The Effect of Paid Sick Leave Mandates on Coverage, Work Absences, and Presenteeism

Abstract

We evaluate the impact of paid sick leave (PSL) mandates on PSL coverage, work absences, and presenteeism (i.e. attending work while sick) for private sector workers in the U.S. Our identification strategy relies on geographic and temporal variation in mandate enactment, as well as within-county variation in the propensity to gain PSL following a mandate. We find that PSL mandates increase coverage rates and work absences for those most likely to gain coverage, and that these effects are larger for women and households with children. We also provide evidence that PSL mandates reduce the rate of presenteeism.

1. Introduction

The United States is one of the only developed countries in the world where workers are not guaranteed access to paid sick leave (PSL) and, as a result, more than one quarter of the nation’s workforce currently lacks PSL coverage (World Policy Analysis Center, 2019; Bureau of Labor Statistics, 2019). Efforts to establish wide-ranging, federally mandated PSL coverage in the U.S. culminated in the reintroduction of the Healthy Families Act to Congress in 2015.¹ The legislation, were it to become law, would create a national standard regulating the provision of PSL benefits for qualified workers. Subsequently, an executive order signed by President Obama that required firms with federal government contracts to provide employees with up to 7 paid annual sick days took effect in 2017. In early 2020, President Trump signed the Families First Coronavirus Emergency Response Act, which temporarily granted PSL coverage to workers in mid-size firms affected by the COVID-19 pandemic.

In the absence of a federal mandate, several states and municipalities have recently enacted legislation requiring employers to provide PSL benefits to their employees.² Alternatively, 19 states have passed legislation preemptively blocking municipalities from enacting local PSL mandates and the expansion of PSL through legislative action has become a contentious political issue. While public opinion polls indicate strong support for mandated PSL benefits, opponents argue that PSL regulations are costly to employers and reduce worker hours by encouraging absenteeism (Jones et al., 2014; Nelson, 2014; Pichler and Ziebarth, 2018).³ Unfortunately, evidence to support a causal link between PSL mandates and absenteeism among U.S. workers is sparse given that the majority of research on the relationship between PSL coverage and labor outcomes has analyzed legislation adopted in European countries. (Henrekson and Persson, 2004; Puhani and Sonderhof, 2010; Ziebarth and Karlsson, 2010; Ziebarth and Karlsson, 2014). These foreign laws are largely dissimilar to the proposed and recently enacted U.S. statutes and serve as poor models for the effects of expanded PSL generosity in the U.S.

Our paper adds to a small, but growing literature examining the employment effects of PSL mandates in the U.S. Specifically, we use the enactment of local and state level mandates to evaluate the relationship between PSL mandates and coverage rates for private sector workers. After establishing that PSL mandates lead to coverage expansions, we then estimate the effect of PSL mandates on work absences. Lastly, we explore whether PSL mandates and the related coverage gains are associated with reductions in “presenteeism” or attending work while sick. Susser and Ziebarth (2017) estimated that approximately 3 million U.S. workers attend work each week while sick and that the overriding concern among those doing so was lost income. Our findings make three contributions to the study of the labor market effects of PSL mandates. We provide the first collective estimates of the coverage effects of PSL mandates in the U.S. at the local and state levels. Understanding how PSL mandates translate into increased coverage is a crucial first step for any analysis of the long-run effects of PSL regulations. Second, we quantify the impact of increased access to PSL on worker absenteeism. The ex-ante effect of access to paid leave on work absences is unclear. The availability of PSL reduces the cost of absenteeism to workers, thus potentially increasing the likelihood of a work absence (Gilleskie, 1998). However, in the presence of a communicable illness, increased coverage could also reduce presenteeism, limiting the spread of illness and leading to fewer work absences (Johns, 2010; Susser and Ziebarth, 2016; Pichler and Ziebarth, 2017; Stearns and White, 2018 ).

We first establish that PSL mandates lead to economically significant increases in coverage using data from the National Health Interview Survey (NHIS) from 2005 to 2018. Relying on geographic and temporal variation in mandate enactment, we estimate difference-in-differences (DD) models comparing outcomes for those living in counties affected by PSL mandates to those living in counties with no mandate in place. Since workers often have PSL benefits in the absence of a legislative mandate, we refine our analysis using a triple-differences (DDD) model that compares within-county changes in coverage rates for those with a low likelihood of pre-mandate coverage to those with a high likelihood of pre-mandate coverage. This method is an improvement over traditional DD models that fail to focus on workers targeted by the legislation. Estimates suggest that PSL mandates increase coverage rates by approximately 20 percentage points for targeted workers, a relative increase of more than 45% from the baseline coverage rate.

We then turn to estimates of the effect of PSL mandates on work absences. Primarily relying on data from the Current Population Survey’s Basic Monthly Files (CPS) and using a similar DDD strategy, we find that workers gaining PSL coverage after a mandate is enacted exhibit increases in own illness-related work absences, however our estimates are somewhat imprecise. We then expand our absence measure to include own illness, childcare problems, or other family and personal obligations, and find much larger effects of PSL mandates on work absences. Estimates from our preferred specification indicate that gaining PSL coverage through a mandate leads to a 3.4 percentage point (61.6%) increase in the likelihood of missing work in the past week. We also show that the magnitude of the estimated effect is larger for women and for parents, consistent with provisions in several of the mandates we study permitting eligible employees to miss work to care for a sick child or family member.⁴

Finally, we use two waves of the of the American Time Use Survey (ATUS) Leave Module to estimate changes in presenteeism associated with PSL mandates for all workers and for workers in industries with historically low PSL coverage rates. Though our estimates are somewhat imprecise due to the relatively small sample sizes of the ATUS leave modules, we provide supporting evidence that gains in PSL coverage lead to substantial reductions in presenteeism.

Our findings have significant implications for the current debate over the expansion of PSL coverage to workers in the U.S. This represents the first study to quantify the collective effects of PSL mandates on coverage and one of the first to use plausibly exogenous variation in coverage rates to determine the effect of mandates on work absences and presenteeism in the U.S. Our results provide timely and compelling evidence for policymakers considering the enactment of similar regulations and inform an active discourse concerning the labor market effects of PSL coverage.

2. Labor Market Effects of Mandated Leave Policies

Since PSL mandates are a relatively new phenomenon in the U.S., studies of the labor market effects of mandated leave policies have generally focused on two related areas: paid leave mandates surrounding childbirth and unpaid access to sick leave. Rossin-Slater et al. (2013), Baum and Ruhm (2016), and Bartel et al. (2018) examined changes in maternal and paternal paid leave following the enactment of California’s 2004 Paid Family Leave program. All studies found that access to paid family leave increased leave-taking on the intensive margin, while Das and Polachek (2015) found persistent negative effects on women’s employment and Curtis et al. (2016) reported increased job churn for young women.

Evidence of the effect of access to unpaid leave on work absences is mixed. The Family and Medical Leave Act (FMLA) of 1993 guaranteed eligible workers access to unpaid, job-protected employment leave for circumstances including a serious health condition that impedes job performance, childbirth, or the care of a close relative with a serious health condition (U.S. Department of Labor, 2016).⁵ Waldfogel (1999) analyzed the effect of the FMLA on work absences, employment, and earnings. She found that the FMLA increased instances of leave-taking, but had no effect on changes in employment or wages. Alternatively, using the FMLA and prior state-level unpaid leave mandates, Baum (2003) reported that unpaid leave had no effect on leave-taking for mothers who had recently given birth, while Han et al. (2009) found that expansions in access to unpaid leave increased leave-taking for both mothers and fathers.

While studies of the labor market effects of unpaid leave policies might provide some indication of the potential response to PSL mandates, the applicability of these studies to the case of PSL for all workers is questionable since work absences related to childbirth are largely planned in advance. Also, as the FMLA covers fewer than half of private sector workers, it is unclear whether these earlier findings of the effects of FMLA on work absences extend to the recent PSL mandates adopted by states and municipalities in the U.S.

Research on the labor market effects of paid leave mandates has generally focused on European countries where mandated benefits have been in place for several years and administrative data on PSL coverage and work absences is more widely available than in the U.S. Henrekson and Persson (2004) concluded that increases in sick leave generosity in Sweden were related to increased absenteeism over an extended period from 1955 to 1999. Puhani and Sonderhof (2010) investigated German legislation that initially reduced and then later expanded opportunities for PSL. The authors reported that decreasing sick pay from 100 percent to 80 percent of wages resulted in a reduction of 2.4 sick days per year on average. Similarly, examining the same reduction in German sick pay, Ziebarth and Karlsson (2010) reported that the share of workers with zero work absences increased between 6 percent and 8 percent, while Ziebarth and Karlsson (2014) found that restoring German sick pay to 100 percent of wages led to a 10 percent increase in work absences.

