The Impact of Medicaid Expansions on Nonprofit Hospitals

1 Introduction

As part of the Obama Administration’s plan to increase health insurance coverage, the 2010 Affordable Care Act (ACA) requires states to either set up a state insurance program or expand their current Medicaid programs to residents with incomes below 138% federal poverty line. Medicaid is co-financed by both states and federal governments through Federal matching funds and state Medicaid funds. The Federal government covers all Medicaid costs for those newly eligible insurers from 2014 to 2016 for states that meet the new ACA requirements. As of 2020, more than 35 states have expanded their Medicaid coverage, and millions of low-income people have gained health insurance coverage from Medicaid.

Medicaid expansions have changed the healthcare market by raising the insurance rate (Courtemanche et al. 2017, Nasseh and Vujicic 2017), increasing hospital access (Finkelstein et al. 2012; Goold et al. 2018), and reducing hospital uncompensated care costs (Blavin 2016; Kanter et al. 2020). As a result, hospitals in the expansion states have higher revenues, lower chances of closure, and better financial performance (Kaufman et al. 2016; Lindrooth et al. 2018). For nonprofit hospitals, the impact of the expansion might be different. The increase in government support in the healthcare industry through Medicaid might crowd out external contributions to nonprofits (Brooks 2000; Warr 1982; Weisbrod and Dominguez 1986).

This article examines the effect of Medicaid expansions on nonprofit hospitals in terms of presence and revenues and explores whether the effect varies by nonprofit hospitals’ type. Following the crowding out theory, we examine the impact of Medicaid expansions on contributions to nonprofit hospitals. A caveat is due here. These contributions might constitute a small percentage of the incoming revenues but remain critical for many nonprofit hospitals in covering their charitable care costs, community service expenses, and other special costs such as research and development. We also study how Medicaid expansions affect the excess margins of nonprofit hospitals, which is an indicator of hospital profitability.

We contribute to the literature in several ways. First, the results suggest that the number of nonprofit hospitals does not change after a state expanded its Medicaid; it should be clear here though that the healthcare system and structure could change in other ways. Second, Medicaid expansions have unexpected consequences on nonprofit general hospitals by significantly reducing revenues coming from contributions, potentially limiting their functionality. Third, this research also suggests potential implications on the substitutional relationships between government and nonprofits. Lastly, using a difference-in-differences identification strategy, we arguably provide causal estimates on the impact of government supports on contributions to a nonprofit sector.

The rest of the article is organized as follows. Sections 2 and 3 review the existing literature on Medicaid expansion and explain the conceptual framework for why Medicaid expansions incur significant changes in the healthcare market and affect nonprofit hospitals differently by its type. Sections 4 and 5 describe the data and empirical model followed by empirical results in Section 6. Finally, Section 7 concludes with discussions of the results and policy implications.

2 Nonprofit Hospitals: Role and Resources

3 Policy Change: the Case of Medicaid Expansion

Medicaid expansions influence hospitals in two potential ways. First, the expansions reduce hospital closures, help rural hospitals (Kaufman et al. 2016) by improving their financial health (Dranove, Garthwaite, and Ody 2017), and decrease the burden of providing uncompensated care by urban hospitals (Kanter et al. 2020). Before Medicaid expansions, hospitals were suffering from a substantial loss due to uncompensated care costs; the estimated loss is around 35 billion dollars in 2008 (Hadley et al. 2008). Nikpay, Buchmueller, and Levy (2015) find that uncompensated care costs were reduced by 15% after Connecticut expanded Medicaid. In another national-level study, Blavin (2016) estimates that, on average, Medicaid expansions reduced individual hospital’s uncompensated care costs by 2.8 million dollars.

Second, Medicaid expansions increase access and utilization of healthcare services. Research almost unanimously finds that Medicaid expansions were associated with increases in medical care accesses (Finkelstein et al. 2012; Gray, Sung, and Richardson 2015; Han, Luo, and Ku 2017). For example, Medicaid expansions increased the likelihood of a doctor visit by 6.6% points and increased the likelihood of having overnight hospital stays by 2.4% points (Wherry and Miller 2016). Similarly, Garthwaite et al. (2017) document a 125.7% increase in Medicaid visits after 12 months of Medicaid expansions among 126 investor-owned acute care facilities.

