What Paid Sick Leave Changes—and What It Cannot
The specific question here is what happens after a paid sick leave mandate expands access: does it merely record more absence, or does it also reduce contagious work and illness in the wider population? The answer matters because leave use, infection activity, and presenteeism are different outcomes. Reading them together lets us identify a real policy effect without claiming that a mandate removes every pressure to work while sick.
The evidence is stronger than a before-and-after story
A simple before-and-after comparison would be difficult to interpret. Illness rates change with weather, circulating viruses, health behavior, workplace conditions, and many other forces. Employment also changes over time. If a city adopts paid sick leave and influenza-like illness later falls, the timing is suggestive, but it does not by itself show that the mandate caused the decline.
The studies summarized in the IZA review use a more informative comparison. Pichler and Ziebarth examine the staggered adoption of paid sick leave mandates in eight U.S. cities and compare those cities with 73 control cities over 2003–2015. Their difference-in-differences analysis uses changes over time in places that adopted a mandate alongside changes in places that did not adopt one during the relevant period. The reported sample contains 48,333 observations. The study estimates that population-level influenza-like illness declined by roughly 6 percent in the first two years after a mandate was adopted.
A second study uses state-level variation in paid sick leave mandates and CDC influenza surveillance data from 2010–2018. The IZA review reports a difference-in-differences analysis with 20,319 observations and an estimated decline of about 11 percent in population-level influenza-like illness in the first post-mandate year. The exact estimates differ across studies, as we would expect when the geography, period, data source, and policy exposure differ. Their value is not that they produce one universal number. It is that both examine policy changes against a comparison that is more informative than watching one treated place in isolation.
The comparison matters because the relevant counterfactual is invisible. We cannot observe the same city in the same week both adopting and not adopting a mandate. A control city or state provides an imperfect stand-in for what might have happened without the policy. Difference-in-differences reasoning therefore asks whether the treated places changed differently from their comparison places after adoption, rather than asking only whether illness fell in the treated places. That design can absorb some common changes—such as a season that raises illness across many places—and some persistent differences between places. It does not make the study experimental, and its credibility still depends on whether the comparison captures the relevant alternative path.
That qualification is important because paid sick leave mandates do not arrive randomly. Jurisdictions that adopt them may differ from those that do not in labor markets, public-health priorities, political conditions, or employer composition. The IZA review characterizes the reported evidence as mostly causal, which is stronger than a simple association, but “mostly causal” is not the same as “free of every assumption.” The estimates should be read as evidence that the policy changes are consistent with a meaningful effect under the studied conditions, not as a guarantee that every jurisdiction would produce the same percentage change.
The mechanism also makes the direction of the result plausible. When a worker lacks paid leave, staying home can mean losing income or risking employment. A mandate can lower that cost, making self-isolation more feasible when the worker is contagious. Fewer infectious workers in shared workplaces can then reduce exposure among coworkers and customers. The effect need not require every worker to change behavior. If enough workers who would otherwise attend while contagious can stay home, population illness activity may fall even while many other workers continue to work through minor symptoms or face barriers to leave.
The review reports complementary evidence on the behavior closer to the employment decision. Maclean and colleagues find that sick-leave use increased by about two days in the first year among private-sector employees newly covered by a mandate. Callison and Pesko report a reduction in “unmet sick leave needs,” meaning fewer respondents said they needed time away for their own illness but could not take it. These findings are consistent with the mechanism rather than contradictory to the infection results. A worker who previously worked through an illness may now take a day away. That appears as more recorded leave, but it can also be part of the reason contagious illness spreads less.
This is the first important interpretive shift: an increase in sick-leave use is not automatically a sign that a policy created waste. It may indicate that previously suppressed, socially valuable absence has become possible. The right evaluation asks what the additional leave replaced. Did it replace productive work that should have continued, or did it replace work performed at low productivity with a risk of infecting others? A leave-days statistic alone cannot answer that question.
There are at least two behavioral responses hidden inside the same increase. A newly covered worker with a contagious respiratory illness may use a day that they previously would have spent at work. That is an increase in absence but a reduction in exposure. Another worker may use more leave because the benefit lowers the cost of staying home when their illness is mild or their work could have continued. That is the moral-hazard concern. The data summarized in the review show that use changes, but they do not assign every additional day to one channel. The economic task is to estimate the balance, not to label the entire increase as either responsible or wasteful.
The timing of the outcome matters as well. The city study reports a decline over the first two years after adoption; the state study reports a first-year decline; the leave-use estimate concerns the first year among newly covered employees. These windows answer related but not identical questions. An immediate change may reflect workers learning the rule or using accumulated needs. A later change may reflect employer adjustment, repeated exposure, or changes in who remains covered. The evidence supports a policy effect in the reported windows, but it does not justify treating one short-term estimate as a permanent equilibrium.