Due to the novelty of PSL mandates in the U.S., few studies have examined labor market effects of paid leave access for U.S workers. Using a survey of employers in San Francisco, Colla et al. (2014) found that the share of firms providing PSL coverage to employees increased from 73% to 91% following the enactment of a PSL mandate in 2007. However, it was not clear how many workers were affected by the policy change. Furthermore, San Francisco employers voluntarily offered PSL benefits at a relatively high rate prior to the mandate, raising concerns over the applicability of these findings to other regions. Using a structural framework simulation, Gilleskie (1998) found that moving from no PSL coverage to full coverage would increase illness-related work absences by 45% per illness episode, but estimates from a reduced form model deviated substantially when she considered alternative policy scenarios that included combinations of PSL mandates and expanded health insurance coverage.

More recently, Ahn and Yelowitz (2016) matched U.S. workers with and without access to PSL on various observable characteristics and found that PSL coverage led to approximately 1.2 additional work absences per year. A concern with this identification strategy is that unobserved differences between those with and without PSL coverage could bias the authors’ estimated effects. Because Ahn and Yelowitz do not rely on the plausibly exogenous (to the worker) effect of PSL mandate enactment, estimates of within-region differences between those with and without PSL coverage are especially susceptible to omitted variable bias. In the same study, Ahn and Yelowitz estimated models that used exogenous measures of regional influenza infection rates to identify the effect of PSL coverage on absenteeism, which resulted in a smaller effect of 0.9 additional work absences per year. Alternatively, Stearns and White (2018) found that PSL mandates in Connecticut, and to a lesser extent D.C., reduced the aggregate rate of illness-related work absences. Finally, Pichler and Ziebarth (2018) found no evidence that U.S. PSL mandates affected employment or wages, though the authors focused on all workers rather than those most likely to gain PSL coverage following the enactment of a mandate. We show below that this distinction has meaningful implications for the magnitude of the estimated effects of PSL mandates.

3. Data

We use several sources of data for our analyses of the effects of PSL mandates on coverage, worker absences, and presenteeism. The first is the National Health Interview Survey (NHIS) maintained by the National Center for Health Statistics (NCHS). The objective of the NHIS is to monitor the health of the U.S. population through the collection and analysis of data on a broad range of health and labor market topics. The NHIS is a cross-sectional household survey with continuous sampling and interviewing throughout the year, which follows a multistage area probability sampling design that permits the representative sampling of households and non-institutional group quarters in the U.S. (e.g. college dormitories). Data are collected through a personal household interview conducted by interviewers employed and trained by the U.S. Census Bureau according to procedures specified by the NCHS (Centers for Disease Control [CDC], 2016).

We rely on restricted access state- and county-identified NHIS data collected between 2005 and 2018 for our analyses. Our sample is restricted to workers between the ages of 18 and 64 who are employed in the private sector and do not report being self-employed at the time of their interview.⁶ We also drop workers who would not qualify for mandated coverage based on firm size and scope of coverage restrictions (see Appendix Table 1). The NHIS includes information on sex, race/ethnicity, education, poverty status, marital status, and health insurance coverage that we use to control for observable differences between individuals in our sample. We construct an indicator that is equal to one for those living in counties that enacted a PSL mandate after the mandate became effective and is equal to zero otherwise.⁷ When a mandate is enacted at the state level, this indicator takes the value of one for all residents of that state.

Our outcome measure for PSL coverage is derived from a question in the NHIS that asks those currently employed, “Do you have paid sick leave on this main job or business?”. We also construct three measures of work absences from a question asking, “During the past 12 months, about how many days did you miss work at a job or business because of illness or injury (do not include maternity leave)?”. From this question, we generate an indicator for whether the respondent reported any work absence in the past twelve months, and two continuous measures of illness/injury work loss days: one for the entire sample (i.e., unconditional days) and another for those reporting at least one illness/injury work loss day in the past year (i.e., conditional days).

Two notable drawbacks to using the NHIS for our analysis of work loss days include challenges related to sample size (approximately 11,000 workers per year remain after sample restrictions) and the somewhat imprecise measure of work absences – workers are likely to recall the number of illness-related work absences over the past 12 months with error. Therefore, we conduct a complementary analysis on the effects of PSL mandates on work absences using data from the Current Population Survey’s Basic Monthly Files (CPS) between 2005 and 2018 and the same sample restrictions described above for the NHIS.⁸ The CPS is administered by the U.S. Census Bureau with a multistage stratified probability sample design that includes approximately 60,000 households per month and is intended to measure characteristics of the U.S. labor force (U.S. Department of Labor, 2006).

The CPS data, with roughly 220,000 workers per year after sample restrictions, allows us to construct multiple measures related to work absences. For those who worked part time during the previous week but typically work full time, the CPS asks the reason for the reduced work hours.⁹ From this information, we generate an indicator equal to one if a respondent missed work in the previous week because of an own illness or health/medical limitation and zero otherwise. We then expand this measure to include own illness or health/medical limitation, childcare problems, and other family/personal obligations; we refer to this second measure as “own illness plus”. Since the CPS asks about work absences in the week prior to the interview rather than in the past 12 months, the potential for recall bias is likely to be reduced for this sample compared to the NHIS sample.

In subsequent analyses, we consider whether PSL mandates reduce presenteeism. A lack of available data has limited research on this topic and, to our knowledge, this represents the first study to examine the relationship between PSL mandates in the U.S. and presenteeism.¹⁰ We combine a recent wave of the ATUS Leave Module, fielded in 2017–2018 and released in 2019, with an earlier version fielded in 2011. The ATUS Leave Module is administered to a subset of all ATUS respondents and contains a series of questions addressing workers’ ability to take leave from their current jobs. We use these questions to generate measures of paid sick leave coverage, an indicator for any past week work absence similar to the “own illness plus” variable from the CPS, and whether a respondent reported the need to take off work in the past week due to their own illness, but did not (i.e., presenteeism).

Finally, to our NHIS, CPS, and ATUS samples, we merge data on the number of physicians per capita (general practitioners, family practitioners, and all MDs), the number of inpatient days per capita, and the number of outpatient physician visits per capita from the Area Health Resource Files (AHRF); county-level Medicare fee-for-service parts A and B per-capita spending from the Centers for Medicare and Medicaid Services; and county-level unemployment rates from the Bureau of Labor Statistics (BLS). We include data on the supply of physicians and on health care expenditures, along with controls for individual insurance coverage, to address concerns with access-related changes and insurance expansions resulting from the implementation of provisions of the Affordable Care Act that occurred contemporaneously with some of the PSL mandates that we study.

Table 1 presents descriptive statistics for our NHIS sample separately for those living in counties that enacted a PSL mandate (treatment counties) and those living in counties with no PSL mandate (control counties). Counties that enacted PSL mandates tended to have more highly educated populations, more racial/ethnic minorities, and more residents with Medicaid coverage. Descriptive information for the CPS sample is included in Appendix Table 2 and exhibits patterns similar to those described for the NHIS sample.

Table 1:

Descriptive Statistics for NHIS Sample

	Treatment Counties	Control Counties
Female	0.474	0.496
Age	39.17	39.92
Black, non-Hispanic	0.108	0.140
Hispanic	0.350	0.152
Other race, non-Hispanic	0.145	0.049
White, non-Hispanic	0.398	0.659
Federal Poverty Level < 1.0	0.087	0.091
Federal Poverty Level 1.0 – 1.99	0.145	0.152
Federal Poverty Level 2.0 – 2.99	0.128	0.153
Federal Poverty Level 3.0 – 3.99	0.100	0.127
Federal Poverty Level 4.0+	0.368	0.330
Federal Poverty Level Missing	0.173	0.148
Less than High School	0.153	0.106
High School Graduate or GED	0.191	0.269
Some College	0.279	0.345
Bachelor’s Degree	0.243	0.198
Graduate or Professional Degree	0.127	0.079
Education Missing	0.007	0.004
Married	0.439	0.455
Widowed	0.014	0.018
Divorced	0.101	0.139
Separated	0.035	0.032
Never Married	0.329	0.273
Living with a Partner	0.079	0.080
Marital Status Missing	0.004	0.002
Private Health Insurance	0.711	0.735
Medicaid Health Insurance	0.058	0.037
Military Health Insurance	0.011	0.022
Other Health Insurance	0.040	0.032
No Health Insurance	0.180	0.181
Health Insurance Missing	0.005	0.003
Observations	31,940	126,717

Notes: Treatment counties include those enacting a PSL mandate between 2007 and 2018, while control counties are those with no PSL mandate in place over this time period. Data are from the NHIS samplefrom 2005 through 2018.