Previous studies show that the profitability of hospitals is determined by many factors. Hospitals with high profitability usually serve fewer patients that are uninsured or insured under public insurance (Bai and Anderson 2016; Ly and Cutler 2018; Moon and Shugan 2020), provide more technologically intensive services (Chandra and Skinner 2012), and face less competition from other hospitals (Moon and Shugan 2020). In addition, these hospitals tend to be larger in size and are more likely to be for-profit hospitals and hospitals with system affiliations (Bai and Anderson 2016). Specifically, Medicaid expansions reduce uncompensated care costs by decreasing the number of uninsured patients. Therefore, Medicaid expansion could increase hospitals’ profitability by increasing revenue and offsetting expenses. Blavin (2016) finds hospitals experience a 3.2 million dollar increase in Medicaid revenue and a 1.1% increase in excess margins. The increases in patients covered by public insurance could also lead to disparities in the distribution of hospital profitability across different types of hospitals. For example, while uncompensated care costs decreased and healthcare utilization increased after the Massachusetts health reform, safety-net hospitals, mainly nonprofit hospitals, actually saw a decline in operating margins because of the low reimbursement rates of Medicaid (Bazzoli and Clement 2014; Leighton et al. 2011; Mohan et al. 2013).

Hospitals have to adapt to the new market with extended insurance coverage, increased service utilization, and reduced uncompensated care costs after Medicaid expansions (Kanter et al. 2020). For-profit hospitals could adapt to the new environment by improving efficiency and providing more profitable healthcare services, incentivized by their goals to generate profits and cut down costs (Sloan 2000). Nonprofit hospitals, on the other hand, might be less responsive to the new changes in the healthcare market due to their commitment to charitable care missions, emphasis on service quality, and funding constraints (Sloan 2000). It is also important to note that Medicaid does not fund nonprofit hospitals directly; Medicaid reimburses nonprofit hospitals for the costs of treating Medicaid patients. Focusing only on nonprofit community health centers, Lam and Grasse (2019) find Medicaid expansions lead to higher excess margins among nonprofit community health centers; however, societal factors such as poverty rate has a large impact on nonprofit hospitals’ financial health. Kanter et al. (2020) find that although Medicaid expansion was associated with some financial relief, hospitals did not use that relief to provide additional community benefits.

Scholars have often used population ecology theory to study how organizations adapt in or to a changing environment (Davis et al. 2003; Fernandez 2008; Frumkin and Andre-Clark 2000; Hager, Galaskiewicz, and Larson 2004; Helmig, Ingerfurth, and Pinz 2014; Sorenson et al. 2006; Twombly 2003). The major tenant of population ecology theory is that organizations occupy certain niches, determined by the domain of their service or targeted customer or geographical outreach (Hannan and Freeman 1977); as such, they might depend on identical resources (financial or non-financial). Any changes in the surrounding environment would impact the availability of these resources and then would drive these organizations to compete. Some organizations may adapt to the environment and new types of organizations may emerge due to a selection process. Organizations that fail to adapt or are with less fit characteristics will eventually be eliminated or die out (Hannan and Freeman 1977). Some organizations are able to develop their niches by exploring new opportunities that are previously overlooked or unexplored by other organizations (Amburgey and Rao 1996). Therefore, organizational characteristics are key in this ‘survival of the fittest.’

Based on population ecology theory, we can differentiate between two different types of organizations: generalist and specialist organizations (Hannan and Freeman 1977). Generalist organizations rely on a wide range of environmental resources and cover wider or multiple domains; these organizations promise to maximize exploration but simultaneously increase risks and incur additional costs to cover the many domains. In contrast, specialist organizations operate in a single domain and rely on limited but specific environmental conditions and resources. Consequently, their ability for exploration is reduced or constrained, but they are more secure, especially since the cost they have to pay to operate in a single domain is limited, and they tend to cater their services to a targeted population. Therefore, Hannan and Freeman (1977) predict that specialists will dominate generalists in stable environments; generalists will succeed in turbulent ones. Generalists can attract customers, respond to competition, and achieve economies of scale (Lancaster 1998). With limited information and heterogeneous consumers’ preferences, a generalist will be better prepared to offer diverse services, meet demands, and survive.

