Mandated minimum nurse-to-patient ratios have been the subject of active debate in the U.S. for over twenty years and are under legislative consideration today in several states and at the federal level.2 A stated intention of minimum ratios is to increase patient welfare through improved healthcare quality.3 Notably, however, most studies have found no or mixed effects of minimum ratios on healthcare quality in hospitals (Cook et al., 2012, Mark et al., 2013, Spetz et al., 2013) which is puzzling given the evidence of large, positive quality returns to nursing time per patient day (Gruber and Kleiner, 2012, Friedrich and Hackmann, 2021).4

The apparent contradiction between the null quality effect of minimum ratios and the large returns to nursing time raises several questions: Do minimum ratio policies lead to crowding out of other inputs due to factor substitution? An increased use of low-skilled nurses? Reductions in length of stay? Hospitals may substitute away from unregulated inputs, hire low-skilled nurses, or discharge patients “quicker and sicker” in response to minimum ratios. Each of these responses may, depending on the production technology, have adverse implications for healthcare quality. Prior literature on factor substitution and the quantity-quality tradeoff in healthcare is limited and the production technology is unique to the sector, therefore these questions must be answered empirically.

In this paper, I use the 1999 California nurse staffing mandate as an empirical setting to study the effects of minimum ratios on input use, capacity, output, costs, and healthcare quality. The mandate required hospitals to meet minimum nurse-to-patient ratios established for each hospital unit by the California Department of Health Services. I combine hospital financial reporting data and administrative patient discharge data with a difference-in-differences research design.

I find that the mandate significantly increased hospitals’ nurse-to-patient ratios and led to limited crowding out of other inputs. However, hospitals responded on other margins: increased use of lower-licensed and younger nurses, reduced capacity by 16 beds (14 percent), and increased bed utilization rates by 0.045 points (8 percent) to 64 percent. The increase in utilization suggests that hospitals were operating with excess bed capacity prior to the mandate and reduced capacity in response to a rise in costs per staffed bed.

Using administrative data on discharges for acute myocardial infarction (AMI), I find that the mandate led to a 5 percent decline in length of stay. Shorter length of stay is used as an indicator for high quality of care because delays and errors in the delivery of care increase length of stay. However, discharging patients “quicker and sicker” may be one way hospitals respond to financial incentives (Morrisey et al., 1988) or capacity constraints (Hoe, 2022). In light of the substantial capacity reduction that I document, I investigate whether the decline in length of stay is indicative of premature discharge or higher care quality. Contrary to the expectations under a “quicker and sicker” hypothesis, I find no effect on the 30-day all-cause readmission rate despite the decline in length of stay. My findings indicate that AMI patients at treated hospitals recovered more quickly following the mandate due to an improvement in care quality per day.

I exploit two institutional features for identification. First, variation in nurse-to-patient ratios across hospitals prior to the mandate created variation in the “bite” of the mandate across hospitals. Hospitals below the mandated threshold were treated. In my main specification, I estimate a difference-in-differences model comparing the outcomes in the general medical/surgical acute care unit (hereafter “acute care unit”) of hospitals initially below and above the mandated minimum ratio threshold in acute care. 5 In a heterogeneity analysis, I exploit the continuity of treatment and show that in line with expectations the treatment effect on labor increases with the gap between the hospital’s initial staffing ratio and the threshold.

Second, the mandated ratios were established at the hospital unit level and created variation in the “bite” of the mandate across hospital units within a hospital. In some hospital units (e.g. general medical/surgical acute care), the majority of hospitals were initially below the unit-specific threshold whereas in other units (e.g. general medical/surgical intensive care), the majority of hospitals were initially above. In California, intensive care units were already subject to minimum nurse-to-patient ratios under state law beginning in the 1976–1977 fiscal year (Spetz et al., 2000).6 In a robustness specification, I estimate my model on outcomes from the intensive care unit as a placebo test and show that there were no significant effects of the mandate in intensive care.

For estimation, I use annual financial data reported for each hospital unit in each hospital between 1990–2016 from the Department of Health Care Access and Information (HCAI) in conjunction with administrative patient discharge data between 1995–2008. The long time frame and granularity of the data allow me to validate my difference-in-differences research design and show several robustness specifications.

The analysis proceeds as follows. First, I find that the mandate had its intended effect on understaffed hospitals’ nurse-to-patient ratios in the acute care unit. I estimate a significant, 0.040 point increase in the nurse-to-patient ratio on a mean of 0.241 (21 percent) for treated hospitals. This implies an additional 58 min of nursing time per patient day.7 I show that roughly 39 min came from Registered Nurses (RNs) and 22 min from lower-licensed Licensed Vocational Nurses (LVNs). I show that substitution away from other labor (aides, physicians) and non-labor (capital, intermediate inputs) inputs was limited. The limited substitutability between nurse and non-nurse labor is consistent with strict scope of practice regulations in California that specify the tasks that each licensed healthcare professional is allowed to perform in the hospital setting. I consequently find that treated hospitals faced a 9 percent increase in the wage bill due to the mandate.

Second, I estimate that the average wage of RNs at treated hospitals declined by 3.3 percent relative to control hospitals. I provide descriptive evidence from several data sources that the wage decline was plausibly due to changes in RN composition towards younger and more recently licensed RNs. I use the National Sample Survey of Registered Nurses to show that RNs employed in California hospitals became younger and more recently licensed than RNs employed at hospitals in other states after the mandate. I use licensing data from the National Council of State Boards of Nursing to show that the changes in composition are consistent with a large growth of new entrants into the California nursing labor market at the time of the mandate. These new entrants came from both the “examined in-state” and “endorsed from out-of-state” channels.

