Effects of US State Preemption Laws on Infant Mortality Rates

Abstract

Studies show that raising the minimum wage in a US state above the federal minimum wage can reduce infant mortality rates in those states. Some states have raised their minimum wage in recent decades, while many others did not, and have prohibited local authorities from doing so by enacting preemption laws. This study investigates how the recent emergence of state preemption laws that remove local authority to raise the minimum wage has affected infant mortality rates. Using county- and state-level data spanning 2001 through 2018, this study models infant mortality rates as a function of minimum wage levels, controlling for confounders. The estimated model, combined with information on the timing, location, and level of preempted minimum wages, is then used to estimate the number of infant deaths that occurred in 2018 that could be attributed to state preemption of local minimum wage increases. In the 9 largest (pop. > 250,000) metro counties most directly affected by state preemption, we estimate that in 2018, 25 infant deaths were attributable to preemption. This equates to a 5.4% reduction in these counties’ infant mortality rate. When considering all large metro counties in preemption states, as many as 605 infant deaths could be attributed to preemption. State preemption laws that remove local authority to enact health-promoting legislation, such as minimum wage increases, are a significant threat to population health. The growing tide of these laws, particularly since 2010, may be contributing to recent troubling trends in US life expectancy.

INTRODUCTION

The growth in US state preemption laws has potentially profound consequences for population health (Carr et al., 2020; Montez, 2020; Pomeranz and Pertschuk, 2017). However, few studies have attempted to quantify these consequences. This study undertakes such an analysis, presenting estimates of the consequences for infant mortality of state preemption laws that remove local authority to raise the minimum wage.

States often adopt preemption laws, which restrict or eliminate lower-level governments’ autonomy in various ways, typically to harmonize state and local laws (Briffault, 2018). The best-known example of preemption has been in the area of tobacco control (Crosbie and Schmidt 2020). However, a newer class of preemption laws intentionally prohibits local authority to address local problems and constituents’ desires, stymies progressive legislation, and protects business interests (Hertel-Fernandez, 2019; Rapoport, 2017; Bottari and Fisher, 2017). Many states have enacted laws preempting local authority across policy domains such as minimum wages, paid sick leave, firearms, pesticide control, and hydraulic fracturing.

We examine minimum wage preemption because of its high prevalence among US states and its salience for a substantial fraction of the US population. Louisiana passed the first minimum wage preemption law in 1995. Colorado’s followed in 1999, and an additional 23 states adopted such laws during 2001-2018. In several states, these laws were passed in response to locally enacted increases in minimum wages (Huizar and Lathrop, 2019). These laws are consequential: approximately 2.6 million workers make at or below the federal minimum wage and another 20.6 million (which represents 30% of all hourly, non-self-employed workers aged 18 and older) make “near minimum wage” (Desilver, 2017).

We focus on the consequences of minimum wage preemption for infant mortality, given existing evidence that increasing the minimum wage reduces infant mortality (Komro et al., 2016) and reduces the prevalence of adverse birth outcomes, including preterm and low-weight births (Wehby et al, 2019; Strully et al., 2010; Andrea et al., 2020), which together constitute the second-leading cause of infant mortality (EPA, 2018). Higher minimum wages are protective partly through lowering financial stress, maternal smoking, and teenage pregnancy, and by increasing access to pre- and postnatal care (Komro et al., 2016; Wehby et al., 2019; Bullinger, 2017). Moreover, given the relatively recent proliferation of preemption laws, a clear “signal” of their health effects might be found in outcomes such as infant mortality that are known to be sensitive to short-term exposures such as economic fluctuations (Dehejia and Lleras-Muney 2004; Orsini and Avenado 2015).

Ours is the first systematic analysis of US state preemption laws on population health. Using 18 years of county-level data, along with state- and county-level data on minimum wages and preemption laws, it has two main aims: (1) estimate the causal effect of minimum wage levels on infant mortality during a period of growing use of preemption legislation, and (2) use the estimated minimum-wage effects to predict the number of infant deaths that can be attributed to preemption.