In this article, a general hospital^[1] is considered a generalist organization that offers a variety of services and serves multiple local communities. A specialty hospital is a specialist organization offering either specific services or particular services to a specific patient group. In terms of a community health system, it provides a variety of health services but tends to be more localized. Potentially, a community health system is a hybrid, combining characteristics of both generalist and specialist organizations. The differences between general hospitals, specialty hospitals, and community health systems are further differentiated by their management structure. Community health systems are multihospital healthcare systems managed or coordinated by a central organization. Coordination among two or more hospitals helps community health systems to adapt faster to environmental changes. The system structure also helps community health systems to buffer the environmental changes. Specialty hospitals target specific types of patients. Therefore, they expose less to the change in the policy environment as well as facing less competition. Unlike community health systems and specialty hospitals, the lack of buffering mechanisms and more intense competition from other hospitals make general hospitals more vulnerable to changes in the healthcare markets. This research examines the heterogeneous effects of Medicaid expansions on different types of hospitals. The focus on these different types, which serve a larger population and provide more comprehensive healthcare services, clearly differentiate this study from Lam and Grasse’s (2019) that relies on nonprofit community health centers.

4 Data

The main nonprofit hospital data used for this research are IRS Business Master File (BMF) extracts and IRS 990 form data. We identified nonprofit hospitals using three specific National Tax Example Entity (NTEE) codes: E21 for community health systems, and E22 and E24 for general hospitals and specialty hospitals. We then calculated the number of nonprofit community health systems, general hospitals, and specialty hospitals between 2003 and 2016 by aggregating hospital-level BMF data at the state-level for each month.^[2] The dataset contains 3053 state-month observations for each type of nonprofit hospital.

We verified the sampled hospitals in two ways. First, we screened the sampled nonprofits using the NAICS code, which is a standard code to differentiate between different industries. Nonprofits that are under the NTEE hospital categories but do not provide health care or are medical hospitals were dropped. Second, we manually screened nonprofits, whose names do not contain words such as “Hospital,” “Health,” “Healthcare,” and “Clinic” to verify the organizational type using the GuideStar website. After dropping organizations that have moved across states and have missing or improper values, the final sample consists of 1765 nonprofit hospitals, which include 306 community health systems, 1354 general hospitals, and 105 specialty hospitals.^[3]

The contributions and financial data primarily come from IRS Form 990 during 2003–2015 fiscal years compiled by National Center for Charitable Statistics (NCCS).^[4] Religious organizations and nonprofits with gross revenues less than 25,000 dollars are not required to fill form 990; therefore, the sample only includes hospitals above that threshold.

State characteristics, such as state average income, poverty rates, local gross product, and population, are compiled from the University of Kentucky Center for Poverty Research (2017) and State Expenditure Report (2003–2015). Following previous practice, nonprofit hospitals in seven states (DE, MA, NY, VT, CA, WA, and MN) and DC are excluded from the sample because these states either have adopted Medicaid expansions before 2014 or have already provided similar coverage to residents with incomes up to 100% of the federal poverty line or greater during 2010–2013 (Lindrooth et al. 2018; Simon, Soni, and Cawley 2017). The final sample consists of 43 states.

5 Statistical Model

Levering the different timings of the Medicaid expansions, we use a difference-in-differences (DD) strategy to compare nonprofit hospital populations, contributions, and excess margins before and after Medicaid expansions (first difference) between states with and without Medicaid expansions (second difference). Specifically, by assuming the outcomes in the expansion and non-expansion states evolve parallelly in the absence of the Medicaid expansions, the DD strategy could guard against bias from unobserved confounding factors.

