportunity in the Workplace. New York: Russell Sage Foundation.
Barnett, R., & Hyde, J. (2001). Women, men work, and family: an expansionist theory. American
Psychologist, 56, 781-796.
Baron, J., & Bielby, W. (1980). Bringing the firms back in: Stratification, segmentation, and the
organization of work. American Sociological Review, 45, 737-765.
Berk, R. (2004). Regression Analysis: A Constructive Critique. Thousand Oaks, CA: Sage.Blank,
R., Danziger, S., & Schoeni, R. (2006). Working and poor: how economic and policy
changes are affecting low-wage workers. Russell Sage Foundation.
Bohen, H.H., & Viveros-Long, A. (1981). Balancing jobs and family life. Do flexible work
schedules help? Philadelphia: Temple University Press.
Bureau of Labor Statistics, US Department of Labor, (April 3, 2009). The Employment Situation
Summary, Economic News Release. http://www.bls.gov/news.release/empsit.nr0.htm.
Bureau of Labor Statistics, US Department of Labor, (2008). Issues in Labor Statistics: Involuntary
Part-time Work on the Rise. http://www.bls.gov/opub/ils/pdf/opbils71.pdf. These trends are
continuing in 2009 (Employment Situation Summary).
Burke, R.J. (1988). Some antecedents and consequences of work-family conflict. Journal of Social
Behavior and Personality, 3(4), 287-302.
Campbell, M. and Moen, P. (1992). Job-Family Role Strain Among Single Mothers of
Preschoolers. Family Relations, 42, 205-211.
Christensen, K. (1998). Countervailing human resource trends in family-sensitive firms. In K.
Barker & K. Christensen (Eds.) Contingent Work: American Employment Relations in
Transition. Ithaca, NY: ILR Press.
precarious work schedules
43
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress. Journal
of Health and Social Behavior, 24, 385-396
Eckenrode, J., & Gore, S. (1990). Stress and coping at the boundary of work and family. In J.
Eckenrode & S. Gore (Eds.) Stress Between Work and Family. New York: Plenum.
Edwards, J. & Rothbard, N. (2005). Work and family stress and well-being: An integrative model
of person-environment fit within and between the work and family domains. In E. Kossek
& S. Lambert (Eds.) Work and Life Integration: Organizational, Cultural, and Individual
Perspectives. Mahwah, NJ: Lawrence Erlbaum Associates.
Fenwick, R. & Tausig, M. (2005). Work schedules and stress. In C. Epstein & A. Kalleberg (Eds.)
Fighting for Time: Shifting Boundaries of Work and Social Life, pps. 77-110. Russell Sage
Foundation.
Frone, M., Russell, M., & Barnes, G. (1996). Work-family conflict, gender, and health-related
outcomes: A Study of employed parents in two community samples. Journal of
Occupational Health Psychology, 1, 57-69.
Frone, M., Russell, M., & Cooper, M. (1997). Relation of work-family conflict to health outcomes:
A four-year longitudinal study of employed parents. Journal of Occupational and
Organizational Psychology, 70, 325-335.
Frone, M., Yardley, U., & Markel, K. (1997). Developing and testing an integrative model of the
work-family interface. Journal of Vocational Behavior, 50, 145-167.
Frye, N.K. & Breaugh, J.A. (2004). Family-friendly policies, supervisor support, work-family
conflict, family-work conflict, and satisfaction: A test of a conceptual model. Journal of
Business and Psychology, 19(2).
Galinsky, E., Hughes, D., & David, J. (1990). Trends in corporate family-supportive policies.
Marriage and Family Review, 15, 88.
Golden, L. (2005). The flexibility gap: Employee access to flexibility in work schedules. In I.U.
Zeytinoglu (Ed.) Flexibility in Workplaces: Effects on Workers, Work Environment and the
Unions. Geneva: IIRA/ILO.
Googins, B. (1991). Work/family Conflict: Private Lives, Public Responses. New York: Auburn
House.
Gottschalk, P., & Danziger, S. (2005). Inequality of wage rates, earnings, and family income in the
United States, 1975-2002. Review of Income and Wealth, 51(2), 231-254.
Grant-Vallone, E. & Donaldson, S. (2001). Consequences of work-family conflict on employee
well-being over time. Work and Stress, 15, 214-226.
precarious work schedules
44
Greene, W.H. (2000). Econometric analysis (4th ed.). Englewood Cliffs, NJ: PrenticeHall.Greenhaus, J. & Beutell, N. (1985). Sources of conflict between work and family roles.