Third, I estimate the effects on capacity, output, and utilization and find that treated hospitals reduced capacity by 16 beds on a mean of 118 beds (14 percent) and increased utilization rates by 0.045 points on a mean of 0.556 (8 percent) almost immediately after the mandate. The increase in utilization to 64 percent among treated hospitals suggests that hospitals were operating with significant excess capacity prior to the mandate.

Finally, I use administrative data on AMI discharges to estimate the effects on the risk-adjusted length of stay, 30-day all-cause readmission, and in-hospital mortality. I find no effect on the in-hospital mortality rate. However, I find a decline in length of stay of 0.281 days on a mean of 6.153 days (5 percent) consistent with descriptive evidence I show from the hospital financial data covering all discharges. I investigate whether the shorter length of stay is indicative of premature discharge or higher care quality. I find that the 30-day all-cause readmission rate was stable despite the decline in length of stay. I conclude that AMI patients at treated hospitals experienced increases in care quality per day which led to quicker recovery times. Importantly, I show that the increase in quality in the long-run is consistent with prior work on the returns to tenure in nursing (Bartel et al., 2014).

I show three robustness checks. First, I extend the pre-period by an additional six years for which I lack data on hospital-level patient severity8 allowing for graphical inspection of pre-trends over a longer period. Second, I repeat the main specification using the intensive care rather than acute care unit of the same sample of hospitals as a placebo test of my findings and estimate null effects for the majority of outcomes. Third, I use a heterogeneity analysis to show that in accordance with expectations the treatment effects on labor are larger for hospitals with the lowest initial ratios prior to the mandate.

My paper relates to several literatures. First, my paper attempts to bridge the gap between prior work on the effects of minimum ratio policies on quality and on the quality returns to nursing. I find that the reduction in length of stay increases over time from 2.6 percent and statistically insignificant within one year of the mandate to 6.9 percent and significant three years after the mandate. These dynamic effects are consistent with estimates of the returns to tenure in nursing measured in length of stay (Bartel et al., 2014) and suggest that the magnitude and significance of the estimated treatment effects depend on the length of the post-mandate estimation period. Prior work on the mandate estimates no or mixed quality effects using a 2004–2006 post-mandate period (Cook et al., 2012, Mark et al., 2013, Spetz et al., 2013). I complement this literature by using a longer post-mandate period from 2004–2008 over which I find positive quality effects.

My findings are therefore consistent with prior evidence on positive quality returns to nursing measured as a decline in length of stay (Bartel et al., 2014) or a decline in readmission with stable length of stay (Friedrich and Hackmann, 2021). At the same time, I find a null effect on in-hospital mortality consistent with Friedrich and Hackmann (2021), who find no effect of a decline in nurse staffing on AMI in-hospital mortality, but distinct from Gruber and Kleiner (2012), who find large increases in in-hospital mortality across conditions. I posit that estimates vary across papers due to differences in the staffing shocks and quality indicators used. In my setting, the incidence of the staffing shock fell on the acute care unit therefore we should expect to observe effects on indicators that are sensitive to acute care staffing. In-hospital mortality is an unlikely candidate because mortality is far more likely to take place in the intensive care unit, where patients in critical condition are stabilized prior to being transferred to acute care.

Second, I contribute more broadly to the literature on the effects of the minimum staffing mandate. As far as I am aware, I provide novel evidence of several responses: the decline in capacity, increase in bed utilization rates, increase in use of younger and more recently licensed RNs, and the limited crowding out of other inputs in response to the mandate. I estimate the cost effects of the mandate to be far smaller than estimated in prior descriptive work (Terasawa, 2016).9

Notably, my identification approach represents an improvement on prior work which has shown little evidence to support research design validity. I provide up to thirteen years of pre-mandate data to allow for graphical inspection of pre-trends, utilize difference-in-differences and event study estimates, and provide several robustness checks.

Third, I contribute to a long literature on hospital production. My finding that hospitals reduced excess capacity in response to an exogenous shock to costs per staffed bed illustrates the hospital’s tradeoff between healthcare access (having a lower probability of turning patients away) and profits (having a lower cost of unused, staffed beds) as modeled in early theoretical literature (Newhouse, 1970). In models of the hospital’s capacity choice, hospitals operate with excess capacity to target a desired probability of turning patients away rather than due to inefficiency (Gaynor and Anderson, 1995).

I corroborate findings that nurse and non-nurse labor have limited substitutability in hospital production (Friedrich and Hackmann, 2021) and complement evidence that hospitals substitute between nurses of different skill levels (Acemoglu and Finkelstein, 2008). The latter finding complements and uncovers relative to prior work (Matsudaira, 2014) that heterogeneity in workforce composition is important to control for when testing for monopsony using labor quantity regulation.

Relatedly, my findings contribute to a broader literature in labor economics on the firm’s responses to labor market regulation. The mandate represents a labor quantity floor which is conceptually similar to minimum wage policies that represent labor price floors. My finding that hospitals hire lower wage nurses (lower-licensed, younger nurses) in response to the mandate is therefore related to prior empirical work that has found changes in workforce composition towards higher-skilled workers following minimum wage policies (Clemens et al., 2021, Gopalan et al., 2021). However, the extent to which my findings are generalizable to other industries is an open question given that substitution patterns across inputs depend on production technology specific to the sector.

The remainder of the paper proceeds as follows. Section 2 describes the institutional context of nursing in the hospital setting and the nurse staffing mandate. In Section 3, I discuss the data and empirical framework. I present the results in Section 4, heterogeneity results in Section 5, and robustness checks in Section 6. Section 7 concludes.