Paid Family and Medical Leave in the United States: A Data Agenda

Overview

The United States continues to be an outlier among developed countries for its lack of a national paid leave program. New parents across our nation struggle to balance the demands of their jobs with the needs of their families. And the necessity of paid leave only escalates as the U.S. population ages and more spouses and parents find themselves in need of support to recover from an illness or injury, or to maintain independence in their communities.

Existing paid leave benefits do not reach everyone. For low-income workers (in the lowest quartile) less than 10 percent have access to paid family leave.1 Too many families are being left out, unable to access benefits that can help them meet the needs of their families and their employers.

Understanding the programs that exist in states is an important step on the pathway to a national program. California, New Jersey, and Rhode Island have had state paid leave programs for years. And since 2017, three new states (New York, Washington, and Massachusetts) and the District of Columbia have all passed paid family and medical leave laws. Important academic research and policy analysis has been done using data from these programs to understand how paid leave affects workers, families, and employers in these states. Yet there are still a number of outstanding questions about how paid family and medical leave functions in the United States.

In a report issued in 2018, the Washington Center for Equitable Growth reviewed the existing evidence and outlined remaining questions comprising a research agenda for paid family and medical leave.2 That report provided a framework for understanding the various channels through which paid leave affects economic outcomes via individual-level health and employment outcomes, firm-level and employee outcomes, and broader macroeconomic impacts.

Compelling evidence exists in each of these channels, suggesting there is an important positive role for paid family and medical leave. But real gaps remain in the knowledge, especially regarding paid leave for one’s own medical needs, as well as caregiving leave. These open questions fall into two broad categories. The first category includes knowledge gaps where data are available that could be marshalled to answer questions. The second, and trickier, category includes knowledge gaps that require the collection of new data or expanded access to existing data to answer research questions.

This follow-up report assesses the state of the data available for answering those remaining questions. It then provides a roadmap for building the necessary data in order to generate evidence-backed policy. The report will focus on a set of key questions raised from the previous report. Specifically, it will examine:

  • The data that currently exist
  • Barriers to accessing those data
  • Pathways to removing those barriers

This report is based on both a survey of the data landscape, as well as the insights shared at a day-long event hosted by the Washington Center for Equitable Growth. Follow-up interviews with participants were then conducted, including with leading scholars of paid family and medical leave, state administrators from states that already have or are in the process of implementing paid leave programs, federal data experts, and researchers in associated fields, including public health and disability policy.

Existing datasets have a lot to offer toward better understanding paid family and medical leave policies, but uncertainty remains about whether data exists to answer critical questions. The content outlined in this report builds on the themes of the event to flesh out a data agenda that could move the issue of paid family and medical leave forward by providing the necessary raw material for answering important unanswered questions.

The remainder of this report is organized into specific topics. It begins with a discussion of the existing types of data available to study paid family and medical leave. Then, in each of the subsequent sections, a different component of paid family and medical leave is examined:

  • Parental leave: leave taken by either parent for bonding with a new baby, whether after a birth or an adoption
  • Medical leave: leave taken to recover from a serious personal illness or injury that requires leave longer than a few days
  • Caregiving leave: leave taken to care for an ill or injured family member other than a new baby
  • Demographic issues across all types of leave
  • Firm effects across all types of leave
  • Interactions with existing state and federal programs across all types of leave

For each type of leave, key questions from the initial report will be highlighted, followed by a short summary of existing data relevant to those questions. Each section includes an analysis of gaps in the existing data and suggestions for future data collection or improvements in data access that would allow for additional research in each area. In conclusion, the report will summarize a data agenda that would provide the tools necessary to strengthen the literature on paid family and medical leave.

Existing data

Existing data

There are three primary sources of data on paid family and medical leave: survey data from nationally representative surveys, administrative data collected by government agencies in the process of administering programs, and a category of other data types that range from smaller nonrepresentative surveys to employer data systems to focus groups. There are valuable data in longitudinal surveys such as the Panel Study on Income Dynamics, or PSID (compiled since 1968 by the University of Michigan’s Institute for Social Research), the National Health and Aging Trends Study, or NHATS (led by Johns Hopkins University Bloomberg School of Public Health since 2011), and the Health and Retirement Survey, or HRS (compiled by the University of Michigan since 1990). All three of these datasets help capture some data on paid leave, as well as long-term health and employment outcomes, by following the same sample of people over an extended period of time.

Yet these surveys have been limited either by a lack of focus on paid leave or the length of time the survey has existed. Neither PSID nor HRS have specific questions about the need for or use of paid leave and while NHATS collects detailed information about the need for caregiving among older adults, it has only been underway since 2011.

Cross-sectional surveys or those that have a shorter follow-up period include the U.S. Census Bureau’s Current Population Survey, or CPS, and the Survey of Income and Program Participation, or SIPP, as well as the U.S. Bureau of Labor Statistics’ National Compensation Survey, or NCS, and the American Time-Use Survey, or ATUS. These federal datasets capture characteristics of the population using paid leave and the take-up rates of programs but cannot capture the long-term effects of program participation on the participants or the unmet need for leave.

CPS has a long history of data collection but is limited to a reference week that might not capture the many ways that workers use short-term leave. ATUS has a detailed leave module from 2011, but outside of the module, the questions associated with leave-taking are limited to employment, earnings, and whether the person has provided eldercare. SIPP includes a fertility module that provides detailed retrospective information on leave taken after the birth of a child but does not contain leave data beyond that of maternity leave for mothers who gave birth to a new baby.

Despite the shortcomings of the available survey data from these large national samples, they offer useful data for analyzing these programs. Importantly, the largest omission in these surveys is data on those who need leave but are not taking it. Survey data can capture individuals who report taking leave and administrative data can capture those individuals who receive benefits while on leave, but both of these sources are inadequate to describe the unmet need for leave.

The Department of Labor Family and Medical Leave Act Worksite and Employee Survey—referred to as the DOL FMLA survey—conducted in 1995, 2000, and 2012 is likely the best source for describing the unmet need for leave among employees. By asking about need for leave regardless of ability to take the leave, these surveys find potentially eligible individuals who are difficult to identify. Survey data can infer the need for parental leave among individuals who have had a child but do not have questions in place to capture potential need to care for a personal illness or injury or to care for another family member. Additional waves of this survey would provide needed insights into those missed by policies and programs meant to support family caregivers.

The second major sources of data on paid leave is administrative data. Administrative data, collected by the government in the process of operating programs and providing benefits, is a rich and currently underutilized resource. States collect information on program applicants and participants as part of running their programs in order to determine eligibility and monitor participation and progress toward desired outcomes. Administrative data may not have the rich detail of some of the survey data available—in particular, it often lacks detailed demographic information if it is not relevant to an eligibility determination process. But these data hold the potential to understand who uses the program and for what purposes.