4. Difference-in-differences Estimates of the Effect of PSL Mandates on Coverage and Work Absences

We begin our empirical analysis by estimating the effects of PSL mandates on coverage and work absences using a difference-in-differences (DD) strategy that exploits the temporal and geographic variation in the timing of the enactment of PSL mandates. The relationship between PSL mandates and our outcomes of interest is formalized as follows:

$${\mathit{Y}}_{ict}=\mathit{\alpha }+\mathit{\gamma }{\mathit{M}\mathit{a}\mathit{n}\mathit{d}\mathit{a}te}_{ct}+\mathit{\beta }{\mathit{X}}_{\mathit{i}\mathit{c}\mathit{t}}+\mathit{\theta }{\mathit{Z}}_{\mathit{c}\mathit{t}}+{\mathit{\delta }}_{\mathit{c}}+{\mathit{\tau }}_{\mathit{t}}+{\mathit{\delta }}_{\mathit{c}}\mathrm{\times }{\mathit{\tau }}_{\mathit{t}}+{\epsilon }_{\mathit{i}\mathit{c}\mathit{t}}$$ (1)

where Y is the outcome of interest for person i in county c at year-quarter t; Mandate is an indicator for the enactment of a PSL mandate in any part of year-quarter t;¹¹ X is a vector of individual characteristics (sex, race/ethnicity, marital status, education, age, health insurance coverage, and industry of employment); Z is a vector of time-varying county-level factors (physician supply, inpatient and outpatient days, FFS Medicare spending, and the unemployment rate); δ_c is a county fixed effect, τ_t is a year-quarter fixed effect, and δ_c × τ_t represents a county-specific linear time trend which is included in our preferred specifications. Equation (1) represents a standard DD analysis where outcomes in our treatment regions (i.e., counties and states enacting a PSL mandate) are compared to control regions that have no PSL mandates in place. Standard errors for our analyses are clustered at the county level.

Our dependent variable, Y_ict in Equation (1) represents one of several possible outcomes. We initially estimate the effect of PSL laws on coverage measured by whether a worker in our sample reports having PSL benefits at their main job. Our hypothesis is that a PSL mandate should increase the share of workers reporting PSL coverage. However, if PSL mandates tend to be enacted in areas where private employers display a high likelihood of offering coverage, then we may find a relatively small effect of the mandate. Additionally, employers may attempt to avoid a mandate by relocating or restructuring their workforce, which would also negate coverage effects of a mandate.

After examining the effect of PSL mandates on coverage, we then estimate the relationship between mandates and work absences using both the NHIS and CPS data. In the NHIS data, our dependent variables include an indicator for any work absence in the past year due to illness or injury, unconditional changes in the average number of work absences, and changes conditional on reporting at least one work absence in the past year. Using the CPS data, we estimate the effect of mandate enactment on the probability of missing any work in the past week for an own-illness or for an own-illness, child care problem, or family/personal obligation (i.e., “own illness plus”).

A potential challenge to the validity of our DD specification is the endogenous adoption of PSL mandates. For example, if a municipality’s population demographics (e.g., share of service industry workers or underlying health of the population) are changing over time in unobserved ways, and this change leads to the enactment of a PSL mandate, then our estimates of the effect of PSL legislation on coverage rates and work absences would be biased. To gauge the threat posed by endogenous policy enactment, we use data from our CPS sample along with proxies for population health to regress PSL adoption on several observable county-level characteristics that would plausibly be related to the enactment of a PSL mandate.¹² Specifically, we examine the association between the enactment of a PSL mandate and county-level estimates of population age, race/ethnicity, education, and income; per capita hospital inpatient days, per capita outpatient visits, and total Medicare spending for parts A and B;¹³ county-level unemployment rates; and industry of employment. While not a definitive test of the exogeneity of the PSL mandates that we study, an inability to explain the variation in the adoption of PSL laws with a rich set of observable population characteristics lends support to our assertion that our identification strategy returns causal estimates of the effect of PSL mandates (Hoynes and Schanzenbach, 2009; Hoynes and Schanzenbach, 2012).

Results of this analysis are presented in Table 2. Column (1), which omits county fixed effects and county time trends, indicates that the share of the population ages 45 and above and average per-capita Medicare spending are negatively associated with the enactment of a PSL mandate, while counties with larger Asian populations were more likely to enact PSL mandates. In addition, several industry coefficients in Column (1) are statistically significant. Column (2) adds county fixed effects, which generally attenuate the associations between industry categories and mandate enactment. Finally, we add county time trends in Column (3) and find that only Medicare spending and the unemployment rate are associated with PSL mandate enactment. If the positive association between Medicare spending and mandate enactment in Column (3) indicates that counties with unobservably sicker populations were more likely to enact a mandate, then estimates of the effect of a PSL mandate on illness-related work absences would be biased. However, we control for Medicare spending in our preferred specifications to minimize any potential bias that might arise. We also note that, despite the inclusion of county fixed effects and county time trends in Column (3), the R-squared of 0.596 indicates that much of the variation in PSL mandate adoption is unexplained by demographic and county-level observables and, therefore, suggests substantial random variation in PSL adoption to be leveraged in our estimation strategy.

Table 2:

Determinants of Paid Sick Leave Adoption

	(1)	(2)	(3)
Female	0.032	−0.013	−0.055
	(0.061)	(0.057)	(0.041)
Age 18–34	Omitted	Omitted	Omitted
Age 35–44	−0.099	0.140**	0.061
	(0.081)	(0.055)	(0.050)
Age 45–54	−0.271***	0.019	0.050
	(0.101)	(0.069)	(0.044)
Age 55–64	−0.342***	−0.112	0.054
	(0.119)	(0.081)	(0.042)
White	Omitted	Omitted	Omitted
Black	−0.096*	−0.262***	−0.058
	(0.055)	(0.098)	(0.057)
Hispanic	0.051	−0.063	−0.089
	(0.046)	(0.108)	(0.071)
Asian	0.612***	0.577***	0.064
	(0.181)	(0.164)	(0.131)
Other Race/Ethnicity	−0.674*	0.067	0.163
	(0.344)	(0.188)	(0.131)
Less than High School	Omitted	Omitted	Omitted
High School	0.030	0.247*	−0.045
	(0.111)	(0.129)	(0.063)
Some College	−0.217*	0.128	−0.043
	(0.130)	(0.129)	(0.064)
College or Greater	0.057	0.268**	−0.038
	(0.106)	(0.109)	(0.068)
Family Income < $25k	Omitted	Omitted	Omitted
Family Income $25k-$49,999	−0.001	−0.021	−0.044
	(0.071)	(0.073)	(0.051)
Family Income $50k-$74,999	−0.008	0.011	0.004
	(0.067)	(0.068)	(0.052)
Family Income > $75k	0.079	0.034	0.049
	(0.064)	(0.069)	(0.048)
Medicare Spending (thousand $)	−0.041***	0.031*	0.055***
	(0.013)	(0.016)	(0.018)
Per Capita Hospital Days	0.012	−0.084*	−0.034
	(0.008)	(0.049)	(0.025)
Per Capita Outpatient Visits	0.001	−0.002	0.004
	(0.001)	(0.003)	(0.003)
Unemployment Rate	0.006	−0.001	−0.034***
	(0.006)	(0.002)	(0.009)
Agriculture Industry	Omitted	Omitted	Omitted
Mining Industry	−0.470*	−0.669**	−0.348
	(0.270)	(0.309)	(0.356)
Construction Industry	−0.413**	−0.220	0.032
	(0.202)	(0.190)	(0.103)
Manufacturing Industry	−0.561***	−0.388**	−0.020
	(0.178)	(0.180)	(0.111)
Transportation & Utilities	−0.349*	−0.160	0.036
	(0.185)	(0.186)	(0.115)
Wholesale Trade Industries	−0.620***	−0.416**	−0.100
	(0.197)	(0.183)	(0.119)
Retail Trade Industries	−0.391**	−0.237	−0.018
	(0.171)	(0.192)	(0.100)
Finance, Insurance, and Real Estate	−0.413*	−0.373*	−0.079
	(0.223)	(0.202)	(0.117)
Business and Repair Services	−0.121	−0.011	−0.023
	(0.228)	(0.195)	(0.108)
Personal Services	−0.385	−0.136	0.127
	(0.237)	(0.200)	(0.118)
Entertainment and Recreation	−0.408*	0.022	0.198
Services	(0.238)	(0.252)	(0.145)
Professional and Related Services	−0.168	−0.222	0.049
	(0.194)	(0.174)	(0.102)
County Fixed Effects	No	Yes	Yes
County Time Trend	No	No	Yes
R2	0.404	0.577	0.596
Observations	16,255	16,255	16,255

Notes: Observations are at the county-year/quarter level for years 2005 through 2018 using data from the CPS sample. All regressions include year/quarter fixed effects and are weighted by county population.Standard errors are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

Along with policy exogeneity, another necessary assumption for the validity of our DD model is that the treatment and control groups would have followed the same trends (in terms of the outcome variables) had the enactment of a PSL mandate not occurred. This assumption is untestable, as it is impossible to observe the treatment group in the untreated state during the post-treatment period. However, evidence that these two groups followed similar trends in the outcome variables in the pre-PSL period lends credence to our estimation strategy. To assess whether trends in our outcomes were similar in the pre-enactment period for our treatment and control counties, we estimate a model similar to Equation (1) that replaces the Mandate term with an interaction between an indicator for whether a county ever enacts a PSL mandate and a series of dummy variables representing sixth-month intervals centered around the mandate’s enactment and extending for 18 months in either direction. A lack of statistical significance on this interaction term in the pre-mandate period indicates that no measurable difference in pre-policy trends exist between the treatment and control counties. We discuss the results of this event-study design along with our estimates of Equation (1) below.