We pay particular attention to rural-urban distinctions, in view of demonstrated differences in trends in health and labor markets according to place (Singh and Siahpush, 2009) and differential exposure to state preemption laws. Prochaska et al. (2020) have recently shown that the life-expectancy payoff to improved social and environmental contexts is larger in urban than in rural areas. Moreover, rural counties have recovered less and more slowly than urban areas since the Great Recession of 2008-2009 (Board of Governors, 2019). Together these facts suggest that the health consequences of minimum wage and its preemption will vary along the rural-urban continuum.

METHODS

Study Sample

We merged multiple data sources on infant deaths, minimum wages, the timing of state preemption laws, and several covariates for 2001 through 2018, as explained below. The merged data, aggregated to the county and calendar-quarter level for all 50 states plus the District of Columbia, contain 222,156 county-quarter observations. The study was determined to be exempt from Human Subjects review by the Syracuse University IRB.

Measures

For infant (< 1 year of age) deaths, we created county-by-quarter counts of deaths using restricted-access National Vital Statistics System (NVSS) death records. Exposure to the risk of death during infancy is represented by the mid-quarter county population of 0-1 year olds, estimated using linear interpolation of annual values found in the CDC’s Bridged-Race Population Estimates files (CDC, n.d.). We divide the population counts by four to approximate the aggregate person-quarters of exposure to the risk of death lived by infants during the indicated time period.

Quarterly data on minimum wage levels came from the Bureau of Labor Statistics, a series compiled by the Washington Center for Equitable Growth (Vaghul and Zipperer, 2016) and our own compilation of state and local statutes. Using data that record the implementation dates for all minimum wage increases, we computed county-level quarterly average minimum wage levels. When a city had a minimum wage that exceeded that of the county in which it is located, we calculated an effective county-level wage, weighting each county component’s minimum wage by its share of the county’s population using 2010 Census population figures as weights.

We include several time-varying covariates commonly used in studies on the effects of minimum wages on health outcomes. At the county level, we controlled for labor force participation and racial ethnic composition. Labor force participation reflects exposure to the minimum wage and also accounts for cyclical economic changes (Van Zendweghe 2017) that are known to influence the prevalence of adverse birth outcomes (Hamersma et al. 2018). Racial differences in infant mortality are well documented (Kandasamy et al. 2020), indicating a need to account for main effects of racial composition. However, studies of race-specific effects of the minimum wage on birth outcomes—i.e., an interaction effect—have produced mixed results (Rosenquist et al. 2020; Wehby et al. 2020).

Some have argued that evaluations of the health effects of minimum wage should be restricted to individuals with low education levels, because minimum wages are most relevant in low-wage labor markets (Leigh et al., 2019; Wehby et al., 2020). We cannot impose such a restriction because the infant death records do not include parents’ education. Instead, we controlled for the educational attainment of mothers of infants bom during the 12-month period represented by the county-level at-risk population. Using CDC birth record files, we included the percentage of mothers with less than high school, high school but fewer than 4 years of college, and 4 or more years of college. Infant mortality counts, exposures, and county-level covariates are merged using the mother’s county of residence. Finally, we account for a county’s place on the rural-urban continuum using the 9-category Rural-Urban Continuum Code (RUCC) produced by the U.S. Department of Agriculture (USDA 2019).

At the state level, we included the annual Tobacco Tax. Safety-net programs are represented by the Medicaid Income Eligibility Level (for a parent with 2 children, expressed as a percentage of the Federal poverty line), the percentage by which the state supplements the Federal Earned Income Tax Credit, and the monthly amounts of Temporary Assistance for Needy Families and Supplemental Nutrition Assistance Program for 3-person families. All dollar-denominated variables were expressed in constant 2019 dollars using the CPI-U series. Details on the sources of all covariates are included in the Supplementary Material.