If administrative data is linked to data on employment and health, which is also captured by other government agencies and programs, then together those data can show the effects of program use as well. Administrative data, particularly for states that have already had paid leave programs in place, is very difficult but not impossible to access. Summary statistics are available online for programs in California, New Jersey, and Rhode Island, but getting individual-level data is nearly impossible either because of privacy protections that limit the disclosure of data, technical barriers caused by antiquated information technology systems, or a combination of the two.

Beyond nationally representative surveys and administrative data lies an assortment of different types of data that vary in their focus, type, and scope. Qualitative studies make use of surveys of small samples for in-depth analysis of experiences with family and medical leave, focus groups of underrepresented groups, data collected on employee performance and retention, among others data sources. These smaller samples and deeper dives into individual or group experiences provide details about specific aspects of caregiving experiences that are difficult to capture in a few survey questions. The small sample sizes limit the potential for analysis and extrapolation to the larger population, but still provide necessary context for understanding the lived experience of people who need leave, their caregivers, families, and employers.

Each of the following sections will describe a different aspect of paid family and medical leave using the data sources that have been used to analyze them and will highlight gaps in the data that could allow for additional analysis and suggest pathways to better data in the future.

Demographics

Demographics

Data used in current research

Inequities in access to government programs are not unusual. Nor are inequities in labor market and health outcomes. Inequities in the U.S. labor market generally and in the distribution of wealth mean that the burden of unpaid leave or job loss resulting from a need for leave may disproportionately affect nonwhite subgroups. Hispanic and African American families have a much lower median household income and considerably lower earnings that would make unpaid leave harder to manage.31

Many researchers are interested in the effect of paid leave programs on different subgroups, but data with sufficient sample sizes for subgroup analysis or detailed demographic data is limited. Some survey data exists on the experience of paid leave participants based on race and education level.32 And some focus groups have targeted particular demographic groups to shed light on the qualitative experience of paid family leave.33 California has even done some market research to look at linguistic and cultural barriers to the use of paid leave benefits.34 These smaller datasets do not allow for broader insights about utilization rates, effects, and variance among different populations.

In 2016, the National Academy of Science conducted a consensus study on family caregiving for older adults with a committee of 19 experts in the fields of caregiving, economics, and long-term services and support. The committee sought to develop a report with recommendations for public- and private-sector policies to support the capacity of family caregivers to perform critical caregiving tasks, to minimize the barriers that family caregivers encounter in trying to meet the needs of older adults, and to improve the health care and long-term services and supports provided to care recipients. The report summarized the state of the evidence in this area and highlighted some key shortcomings. The report found that:

Due to resource constraints, all the surveys that are relevant to family caregiving are limited in size, which in turn limits subgroup analyses. No current survey has sufficient power to assess the needs and experiences of older adults and their caregivers by all of the varied subgroups of interest, including those defined by race and ethnicity, rural residence, or sexual orientation.35

While this only considered caregiving for older adults, it holds true for other components of paid leave as well.

There has recently been some additional progress made in studying racial and ethnic disparities in paid leave access and utilization. Health economists and medical and policy scholars Ann Bartel and Soohyun Kim at Columbia University, Jaehyun Nam at Pusan National University, Stanford’s Maya Rossin-Slater, the University of Virginia’s Christopher Ruhm, and Jane Waldfogel at Columbia recently completed an analysis of the American Time Use Survey leave module, the Current Population Survey’s Annual Social and Economic Supplement, the National Study of the Changing Workforce, and the Survey of Income and Program Participation to find that, in particular, Hispanic workers have lower rates of paid leave access than white non-Hispanic survey participants.36 This kind of work is a promising approach to tackling subgroup analysis using representative surveys.

Data needs and a path forward

Understanding the different needs of U.S. subpopulations and how they vary can have meaningful implications for the outcomes of the program. Without understanding which subpopulations are over- or underrepresented in take-up rates for paid leave programs, how barriers vary for these subpopulations, and the ways a program could be tailored to be more accessible, disparities can persist without recognition and programs cannot be improved.

Administrative data for paid leave does not provide detailed demographic data. States with existing programs do have data going back years—California for more than a decade—but the agencies administering these programs do not collect detailed demographic data in order to impose the minimum burden on applicants, while compiling enough data to run the program effectively. In this case, the needs for minimizing burden and maximizing privacy are at odds with the desire of researchers for rich data.

For states that have existing paid leave programs, characteristics such as gender, race and ethnicity, and age are not recorded in public data released on programs. These programs collect data on age using date of birth and on gender, though none ask about race and ethnicity. To understand program effects at the subgroup level, additional data linkages must be made or techniques such as imputation used to make finer levels of analysis possible.

While imputing race is not common in the existing academic literature on paid leave, methods for imputing race have been tested in other areas and could be brought to future paid leave research on the implications for subgroups. Race imputation based on surnames has had mixed results, with more success for some groups than others. The Rand Corporation developed a method to integrate surnames and geocodes for race/ethnic coding, which improved accuracy for African Americans and Asians, but not for Native American or multiracial individuals.37

Other researchers have been more successful identifying Hispanic individuals using surname-based imputation.38 Still others use preferred language for receiving materials and residence location to impute race and ethnicity.39 These techniques have not been validated for datasets of paid leave participants but could offer an alternative to increased burden on program participants tied to asking for this information to be reported. Testing these methodologies could offer useful tools for researchers who want to work with administrative data to understand the impacts of programs for different subgroups.

In addition to better understanding the way U.S. subpopulations interact with existing programs, building a foundation for tailoring programs in the future requires additional qualitative data on:

  • The program users’ knowledge of the program
  • The program users’ experiences in the program
  • The impacts upon the program users’ employment status
  • The ways in which paid leave benefits helped program users economically, physically, and mentally

This type of qualitative data could be collected through interviews and focus groups focusing on including subgroups where existing data is not sufficient to analyze their experience of the program.

Firm-level effects

Firm-level effects

Data used in current research

A few notable surveys, particularly the Department of Labor Family and Medical Leave Act survey, have gathered data on the impact of leave-taking on firms, whether paid or unpaid. A key question about the feasibility of paid leave remains the return on investment or potential negative effects on businesses, particularly small businesses, yet data about turnover rates and productivity are limited. Surveys that currently exist ask for the self-reported effects of paid leave on costs for the business, perceptions about abuse of the program, and about coordinating paid leave with other benefits offered by the employer.40

The DOL FMLA survey includes data on the self-reported effects of providing paid leave on firms, including costs associated with administering the program and how employers cover work when employees are on leave. This survey is probably the most comprehensive survey of effects on employers, with interviews from more than 1,800 worksites, though it does not specifically differentiate between unpaid and paid leave.

