A Better Test for Whether Leave Policy Works

Paid sick leave policy is often judged by a single number: how many workers are covered, or how many days are taken. The harder question is why presenteeism can persist after access expands, and how to judge a system that improves some choices while leaving others unchanged. The answer developed across this series is a chain: coverage creates an option, usability determines whether a worker can act on it, benefit design changes the price and timing of absence, and institutional architecture decides who carries the risk.

The full chain is where the gap appears

Formal access is the beginning of protection, not its endpoint. A worker must be included by the rule, know that the benefit exists, understand how to claim it, and be able to absorb whatever cost remains. The workplace must also make absence possible in practice. If a worker is afraid of losing hours, cannot afford a partial wage replacement, or believes that too much work will remain for the team, the entitlement has not disappeared—but its practical value has been reduced.

That is why coverage statistics and presenteeism cannot be treated as mirror images. The IZA review reports that the share of U.S. private-sector jobs with paid sick leave rose from 63 percent in 2010 to 77 percent in 2023. The BLS 2024–25 survey reports 71 percent access among wage and salary workers at their sole or main job, compared with 66 percent in 2017–18. These measures use different populations and frames, so they should not be combined into one trend. Their shared implication is more modest and more useful: access has expanded, but it still does not reach everyone or guarantee that covered workers can safely use it.

The remaining gap is visible in the BLS measure of unmet need. Eight percent of wage and salary workers needed leave during an average month but did not take it. That is not an eight-percent estimate of presenteeism; some people may have delayed care, used another arrangement, or needed leave for a non-illness reason. But the reasons for non-use identify the pressures that can keep illness at work: 29 percent of those with unmet need cited too much work, 14 percent feared negative employment consequences or had a request denied, and 13 percent could not afford the loss in income. Access changes the option set, but those constraints determine whether the option is usable.

The option set is also distributed unevenly. Recent BLS figures show much higher access among full-time than part-time workers, among management and financial occupations than service occupations, among public-sector than private-sector workers, and among workers able to work at home than those unable to do so. The IZA review’s earlier U.S. benchmark shows a strong wage and firm-size gradient as well. These comparisons are descriptive rather than a complete causal diagnosis, but they establish that the same legal entitlement will not have the same reach in every job.

The next link is benefit design. A waiting day raises the cost of the first day away. Medical certification raises the cost of proving that absence is justified. A lower replacement rate makes each absent hour more expensive to the worker. A maximum duration places a limit on protection when a sickness spell continues. Each tool can reduce unnecessary absence, but each can also push a genuinely ill worker toward work, delay care, or return before recovery. The relevant question is therefore not whether a system contains incentives. It is where those incentives fall in the illness spell.

The architecture then determines how those rules are delivered. An individual credit account can make a short absence easy when a worker has a balance, using its future value to reduce the need for formal monitoring.