Introduction
At Workpartners®, we use advanced data analytics to support our suite of absence management services and provide actionable insights to our clients.
Problem
Between 2018 and 2022, New York, Washington, Massachusetts, Connecticut, and Washington, DC, all implemented paid Family and Medical Leave Act (FMLA) policies. These additions meant that seven states had implemented paid FMLA policies, with seven more states mandating coverage or planning to do so between 2022 and 2026. One of our large government clients determined that it needed a statistical model to predict FMLA rates and duration if it implemented a paid leave policy. Our client has clients across multiple states; thus, the states’ laws would have a varying impact on their workforce. Data analysis compared FMLA use across states and clients with and without a paid FMLA policy.
Solution
Using data from its Research and Reference Database (RRDb), Workpartners’ Analytics team developed statistical models to estimate the number of employees who would take leave and the number of days employees would be absent given paid or unpaid FMLA leave. (The RRDb is a de-identified aggregation of more than 5 million commercially employed individuals. It contains industry, demographics, workforce data, absence, disability, and health plan data at a person level.)
The logistic and gamma models controlled for the following factors:
- Industry
- Gender
- Age
- Area deprivation index (ADI), which is used as a proxy for salary and race
- Part-time versus full-time employees
- Exempt versus nonexempt employees
The models used data from 2017 to 2022 for more than 700,000 employees ages 19 to 69. The information was derived from 15 national clients with employees in all 50 states in industries including technology, academia, health care, and transportation.
Results
Workpartners’ analysis showed that, in the year after a paid FMLA program was introduced, there was a 15 percent increase in claims, and they were two days longer (on average) than pre-program claims. In the second year, however, claim rates and length returned to the baseline, possibly indicating a natural fluctuation in the number of claims.
The results were largely repeated across the controlling factors that our client wanted to examine. Claim rates increased in year one and decreased the next year based on gender, full-time versus part-time status, exempt versus nonexempt status, and age. The analysis of ADI found that, while claims and lost days were highest in less disadvantaged neighborhoods (where 85 percent of out client's employees live), rates declined to the baseline in year two. In more advantaged areas, the rate of claims increased in year two. In the highest cohort, days lost increased substantially.
The analysis of FMLA claims and days lost across states with and without paid FMLA leave provided several insights.
- In states with paid FMLA leave, both men and women took two more days per leave than in states without the benefit.
- Full-time employees in states with paid FMLA had leaves that were, on average, two days longer than employees in states without the benefit. Part-time employees had leaves that were one day shorter.
- Exempt employees in states with paid FMLA leave took more than twice as many days as workers in states without paid leave. Nonexempt employees had almost equal rates.
- Among different age groups, ADI quartiles, and industries, those living in states with paid FMLA leave took more FMLA leave days than those living in non-paid leave states.
The final analysis quantified the increase in lost workdays and cost that our client would likely incur if a paid policy was introduced into their current benefit offering.
Conclusion
In addition to building statistical models to estimate impact of an initial leave policy design, Workpartners developed a simulator tool for our client's team to utilize that would quantify estimates when demographic and policy parameters were modified. This was a useful tool for the team that needed to create a budget for this new leave policy. This information was impactful because it pertained to other company specific policies and the overarching impact of case management, with many policies often running concurrently.