Similarly, there are other studies that use survey data from various employers to gauge attitudes toward paid leave and impacts of programs where they exist. These surveys vary in size but capture some data on employers.41 The Society for Human Resource Management and the Family and Work Institute currently field a large survey of 920 employers that includes data about what type of leave they offer but does not capture some of the firm-level outcomes. The survey captures the rationale for offering programs that are not mandated and self-reported obstacles to offering benefits. In 2016, Deloitte, a global consulting firm, conducted a survey of 1,000 employees to gauge the impacts of parental leave, and similarly, Pivotal Ventures, the executive office of Melinda Gates, commissioned Ernst & Young LLP, a professional service organization, to survey more than 1,500 human resource decisionmakers and more than 3,000 employees on the role of paid family and medical leave in workplaces and the impacts it has.42

Smaller samples of employers have also been used to study the effects of paid leave programs. In 2016, the scholars Bartel, Rossin-Slater, Ruhm, and Waldfogel used a survey of small and medium businesses in Rhode Island to study firm characteristics and productivity, employee life events and work flow, and employer-provided benefits. This survey also collected data from Massachusetts and Connecticut employers over the same time period to compare responses within similar states that do not have a paid leave benefit. Similarly, Sharon Lerner and Eileen Appelbaum at the Center for Economic and Policy Research conducted similar surveys in New Jersey based on 18 in-depth interviews with employers in a variety of industries.43

Data needs and a path forward

Considering the data currently available, the large samples available through DOL FMLA employer surveys or the Society for Human Resource Management and the Family and Work Institute survey seem to provide the most promising potential resource for robust data on employers. If questions about effects on outcomes for employers around cost, turnover, and leave-taking behaviors of employees could be introduced into questionnaires for future iterations of the survey, then they could reach a larger sample than fielding new samples of willing employers.

Beyond survey data, there are beginning to be attempts to tap into administrative data collected by employers rather than the state. One approach in this area was described at the convening of experts leading up to this report. Panorama, a Seattle-based think tank, hopes to make progress in collecting employment data directly from employer tracking systems rather than from surveys, and promote public reporting of data related to paid leave. While Panorama’s current reach is small, at about eight companies from different industries and of different sizes, the organization hopes to grow the sample size over time. 44

This effort is promising in its potential to create data systems that could give researchers access to data on employment and leave-taking but a much larger undertaking would be necessary to compile robust data on firm effects comparable to the reach of the DOL FMLA survey. If this preliminary effort is promising, then this type of work could lead to a blueprint for approaching employers and building a larger dataset that would provide a more direct insight into firm behavior than self-reported surveys. But considering the metrics and interoperability of data systems across employers, it will be challenging to create a larger dataset.

Interactions with other federal and state government programs

Interactions with other federal and state government programs

Data used in current research

Any serious consideration of providing paid family and medical leave at the federal level raises questions about how such a program could be implemented. The literature on interactions with other programs is not as rich as on the effects of programs on leave-takers and care recipients. Examining these questions now would facilitate more effective conversations about the logistics of launching a large new federal program. Unpacking potential interactions between existing benefit programs and a paid leave program requires considering several major federal and state programs.

Some attempts have been made to study and document these effects. In their paper, Massachusetts Institute of Technology economist David Autor and his co-authors use data from the CPS Annual Social and Economic Supplement and short-term disability coverage rates from the U.S. Bureau of Labor Statistics to examine the relationship between private short-term disability insurance and long-term Social Security Disability Insurance.45 The authors were unable to estimate the impact on SSDI based on their empirical approach, but their work offers a valuable springboard for future study. Studies of the interactions between paid leave and other federal programs such as Temporary Assistance to Needy Families, the Supplemental Nutrition Assistance Program, and the Women, Infants, and Children program are limited. But both SIPP and CPS have been used to probe interactions between births and program utilization.46 These types of studies could be built upon to investigate the interaction with paid leave by studying utilizations rates prior to and after the implementation of paid leave, especially in states where programs will begin providing benefits in the next few years.

Evidence around the interactions between family caregiving and long-term care needs is mixed and largely based off of small samples with limited generalizability. Having a family caregiver is associated with fewer and shorter hospital stays for older adults, according to surveys such as Women’s Health and Aging Study and its accompanying Caregiving Survey.47 A relatively small sample of 156 patient-caregiver pairs was used to find that caregiving is associated with delayed or otherwise problematic hospital discharges and readmissions.48 But a sample from the Chicago Health and Aging Project found caregiving associated with increased hospitalization.49 Then, there are the randomized controlled trials that have been conducted, which demonstrate that when older adults’ caregivers receive sufficient support, hospital readmissions and expenditures for emergency room visits decline and nursing home placement is delayed.50

Each of these studies is representative of a smaller-scale study of effects of caregiving, largely not specific to paid leave, and its effects on medical care and hospitalization. While exploring these relationships is important to understanding potential implications for programs, the scale of the studies limit generalizability. Additionally, the evidence is not yet strong enough to demonstrate cost-savings or even whether promoting caregiving results in positive outcomes for the health care system, or to complete a cost/benefit analysis of caregiving for long-term services and supports.

Data needs and a path forward

To better understand the implications of paid family and medical leave on interactions between these large programs, additional research needs to be done that considers take-up rates and costs associated with the provision of paid leave. The Survey on Income and Program Participation is well-positioned to study trends in take-up rates among different programs but does not have a strong focus on leave-taking outside of maternity-related leave. Administrative data can be a better source of understanding the effects of new programs on existing programs. This could be particularly useful in states that are still implementing paid leave programs, where a baseline could be set prior to the beginning of benefit distribution.

Unlike in survey data, underreporting of the use of programs or confusion about which program a person is using is not a challenge. Accessing administrative data that includes enrollment information about programs such as Temporary Assistance for Needy Families, Social Security Disability Insurance, or Supplemental Security Income will be a challenge but would provide useful insights into caseloads and potential interactions between paid leave and other benefit programs. Barriers to this type of linked administrative data are discussed in the following section.

Policy components for paid leave programs

Policy components for paid leave programs

Data used in current research

There have been some studies of how paid leave policy elements—for example, the inclusion of statutory job protection so that leave-takers cannot be fired for taking leave—affect labor market outcomes. Additional examples include how the level of wage replacement or the portion of a person’s salary provided as a benefit affects take-up rates for the benefit or how the maximum length of leave available affects take-up rates, job retention, and health outcomes. Some studies offer promising approaches for replication in other contexts and for further examination.

The inclusion of job protection so that workers who take leave cannot be fired for taking unpaid leave they are eligible for is a provision that is included in the federal Family and Medical Leave Act and in some state programs but not others. Understanding the impact of whether job protection is included is an important component of designing a policy. One study of the effect of job protection used data from the publicly available U.S. Census Bureau’s Longitudinal Employer-Household Dynamics database, which includes employment data and data on industry dynamics and individual employment trajectories, to compare female employment within firms in California to those outside of California to identify effects of paid leave with and without job protection on female labor market outcomes.51

In 2017, Sara Bana and Kelly Bedard at the University of California, Santa Barbara and Stanford’s Maya Rossin-Slater used 10 years of linked administrative data from California to estimate the effects of the level of weekly benefits for women and their labor market outcomes, finding no evidence that a higher weekly benefit amount increases the length of leave taken or leads to adverse future labor market outcomes for mothers.52 A U.S Government Accountability Office study cited research that indicated that job protection reduced the amount of time women spent out of the labor force.53 The variation that exists in states and will continue to grow as additional states begin providing benefits of different generosities for different amounts of time to different groups of people allows for tests of the impacts of these differences on important metrics such as take-up rates, job retention or workforce participation, and health outcomes.