Table 3 presents estimates of the effect of PSL mandate enactment on the probability of reporting PSL coverage. Results in the first column of Table 3 are from a specification that omits controls for industry of employment, time-varying county characteristics, and county time trends. Column (1) indicates that the enactment of a PSL mandate results in an increase in coverage of 9.6 percentage points compared to counties with no mandate in effect. Based on a mean coverage rate of 57.2% in our sample, this represents a 16.8% increase in access to PSL resulting from the enactment of a mandate. The specification in Column (2) adds separate controls for the 22 industry categories listed in Appendix Table 3. The inclusion of industry controls has little effect on the magnitude of the estimate. Column (3) adds our county level controls that include per capita measures of physician supply, health care utilization, Medicare FFS per capita spending, and the county unemployment rate. Our estimate remains largely unchanged after the addition of these county controls. Finally, Column (4) adds county time trends to further control for potential policy endogeneity or anticipation effects. Here our coefficient estimate is slightly smaller than the estimates in Columns (1) through (3) and suggests a relative increase in PSL access of nearly 15%.

Table 3:

Difference-in-difference Estimates of the Effect of Paid Sick Leave Mandates on Paid Sick Leave Coverage

	(1)	(2)	(3)	(4)
Mandate	0.096***	0.097***	0.097***	0.084***
	0.011)	(0.012)	(0.012)	(0.012)
Industry Controls	No	Yes	Yes	Yes
County Controls	No	No	Yes	Yes
County-Time Trend	No	No	No	Yes
Mean PSL Coverage	0.572	0.572	0.572	0.572
Observations	156,735	156,735	156,735	156,735

Notes: All regressions include controls for sex, race, marital status, education, age, health insurance coverage, U.S. born, and county and year-quarter fixed effects. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits,and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the NHIS sample from 2005 through 2018. Standard errors are in parentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

Figure 1 presents the event study estimates of the effect of a PSL mandate on coverage from a specification similar to that in Column (3) of Table 3. We see no evidence of a differential trend in PSL coverage between treatment and control counties in the 18 months prior to a mandate’s enactment. However, in the first six months following mandate enactment we see an increase in coverage of approximately 7 percentage points relative to the six months prior to enactment and this effect continues to grow over the next two six month periods. Two explanations for the steady increase in PSL coverage after mandate enactment depicted in Figure 1 include the potential that not all firms immediately complied with the mandates (Colla et al., 2014) and, because workers must accrue sick leave benefits over time, respondents may report coverage with a delay. Despite relatively high levels of existing PSL coverage in counties enacting a mandate, estimates in Table 3 and Figure 1 indicate that PSL mandates lead to higher coverage rates.

Figure 1: Event Study Estimates of the Effect of Paid Sick Leave Mandates on Coverage Paid Sick Leave Coverage.

Notes: Estimates are from a modified version of Equation (1) that allows the effect of mandate enactment to vary over time. Month 0 represents the month-year that the PSL mandate took effect in treatment counties. Data are from the NHIS sample from 2005 through 2018. Standard errors are clustered at the county level.

We also use a specification similar to that in Column (4) of Table 3 to explore the possibility of heterogeneous impacts of PSL mandates in Table 4. Here we split the sample into 6 subgroups and examine the effect of PSL mandates separately for: women, men, white non-Hispanics, nonwhites or Hispanics, and workers in industries with historically low and historically high coverage rates.¹⁴ Estimates in Table 4 suggest that women see slightly larger gains in PSL coverage as the result of mandate enactment compared to men. Coverage for women increases by 10.9 percentage points (18.8%) compared to 8.9 percentage points for men (15.8%). White, non-Hispanic workers see an 8.0 percentage point increase in coverage as a result of a mandate compared to an 11.4 percentage point increase for non-white or Hispanic workers. Lastly, Columns (5) and (6) indicate that, as expected, those working in industries where coverage rates were low prior to a mandate see a large, 14.4 percentage point (36.2%) gain in coverage following the enactment of a mandate, while those in industries with high rates of PSL coverage experience much smaller gains.

Table 4:

Difference-in-difference Estimates of the Effect of Paid Sick Leave Mandates on Paid Sick Leave Coverage by Subgroup

	Women	Men	White, Non-Hispanic	Non-White or Hispanic	Low PSL Industries	High PSL Industries
	(1)	(2)	(3)	(4)	(5)	(6)
Mandate	0.109***	0.089***	0.080***	0 114***	0 144***	0.070***
	(0.014)	(0.013)	(0.011)	(0.017)	(0.014)	(0.013)
Industry Controls	Yes	Yes	Yes	Yes	Yes	Yes
County Controls	Yes	Yes	Yes	Yes	Yes	Yes
County-Time Trend	Yes	Yes	Yes	Yes	Yes	Yes
Mean PSL Coverage	0.580	0.565	0.603	0.524	0.398	0.695
Observations	77,144	79,591	95,183	61,552	64,995	91,740

Notes: All regressions include controls for sex, race, marital status, education, age, health insurance coverage, U.S. born, and county and year-quarter fixed effects. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Low (high) PSL industries are defined as those industries where fewer (more) than the overall mean share of workers report PSL coverage prior to the enactment of a PSL mandate. Data are from the NHIS sample from 2005 through 2018. Standard errors are in parentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

We next turn to the second phase of our analysis, which includes estimates of the effect of PSL mandates on work absences in the past 12 months due to illness or injury using the NHIS sample in Table 5. We estimate separate regressions for any illness-related absence in the past 12 months, and unconditional and conditional illness-related work loss days. All columns in Table 5 include industry and county controls, while the even numbered columns omit the county-time trend term. Column (1) indicates that PSL mandates are associated with small, but statistically significant increases in the likelihood of reporting at least one past year work absence. However, when county-time trends are included in Column (2), the estimate is attenuated and no longer statistically significant. We observe similar patterns for unconditional and conditional work loss days in Columns (3) through (5).

Table 5:

Difference-in-difference Estimates of the Effect of Paid Sick Leave Mandates on Annual Work Absences

	Any Absences		Unconditional Work Loss Days		Conditional Work Loss Days
	(1)	(2)	(3)	(4)	(5)	(6)
Mandate	0.026***	0.011	0.418**	0.066	0.439	0.057
	(0.011)	(0.012)	(0.200)	(0.289)	(0.460)	(0.621)
Industry Controls	Yes	Yes	Yes	Yes	Yes	Yes
County Controls	Yes	Yes	Yes	Yes	Yes	Yes
County-Time Trend	No	Yes	No	Yes	No	Yes
Outcome Mean	0.446	0.446	3.475	3.475	7.791	7.791
Observations	157,826	157,826	157,826	157,826	70,396	70,396

Notes: All regressions include controls for sex, race, marital status, education, age, health insurance coverage, U.S. born, and county and year-quarter fixed effects. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the NHIS sample from 2005 through 2018. Standard errors are in parentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

5. Triple Difference Estimates using Variation in Pre-Mandate Coverage Rates

Overall, estimates in Tables 3 and 4 consistently indicate that PSL mandates increase coverage rates for workers in affected counties compared to workers in counties with no mandate enactment. However, since a large share of workers in our sample reported coverage in the absence of mandate, our DD results likely underestimate the coverage gain for those workers who were most affected by a PSL mandate (i.e., those workers who lacked coverage in the absence of a mandate). This notion is supported by the much larger effect we estimate for workers in low-PSL industries. To shed additional light on the effect of a PSL mandate on coverage rates and work absences for those most likely to benefit from the legislation, we modify our empirical strategy and estimate a triple-difference (DDD) model that allows the effect of PSL mandates to differ by the probability that a worker lacks coverage prior to the mandate taking effect. The first step in our DDD analysis is to estimate the probability that a worker in our sample has PSL coverage using data from time periods before mandate enactment in our treatment counties or before a randomly assigned pseudo-enactment date for control counties.¹⁵ We use a logistic regression model that includes the same individual demographic characteristics found in Equation (1), as well as income, firm size, and codes for specific industry of employment.¹⁶ Coefficient estimates from the prediction model are presented in Appendix Table 4. As expected, PSL coverage in the absence of a mandate is positively associated with income and education. Workers in the agricultural, construction, and food service industries are particularly unlikely to report coverage. Figure 3 provides a visual depiction of the predicted PSL probabilities for the NHIS sample. The distribution of predicted probabilities is somewhat bimodal with a large share of the sample centered around 0.8 and another smaller share centered around 0.16. To gauge the accuracy of our prediction model, we include a conditional density plot in Figure 4 that compares actual PSL coverage to predicted coverage. Figure 4 indicates that those with a low predicted probability of coverage overwhelmingly lack coverage in the absence of a mandate. As the predicted probability of coverage rises, so too does the share of the sample reporting that they have coverage at their current job. The conclusion that we draw from the visual evidence in Figure 4 is that our prediction model is a useful tool to categorize individuals in our sample by the likelihood that they are directly affected by a PSL mandate.

Figure 3: Distribution of Predicted Probabilities of Paid Sick Leave Coverage.

Notes: Histograms depict estimates from a logistic regression model of the probability that an individual reports paid sick leave coverage in the pre-mandate period. Models include controls for all individual and county characteristics listed in Appendix Table 4. Data are from the NHIS sample from 2005 through 2018.

Figure 4: Actual Versus Predicted Pre-Mandate Paid Sick Leave Coverage.