The temporary federal response to COVID-19 illustrates both the potential and the limit of emergency access. The review reports that at least eight million employees used emergency sick leave during 2020, while an estimated 15 million people per week still reported unmet sick-leave needs. The emergency policy was unusual in scope, timing, and funding, so it should not be treated as a clean estimate of a permanent mandate. But the coexistence of substantial use and substantial unmet need is revealing. Expanding access can move behavior in the desired direction without reaching everyone who needs protection.
A partial effect is still a real effect
The reported results should be understood as a chain with losses at each link. A law may expand formal eligibility, but some workers remain uncovered. Covered workers may not understand the rule, may lack enough balance, or may fear requesting leave. Among those who can take leave, the replacement rate and duration may still make staying home costly. And among workers who stay home, the effect on community illness depends on whether the illness is contagious, when contagiousness occurs, and how many people they would otherwise have encountered.
That chain explains why lower infection activity and residual unmet need can coexist. A mandate can help the workers who gain access and use it while leaving other workers constrained by part-time status, high-contact occupations, workload, or income loss. It can reduce the average probability that a contagious worker is present without making the probability zero. It can also increase leave use among newly covered workers without increasing unnecessary absence by the same amount.
The average effect can conceal this uneven response. Suppose a mandate reaches many workers in offices where work can be reorganized, but reaches fewer workers in jobs that require close contact and physical presence. A population illness measure may improve because some transmission chains are interrupted, while the workers with the fewest alternatives continue to face high exposure. Conversely, a small number of workers in a dense workplace may have effects that are larger than their share of the workforce suggests, because one decision not to attend can protect many contacts. The available evidence does not let us calculate that distribution for every occupation, but it explains why coverage, job context, and outcome measurement must remain connected.
Residual unmet need is therefore not evidence that the mandate did nothing. It is evidence that the mandate did not close every route from illness to work. The same policy can reduce infection activity at the population level and leave a practical protection gap for a subset of workers. Treating the two findings as mutually exclusive would impose an all-or-nothing standard that the mechanism does not support.
The outcomes are related, but each answers a different question:
- Population influenza-like illness asks whether illness activity changed across a place or group. It captures an outcome consistent with reduced transmission, but it does not identify every individual behavior that produced it.
- Sick-leave use asks whether workers took more time away. It shows a behavioral response to access, but it does not tell us whether each absence was necessary or contagiousness was involved.
- Unmet sick-leave need asks whether people who needed time away could not take it. It identifies a constraint, but it does not establish what they did instead.
- Presenteeism asks whether a person worked while sick and what happened to their productivity, safety, or contacts. The studies summarized here do not provide one universal individual-level measure of that whole experience.
Keeping these outcomes separate prevents two opposite errors. The first is to dismiss a mandate because leave use rises. That treats every absence as a social loss and ignores the possibility that the policy replaces contagious work. The second is to claim that any decline in population illness proves that all forms of presenteeism have been solved. That treats a population proxy as a complete account of individual decisions and overlooks the workers whose unmet needs remain.
The evidence is also mainly U.S.-based and focused on short-term paid sick leave. The review emphasizes that the U.S. expansion studied in this literature often uses individualized sick-time credit accounts. The result therefore travels most confidently to questions about short-term access and contagious illness under similar institutional conditions. It says less about long-term sickness, countries with different social insurance arrangements, self-employed workers, or workplaces where the main constraint is not income but informal retaliation or a lack of replacement staff.
Nor does the evidence identify one universally optimal mandate. The causal studies can show that a particular expansion was followed by changes consistent with lower illness activity and greater leave-taking. They do not, by themselves, establish the best waiting period, replacement rate, monitoring rule, or maximum duration. Those parameters determine who can act on the entitlement and what kind of absence the system encourages. That is why an estimate of an average mandate effect is not the end of the policy question.
A good interpretation therefore has three layers. First, state the observation precisely: the reviewed studies report reductions in population influenza-like illness after certain U.S. mandates, increases in leave use among newly covered workers, and reductions in reported unmet needs. Second, state the mechanism that connects the observation to the policy: more usable leave can make self-isolation economically possible and reduce exposure. Third, state the boundary: the evidence supports a partial reduction in contagious presenteeism, not the elimination of working while sick or a guarantee that every worker can use leave.
That conclusion is more useful than either a blanket endorsement or a blanket rejection. It explains why paid sick leave can improve public health even when absence rises, and why a successful mandate can still leave a meaningful access problem. The remaining question is no longer merely whether mandates work. It is how their design changes the price of staying home at different moments of an illness—and whether the rules deter unnecessary absence by pushing genuine illness back into the workplace.