Design elements such as those discussed above may have important implications for the take-up of paid leave benefits, but program awareness is another key element driving uptake. Little is known about how to best ensure program awareness, and states with existing programs generally have not had the resources to test different approaches or to do sustained outreach. In this area, data on eligible individuals compared to individuals actually taking leave is the first step to identifying unmet needs where program awareness or other barriers exist. Then, testing outreach campaigns, anything from small so-called A/B testing of materials provided to new mothers, to large advertising campaigns would provide important information to other states trying to tackle similar program awareness challenges.

Some survey data exists that has targeted program awareness. Following the implementation of Rhode Island’s Temporary Caregiver Insurance program, a survey was conducted to gauge program awareness among workers who were eligible for the benefit and workers who were not.54 A similar survey was conducted in New Jersey to gauge public awareness of the benefit, finding awareness low, particularly among nonwhite adults, young adults, retired adults, and adults earning less than $50,000 per year.55 A survey in California even targeted awareness among parents of chronically ill children, a likely underrepresented group in leave-taking given the low take-up rate of nonbonding leave.56 Additional survey data need to be collected to better understand program awareness, particularly among groups that data like those in the DOL FMLA survey indicate have a need for leave but are not appearing in the utilization data for state programs.

Data needs and a path forward

These types of studies of program structure and benefit levels are rare but can be instrumental in policy conversations about the development of new programs for states considering paid family and medical leave legislation. They also are essential for deliberation about a federal paid family and medical leave program. Design questions will become increasingly critical as evidence for the need for paid leave becomes increasingly clear and a consensus for providing leave through a national program grows. Variation between states and within states where cities or counties are building on state programs provide exciting possibilities for examining how the structure and details of a program affect participation and outcomes.

Take-up rates are a major issue with regard to paid family and medical leave. While take-up rates among women for bonding after the birth of a child are much higher than for other types of leave, understanding who takes up the program is less clear. Take-up rates are much lower for fathers, adoptive parents, caregivers for other family members, and those who need leave to recover from a serious injury or medical condition. And there is still a lack of clarity about who is likely to take up the program, who is not, and why.

Comparing take-up rates in states when program components are changed—for example, when a benefit rate is increased or the length of leave is increased—provides a useful opportunity to study the effect of policy design on take-up rates. Comparing take-up rates in states with varying policy designs also can be useful, but to better understand take-up rates and target outreach to improve take-up rates, detailed demographic data about the characteristics of program participants and eligible individuals who opt not to participate is necessary.

State administrative data, however, does not capture detailed demographic information when an individual applies for benefits and also does not capture information about eligible individuals who do not apply. Using survey data to understand eligibility for benefits for individuals who have recently had a baby is the easiest way to understand the universe of possible beneficiaries and with detailed demographic information about program participants, could be used to understand what characteristics drive take-up of the program.

Implementation and policy design around paid medical leave for a personal illness pose unique challenges, and there are not currently good data available to help better understand the policy design decisions that goes into this type of leave. For the birth of a child or an adoption, there is a clear and discrete event that is tied to the generation of a birth certificate or paperwork for verification purposes. For medical claims, adjudicating the severity and appropriate length of the benefit would be much more complicated.

States with existing temporary disability insurance programs have varying adjudication processes in place, and studies that compare the costs and benefits of each type of system could guide the development of a national program. In California, for example, medical certification is provided directly to the state from a licensed medical professional. Practitioners must provide the state with a diagnosis and an International Classification of Diseases, or ICD code—a system that is used to track and understand the clinical needs of patients. Medical professionals who submit documentation to the state also must provide an anticipated date when the individual is likely to be able to return to work.57

In Rhode Island, the process is similar but the health care industry’s Official Disability Guidelines are used to determine the duration of leave. And in New Jersey, a health care provider must certify the date, if known, on which the serious health condition commenced and the probable duration, though the ICD and ODG codes are not required. Understanding if these systems result in different prescribed lengths of leave and then looking at employment outcomes could shed light on how policy designs affect labor force attachment.

First, access to administrative data where these codes are captured would need to be achieved. For the existing programs in California, New Jersey, and Rhode Island, data on ICD and ODG codes are not captured in the publicly available data on the programs, though they would be necessary for implementing such a program and so must be captured in the administrative data used by the state programs. Gaining access to this data at an individual level would allow for a better understanding of how policy design decisions affect outcomes for medical leave.

Barriers to data collection and access to data

  Barriers to data collection and access to data

Throughout this report, there have been references to barriers to progress on outstanding questions that stem from the lack of data currently available. These barriers take several forms: those created by policies, those that are the result of technological shortcomings, and those caused by a lack of data.

As a precursor to accessing data, there must be knowledge of the data that exists. Currently, despite the variety of surveys fielded from different sources that cover a range of relevant topics, there is not a directory to examine what variables have already been collected. There is a host of data available, but knowing it exists and how to access it can pose the first major barrier to research by taking valuable time to find out what exists before it can be used to answer new questions.

The spreadsheet associated with this report is a first step toward creating a directory of data resources, but it is not exhaustive. Once surveys have been identified, researchers must still gain access to data. Some data are available in public use files but are limited for privacy and security reasons. Getting individual-level data often requires a lengthy application process, negotiation of a data sharing agreement, approval from an Institutional Review Board—an ethics organization for approving research—and potentially the ability to access a secure facility such as the U.S. Census Bureau’s Research Data Center. Steps that ensure privacy and security of the data are essential, but it should be easier to understand what steps need to be taken.

For state administrative data, existing programs publish aggregate data about their programs online and through annual reports. This gives access to aggregate utilization rates, breakdowns by types of leave, parental leave, and length of leave. California even has a useful data portal to look at and download monthly aggregate program data on paid leave.58 Rhode Island makes monthly summary reports available publicly on its temporary disability insurance site.59 And New Jersey provides data in annual reports available on its Division of Temporary Disability and Family Leave Insurance site.60 This aggregate data is useful in tracking utilization rates. But without individual-level data, researchers are unable to test for patterns in use based on other factors such as demographic characteristics or firm type, or to look at outcomes over time.

Accessing individual-level data seems to be nearly impossible. For existing administrative data, attempts to access the data have been limited. Representatives from neither Rhode Island nor New Jersey could identify any instances of successfully completing data sharing agreements with outside research institutions. Rhode Island has released administrative data for a report on short-term disability and return to work.61 Yet the process for that release is no longer viable, according to state sources. Rhode Island also had worked with the University of Rhode Island on a survey about awareness and access to the state’s Temporary Caregiver Insurance program after it was enacted, but no administrative data was shared for the purposes of that research.62

Even California has had limited success working with academic research partners. The scholars Bana, Bedard, and Rossin-Slater used 10 years of linked administrative data from California to estimate the impacts of benefits,63 but the policy of the state is that individual-level data cannot be released because of confidentiality standards set by the department and their Information Security Office. This means future use of these data may not be possible.