Statistical Analysis

The “difference in differences” approach often used to evaluate programmatic effects cannot be used to determine the impact of state preemption of minimum wage changes on infant mortality, because the preemption “treatment” actually prevents, rather than introduces, an increase in a jurisdiction’s minimum wage (Wing et al., 2018). Therefore, we implement the following two-step approach. First, using our county-quarter data, we estimate a regression model to assess the effect of minimum wage levels on the count of infant deaths. This fixed-effects negative-binomial regression expresses the expected count of infant deaths as the product of the instantaneous mortality rate—using a log-linear regression expression—and the mid-quarter exposure described previously (Aheam et el., 2008). The model includes the time-varying county-level and state-level covariates as well as fixed effects for (1) all possible combinations of state and RUCCs, (2) calendar year and quarter, and (3) year-by-RUCC interactions, allowing for differing trends in infant mortality according to type of place. The model is estimated using Stata’s nbreg procedure (StataCorp, 2017).

In the second step, we use the minimum wage effects from step one to predict the number of infant deaths that might be attributable to minimum wage preemption. Specifically, we assume a counterfactual increase in the minimum wage if the preemption constraint were to be removed and then use our model to estimate the counts of infant deaths. The difference between the “factual” and the “counterfactual” counts and rates constitutes the “preemption effect.”

RESULTS

Descriptive Analysis

Table 1 shows the distributions of counties, population, average nominal minimum wages, and unadjusted infant mortality rates (IMRs) during the study period, 2001-2018. These IMRs are computed in the conventional way, i.e., an annual count of infant deaths per 1,000 births. The first row shows that majorities of counties and of the population are located in states that have preempted minimum wage increases. Moreover, minimum wages are lower, while infant mortality rates are higher, in counties located in minimum-wage preemption states. Overall, as well as within preemption categories, IMRs are lowest in the largest metro counties (t-tests of pairwise differences in IMRs between the largest metro counties and all other categories reject the null hypothesis in every case).

Table 1:

Distribution of counties, population, minimum wages and infant mortality rates by rural-urban continuum (RUCC)

				Percentage of Counties		Percentage of 2010 Population		Average Minimum Wage, 2001-2018 ($)			Average Infant Mortality Rate, 2001-2018
RUCC	County characteristics	% of counties	% of 2010 pop.	In preemption states	In nonpreemption states	In preemption states	In nonpreemption state	All	In preemption states	In nonpreemption states	All	In preemption states	In nonpreemption states
--	U.S. Total	--	--	69.3	30.7	55.1	44.9	6.99	6.59	7.47	6.3	7.0	5.6
1	Metro area; 1 m+ pop	13.9	54.6	12.4	17.5	44.7	66.8	7.12	6.63	7.52	6.0	6.8	5.3
2	Metro area; 250k-1 m pop	12.2	21.2	13.0	10.4	25.0	16.5	6.92	6.58	7.54	6.6	7.0	5.9
3	Metro area; < 250k pop	11.5	9.2	10.8	13.1	10.9	7.2	6.81	6.57	7.26	6.8	7.2	6.0
4	Urban area; 20k+ pop; adjacent to metro area	6.9	4.4	7.1	6.5	5.6	2.9	6.79	6.58	7.30	6.8	7.1	6.0
5	Urban area; 20k+ pop not adjacent to metro area	3.0	1.6	2.8	3.5	1.7	1.5	6.76	6.58	7.03	6.7	7.1	6.0
6	Urban area; 2,500-19,999 pop adjacent to metro area	19.2	4.8	21.5	13.9	6.8	2.4	6.63	6.51	7.04	7.2	7.5	6.2
7	Urban area; 2,500-19,999 pop not adjacent to metro area	13.9	2.7	13.9	14.0	3.4	1.8	6.68	6.53	7.01	7.0	7.3	6.4
8	Rural or < 2,500 urban pop adjacent to metro area	6.9	0.7	6.8	7.2	0.9	0.5	6.57	6.44	6.88	7.2	7.4	6.8
9	Rural or < 2,500 urban pop not adjacent to metro area	12.5	0.8	11.8	13.9	1.1	0.5	6.61	6.52	6.85	7.4	7.4	7.3