Notes: The predicted probability of paid sick leave is derived from a logistic regression model using time before county mandate adoption and includes controls for all individual and county characteristics listed in Appendix Table 4. Data are from the NHIS sample from 2005 through 2018.

After obtaining the predicted probability of pre-mandate PSL coverage, we estimate the following DDD specification:

$$Y_{\mathit{i}\mathit{c}\mathit{t}}=\stackrel{\prime }{\mathit{\alpha }}+{\stackrel{\prime }{\mathit{\gamma }}}_{\mathrm{1}}\mathit{M}\mathit{a}\mathit{n}\mathit{d}\mathit{a}t{\mathit{e}}_{\mathit{c}\mathit{t}}+{\stackrel{\prime }{\mathit{\gamma }}}_{\mathrm{2}}\mathit{p}\mathrm{(}\mathit{N}\mathit{o}\mathit{P}\mathit{S}\mathit{L}{\mathrm{)}}_{\mathit{i}\mathit{c}}+{\stackrel{\prime }{\mathit{\gamma }}}_{\mathrm{3}}\mathit{M}\mathit{a}\mathit{n}\mathit{d}\mathit{a}t{\mathit{e}}_{\mathit{c}\mathit{t}}\mathrm{\times }\mathit{p}\mathrm{(}\mathit{N}\mathit{o}\mathit{P}\mathit{S}\mathit{L}{\mathrm{)}}_{\mathit{i}\mathit{c}}\mathrm{+}\stackrel{\prime }{\mathit{\beta }}{\mathit{X}}_{\mathit{i}\mathit{c}\mathit{t}}+$$ (2)

Equation (2) is similar to Equation (1) but allows the effect of PSL mandate enactment to vary by the probability of PSL coverage prior to mandate enactment. For ease of interpretation, we use the inverse of the predicted probabilities in Equation (2) so that the coefficient of interest, ${\stackrel{\prime }{\gamma }}_3$, measures the impact of a PSL mandate on an individual who is predicted to lack pre-mandate PSL coverage. The coefficient ${\stackrel{\prime }{\mathit{\gamma }}}_{\mathrm{1}}$ represents the effect of mandate enactment on those with PSL coverage prior to the mandate (i.e., predicted probability of lacking PSL access equals zero). The coefficient ${\stackrel{\prime }{\gamma }}_2$ is then the difference in outcomes for those with no PSL benefits compared to those with PSL benefits when no mandate is in place. We also include two-way interactions between the probability of lacking PSL coverage and year-quarter fixed effects and between the probability of lacking PSL coverage and county fixed effects. These interaction terms allow for additional flexibility between the effect of predicted coverage on outcomes over time and across counties.

Notably, the coefficients from our DD and DDD models are not directly comparable. The DD coefficient of interest in Equation (1) measures the effect of a PSL mandate on workers in treatment counties compared to workers in control counties (in other words, this is an intent-to-treat estimate). The DDD coefficient of interest in Equation (2) estimates the effect of a PSL mandate on workers in treatment counties who are more likely to gain coverage compared to those in treatment counties who are less likely to gain coverage (i.e., a treatment-on-the-treated effect).

Table 6 displays results from our DDD model on the effect of PSL mandates on coverage for those most likely to be affected by a mandate. Column (1) contains results from a specification that includes industry controls, but omits county controls and county time trends. Column (2) repeats the analysis with the addition of county controls, while Column (3) adds county time trends. Finally, Column (4) in Table 3 drops individuals in the middle two quartiles of predicted PSL coverage and compares the effect of a PSL mandate for those in the top quartile of the predicted probability distribution to those in the bottom quartile. All four of the columns in Table 6 suggest that PSL mandates result in large and statistically significant increases in coverage for those most likely to be affected by a mandate with estimates ranging from 16 to 19 percentage points (28% to 33% gains relative to pre-mandate mean coverage rates). Column (4) indicates that, compared to the top quartile of the predicted probability distribution, a mandate increases coverage rates by 20.7 percentage points (46.1%) for workers most likely to lack pre-mandate coverage. The coefficient on the p(NoPSL) term suggests that our predication model is successful at identifying individuals who, in the absence of a mandate, would be especially likely to lack PSL coverage. However, we hesitate to draw firm conclusions about our prediction model from this coefficient since estimates rely on out-of-sample predictions as no individual in our sample actually has a predicted probability of lacking PSL coverage that is equal to zero. The small and generally statistically insignificant coefficient estimates on the Mandate term are expected given that this represents the effect of a mandate on those with a high likelihood of pre-mandate coverage.

Table 6:

Triple Difference Estimates of the Effect of Paid Sick Leave Mandates on Paid Sick Leave Coverage

	(1)	(2)	(3)	(4)
Mandate x p(NoPSL)	0.189***	0.188***	0.160***	0.207***
	(0.032)	(0.032)	(0.031)	(0.023)
Mandate	0.025*	0.025*	0.022	0.004
	(0.014)	(0.014)	(0.015)	(0.013)
p(NoPSL)	−1.008***	−1.079***	−1.074***	−1.235***
	(0.015)	(0.023)	(0.023)	(0.026)
Industry Controls	Yes	Yes	Yes	Yes
County Controls	No	Yes	Yes	Yes
County-Time Trend	No	No	Yes	Yes
Top-to-Bottom Quartile	No	No	No	Yes
Mean PSL Coverage	0.572	0.572	0.572	0.449
Observations	156,735	156,735	156,735	78,367

Notes: All regressions include controls for sex, race, marital status, education, age, health insurance coverage, U.S. born, and county and year-quarter fixed effects. The probability of lacking PSL coverage was determined through a series of logistic regression models (see Appendix Table 4 for details). Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capitain patient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the NHIS sample from 2005 through 2018. Standard errors are inparentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

After confirming that PSL mandates increase access to paid leave and that the effect is larger for those most likely to gain coverage, we again turn to the effect of PSL mandates on work absences by re-estimating our DDD model in Equation (2) using the same work absence measures presented in Table 5. For each outcome, we report results from specifications that include industry and county controls, that add a county-time trend, and that compare those with the highest likelihood of gaining coverage (top quartile) to those with the highest likelihood of existing coverage prior to a mandate (bottom quartile). Results are presented in Table 7 and generally indicate positive, but imprecisely estimated, effects of PSL mandates on work absences. The first three columns of Table 7 contain estimates for the effect of PSL mandates on the likelihood that a respondent reported an illness/injury-related work absence in the past 12 months, while Columns (4) through ( 9)include estimates of changes in the number of illness-related work loss days. Estimates of the effect of a PSL mandate on any illness-related work absence are positive, though modest in effect size and not statistically significant. PSL mandates increase the unconditional average number of work loss days by 0.74 off of a mean of 3.5, a 21.3% increase. When comparing those most likely to gain PSL as a result of a mandate to those least likely to gain coverage, the estimate increases to 1.2 days (37.9%). However, in both cases, the coefficient estimates, while sizable in magnitude, are not statistically significant. We observe a similar pattern in Columns (7) through (9), which restrict the sample to individuals reporting at least one illness-related work loss day in the past 12 months. For this group, PSL mandates increase average work loss days by 1.34 (17.2%).

Table 7:

Triple Difference Estimates of the Effect of Paid Sick Leave Mandates on Annual Work Absences

	Any Absences			Unconditional Work Loss Days			Conditional Work Loss Days
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Mandate x p(NoPSL)	0.003	0.011	0.027	0.971	0.741	1.237	1.763	1.340	2.165
	(0.031)	(0.032)	(0.034)	(0.860)	(0.898)	(0.982)	(2.136)	(2.216)	(2.531)
Mandate	0.021	0.003	0.002	−0.005	−0.307	−0.396	−0.247	−0.690	−1.031
	(0.015)	(0.017)	(0.018)	(0.339)	(0.385)	(0.480)	(0.746)	(0.860)	(1.013)
p(NoPSL)	−0 149***	−0.113*	−0.318**	3.527*	3.555*	1.007	6.920*	8.229**	2.725
	(0.056)	(0.058)	(0.087)	(1.872)	(1.883)	(2.883)	(3.383)	(3.437)	(4.273)
Industry Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
County Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
County-Time Trend	No	Yes	Yes	No	Yes	Yes	No	Yes	Yes
Top-to-Bottom Quartile	No	No	Yes	No	No	Yes	No	No	Yes
Mean Absences	0.446	0.446	0.445	3.475	3.475	3.265	7.791	7.791	7.337
Observations	157,826	157,826	78,970	157,826	157,826	78,970	70,396	70,396	35,138

Notes: All regressions include controls for sex, race, marital status, education, age, health insurance coverage, U.S. born, and county and year-quarter fixed effects. The probability of lacking PSL coverage was determined through a series of logistic regression models (see Appendix Table 4 fordetails). Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the NHIS sample from 2005 through 2018. Standard errors are in parentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

One additional element of the estimates in Table 7 is worth noting. As previously discussed, the Mandate term the DDD specification represents the effect of PSL mandate enactment on work absences for those with a high likelihood of existing PSL coverage. For both intensive margin measures of work loss days in Table 7, estimates of this term are negative, and though smaller in magnitude than the effect on those gaining coverage and not statistically significant, provide some suggestive evidence that those with existing PSL coverage experience fewer illness-related work loss days after mandate enactment. We return to this issue below when examining the effect of PSL mandates on presenteeism.