Moreover, even if in the future, procedures were put in place to allow for access to data for research purposes, another barrier to access would be technical. In California, Rhode Island, and New Jersey the systems for paid leave are built at least in part on the Common Business-Oriented Language, or COBOL, a programing language designed in the 1950s and 1960s. The challenges of using this system are two-fold. First, it is a legacy system for which engineers and programmers are few and far between, making them expensive and leaving few resources for nonessential tasks. Second, the system is difficult to alter. The program in New Jersey has begun a modernization of their system because the current model does not allow for even basic changes to the benefit levels or eligibility criteria, but upgrades take time and not every state will allocate resources to this endeavor. Hopefully, these technical challenges will be removed as new systems are built that are more flexible, have detailed data inventories, and provide the ability to generate reports for research purposes.

Progress on administrative data

Progress on administrative data

Important data are currently locked up in systems that are difficult to use and the barriers to achieving a data sharing agreement to access such data are significant. As new states pass paid family and medical leave laws, lobbying state legislatures to include funds for and legislative language about data infrastructure, security, and usage for program integrity and academic research could lower some of these barriers. New programs have the opportunity to avoid some of the pitfalls, especially with regard to information technology barriers with which states with existing programs struggle.

A useful addition to our spreadsheet would be a scaffolding for states that incorporates the type of language and system structures that have allowed for easier access to data for research purposes in other contexts. What type of language have states used in other programs where researchers can more easily access data? Are there ways states should think about contracting for IT systems that incorporates better data hygiene and data mapping? What would states with existing programs like to change about their IT systems, or are there best practices that could be shared with other states to maximize efficiencies in new systems?

In the meantime, a promising pathway to increase access to and use of existing administrative data is to create a “blueprint” for researchers similar to the one described by U.S. Department of Labor economists Christopher McLaren and Elliot Schreur and envisioned by the spreadsheet associated with this report to outline the process for acquiring data from specific state agencies.64 Agencies that collect data on paid leave do not generally track employment and health outcomes, so data would have to be linked across multiple agencies in order to look at some of the outcomes of interest, requiring additional agreements and negotiations. While linking data owned by different agencies presents more of a challenge than for data owned by a single agency, such work makes it possible to create a much richer dataset for analysis of these programs.

The most promising potential for answering outstanding research questions is through building data linkages that connect paid family leave data to other administrative data on employment and health. To do so is a long, time-consuming process. Time is needed to build relationships and get buy-in from multiple agencies, to ensure the data is in the necessary condition for creating linkages, and for the actual linking of the data.

Yet a telling example of this approach in action comes courtesy of Washington state and was described at the Washington Center for Equitable Growth convening in 2018 on paid leave. In partnership with researchers at the University of Washington, the state of Washington’s Department of Social and Health Services Research and Data Analysis team developed the Washington Merged Longitudinal Administrative Database. WMLAD was created to examine the effects of the Seattle $15 minimum wage ordinance. While use of WMLAD is restricted to the University of Washington research team’s work on the minimum wage, this or similar data could also potentially be used to examine the effects of the new state paid leave program and other state or local policies. WMLAD links unemployment insurance data and state administrative voter, licensing, social service, income transfer, and vital statistics records to create a dataset with information on employment and earnings data, along with information on age, sex, race and ethnicity, public assistance receipt, household membership, and demographics for most residents of Washington.65

Creating this dataset has been years in the making. It required building a strong, reciprocal relationship between the University of Washington researchers and the state’s Research and Data Analysis staff. The RDA staff members were experienced at merging their internal data with Unemployment Insurance records. The University of Washington team helped to secure confidential linkable data from other agencies and to work with the RDA team on techniques for matching households. To get to this point, data sharing agreements to grant researchers access to the administrative data and state Institutional Review Board approval were necessary.

It took nearly 4 years into the project before the data were available to the University of Washington team. While its use is restricted right now, it is this kind of data infrastructure that would allow researchers to potentially answer questions about paid leave and the effects of paid leave on other programs, employment and earnings, and health outcomes in Washington state.

Washington state had an advantage in the existence of the RDA and the technical expertise that allowed for data linkages to be made. Some states will not have teams dedicated to data infrastructure and maintenance, and creating linkages such as those in the WMLAD database will be more difficult. Progress is measured in years, not months, and merging public records in this way necessarily requires state-of-the-art data security. But this type of data infrastructure could make answering some of the more difficult questions about the effects of a policy such as paid family and medical leave possible.

A key take-away that Heather Hill, one of the researchers on the WMLAD team, emphasized was that relationships with state agencies need to be reciprocal. Researchers should try to find a way for their work or expertise to be useful to the state agency rather than simply requesting data access. This approach can help build stronger relationships and improve the likelihood that access will be granted.

The path forward

This report has covered many challenges and gaps that currently exist in the data around paid leave. There are plenty of useful datasets that researchers have used to answer important questions about paid leave, and continuing to use those datasets in creative ways will allow for future research to progress but there is more that can be done. There are several key actions that could be taken to ensure the development of better data on paid family and medical leave.

Researchers can:

  • Advocate with administrators in states with existing programs to create processes that allow for use of administrative data in analyzing paid leave programs. While privacy should always be a concern, there are ways to protect privacy and still allow for research to better understand the impacts of these programs.
  • Build on access to administrative data through data linkages with health and employment data to begin analyzing the long-term effects of these programs.
  • Advocate with states just rolling out paid leave programs to incorporate a research agenda on the success and effects of their programs to ensure researchers gain access to new administrative data sources early in the process and that technological barriers can be overcome during the developmental stages of the program.
  • Advocate for additional waves of the DOL FMLA survey to capture the unmet needs for leave, particularly among those needing to recover from an illness or injury or to care for a family member other than a new infant.
  • Collect additional survey data to better understand program awareness, particularly among groups that data in the DOL FMLA survey indicate have a need for leave but are not appearing in the utilization data for state programs.
  • Expand our spreadsheet to include guides for researchers interested in accessing administrative data. This guide could include:
    • An expansion of the inventory in this spreadsheet to create an easily accessible directory of available data for researchers to use in designing studies of paid family and medical leave
    • A step-by-step process for requesting data from states, particularly including contact information for the person or office responsible for executing data sharing agreements
    • A sample Memorandum of Understanding or Data Use Agreement that had successfully been used to access administrative data
    • Resources states could use to clean their data and prepare it for sharing
    • Model legislative or regulatory language that states could use to ensure proper and secure sharing of data in new programs as they are implemented so that researchers could advocate for better access going forward
  • Test and validate techniques such as race and ethnicity imputation with paid leave administrative data.
  • Advocate for the inclusion of specific questions about paid leave in surveys such as the Survey on Income and Program Participation and additional waves of the American Time-Use Survey’s leave module.