To examine the impact of Medicaid expansions on the number of nonprofit hospitals, we specify the DD model as below:

NH _st is the outcome variable, which is the number of nonprofit hospitals in time t for state s. We also conduct separate estimations for nonprofit community health systems, general hospitals, and specialty hospitals. Expansion_st is a dummy indicator that equals one if state s has adopted Medicaid expansion at time t. X is a set of the state-level controls mentioned in the last section. δ a is the DD estimator that identifies the causal impact of Medicaid expansions on states’ nonprofit hospital population and is the coefficient of interest. α _s and α _t are the time and state fixed effects to account for annual fluctuations and across state differences in the outcome variables. Given that the hospital population is a count variable, we estimate Eq. (1) with Poisson regressions as the preferred estimates, in which the outcome is the number of nonprofit hospitals and the coefficient for the total population is restrained to 1 for each analysis. This allows an interpretation of Poisson estimates as the number of hospitals per million population.

For hospital-level analysis of contributions and excess margins, we employed a similar DD specification as below:

where Y _it is the outcome variable for nonprofit hospital i at time t, which includes the IHS transformed contributions and the excess margin for each sampled nonprofit hospital. Expansion_it is a hospital-level treatment dummy for whether the hospital is located in a state with Medicaid expansion at time t. X is a set of state-level characteristics mentioned above and previous years’ fundraising expenditures at the hospital-level. We also include year and month fixed effects ( θ t ) to account for seasonal and nonlinear time patterns. The hospital fixed effects ( θ i ) enable us to control for hospital time-invariant unobserved variables.

6 Results

Table 2 presents the estimated impact of Medicaid expansions on nonprofit hospital populations using BMF data. Columns 1 to 3 report estimates using OLS, while column 4 reports Poisson estimators. However, the estimates in Table 2 are too imprecise to suggest any changes in nonprofit hospital populations for all types of nonprofit hospitals after Medicaid expansions.

Table 3 presents the estimates for contributions using Eq. (2). Specifically, results in column 1 are estimated without the hospital and state-level controls. We add both hospital- and state-level controls in the models to avoid biases due to time-varying observed factors; the results are presented in column 2. To release the assumption that there are no time-varying unobserved variables, we also add hospital specific time trends among models in column 3. Finally, we present results from the preferred models using PPML in column 4.

Panel A reports DD estimates for all hospitals. Although all coefficients are negative, only the coefficient in column 3 is statistically significant at the 10% level, which indicates Medicaid expansion had little or no impact on the contributions for nonprofit hospitals overall. For community health systems, coefficients in panel B suggest Medicaid expansions positively affect contributions received by nonprofit community health systems, but the estimates are not consistent across models, which limits the ability to make inferences based on the estimates. For general hospitals, all coefficients, except for the DD estimates without hospital- and state-level controls, are consistent and statistically significant at the 10% level. Notably, Medicaid expansions reduced contributions to general hospitals by around 20–32% (i.e.,  e ( − 0.22 ) − 1 ≈ − 0.20 e ( − 0.38 ) − 1 ≈ − 0.32 ). Before Medicaid expansions, these nonprofit general hospitals averaged a recipient of 2.62 million dollars contributions. A 20% reduction, according to the PPML estimate in column 4, implies a drop of contributions by around 0.60 million for nonprofit general hospitals. We can also interpret the results in terms of the percentage decreases in hospital revenue. On average, contribution accounted for about 2.1% of nonprofit general hospitals’ revenue. Based on this number and the results in Table 3, we find that Medicaid expansions reduced nonprofit general hospitals’ revenue by 0.48%. Although the decrease in revenue in those hospitals is small, as mentioned before, it could affect their provision of charitable care and research and development projects. Finally, we also find specialty hospitals’ contributions were little affected by Medicaid expansion based on results in panel D.

Finally, Table 4 reports the OLS estimates for excess margins. Panel A shows the results for all hospitals. Panels B through D report estimates for each type of nonprofit hospital separately. However, all the coefficients are small and statistically insignificant, which implies Medicaid expansions had little or no impact on hospital excess margins. Even for nonprofit general hospitals, the excess margins were not affected because the reduction in the contributions after Medicaid expansions only accounts for 0.48% of total revenue.