Overall, results in Table 7 are generally indicative of a positive relationship between PSL mandates and work absences for those likely to gain coverage. However, the relatively small sample size in the NHIS means that our estimates lack precision and we are hesitant to draw firm conclusions from these findings alone. To address this concern, we supplement our analysis of work absences from the NHIS with data from the CPS Basic Monthly Files. While the CPS significantly increases our sample size, the survey does not include information on whether respondents have paid sick leave coverage at their current place of employment. Therefore, we use the NHIS to calculate the mean rate of PSL coverage in the absence of a mandate for 30 separate groups defined by age (18–24, 25–34, 35–44, 45–54, and 55–64), education (high school or less, some college, and college graduate), and sex. We then estimate the following modified version of Equation (2) using the CPS sample and the PSL coverage rates for each of the 30 groups:

$$Y_{\mathit{i}\mathit{c}\mathit{t}}=\tilde{\mathit{\alpha }}+{\tilde{\mathit{\gamma }}}_{\mathrm{1}}\mathit{M}\mathit{a}\mathit{n}\mathit{d}\mathit{a}t{\mathit{e}}_{\mathit{c}\mathit{t}}+{\tilde{\mathit{\gamma }}}_{\mathrm{2}}\mathit{M}\mathit{a}\mathit{n}\mathit{d}\mathit{a}t{\mathit{e}}_{\mathit{c}\mathit{t}}\mathrm{\times }\mathit{p}\mathrm{(}\mathit{N}\mathit{o}\mathit{P}\mathit{S}\mathit{L}{\mathrm{)}}_{\mathit{g}}+\tilde{\mathit{\beta }}{\mathit{X}}_{\mathit{i}\mathit{c}\mathit{t}}+\tilde{\mathit{\theta }}{\mathit{Z}}_{\mathit{c}\mathit{t}}+{\tilde{\mathit{\delta }}}_{\mathit{c}}+{\tilde{t}}_{\mathit{t}}+$$ (3)

Where p(NoPSL)_g is the mean probability that an individual in each of the 30 categories lacks pre-mandate coverage, λ_g is an indicator for each of the 30 groups¹⁷, and all of the remaining variables are as previously defined. In this specification, ${\tilde{\gamma }}_2$, measures the effect of a PSL mandate on those most likely to gain coverage. The dependent variables in our CPS estimates are described in detail above and include the probability of a work absence in the prior week due to a respondent’s own illness or own illness, childcare problems, and other family/personal obligations (“own illness plus”). Standard errors for estimates derived from Equation (3) are bootstrapped using 500 replications to account for the two-sample nature of the estimation strategy and clustered at the county level.

Results for the effect of PSL mandates on work absences using the CPS sample are presented in Tables 8 and 9. Odd numbered columns report estimates from a specification that includes industry and county controls, while even numbered columns add a county-time trend. We begin by discussing estimates from a DD model similar to Equation (1) in Table 8. Regardless of the regression specification or the outcome we examine in Table 8, we find positive coefficients that are small in magnitude and statistically insignificant. Estimates from our DDD models in Equation (3) that distinguish between those likely to gain coverage as a result of the mandate and those with existing coverage are reported in Table 9. Here, mandates increase the probability of a prior week work absence due to own illness for those most likely to gain coverage by 0.4 percentage points compared to those with existing coverage and, though the effect is large in relative magnitude (a 24% increase), the outcome is relatively rare and our estimate is not statistically significant. In Columns (3) and (4), we replace the dependent variable with the more encompassing “own illness plus” measure. Here we find a large and statistically significant increase in the probability of missing work in the prior week for those most likely to gain coverage. Compared to those with a high likelihood of pre-mandate PSL coverage, those gaining coverage as the result of mandate increase the probability of a prior week work absence by 3.4 percentage points or 61.6%. Once again, we find evidence that those likely to have had existing coverage are less likely to report a work absence following the enactment of a mandate.

Table 8:

Difference-in-difference Estimates of the Effect of Paid Sick Leave Mandates on Prior Week Work Absences

	Own Illness		Own Illness Plus
	(1)	(2)	(3)	(4)
Mandate	0.0008	0.0002	0.0001	0.0003
	(0.0005)	(0.0006)	(0.0015)	(0.0013)
Industry Controls	Yes	Yes	Yes	Yes
County Controls	Yes	Yes	Yes	Yes
County-Time Trend	No	Yes	No	Yes
Mean Share Absent	0.0168	0.0168	0.0554	0.0554
Observations	3,054,904	3,054,904	3,054,904	3,054,904

Notes: All regressions include controls for sex, race, marital status, education, age, U.S. born, and county and year-quarter fixed effects. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the CPS sample from 2005 through 2018. Boot strapped standard errors (500 replications) are in parentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

Table 9:

Triple Difference Estimates of the Effect of Paid Sick Leave Mandates on Work Absences

	Own Illness		Own Illness Plus
	(1)	(2)	(3)	(4)
Mandate x p(NoPSL)	0.0038	0.0040	0.0332***	0.0341***
	(0.0038)	(0.0035)	(0.0064)	(0.0055)
Mandate	−0.0005	−0.0010	−0.0103***	−0.0105***
	(0.0008)	(0.0016)	(0.0016)	(0.0027)
Industry Controls	Yes	Yes	Yes	Yes
County Controls	Yes	Yes	Yes	Yes
County-Time Trend	No	Yes	No	Yes
Mean Share Absent	0.0168	0.0168	0.0554	0.0554
Observations	3,054,904	3,054,904	3,054,904	3,054,904

Notes: All regressions include controls for sex, race, marital status, education, age, U.S. born, and county and year-quarter fixed effects. The probability of lacking PSL coverage was determined using age-sex-education-specific coverage rates from the NHIS. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors,family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the CPS sample from 2005 through 2018. Boot strapped standard errors (500 replications) are in parentheses and areclustered at the county level.

^*p<0.1

^**p<0.05

^***p<0.01

Figure 2 provides event study estimates corresponding to the DDD results in Table 9, Column (3). We see no indication of differential changes in the probability of missing work in the past week for those most likely to gain coverage compared to those least likely to gain coverage in the 18 months leading up to mandate enactment. The likelihood of a prior week work absence increases slightly for those gaining coverage within the first 6 months that the mandate is in effect, and then rises again between 12 and 18 months from mandate enactment. This pattern of an increasing effect over time observed in Figure 2 is consistent with stipulations in all PSL mandates requiring workers to accrue leave incrementally. For example, a worker gaining PSL coverage through New York City’s 2014 mandate earns 1 hour of paid leave for every 30 hours worked.¹⁸

Figure 2: Event Study Estimates of the Effect of Paid Sick Leave Mandates on the Probability of a Prior Week Work Absence due to Own-Illness, Child Care Problems, or Other Personal/Family Obligations.

Notes: Estimates are from a modified version of Equation (3) that allows the effect of mandate enactment to vary over time. Month 0 represents the month-year that the PSL mandate took effect in treatment counties. Data are from the CPS sample from 2005 through 2018. Standard errors are bootstrapped using 500 replications and clustered at the county level.

In an attempt to substantiate and further examine the large effect of PSL mandates on missing work reported in Table 9, we re-estimate the relationship between mandates and “own illness plus” separately for women and men and by the number of children present in the household in Table 10. We find that PSL mandates have a much larger effect on the probability of missing work in the prior week for women gaining coverage (a 6.9 percentage point increase) compared to men (a 0.85 percentage point increase). Columns (3) through (6) estimate separate versions of Equation (3) by the number of children in the worker’s household. Point estimates of the effect of PSL mandates on prior week work absences for those gaining coverage are approximately four times larger for households with children than for childless households. For example, PSL mandates increase the probability of a previous week work absence by 1.5 percentage points (42.0%) for those with no children and by 5.9 percentage points (75.9%) for those with at least one child in the household.

Table 10:

Triple Difference Estimates of the Effect of Paid Sick Leave Mandates on Work Absences due to Own-Illness, Child Care Problems, or Other Personal/Family Obligations by Sex and by Number of Children Residing in Home

	Women	Men	No Children	At Least 1 Child	At Least 2 Children	At Least 3 Children
	(1)	(2)	(3)	(4)	(5)	(6)
Mandate x p(NoPSL)	0.0691***	0.0085	0.0154*	0.0592***	0.0547***	0.0450***
	(0.0080)	(0.0069)	(0.0081)	(0.0086)	(0.0083)	(0.0103)
Mandate	−0.0205***	−0.0024	−0.0051**	−0.0179**	−0.0139*	−0.0192
	(0.0054)	(0.0033)	(0.0020)	(0.0072)	(0.0074)	(0.0175)
Industry Controls	Yes	Yes	Yes	Yes	Yes	Yes
County Controls	Yes	Yes	Yes	Yes	Yes	Yes
County-Time Trend	Yes	Yes	Yes	Yes	Yes	Yes
Mean Share Absent	0.0915	0.0221	0.0367	0.0780	0.0849	0.088
Observations	1,464,205	1,590,699	1,676,580	1,378,324	818,788	286,430

Notes: All regressions include controls for sex, race, marital status, education, age, U.S. born, and county and year-quarter fixed effects. The probability of lacking PSL coverage was determined using age-sex-education-specific coverage rates from the NHIS. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details). County controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the county unemployment rate. Data are from the CPS sample from 2005 through 2018. Boot strapped standard errors (500 replications) are in parentheses and are clustered at the county level.