Academy of Management Review, 10, 76-88.
Greenhaus, J. & Powell, G. (2006). When work and family are allies: A theory of work-family
enrichment. Academy of Management Review, 31, 72-92.
Grzywacz, J.G., Almeida, D.M., & McDonald, D.A. (2002). Work-family spillover and daily
reports of work and family stress in the adult labor force. Family Relations, 51, 28-36.
Grzywacz, J. G. & Bass, B. (2003). Work, family, and mental health: Testing different models of
work-family fit. Journal of Marriage and Family, 65, 248-262.
Hacker, J. (2006). The Great Risk Shift: The assault on American jobs, families, health care, and
retirement and how you can fight back. New York: Oxford University Press.
Henly, J., & Lambert, S. (2005). Nonstandard work and child care needs of low-income parents. In
S. Bianchi, L. Casper, & R. King (Eds.) Work, Family, Health, & Well-being, pps. 472-491.
Mahwah, NJ: Lawrence Erlbaum Associates.
Henly, J. R., Shaefer, H.L., & Waxman, R.E. (2006). Nonstandard work schedules: Employer- and
employee-driven flexibility in retail jobs. Social Service Review, 80, 609–634.
Herzenberg, S., Alic, J. & Wial, H. (1998). New rules for a new economy. Ithaca; Cornell
University Press.
Heymann, J. (2000). The Widening Gap. New York, NY: Basic Books.
Jacobs, D. (1994). Organizational theory and dualism: Some sociological determinants of spot and
internal labor markets. Research in Social Stratification and Mobility, 13, 203-235.
Kahn, R.L., Wolfe, D.M., Quinn, R.P., Snoek, J.D., & Rosenthal, R.A. (1964). Organizational
Stress: Studies in Role Conflict and Ambiguity. New York: Wiley.
Kalleberg, A.L. (2009). Precarious work, insecure workers: Employment relations in transition.
American Sociological Review, 74, 1-22.
Kalleberg, A., Reskin, B., & Hudson, K. (2000). Bad jobs in America: standard and nonstandard
employment relations and job quality in the United States. American Sociological Review,
65, 256-278.
Kelloway, E., Gottlieb, B., & Barham, L. (1999). The source, nature, and direction of work and
family conflict: A longitudinal investigation. Journal of Occupational Health Psychology,
4, 337-346.
precarious work schedules
45
Kossek, E.E. (2005). Workplace policies and practices to support work and families. In S. Bianchi,
L. Casper & R. King (Eds.) Work, Family, Health, and Well-being, pps. 97-116. Mahwah,
New Jersey: Lawrence Erlbaum.
Kossek, E., E., Lautsch, B., & Eaton, S. (2005). Flexibility enactment theory: Implications of
flexibility type, control, and boundary management for work-family effectiveness. In E.
Kossek & S. Lambert (Eds.) Work and Life Integration: Organizational, Cultural, and
Individual Perspectives. Mahwah, NJ: Lawrence Erlbaum Associates.
Lambert, S. (1990). Processes linking work and family: A critical review and research agenda.
Human Relations, 43, 239-257.
Lambert, S.J. (1999). Lower-wage workers and the new realities of work and family. Annals of the
American Academy of Political and Social Science, 562(March), 174-190.
Lambert, S. (2008). Passing the buck: Labor flexibility practices that transfer risk onto hourly
workers. Human Relations, 61(9), 1203-1227.
Lambert, S. (2009). Making a difference for hourly employees. In A.C. Croutere & A. Booth
(Eds.), Work-Life Policies. Washington DC: Urban Institute Press.
Lambert, S., & Waxman, E. (2005). Organizational stratification: Distributing opportunities for
work-life balance. In E. Kossek & S. Lambert (Eds.) Work and Life Integration:
Organizational, Cultural, and Individual Perspectives. Mahwah, NJ: Lawrence Erlbaum
Associates.
Lein, L., Benjamin, A., McManus, M., & Roy, K. (2005). Economic roulette: When is a job not a
job? Community, Work, & Family, 8, 359-378.
Major, V.S., Klein, K., & Ehrhart, M. (2002). Work time, work interference with family, and
psychological distress. Journal of Applied Psychology, 87, 427-436.
Marks, S. (1977). Multiple roles and role strain: Some notes on human energy, time, and
commitment. American Sociological Review, 42, 921-936.
Matthews, L. Conger, R., & Wickrama, K.A.S. (1996). Work-family conflict and marital quality:
Mediating processes. Social Psychology Quarterly, 59, 62-79.