Funders can:

  • Ensure researchers pursuing data sharing agreements and data linkages across agencies are able to obtain long-term funding. Negotiating these agreements and accessing the data will take time—much longer than a year—and without sustained funding opportunities, it is unlikely that they will find success.
  • Consider funding questions or modules for large representative surveys such as the Panel Study of Income Dynamics.
  • Develop contacts in state legislatures and agencies to help advocate for proactive planning around data sharing in new programs.

Policymakers can:

  • Ensure state and federal agencies have the infrastructure in place to ensure privacy while allowing for data sharing for research purposes.
  • Remove legislative barriers to creating data sharing agreements.
  • Support and fund additional waves of important surveys that collect data on paid family and medical leave and advocate for inclusion of questions on paid family and medical leave.

Conclusion

The time has come to build on the strong foundational research in the area of paid leave across the United States. Exciting work in data linkages, building relationships with private employers willing to share data on their workforce and the implementation of paid leave policies, and detailed investigations of the effects of population subgroups who use paid leave are being done. This work will continue to build on the evidence for this policy intervention but data needs represent a major challenge to furthering a research agenda.

Administrative data from state programs could provide a wealth of knowledge. Developing better systems at the state level and allowing researchers to access data should be a priority. By linking administrative data on paid leave with other state administrative data, it would be possible to track employment outcomes, health outcomes, and use of other social programs.

There is also a need to focus on collecting more detailed data on medical and caregiving leave. Available survey data does not capture sufficient detail to understand the types of conditions during which individuals need leave and the variation in outcomes that results from these varying needs. Survey data need to focus on collecting more data on medical and caregiving leave to build on the foundation of research that currently exists and catch up with the research on parental leave around health and employment effects. Addressing this gap with new data would facilitate better understanding of the needs of caregivers and the ways that paid leave programs can best address those needs to promote better outcomes.

The need for paid family and medical leave is clear. As more states consider or implement programs to provide this benefit and as a federal program is considered, the research powered by the types of data discussed in the report will help support better policy decisions, will help policymakers better understand the impacts of their design decisions, and will support future improvements to existing programs, ultimately improving the health and employment outcomes of countless families.

About the author

Amy Batchelor is an adjunct instructor at the Columbia University School of Social Work. She is a former Presidential Management Fellow and examiner at the Office of Management and Budget, where she worked on policy and budgetary issues related to the U.S. Department of Labor, including paid family and medical leave. Prior to her work at the Office of Management and Budget, she supported the 2016 National Academies consensus study on Families Caring for an Aging America. She holds a master’s degree from Columbia University School of Social Work and a bachelor’s degree from the George Washington University.

Acknowledgements

The author would like to thank Elisabeth Jacobs and Alix Gould-Werth for all their support in the development and execution of this report; Ed Paisley and the Washington Center for Equitable Growth Editorial team for their excellent editorial and design support; Jill Gutierrez, Heather Hill, Chris McClaren, Amy O’Hara, Ray Pepin, Kimble Snyder, Jane Waldfogel, and Jennifer Wolff for sharing their experience and insights; and all the participants in the Advancing the Evidence on Paid Family + Medical Leave event on October 24, 2018.

Glossary of data sets

Listed in alphabetical order by acronym or the full name of the database

ACS—American Community Survey (U.S. Census Bureau)

ATUS—American Time-Use Survey (U.S. Bureau of Labor Statistics)

CPS—Current Population Survey (U.S. Census Bureau)

DOL FMLA—Department of Labor Family and Medical Leave Act Worksite and Employee Survey (U.S. Department of Labor)

HRS—Health and Retirement Survey (University of Michigan)

LEHD—Longitudinal Employer-Household Dynamics (U.S. Census Bureau)

MarketScan® Health and Productivity Management (IBM MarketScan Research Databases)

MarketScan® Commercial Claims and Encounters (IBM MarketScan Research Databases)

National Alliance for Caregiving survey (National Alliance for Caregiving)

National Longitudinal Survey of Youth (U.S. Bureau of Labor Statistics)

NSOC—National Study of Caregiving (Johns Hopkins University Bloomberg School of Public Health)

National Study of the Changing Workforce (Society for Human Resource Management)

NCS—National Compensation Survey (U.S. Bureau of Labor Statistics)

NHATS—National Health and Aging Trends Study (Johns Hopkins University Bloomberg School of Public Health)

NSAF—National Survey of America’s Families (The Urban Institute’s Assessing the New Federalism research project)

PSID—Panel Study on Income Dynamics (University of Michigan’s Institute for Social Research)

SSA-DER—Social Security Administration Detail Earnings Records (U.S. Social Security Administration)

SIPP—Survey of Income and Program Participation (U.S. Census Bureau)

Social Security Disability Insurance administrative data (Social Security Administration)

Supplemental Security Income administrative data (Social Security Administration)

Supplemental Nutrition Assistance Program administrative data (U.S. Department of Agriculture)

Temporary Assistance for Needy Families administrative data (U.S. Department of Health and Human Services)

Viewpoints on Paid Family and Medical Leave (Ernst & Young LLP commissioned by Pivotal Ventures)

Vital Statistics (The Centers for Disease Control and Prevention)

WMLAD—Washington Merged Longitudinal Administrative Database (Washington state’s Department of Social and Health Services Research and Data Analysis)

Women’s Health and Aging Study (John Hopkins University’s Center on Aging and Health)

Women, Infants, and Children program administrative data (U.S. Department of Agriculture)

Work, Family, Community Nexus survey (Institute on Urban Health Research at Northeastern University and Institute for Health and Social Policy at McGill University)

End Notes

1 Bureau of Labor Statistics, National Compensation Survey (U.S. Department of Labor, 2018), “Table 32. Leave benefits: Access, private industry workers,” available at https://www.bls.gov/ncs/ebs/benefits/2018/employee-benefits-in-the-united-states-march2018.pdf.

2 Elisabeth Jacobs, “Paid Family and Medical Leave in the United States” (Washington: Washington Center for Equitable Growth, 2018).

3 Maya Rossin-Slater, Christopher J. Ruhm, and Jane Waldfogel, “The Effects of California’s Paid Family Leave Program on Mothers Leave Taking and Subsequent Labor Market Outcomes,” Journal of Policy Analysis and Management 32 (2) (2011): 224–245.

4 Linda Houser and Thomas P. Vartanian, “Pay Matters: The Positive Economic Impacts of Paid Family Leave for Families, Businesses and the Public” (New Brunswick, NJ: Center for Women and Work, 2012).

5 Linda Laughlin, “Maternity Leave and Employment Patterns of First-Time Mothers: 1961-2008” (Washington: U.S. Census Bureau, 2011), available at https://www.census.gov/prod/2011pubs/p70-128.pdf.