^*p<0.10

^**p<0.05

^***p<0.01

In sum, we find that PSL mandates are associated with large gains in coverage for targeted workers and that the distinction that we make between DD estimates of the intent-to-treat effect and DDD estimates approximating the effect of treatment on the treated has meaningful implications. We generally find that own illness-related work absences increase for targeted workers; however, though sizable in magnitude, these changes are often not statistically significant. Expanding our work absence definition to include own illness, childcare problems, and other family/personal obligations, we find economically and statistically significant impacts of PSL mandates on absenteeism. Our findings comport with earlier evidence showing that a large share of those using PSL benefits do so for reasons other than their own illness (Susser and Ziebarth, 2016). Furthermore, all mandates in our sample explicitly allow for paid leave for reasons other than own illness, such as caring for a sick spouse or child and, in some cases, extended family.

6. Paid Sick Leave Mandates and Presenteeism

We have previously drawn attention to the fact that illness-related work absences appear to decline for those with existing PSL coverage after mandate enactment. We now test a potential explanation for this finding: that workers gaining PSL coverage are less likely to attend work while sick. To do so, we use data from the 2011 and 2018 ATUS Leave Modules and estimate models similar to Equation (1). Unlike our NHIS and CPS samples, we do not have information on a respondent’s county of residence in the ATUS Leave Module.¹⁹ Therefore, we focus our presenteeism analysis on state-level PSL mandates enacted by California, Connecticut, D.C., Maryland, Massachusetts, New Jersey, Oregon, Rhode Island, Vermont, and Washington. We then drop states with county or city mandates, but no state mandate (Illinois, Minnesota, New York, and Pennsylvania). For comparison to our earlier results, we also present estimates of changes in PSL coverage and a measure similar to the “own illness plus” variable constructed using the CPS data.²⁰ Due to the size of our ATUS sample, we are unable to estimate DDD models with any real precision, and instead provide DD estimates for workers in all industries and workers in low PSL industries.

Results from this analysis are include in Table 11. Column (1) includes estimates of the effect of PSL mandates on coverage among workers in the ATUS sample. Overall, we find that PSL mandates increase coverage by 8.0 percentage points for all workers and by 11.9 percentage points for workers in low PSL industries, though only the first estimate is statistically significant. Reassuringly, these effects are quite similar to coverage estimates from our difference-in-differences models reported in Tables 3 and 4. While there may be some indication that PSL mandates are positively related to work absences for workers in low PSL industries in our ATUS sample, estimates in Column (2) are imprecise and lack statistical significance. Finally, Column (3) contains our estimates of the effect of PSL mandates on presenteeism. Panel A includes workers in all industries and the estimated effect, while negative, is small and statistically insignificant. However, when we restrict our sample to workers in low PSL industries in Panel B, we find that mandates are associated with a 4.5 percentage point reduction (p<0.10) in the likelihood that a respondent attended work in the past week while sick. Overall, results from our analysis of PSL mandates and presenteeism complement earlier work showing PSL mandates reduce the spread of contagious illness (Pichler and Ziebarth, 2017) and that Connecticut’s PSL mandate reduced illness-related work absences (Stearns and White, 2018). Connecticut’s mandate applied only to workers in service industries and those industries fall into our classification of low PSL industries.

Table 11:

Paid Sick Leave and Presenteeism

	Paid Sick Leave	Own Illness Plus	Presenteeism
	(1)	(2)	(3)
Panel A: All Industries
Mandate	0.080***	−0.019	−0.018
	(0.027)	(0.017)	(0.013)
Outcome Mean	0.575	0.064	0.014
Observations	9,707	8,790	9,486
Panel B: Low PSL Industries
Mandate	0.119	0.029	−0.045*
	(0.073)	(0.045)	(0.026)
Outcome Mean	0.351	0.028	0.027
Observations	1,904	1,660	1,863
Industry Controls	Yes	Yes	Yes
State Controls	Yes	Yes	Yes

Notes: All regressions include controls for sex, race, marital status, education, age, U.S. born, and state andyear fixed effects. Industry controls include indicators for 22 separate industry categories (see Appendix Table 3 for details and categorization of low PSL industries). State controls include number of medical doctors, family doctors, and general practitioners per capita, per capita inpatient days and outpatient visits, and Medicare Parts A and B expenditures, and the state unemployment rate. Data are from the 2011 and 2018 ATUS Leave Module sample. Standard errors are in parentheses and are clustered at the state level.

^*p<0.10

^**p<0.05

^***p<0.01

7. Falsification Analyses

As a final test of the validity of our estimates, we conduct a series of falsification analyses of the effect of PSL mandates on our four main outcomes: PSL coverage; the probability of missing any work in the previous work week due to own illness; the probability of missing any work in the previous work week due to own illness, child care problems, or other personal/family obligations (“own illness plus”); and the probability of attending work while sick (i.e., presenteeism). We use our DDD model to estimate all falsification tests with the exception of the presenteeism outcome for which we use our DD model and restrict the sample to those working in low-PSL industries. For each outcome we conduct a randomization inference procedure similar to those described by Conley and Taber (2011), Kaestner (2016), and Mazumder and Miller (2016) that involves randomly assigning treatment status to a group of placebo counties or states (i.e., counties or states that never enacted a PSL mandate) and then comparing changes for workers in these pseudo-treatment regions to workers in the remaining control regions. We carry out this falsification test 100 times for each of the four outcomes and plot the t-statistics of the coefficient estimates in Figure 5 along with the t-statistics of our actual estimates. For PSL coverage, work absences due to “own illness plus”, and presenteeism, t-statistics from the actual estimates are at or near the top of the distributions of placebo estimates. The t-statistic from the actual estimate of the effect of PSL mandates on work absences due to own illness is near the bottom of the placebo distribution and, accordingly, was not statistically significant in our DDD model. These falsification analyses confirm that the likelihood of our estimated effects of PSL mandates on coverage, work absences, and presenteeism occurring by chance is low and thus lend further support to the validity of our results.

Figure 5: Falsification Test Using Counties and States with No PSL Enactment.

Notes: Actual estimates denoted by the black diamonds are from models of PSL coverage using the NHIS sample, work absences using the CPS sample, and presenteeism using the ATUS sample. Placebo estimates denoted by empty circles were derived from a series of models that assigned false county treatment status and mandate enactment dates to a random selection of control counties or states. The process was repeated 100 times for each outcome changing the pseudo-treatment counties/states and pseudo-enactment dates in each iteration. Standard errors are clustered at the county level for the NHIS and CPS samples and the state level for the ATUS sample. Standard errors are bootstrapped using 500 replications for analyses using the CPS sample.

8. Discussion

Legislative mandates for paid sick leave (PSL) benefits are a relatively recent phenomenon in the U.S. Beginning with San Francisco in 2007, several cities, counties and states now require employers to provide their workers with paid leave for illness or injury (National Partnership for Women & Families, 2019). Despite strong public support for expanding PSL coverage, legislators in 19 states have banned municipal-level PSL mandates citing concerns over costs to businesses and rising rates of employee absenteeism (Pichler and Ziebarth, 2018). Evidence to support these concerns, however, is scant due to the lack of research into the labor market effects of PSL mandates in the U.S.

Our goal in this paper was to first establish a link between the enactment of a PSL mandate and rates of worker coverage. Since many employers provide PSL benefits in the absence of a mandate, the effect on coverage is not evident ex-ante and represents a crucial first step to quantifying the labor market effects of PSL mandates. Results from a standard difference-in-differences model comparing workers in counties that enact mandates to workers in counties without mandates indicate that, overall, PSL mandates increase access to paid leave by 8.4 percentage points (14.7%). We find larger effects for women, non-white and Hispanic workers, and those working in industries with low levels of PSL coverage prior to the mandate. This modest association between PSL mandates and coverage gains can explain why studies of the supply-side labor market effects of PSL mandates in the U.S. have found minimal impacts on employment and wages (Ahn and Yelowitz, 2015; Pichler and Ziebarth, 2018). However, when we modify our empirical strategy to focus on coverage changes among workers who likely lacked PSL coverage prior to a mandate, we find that mandates increase coverage by 16.0 percentage points, a relative increase of nearly 28%. Our application of this triple-difference strategy is novel to the literature on PSL and highlights the importance of concentrating on the population targeted by the mandates.