McLoyd, V.C., Toyokawa, T., Kaplan, R. (2008) Work demands, work-family conflict, and child
adjustment in African-American families. Journal of Family Issues, 29(10), 1247-1267.
Moen, P., & Yu, Y. (1999). Having it all: Overall work/life success in two-earner families.
Research in the Sociology of Work, 7, 109-139.
Moss, P., Salzman, H., & Tilly, C. (2005). When firms restructure: Understanding work-life
outcomes. In E. Kossek & S. Lambert (Eds.) Work and Life Integration: Organizational,
Cultural, and Individual Perspectives. Mahwah, NJ: Lawrence Erlbaum Associates.
precarious work schedules
46
Netemeyer, R.G., Boles, J.S., & McMurrian, R. (1996). Development and validation of work-family
conflict and family-work conflict scales. Journal of Applied Psychology, 81, 400-410.
Osterman, P. (1999). Securing Prosperity: The American Labor Market: How It Has Changed and
What to Do About It. Princeton: Princeton University Press.
Padovic, I., & Reskin, B.F. (2002). Women and Men at Work (2nd ed.). Thousand Oaks, CA: Pine
Forge Press.
Perry-Jenkins, M. (2005). Work in the working-class: Challenges facing families. In S. Bianchi, L.
Casper, & R. King (Eds.) Work, Family, Health, & Well-being, pps. 453-471. Mahwah, NJ:
Lawrence Erlbaum Associates.
Perry-Jenkins, M. Repetti, R. & Crouter, A. (2000). Work and families in the 1990s. Journal of
Marriage and the Family, 62, 981-998.
Pleck, J.H., Staines, G.L., & Lang, L. (1980). Conflicts between work and family life. Monthly
Labor Review, 103(3), 29-32.
Pleck, Joseph H. 1995. "Introduction. Work Roles, Family Roles, and Well-Being: Current
Conceptual Perspectives." Pp. 17-22 in The Work and Family Interface: Toward a
Contextual Effects Perspective, edited by G. L. Bowen and J. F. Pittman. Minneapolis:
National Council on Family Relations.
Presser, H.B. (2003). Working in a 24/7 Economy: Challenges for American Families. New York:
Russell Sage Foundation.
Presser, H.B. & Cox, A.G. (1997). The work schedules of low-educated American women and
welfare reform. Monthly Labor Review, April, 25-34.
Rayman, P.M., & Bookman, A. (1999). Creating a research and public policy agenda for work,
family and community. Annals of the American Academy of Political and Social Sciences,
562, 191-211.
Rogers, W. H. (1993). Regression standard errors in clustered samples. Stata Technical Bulletin, 13,
19–23.
Ross, L.T. & Hill, E.M. (2000). The Family Unpredictability Scale: Reliability and Validity.
Journal of Marriage and Family, 62, 549-562.
Roy, K., Tubbs, C., & Burton, L. (2004). Don’t have no time: Daily rhythms and organization of
time for low-income families. Family Relations, 52, 168-178.
Ruderman, M., Ohlott, P., Panzer, K., & King, S. (2002). Benefits of multiple roles for managerial
women. Academy of Management Journal, 45, 369-386.
precarious work schedules
47
Scott, E.K., London, A.S., Hurst, A. (2005). Instability in patchworks of child care when moving
from welfare to work. Journal of Marriage and Family, 67, 370–386
Staines, G.H. & Pleck, J.H. (1983). The Impact of Work Schedules on the Family. Ann Arbor, MI:
Institute for Social Research, University of Michigan.
Swanberg, J.E. (2005). Job-family role strain among low-wage workers. Journal of Family and
Economic Issues, 26, 143-158.
Swanberg, J.E., James, J.B., Mamta, U.O., Werner, M., & McKechnie, S.P. (2008). CitiSales Study:
Jobs that Work. www.citysalesstudy.com.
Tilly, C. (1996). Half a job: bad and good part-time jobs in a changing labor market. Philadelphia:
Temple University Press.
Voydanoff, P. (2004). The effects of work demands and resources on work-to-family conflict and
facilitation. Journal of Marriage and Family, 66, 398-412.
Voydanoff, P. (2005). Work demands and work-to-family conflict and family-to-work conflict:
Direct and indirect relationships. Journal of Family Issues, 26, 707-726.
Warren, J. A., & Johnson, P. J. (1995). The impact of workplace support on work family role strain.
Family Relations, 44, 163-169.
