A Summary of “The Incidence of Mandated Maternity Benefits”
This is a series of summaries of papers that heralded the Credibility Revolution in Empirical Economics. It is my attempt to appreciate these works, catalog them for posterity/inspiration/review, and start thinking about Economics beyond the classroom.
In this 1994 American Economic Review article, Jonathan Gruber explores the labor market effects of maternity benefits.
It is remarkable to write the sentence above and realize a paper at once intersects public, labor and health economics. The paper is bound to show up on the reading list of a first-semester course in either of these fields. No surprise that it is Gruber’s most widely cited academic paper.
First, the policy context: until the 1970s, insurers treated maternity on par with preexisting conditions and disabilities. This meant they had latitude to offer inconsistent and incomplete coverage for expecting mothers. States, and later the federal government, stepped in to legislate a standard of maternal care for all insurance products. It was not that the benefits of maternity care were unknown: insurance companies could get away with it.
There was also (and/or still is?) a debate within the economics profession what to make of health insurance mandates. Economic theory views mandates as an implicit benefits tax: workers value the benefit at the level it costs the employer and wages adjust accordingly. However, if workers do not fully value the benefit or wages are not allowed to adjust (think minimum wage or antidiscrimination), a mandate can incur a deadweight loss akin to one generated by an explicit tax. Gruber provides evidence that workers fully value maternity benefits.
Gruber uses the 1979 Census Population Survey (CPS) dataset to focus on two fairly obvious groups: women of childbearing age (defined as those between 20 and 40 years old) and their husbands. By 1979, 23 states had passed legislation that forbade discrimination of pregnant women. Around the same time, the federal Pregnancy Discrimination Act was also passed. Both state mandates and the federal mandate are studied for their effects on labor market outcomes.
An issue in evaluating mandates (pre-ACA rollout) is public data do not feature detailed information on the benefits covered. Gruber’s work is famous in part for getting around this issue and articulating an empirical strategy based on difference-in-differences-in-differences (DDD) to study mandates:
- find a mandate that targets an identifiable group,
- identify average cost parameters and public-use data for a rich set of regression control and
- use these over a time period with sufficient within-state variation in policy for a given outcome.
Almost every paper evaluating mandates since has followed this framework. Combined with falsification tests, the framework provides compelling answers.
Results
The DDD framework is expressed in a regression as follows:
State-level: W is an outcome for individual i living in state j (treatment state versus control state) in time t (1974 versus 1978). It is regressed on X individual-level controls from the CPS, state and time effects, interaction terms with TREAT being a dummy for being in the treatment group.
Gruber chooses New Jersey, New York and Illinois as the treatment group and Connecticut, Massachussets, North Carlona, Ohio and Idaho as the control. Note that the three states have different implementation dates. Gruber makes a modeling choice to take 1978 as the “on” switch for policy.
Federal-level: Similar to above, except all states that did not pass a mandate prior to the federal one comprise the treatment group. Thus, t is 1978 and 1981, where the latter is the year by which almost all states had their own mandates.
The state-level analyses show that married women between the ages of 20 and 40 shoulder the largest burden of the mandate through a greater wage cut, relative to the wages of those in the control group. Reductions in the hourly wage combined and likelihood of employment combined with an increase in labor input illustrates employers’ response to mandates.
A mandate such as one for maternity benefits increases the fixed costs of hiring an individual worker. Employers are less likely to hire a worker and instead expect more hours and input from those currently employed. Thus, the post-mandate relative cost of employing a particular group (in this case married women aged 20 to 40) does not change.
This finding also holds for the federal level, albeit at different statistical significance levels. Results are below:
(I’m glossing over the individual parametrization portion since the main takeaways of the paper do not rely on them too much. I highly recommend reading it nonetheless.)
Twenty four years since the publication of the paper, mandates remain a relevant topic. This is because, to ensure a standard level of care, the ACA requires certain Essential Health Benefits in all health insurance plans. Given the attempts to repeal and replace the ACA, the future of the Benefits is uncertain. If we return to a policy scene dominated entirely by state mandates, it is helpful to know the effects of mandates. Maternity care is fairly price inelastic; it is the sort of necessary case you cannot easily refuse if its price goes up. And workers appear to value the mandate enough to bear its costs through depressed wages and increased input. Gruber provides estimates of the pass-through rates of a dollar increase in mandates: employers pass at least 100% of the burden of every dollar spent to employees.