The Infant-Mortality Response to Changes in the Minimum Wage

We conducted three preliminary analyses to evaluate the robustness of our data and assumptions. We first tested for differences in the effect of minimum wages on IMR according to whether a state ever adopted a preemption law. We failed to reject the null hypothesis of no difference in response, supporting the use of all counties in all states in our analyses. Second, we examined the missingness of mothers’ education. Approximately 7% of county-quarter observations are missing mothers’ education. This is due almost entirely to the change from the 1989 to 2003 standard birth certificate form and inconsistencies in the timing of the adoption of the new form across reporting areas. We tested for differences in minimum wage effects by the presence or absence of mothers’ education data, but failed to reject the null hypothesis (for details, see the Supplementary Material). After excluding observations missing mothers’ education, our analytic sample contains 205,787 observations. Third, we assessed whether the minimum wage effect estimated without regard to rural-urban county type is consistent with prior studies. In a model (not shown) that imposes homogeneous minimum wage effects across all nine rural-urban county types, we found that an increase in minimum wage predicts a decline in infant mortality (net of the controls described earlier) and the magnitude of our estimated minimum wage effect accords with previous studies cited above. Specifically, the mortality rate ratio (MRR) = 0.987 (95% CI = 0.979, 0.984) implies that each dollar increase in minimum wage reduces the IMR by 1.3%.

Our final model, which addresses our first aim, incorporates an interaction between the minimum wage and a county’s RUCC, given differences in IMRs and minimum wages by metro status as shown in Table 1, plus all controls described earlier. Minimum wage effects from the final model, expressed as MRRs, are shown in Table 2. In seven of the nine RUCCs, the estimated effect of raising the minimum wage is to reduce infant mortality. However, in only the largest metro counties (RUCCs 1 and 2) is this reduction statistically significant (p < 0.05). In those cases, a one dollar increase in minimum wage is predicted to reduce the infant mortality rate by 1.5 to 1.8 percent. Because we find significant minimum wage effects in only the largest two county categories, we use only those counties to estimate infant deaths attributable to preemption.

Table 2.

Multiplicative effect of a 1-dollar increase in minimum wage on the infant mortality rate, by rural-urban continuum code

RUCC	Mortality rate ratio (MRR)a	95% confidence interval for MRR
1	0.985	(0.975,0.995)
2	0.982	(0.965,0.999)
3	0.989	(0.965,1.014)
4	1.007	(0.969,1.047)
5	0.967	(0.907,1.032)
6	0.994	(0.956,1.033)
7	1.037	(0.988,1.088)
8	0.980	(0.872,1.102)
9	0.935	(0.850,1.030)

Note: RUCC=rural-urban continuum code (see Table 1 for a description).

^aMortality Rate Ratio is the multiplicative effect of a 1-dollar increase in minimum wage on the infant mortality rate, for a particular RUCC category, controlling for confounders.

Although not the central focus of our analysis, we note that other safety-net factors such as Medicaid, EITC, TANF and SNAP—all of which have been shown by past studies to influence health outcomes (e.g., Komro et al. 2019; Strully et al. 2010; Almond et al. 2011)—are insignificant in our analysis, net of the other controls (see the Supplementary Material).

Infant Deaths Attributable to Preemption

To estimate the infant mortality consequences of preemption, we assume a counterfactual value of the minimum wage that would be imposed in a preempted county if the preemption constraint were removed. We then predict the expected count of infant deaths during 2018 in the presence of the assumed higher minimum wage, limiting our focus to 541 counties with RUCCs 1 or 2, the categories for which we found statistically significant minimum wage effects.