We then turn to the relationship between PSL mandates and worker absenteeism. Using two different datasets and several definitions of leave taking, our difference-in-differences models suggest that PSL mandates are weakly positively associated with work absences. In an effort to refine our analysis, we again estimate a triple-difference model that allows us to observe the effects of PSL mandates on those most likely to gain coverage compared to those with a high likelihood of pre-mandate coverage. We find only a modest positive relationship between PSL mandates and illness-related work absences, however our estimates are somewhat imprecise. Alternatively, when we expand our outcome measure to a prior week work absence due to own illness, childcare problems, or other personal/family obligations, we find a large and statistically significant impact of PSL mandates on absences for those most likely to gain coverage; PSL mandates increase this type of leave taking by 3.4 percentage points (61.5%). Additionally, we show that the effect of PSL mandates on work absences is larger for women gaining coverage than for men and for those with children present in the household.

Finally, we provide the first estimates of the effect of PSL mandates on presenteeism in the U.S. using data on workers’ leave taking behaviors. We find that PSL mandates reduce the likelihood that an individual attended work in the past week while sick and that the magnitude of this effect grows when we focus on those in industries where workers have historically lacked PSL coverage.

Appendix

Appendix Table 1:

Municipal and State Paid Sick Leave Mandates, 2005 – 2018

Municipality or State	Effective Date	Scope of Coverage (firm size)	Estimated Population Gaining PSL	Cumulative Population Gaining PSL
San Francisco, CA	Feb.2007	-	59,000	59,000
Washington DC	Nov. 2008	-	220,000	279,000
Connecticut	Jan. 2012	50+	200,000	479,000
Seattle, WA	Sept. 2012	5+	150,000	629,000
Portland, OR	Jan. 2014	6+	120,000	749,000
Jersey City, NJ	Jan. 2014	-	38,000	787,000
New York, NY	Apr. 2014	6+	1,200,000	1,987,000
Newark, NJ§	May 2015	-	42,500	2,029,500
Irvington, NJ§; Passaic, NJ; East Orange, NJ§; Paterson, NJ	Jan. 2015	-	51,300	2,080,800
Oakland, CA; Montclair, NJ§; Trenton, NJ	Mar. 2015	-	74,800	2,155,600
Philadelphia, PA	May 2015	10+	200,000	2,355,600
Bloomfield, NJ§	Jun.2015	-	7,300	2,362,900
California	Jul. 2015	-	6,900,000	9,262,900
Massachusetts	Jul. 2015	11+	900,000	10,162,900
New Brunswick, NJ	Jan. 2016	-	9,500	10,172,400
Oregon	Jan. 2016	10+ (6+ in Portland)	473,000	10,645,400
Tacoma, WA	Feb. 2016	-	40,000	10,685,400
Elizabeth, NJ	Mar. 2016	-	19,000	10,704,400
Plainfield, NJ; San Diego, CA	Jul. 2016	-	441,000	11,145,400
Montgomery County, MD	Oct. 2016	5+	90,000	11,235,400
Morristown, NJ; Spokane, WA; Vermont	Jan. 2017	-	103,000	11,338,400
Arizona; Chicago, IL; Cook County, IL; Minneapolis, MN; St. Paul, MN	Jul. 2017		1,965,000	13,303,400
Washington	Jan. 2018	-	1,000,000	14,303,400
Maryland	Feb. 2018	15+	750,000	15,053,400
Rhode Island	Jul. 2018	18+	100,000	15,153,400
Austin, TX; New Jersey	Oct. 2018	-	1,423,000	16,576,400

Notes: Effective dates and population estimates were collected from the National Partnership for Womenand Families (2019).

The § symbol indicates that these municipalities are in Essex County, NJ, which we consider to have enacted a PSL mandate beginning with Newark in the 2^nd quarter of 2015. Additional scope of coverage limitations apply for several mandates (see https://www.nationalpartnership.org/our-work/resources/economic-iustice/paid-sick-davs/paid-sick-davs-statutes.pdf for further details).

Appendix Table 2:

Descriptive Statistics for CPS Sample

	Treatment Counties	Control Counties
Female	0.474	0.483
Age	39.15	40.06
Black, non-Hispanic	0.109	0.102
Hispanic	0.276	0.139
Other race, non-Hispanic	0.128	0.084
White, non-Hispanic	0.487	0.675
Less than High School	0.119	0.079
High School Graduate or GED	0.241	0.294
Some College	0.263	0.313
College Degree	0.378	0.314
Married	0.496	0.543
Widowed	0.012	0.014
Separated/Divorced	0.110	0.133
Never Married	0.382	0.310
Observations	1,124,986	1,929,918

Notes: Treatment counties include those enacting a PSL mandate between 2007 and 2018, while control counties are those with no PSL mandate in place over this time period. Data are from the CPS sample from2005 through 2018.

Appendix Table 3:

NHIS Industry Classifications

*Agriculture, Forestry, Fishing, and Hunting Industries

Mining Industries

Utilities Industries

Construction Industries

Manufacturing Industries

Wholesale Trade Industries

Retail Trade Industries

Transportation and Warehousing Industries

Information Industries

Finance and Insurance Industries

Real Estate and Rental and Leasing Industries

Professional, Scientific, and Technical Services Industries

Management of Companies and Enterprises Industries

*Administrative and Support and Waste Management and Remediation

Education Services Industries

Health Care and Social Assistance Industries

*Arts, Entertainment, and Recreation Industries

*Accommodation and Food Services Industries

*Other Services (expect Public Administration Industries)

Public Administration Industries

Armed Forces

Industry Unknown

Notes

^*indicates that the industry is included in the “Low PSL” sample. These are industries in which fewer than the overall mean share of workers report PSL coverage prior to the enactment of a mandate.Industries with no asterisk are included in the “High PSL” sample.

Appendix Table 4:

Characteristics Associated with Pre-Mandate Paid Sick Leave Coverage

	Coefficient	Standard Error
Demographics
Intercept	0.358	0.277
<50% FPL	−0.786***	0.084
50–74% FPL	−0 791***	0.084
75 to 99% FPL	−0.538***	0.070
100–124% FPL	−0.246***	0.063
125–149% FPL	−0.264***	0.061
150–174% FPL	−0.234***	0.060
175–199% FPL	0.061	0.059
200–249% FPL	0.013	0.047
250–299% FPL	0.102**	0.047
300–349% FPL	0.178***	0.050
350–399% FPL	0.279***	0.053
400–449% FPL	0.331***	0.055
450–499% FPL	0.250***	0.058
>=500% FPL	0.413***	0.036
Missing FPL	Omitted	Omitted
Married	−0.376**	0.190
Widowed	−0.188	0.207
Divorced	−0.211	0.192
Separated	−0.128	0.197
Never Married	−0.237	0.191
Living with a Partner	−0.355*	0.194
Maritai Status Missing	Omitted	Omitted
White, non-Hispanic	−0.220***	0.045
Black, non-Hispanic	0.096*	0.050
Hispànic	0.106**	0.044
Other race, non-Hispanic	Omitted	Omitted
Less Than High School	−0.485***	0.053
High School Graduate or GED	−0.252***	0.047
Some College or Associates Degree	−0.262***	0.044
Bachelor’s Degree	0.057	0.044
Graduate or Professional Degree	Omitted	Omitted
Female	−0.086***	0.023
US Born	0.151***	0.031
Health Insurance
Private Health Insurance	1 170***	0.133
Medicaid Health Insurance	−0.302**	0.140
Military Health Care	0.086	0.138
State-sponsored Health Pian	−0.385**	0.151
Other Government Pian	−0.390***	0.150
Single Service Pian	0.175***	0.055
No Coverage of Any Type	−0.559***	0.135
Health Insurance Missing	Omitted	Omitted
Industry
Agriculture, Forestry, Fishing, and Hunting	−0.931***	0.206
Mining	0.168	0.207
Utilities	0.803***	0.175
Construction	−1 179***	0.073
Manufacturing	−0.051	0.068
Wholesale Trade	0.096	0.084
Retail Trade	−0.237***	0.067
Transportation and Warehousing	−0 199***	0.076
Information	0.188**	0.088
Finance and Insurance	0.608***	0.079
Reai Estate and Rental and Leasing	−0.533***	0.089
Professional, Scientific, and Technical Services	0.081	0.072
Management of Companies and Enterprises	0.469	0.364
Admini strati ve and Support and Waste Management	−0.715***	0.074
Education Services	−0.073	0.083
Health Care and Social Assistance	0.186***	0.067
Arts, Entertainment, and Recreation	−0 771***	0.091
Accommodation and Food Services	−1 047***	0.075
Other Services (except Public Admini strati on)	−0 744***	0.075
Public Administration	0.596**	0.254
Armed Forces	−0.183	0.914
Unknown	Omitted	Omitted
Employer Size	0.0005	0.0008
Coimty controls Per Capita Physicians	36.664***	8.922
Per Capita General Practitioners	−309.790	605.029
Per Capita Family Practitioners	−617.537	649.803
Per Capita Hospital Days	0.074*	0.040
Per Capita Outpatient Days	−0.038***	0.012
Medicare Spending	−0.000	0.000
Unemployment Rate	−0.021***	0.005
Observations	51,589

Notes: Estimates are from a logistic regression model with paid sick leave coverage as the dependent variable using data prior to mandate enactment. In addition to the characteristics listed in the table, the regression also contains individual age dummies, which we omit to conserve space. Data are from the NHISsample.

^*p<0.10

^**p<0.05

^***p<0.01