6 Tanya S. Byker, “Paid Parental Leave Laws in the United States: Does Short-Duration Leave Affect Women’s Labor-Force Attachment?” American Economic Review 106 (5) (2016): 242–46.

7 Emma Tominey, “Maternity Leave and the Responsiveness of Female Labor Supply to a Household Shock” (Bonn, Germany: Institute for the Study of Labor, 2013), available at http://ftp.iza.org/dp7462.pdf.

8 Gretchen Livingston, “The Link Between Parental Leave and the Gender Pay Gap” (Washington: Pew Research Center, 2013).

9 YoonKyung Chung and others, “The Parental Gender Earnings Gap in the United States” (Washington: Center for Economic Studies, 2017), available at https://www2.census.gov/ces/wp/2017/CES-WP-17-68.pdf.

10 Claudia Goldin, “How to Achieve Gender Equality,” Milken Institute Review (3) (2015): 24–33.

11 Maya Rossin-Slater, “The effects of maternity leave on children’s birth and infant health outcomes in the United States,” Journal of Health Economics 30 (2) (2011): 221–239, available at https://doi.org/10.1016/j.jhealeco.2011.01.005.

12 Office of Labor Standards Enforcement, Paid Parental Leave Ordinance (City of San Francisco, n.d.), available at https://sfgov.org/olse/paid-parental-leave-ordinance.

13 EY, “Viewpoints on paid family and medical leave” (2017), available at https://www.ey.com/Publication/vwLUAssets/EY-viewpoints-on-paid-family-and-medical-leave/$FILE/EY-viewpoints-on-paid-family-and-medical-leave.pdf

14 Jacob Klerman, Kelly Daley, and Alyssa Pozniak, “Family and Medical Leave in 2012: Technical Report” (Cambridge, MA: Abt Associates, 2012).

15 Bureau of Labor Statistics, “Employee access to disability insurance plans,” The Economics Daily, October 26, 2018, available at https://www.bls.gov/opub/ted/2018/employee-access-to-disability-insurance-plans.htm;
Bureau of Labor Statistics, “Table 22: Short-term disability plans: Method of funding, private industry workers” (Washington: U.S. Department of Labor, 2017), available at https://www.bls.gov/ncs/ebs/benefits/2017/ownership/private/table22a.pdf

16 Kristen Monaco, “Disability insurance plans: trends in employee access and employer costs,” Beyond the Numbers: Pay and Benefits 4 (4) (2015); Bureau of Labor Statistics, https://www.bls.gov/opub/btn/volume-4/disability-insurance-plans.htm

17 A.Z. Fu and others, “Absenteeism and short-term disability associated with breast cancer,” Breast Cancer Research and Treatment (2011): 130–235, available at https://doi-org.ezproxy.cul.columbia.edu/10.1007/s10549-011-1541-z

18 See, for example, Allison Earle, S. Jody Heyman, and John Z. Ayanian, “Work Resumption after Newly Diagnosed Coronary Heart Disease: Findings on the Importance of Paid Leave,” Journal of Women’s Health 15 (4) (2006).

19 Annette M. Bourbonniere and David R. Mann, “Benefit Duration and Return to Work Outcomes in Short Term Disability Insurance Programs: Evidence from Temporary Disability Insurance Program,” Journal of Occupational Rehabilitation (2018), available at http://doi.org/10.1007/s10926-018-9779-5.

20 V. A. Freedman and B. C. Spillman, “Disability and care needs among older Americans,” Milbank Quarterly 92 (3) (2014): 509–541.

21 National Alliance for Caregiving, “Rare Disease Caregiving in America” (2018), available at https://www.caregiving.org/wp-content/uploads/2018/02/NAC-RareDiseaseReport_February-2018_WEB.pdf.

22 Brian Morefield and others, “Leaving it to the Family: The Effects of Paid Leave on Adult Child Caregivers” (Washington: U.S. Department of Labor, 2016), available at https://www.dol.gov/asp/evaluation/completed-studies/Paid_Leave_Leaving_it_to_the_family_Report.pdf.

23 Richard W. Johnson and Anthony T. Lo Sasso, “The Impact of Elder Care on Women’s Labor Supply,” Inquiry 43 (3) (2006): 195–210.

24 Bureau of Labor Statistics, Unpaid Eldercare in the United States – 2015-16: Data form the American Time Use Survey (U.S. Department of Labor, 2017), available at https://www.bls.gov/news.release/pdf/elcare.pdf.

25 S. Ettner, “The Impact of “Parent Care” on Female Labor Supply Decisions,” Demography 32 (1) (1995): 63–80, available at http://www.jstor.org/stable/2061897.

26 E. Dentinger and M. Clarkberg, “Informal Caregiving and Retirement Timing among Men and Women: Gender and Caregiving Relationships in Late Midlife,” Journal of Family Issues 23 (7) (2002): 857–879.

27 Alison Earle and Jody Heymann, “Protecting the health of employees caring for family members with special health care needs,” Social Science & Medicine 73 (1) (2011): 68–78.

28 M. A. Schuster and others, “Awareness and use of California’s Paid Family Leave Insurance among parents of chronically ill children,” JAMA 300 (9) (2008): 1047–55.

29 National Academies of Sciences, Engineering, and Medicine, Families caring for an aging America (Washington: The National Academies Press, 2016).

30 The Commission on Evidence-Based Policy Making, “The Promise of Evidence-Based Policymaking” (2017), available at https://www.cep.gov/content/dam/cep/report/cep-final-report.pdf.

31 U.S. Census Bureau, Real Median Household Income by Race and Hispanic Origin: 1967 to 2017 (U.S. Department of Commerce, September 2018); U.S. Census Bureau, Current Population Survey, 1968 to 2018 Annual Social and Economic Supplements (U.S. Department of Commerce, March 2018); Bureau of Labor Statistics, Labor force characteristics by race and ethnicity, 2017 (U.S. Department of Labor, 2018), available at https://www.bls.gov/opub/reports/race-and-ethnicity/2017/pdf/home.pdf.

32 Juliana Horowitz and others, “Americans Widely Support Paid Family and Medical Leave, but Differ Over Specific Policies Personal experiences with leave vary sharply by income” (Washington: Pew Research Center, 2017), available at http://assets.pewresearch.org/wp-content/uploads/sites/3/2017/03/22152556/Paid-Leave-Report-3-17-17-FINAL.pdf.

33 Russell Tisinger and others, “Understanding Attitudes on Paid Family Leave: Discussions with Parents and Caregivers in California, New Jersey and Rhode Island” (Washington: L&M Policy Research for the U.S. Department of Labor, 2016); Available at: https://www.dol.gov/asp/evaluation/completed-studies/Paid_Leave_AwarenessBenefitsBarriers.pdf.

34 California Employment Development Department, “Paid Family Leave Market Research” (2015), available at https://www.edd.ca.gov/disability/pdf/Paid_Family_Leave_Market_Research_Report_2015.pdf.

