The incidence of Mandated Maternity Benefits By Jonathan Gruber, AER vol 84. No. 3 pp 622-641 Gruber AER 1994 • “The Incidence of Mandated Maternity Benefits” – Maternity Mandates – enacted in several states and then at the federal level to force employers to offer comprehensive childbirth coverage in health insurance for all women. It, therefore, raises the costs of hiring childbearing age women. – It’s an attractive way for Governments to engage in social policy since they do not have to finance it. – But it may have other effects that are unintended. • This paper: – exploits the enactment of state and federal maternity mandates (“natural experiments”) to estimate the response of the labor market. – Data: Current Population Survey (CPS) Gruber AER 1994 • Mandates (relative to public provision) may be more efficient because they are benefit taxes i.e. the beneficiaries pay (part) of the mandate through for example lower wages or increased in labor supply reducing the deadweight loss associated with them. • If valuation of the benefits is full, then there is no deadweight loss associated with mandates because employers transfer the benefits to lower wages. Gruber AER 1994 • Group-specific mandates may have inefficiencies and result in increased unemployment for that group even in the presence of full valuation from the employees due to impediments to lower wages accordingly: – Antidiscrimination laws (prohibiting differential pay for the same job and differential promotional decisions across groups) – Within firm antidiscriminatory rules – If the group is composed of minimum-wage earners there is no possible downward adjustment. Summers (1989) “It is possible that mandated benefit programs can work against the interests of those who most require the benefit being offered” because they introduce differences in the hiring costs of these groups relative to other groups in the population. Gruber AER 1994 • Possible effects of the mandates: – Lower wages for this group. If wages of these women are lowered in the same amount as the benefit then beneficiaries of the policy “pay” the full cost of the coverage and there should be no effects on employment (in this case the policy would be efficient, there is no DWL). – Lower labor demand for this group if wages are not able to adjust • Results: finds substantial shifting of these costs to wages for the target group Gruber AER 1994 • 1975-1979 – 23 states passed laws prohibiting treating pregnancy differently from “comparable illnesses” in health insurance benefits • October 1978 – Pregnancy Discrimination Act (PDS) Federal Law prohibiting “differentiated treatment of pregnancy in the employment relationship” Gruber AER 1994 • Using the 1977 National Medical Care Expenditure Survey, restricting to 2,900 women of ages 20-40 who were covered through employment based group health insurance (in own name or family member) Gruber finds that around 20% of women did not have coverage for maternity benefits while they had coverage for “inhospital room and board” and “other services” Gruber AER 1994 Gruber AER 1994 • In this paper he estimates whether there is shifting of the costs of the mandates to the wages of the affected group. If there is no shifting then either: – The group that benefits from the mandate does not value it – Or there are impediments to the adjustment This would imply efficiency costs of the policy. Gruber AER 1994 • Empirical strategy. The objective is to identify the effect of laws passed by certain states (experimental states) on certain individuals (the treatment group). For that effect we need to control for any systematic effects to the treatment group that may be correlated with the laws but are not due to the law. – year effects – capture national trends – State effects – capture differences in earnings across states – State-year effects – capture state specific shocks which may be correlated with the passage of the laws. The exercise thus compares treated individuals in the experimental states with control individuals in those same states and compare their relative outcomes (before and after) with those in nonexperimental states. These are differences-in-differences-indifferences estimates (DDD). Gruber AER 1994 • The treatment group = insured workers that may be parents. • Control group = individuals unaffected by the law. • Identifying assumption: there must not be a contemporaneous shock affecting the relative outcomes of the treatment group in the same state-years as the law. Gruber AER 1994 Because no information on the CPS on health insurance coverage before 1979: 1) use as treatment group: married women of childbearing age (20-40 years old) and as control group: All individuals over 40 and single males 20-40. (single women 20-40 and married males 20-40 are dropped from the sample because may be either treated/control) 2) Use other data with information on insurance coverage to estimate the likelihood of insurance coverage and the type of insurance coverage that they would receive and assign to each individual an estimate of the cost of the mandate. [we will not see this part of the analysis in class] Gruber AER 1994 Experiments: a) Effects of the state mandates - Can only use those states that were identifiable in CPS before 1977 (state laws 76-77). Focus on 3 treated states (out of 23) and small set of control states. Use two years before laws (74,75) and two years after laws (77-78). On average wages fell more in experimental states than in non-experimental states. Are the mandates responsible for all or part of these drop? Gruber AER 1994 Gruber AER 1994 Experiments: b) Effects of the federal law – Because later we can identify all states from the CPS, more states can be included in the analysis. 1978-1979 as the “before period” and 19811982 as the “after period”. experimental states: those that had not enacted any state mandate by Jan 1977 control states: all those that did pass state mandates by Jan 1977 Gruber AER 1994 – raw DDD Experiments: a) Raw DDD say there is a 5.4 percent fall in the relative wages of 20-40 year old married women of childbearing age in the states that passed the law compared to the change in real wages in non-experimental states. Suggests the affected group bears part/all of the cost of the mandates. Gruber AER 1994 – raw DDD Experiments: a) DDD in a regression framework (allows to control for other observables that affect the outcome). i=individual, j=0,1 (states), t=0,1 (time before and after the law) W is the log real hourly wage (b8 captures the DDD estimate): Wijt b1 X ijt b 2 t b 3 j b 4TREATi b 5 j t b 6 t TREATi b 7 j TREATi b8 j t TREATi Gruber AER 1994 – raw DDD b8<0 means that there was shifting of the costs of the law to the treatment group in the form of low wages. The effect on # hours and employment are uncertain: a) If employers shift all cost to the worker and there is full valuation =>no change in the # hours worked nor in the probability of employment. Otherwise: b) employers may want to increase hours of work i.e. reduce part-time jobs and reduce employment for that group. This is what he finds. c) as some part-time workers do not have health insurance, there may be a shift towards part-time jobs for this group i.e. lower hours and higher employment Gruber AER 1994 – raw DDD Wages fell by 4.3% Smaller than raw DDD but similar meaning that it is similar to an experim ent Hours increase by 4.9% Employ ment decreas es by 1.6%