Nine Known Counterfactual Counties

For nine of these counties, we need not assume what the counterfactual minimum wage would be; in those counties, a law raising the minimum wage to a specified level was passed, but its implementation was blocked or reversed by a state preemption law. Although our calculations for these nine counties can be viewed as a lower bound estimate of preemption effects on IMR, their importance for our analysis is to provide evidence on the magnitude of response that could be expected more generally given plausible levels of minimum wage increases.

The nine “known counterfactual” counties had an average IMR that was slightly smaller than the other 532 preempted metro counties during 2001-2018 (6.64 versus 6.99; p=0.052 for a test of differences). The average 2018 minimum wage in these nine counties was $7.74 and the average preempted minimum wage was $9.99. Based on these minimum wage levels, our model predicts that these nine counties would have experienced 468 infant deaths in 2018 instead of the 493 they actually did, which is a reduction of 25 deaths (equivalent to a 5.4% reduction in their IMR) had they been permitted to raise their minimum wage to the desired level. This preemption effect ranges from a reduction of 2.3% to a 7.6% in IMR across the nine counties, reflecting variation across them in factors such as the prevailing and preempted minimum wage levels as well as the prevailing IMR. Information about these nine counties, including our calculated preemption effect, is included in the Supplementary Material.

541 Large Metro Counties

With these predictions for the nine “known counterfactual” counties serving as context, we turn to the full set of large metro counties that face a minimum wage preemption constraint (n=541). Some of those counties would raise their minimum wage if permitted to—indeed, many have passed “living wage” resolutions that indicate local support of higher wages, but that are not binding on private-sector employers (Swartz and Vasi 2011). However, rather than attempt to predict where and by how much a currently preempted county would increase its minimum wage if the preemption constraint were removed, we present, for these counties as a whole, the preemption effects associated with a range of possible counterfactual minimum wage increases.

Table 3 shows the results. If the 541 counties had been allowed to raise their minimum wage to just $8.75, we estimate this could have reduced the number of infant deaths in these counties in 2018 from 9438 to 9045, a reduction of 393 deaths, corresponding to a 4.2% reduction in the IMR. If they had raised to raise the wage to the same level ($9.99) as the nine known counterfactual counties, we estimate this could have prevented 605 infant deaths. If they had raised their wage to $15.00, it might have prevented 1419 infant deaths.

Table 3:

Effect of Minimum Wage Preemption on Infant Mortality in 541 Large Metro Counties

Counterfactual Minimum Wage	Baseline IMR	Counterfactual IMR	% Reduction in IMR	Infant Deaths Attributable to Preemption
$8.75	6.1	5.8	4.2	393
$9.99	6.1	5.7	6.4	605
$10.60	6.1	5.6	7.5	708
$11.70	6.1	5.5	9.4	892
$12.80	6.1	5.4	11.4	1071
$13.90	6.1	5.3	13.2	1247
$15.00	6.1	5.2	15.0	1419

DISCUSSION

Legal and public health scholars have argued that the recent surge of state preemption laws may have deleterious consequences for population health (Carr et al., 2020; Crosbie and Schmidt 2020; Montez, 2020; Pomeranz and Pertschuk, 2017). To the best of our knowledge, ours is the first study to quantify those consequences using an established population health indicator—the infant mortality rate. We find that among nine counties whose 2018 minimum wage would have been higher were it not for a state preemption law, IMRs would have been more than 5% lower on average, had the minimum wage increases been enacted. These nine counties are representative of a larger set of 541 large metro counties located in preemption states. Among that larger group we estimate that the average reduction in EVER ranges from about 4% to nearly 15% across a range of minimum wage values that fall within the current policy landscape. There is, of course, a range of predicted IMR responses to minimum wage increases, reflecting both measured and unmeasured state- and county-specific factors that influence a local area’s IMR. However, our range of estimated responses is reasonably narrow, with a 95^th percentile that is about 5 percentage points higher than its 5^th percentile.