Williams, R.L. (2000). A note on robust variance estimation for cluster-correlated data. Biometrics,
56, 645-646.
precarious work schedules
48
Appendix A
Indices from Employee Survey
Schedule Unpredictability
Schedule unpredictability is constructed by combining responses to an item about the number of
days of advance notice respondents receive regarding their work schedule and respondents’ scores
on an index of hours/timing predictability.
The measure of advance notice asks: “Usually, how many days in advance do you know your
schedule?” Answers were provided by respondents in days, and then recorded by interviewer as “03 days”, “4-7 days”, “8-14 days”, “15-21 days”, “more than 21 days”. These responses were
collapsed into a dichotomous indicator, one week or less versus more than one week.
The measure of hours/timing predictability is based on a three-item index (alpha = 0.65) that
captures the extent to which respondents agreed or disagreed that they can anticipate and/or count
on the number and timing of hours that they will be working week-to-week. The respondents are
asked how much they agree (strongly agree, agree, disagree, or strongly disagree?) with the
following three statements:



You can easily anticipate what days and times you’ll be working week-to-week.
Most weeks, you can count on getting the number of hours you want.
Most weeks, you can count on working the days and shifts you want.
Based on these two measures (advance schedule notice and hours/timing predictability), a measure
of schedule unpredictability was constructed that has three possible values including (1) more than
one week notice of work schedule and a score in the top half of the median split of the 3-item index
of hours/timing predictability, (2) one week or less notice of work schedule and a score in the top
half of the median split of the 3-item index of hours/timing predictability, and (3) one week or less
notice of work schedule and in the bottom half of the median split of the 3-item index of
hours/timing predictability.
Lack of Scheduling Input
ALPHA = 0.82
For each of the following, please tell me whether you feel you have a lot of input, some input, a
little input, or no input at all.




The days you have off each week.
The days you work each week.
When you begin and end each work day.
The total number of hours you work each week.
General Work-to-Family Conflict
ALPHA = 0.87
precarious work schedules
49
Would you say [this] happens all of the time, most of the time, some of the time, hardly any of the
time, or none of the time?





The demands of your work interfere with you personal or family time.
Your work schedule makes it difficult to fulfill your personal or family responsibilities.
Things at home do not get done because of the demands of your job.
Your work produces strain that makes it difficult to fulfill personal or family duties.
You have to make changes to your plans due to work related duties.
Note: Adapted from Netermeyer, R.G., Boles, J.S., & McMurrian, R. (1996). Development and Validation of WorkFamily Conflict and Family-Work Conflict Scales. Journal of Applied Psychology, 81(4) 400-410.
Work Interferences in Nonwork Activities
ALPHA = 0.84
Given how far in advance you generally know your upcoming work schedule, please tell me
whether you feel you have more than enough time, just enough time, or not enough time to organize
each of the following.




Scheduling a doctor’s appointment for yourself, a child or someone else.
Planning activities with your friends.
Planning a family outing.
Arranging to cook a meal at home.
Note: We developed this scale by “unpacking” a single question that Swanberg et al. (2008) used to assess the extent to
which workers in a drug-store chain reported sufficient time to “plan personal, family, or other responsibilities.”
Perceived Stress
ALPHA = 0.82
For each question, please tell me how often you have felt or thought this way over the course of the
last month (very often, fairly often, sometimes, almost never, or never).








How often have you been upset because of something that happened unexpectedly?
How often have you felt that you were unable to control the important things in your life?
How often have you felt nervous and stressed?
How often have you felt confident about your ability to handle your personal problems?
How often have you found that you could not cope with all of the things that you had to do?
How often have you felt that you were on top of things?
How often have you been able to control the way that you spend your time?
How often have you felt difficulties were piling up so high that you could not overcome
them?
Note: Adapted from the 14-item Perceived Stress Scale: Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global
measure of perceived stress. Journal of Health and Social Behavior, 24, 385-396.
Supervisor Support (emotional support subscale)
precarious work schedules
50
ALPHA = 0.83
With each of the following statements, would you strongly agree, agree, disagree, or strongly
disagree?




My supervisor is willing to listen to my problems in juggling work and nonwork life.
My supervisor takes the time to learn about my personal needs.
My supervisor makes me feel comfortable talking to him/her about my conflicts between
work and nonwork.
My supervisor and I can talk effectively to solve conflicts between work and nonwork issues.
Note: Hammer, L. B. Kossek, E. E., Yragui, N. L., Bodner, T. E. & Hanson, G. C. (in press). Development and
Validation of a Multidimensional Measure of Family Supportive Supervisor Behaviors (FSSB). Journal of Applied
Psychology.