35 National Academies of Sciences, Engineering, and Medicine, Families caring for an aging America.

36 Christopher F. McLaren and Elliot J. C. Schreur, “Improving Access to Data for Disability-Related Topics” (Washington: National Academy of Social Insurance, forthcoming in 2019).

37 M. Elliott, “Presentation to the IOM Committee on Future Directions for the National Healthcare Quality and Disparities Reports: Use of indirect measures of race/ethnicity to target disparities,” March 12, 2009 (Newport Beach, CA, 2009).

38 C. Eicheldinger and A. Bonito, “More accurate racial and ethnic codes for Medicare administrative data,” Health Care Financing Review 29 (3) (2008): 27–42.

39 Lisa LeRoy and others, “Research Addressing the HHS Strategic Framework on Multiple Chronic Conditions Understanding Disparities in Persons with Multiple Chronic Conditions: Research Approaches and Datasets” (Washington: U.S. Department of Health and Human Services, 2013), available at https://aspe.hhs.gov/system/files/pdf/53546/rpt_ResearchAddressing.pdf.

40 Eileen Appelbaum and Ruth Milkman, “Leaves That Pay: Employer and Worker Experiences with Paid Family Leave in California” (Washington: Center for Economic Policy Research, 2011).

41 Ann P. Bartel and others, “Employer Attitudes to Paid Family Leave” (Stanford, CA: Stanford University, 2017), available at https://web.stanford.edu/~mrossin/Bartel_et_al_EmployerAttitudesReport_Aug2017.pdf.

42 Deloitte, “Parental Leave Survey” (2016), available at https://www2.deloitte.com/content/dam/Deloitte/us/Documents/about-deloitte/us-about-deloitte-paternal-leave-survey.pdf; EY, “Viewpoints on paid family and medical leave.”

43 Sharon Lerner and Eileen Appelbaum, “Business as Usual: New Jersey Employers Experience with Family Leave Insurance” (Washington: Center for Economic and Policy Research, 2014), available at https://www.demos.org/sites/default/files/publications/nj-fli-2014-06.pdf.

44 The Paid Leave Project, “Emerging Business Trends in Family and Medical Leave” (2018), available at http://www.paidleaveproject.org/wp-content/uploads/2018/03/Panorama_PaidLeaveReport-FINAL.pdf.

45 David Autor and others, “How does Access to Short Term Disability Insurance Impact SSDI Claiming?” (Cambridge, MA: NBER Disability Research Center, 2013), available at http://projects.nber.org/projects_backend/drc/papers/odrc13-09.

46 Lindsay M. Monte, “In the Absence of Leave: The Financial Coping Strategies of Disadvantaged New Mothers in the Great Recession,” Poverty and Public Policy 7 (4) (2015): 420–35; Heather D. Hill, “Welfare as Maternity Leave? Exemptions from Welfare Work Requirements and Maternal Employment” Social Service Review 86 (1) (2012): 37–67.

47 G. Picone, R. M. Wilson, and S. Chou, “Analysis of hospital length of stay and discharge destination using hazard functions with unmeasured heterogeneity,” Health Economics 12 (12) (2003): 1021–1034; J. L. Wolff, and J. D. Kasper, “Informal caregiver characteristics and subsequent hospitalization outcomes among recipients of care,” Aging Clinical and Experimental Research 16 (4) (2004): 307–313.

48 K. Schwarz and C. Elman, “Identification of factors predictive of hospital readmissions for patients with heart failure,” Heart & Lung: The Journal of Acute and Critical Care 32 (2) (2003): 88–99.

49 X. Dong and M. A. Simon, “Elder abuse as a risk factor for hospitalization in older persons,” JAMA Internal Medicine 173 (1) (2013): 911–917.

50 National Academies of Sciences, Engineering, and Medicine, Families caring for an aging America.

51 Natasha Sarin, “The Impact of Job-Protected Leave on Female Leave-Taking and Employment Outcomes” (Cambridge, MA: Harvard University, 2017), available at https://scholar.harvard.edu/files/nsarin/files/20170821_jobprotection_final.pdf.

52 Sarah Bana, Kelly Bedard, and Maya Rossin-Slater, “The Impacts of Paid Family Leave Benefits: Regression Kink Evidence from California Administrative Data.” Working Paper (Cambridge, MA: National Bureau of Economic Research, 2018).

53 U.S. Government Accountability Office, Women and Low-Skilled Workers: Efforts in Other Countries to Help These Workers Enter and Remain in the Workforce (2007).

54 Barbara Silver, Helen Mederer, and Emilija Djurdjevic, “Launching the Rhode Island
Temporary Caregiver Insurance Program (TCI): Employee Experiences One Year Later” (Kingston, RI: University of Rhode Island, 2016), available at http://www.dlt.ri.gov/tdi/pdf/RIPaidLeaveFinalRpt0416URI.pdf.

55 Linda Houser and Karen White, “Awareness of New Jersey’s Family Leave Insurance Program Is Low, Even As Public Support Remains High and Need Persists” (New Brunswick, NJ: Center for Women and Work at Rutgers, The State University of New Jersey, 2012).

56 Mark A. Schuster and others, “Awareness and Use of California’s Paid Family Leave Insurance Among Parents of Chronically Ill Children,” JAMA 300 (9) (2008): 1047–1055.

57 “Report Fraud,” available at http://www.edd.ca.gov/Disability/Report_Fraud.htm (last accessed October 2015).

58 State of California Employment Development Department, “Paid Family Leave (PFL) – Monthly Data” (n.d.), available at https://data.edd.ca.gov/Disability-Insurance/Paid-Family-Leave-PFL-Monthly-Data/r95e-fvkm/data.

59 Rhode Island Department of Labor and Training, “Temporary Disability Insurance/Temporary Caregiver Insurance: Statistics” (n.d.), available at http://www.dlt.ri.gov/tdi/.

60 New Jersey Department of Labor and Workforce Development, Office of Research and Information, “Program Statistics” (n.d.), available at https://myleavebenefits.nj.gov/labor/myleavebenefits/about/stats/.

61 Bourbonniere and Mann, “Benefit Duration and Return to Work Outcomes in Short Term Disability Insurance Programs.”

62 Silver, Mederer, and Djurdjevic, “Launching the Rhode Island Temporary Caregiver Insurance Program (TCI).”

63 Bana, Bedard, and Rossin-Slater, “The Impacts of Paid Family Leave Benefits.”

64 Ann P. Bartel and others, “Racial and ethnic disparities in access to and use of paid family and medical leave: evidence from four nationally representative datasets,” Monthly Labor Review: Bureau of Labor Statistics (2019), available at https://www.bls.gov/opub/mlr/2019/article/racial-and-ethnic-disparities-in-access-to-and-use-of-paid-family-and-medical-leave.htm.

65 Jennifer Romich and others, “The Washington State Merged Longitudinal Administrative Database,” International Journal of Population Data Science 3 (5) (2018).