Even a modest reduction in infant mortality is economically meaningful. A single “statistical life” saved is valued by as much as $9.6 million (in 2016 dollars) by US government agencies (US DoT, 2016). Furthermore, state preemption of local minimum wage increases has disproportionately occurred in high-poverty states (Huizar and Lathrop, 2019), thereby compounding disadvantage. Thus our findings concur with others who have argued that the growing use of preemption is a threat to population health (Crosbie and Schmidt 2020; Pomeranz and Pertshuk, 2017).

Our findings also point to the growing importance of US state policy contexts on population health. Since the 1970s, the policy context in which Americans reside is increasingly determined by their state of residence (Grumbach, 2018). States have captured greater policymaking authority through the transfer of decision making from the federal to state governments (often termed devolution) and the removal of certain authorities from local governments using preemption (Montez, 2020). The increasing use of preemption laws to stymie localities from addressing local problems has occurred without much attention from health researchers or the public. However, the response to the COVID-19 pandemic has brought the issue to the forefront. In 2020, several state governments—generally through executive order rather than legislation—have preempted efforts by counties or cities to adopt “stay at home” orders, social distancing, or a requirement to wear face masks as tools to control the spread of coronavirus (Local Solutions Support Center, 2020).

Our findings call attention to a tradeoff inherent in state preemption of local authority. Raising the minimum wage can improve several measures of population health (Komro et al., 2016; Wehby et al., 2019; Bullinger, 2017; Van Dykea et al., 2018), albeit at the cost of possible job losses. Some studies have found that higher minimum wages have small or nonexistent employment effects (Dube et al, 2010), and some research has shown that a minimum wage increase may induce job losses at wages less than the new minimum, but stimulate the creation of new jobs at wages above the new minimum (Cengiz et al., 2019). Keeping the minimum wage low may protect business profits and keep prices lower for consumers, but our results suggest that the tradeoff in human lives is steep.

Our results must be considered in light of certain limitations. First, as our study is ecological, we are unable to determine whether the infants’ survival is directly linked to wage levels. Future epidemiological research based on individual-level data could illuminate this issue. Second, although we have attempted to rid our estimates of omitted-variable bias using extensive controls, including spatial and temporal fixed effects, there could still be omitted variables that are correlated with both minimum wages and infant deaths. For example, patterns of hospital closings are known to have varied by rural-urban status in recent years (Wishner et al. 2016), and changed access to health services could be implicated in the association between minimum wages and infant mortality. Nevertheless, numerous studies using quasi-experimental and other robust causal methods conclude that raising the minimum wage improves birth outcomes (Wehby et al., 2019; Andrea et al., 2020). Additionally, while we have controlled for the racial composition of the at-risk population, we have not investigated whether there are race-specific responses to minimum wage changes, nor by implication, whether there could be race-specific responses to state preemption laws. A more disaggregated analysis of these questions is clearly warranted, and is a promising topic for future research. Finally, we are unable to control for maternal residential mobility. Studies suggest that between 9 and 32% of women move during pregnancy, although most stay in the same county (Amoah et al. 2018; Fell, Dodds, and King 2004; Miller, Siffel, and Correa 2010).

CONCLUSIONS

Public health and legal scholars have raised concerns that the rising tide of state preemption laws that prohibit or restrict local authority to address local problems may adversely affect population health. This study supports those warnings. It focused on one type of preemption (the removal of local authority to raise the minimum wage) and one health indicator (infant mortality), finding compelling evidence that fewer infant deaths would have occurred in the absence of preemption. The consequences of preemption on population health are likely much larger given the sheer number of domains in which preemption laws have been passed and their widespread effect on various health outcomes.

• 25 US states have preempted local authority to increase the minimum wage.

• These “state preemption laws” may have contributed to infant deaths.

• Each dollar of minimum wage may reduce infant deaths by 1.5[isp-doubt]-1.8% in metro counties.

• The 9 counties that rescinded minimum wage increases may have had 25 attributable deaths. The 541 counties that rescinded or blocked increases may have had 600+ attributable deaths.