A two-way fixed effect difference-in-differences model can be biased when the treatment is assigned at different times (Goodman-Bacon 2018). Therefore, we also estimate the event study version of the DD models by adding a series of lead and lag dummies for the Medicaid expansion dummies in Eqs. (1) and (2). These dummies will capture the differences in the outcome variables before and after the Medicaid expansions after adjusting for controlled time-variant variables and any unobserved time-invariant variables. This approach enables us to track changes in hospital populations, contributions, and hospital excess margins over time. It could also confirm the assumption that the parallel trends assumption is not violated (Goodman-Bacon and Schmidt 2019; Lafortune, Rothstein, and Schanzenbach 2018). In the event study analysis, we bin the observations within each tax year and centralize the time at the year before the implementations of Medicaid expansions. The coefficients and the 95% confidence interval for the lead and lag dummies using event study graphs are also reported.

Figure 1 reports event study results for nonprofit hospital populations using IRS BMF data. The figure confirms the no parallel trends assumption because all coefficients on the lead dummies, except for the two lead dummies for nonprofit general hospitals, are not statistically distinguishable from zero at the 5% level.

Figure 1:

Differences in the number of nonprofit hospitals between the expansion states and the non-expansion states.

Notes: Expansion states are AR, AZ, CO, CT, DE, HI, IA, IL, KY, MA, MD, MI, ND, NH, NM, NV, NY, OH, OR, RI, VT, and WV. Observations are centralized at the expansion year. The dots and vertical lines present the point estimates and the 95 confidence intervals, respectively.

Sources: Authors’ calculations based on NCCS Data.

Figure 2 presents the event study graphs for hospital contributions by hospital type. We do not find any pre-existing trends before the year of Medicaid expansions. Overall, contributions for nonprofit hospitals decreased by around 30% after Medicaid expansions as indicated by panel A of Figure 2, which is similar to the coefficient in Table 3.

Figure 2:

Differences in contributions between the expansion states and the non-expansion states.

Notes: Expansion states are AR, AZ, CO, CT, DE, HI, IA, IL, KY, MA, MD, MI, ND, NH, NM, NV, NY, OH, OR, RI, VT, and WV. Observations are centralized at the expansion year. The dots and vertical lines present the point estimates and the 95 confidence intervals, respectively.

Sources: Authors’ calculations based on NCCS Data.

Finally, we present the event study figures for excess margins in Figure 3. Again, the event graphs imply there are no differences in the excess margins between nonprofit hospitals in the expansion and non-expansion states both before and after Medicaid expansions.

Figure 3:

Differences in the excess margin between the expansion states and the non-expansion states.

Notes: Expansion states are AR, AZ, CO, CT, DE, HI, IA, IL, KY, MA, MD, MI, ND, NH, NM, NV, NY, OH, OR, RI, VT, and WV. Observations are centralized at the expansion year. The dots and vertical lines present the point estimates and the 95 confidence intervals, respectively.

Sources: Authors’ calculations based on NCCS Data.

To avoid the concern that the negative impact of Medicaid expansions on nonprofit general hospitals’ contribution is found by chance, we conduct a permutation test using Eq. (2). First, we randomly assign Medicaid expansions status to 43 states in the sample. Second, we estimate the placebo effect of Medicaid expansion on nonprofit general hospitals using the assigned expansion status with Eq. (2). We also store the results. Third, we redo steps one and two for 10,000 times. Lastly, we calculate the randomization inference p-value and plot the frequency and density of coefficients for placebo tests following Pfeifer, Reutter, and Strohmaier (2020) in Figure 4. The solid vertical line indicates the effects found in the original data, and the vertical dashed line indicates the 5th and 95th percentile of the 10,000 coefficients from the placebo samplings. As shown in the graph, the effects lie on the left tail of the distribution and the randomization p-value for the main results is 0.03, which suggests they are unlikely to be coincidental.

Figure 4:

Permutation tests for the impact of Medicaid expansions on general nonprofit hospitals.

Notes: These are FE regressions using the IRS 990 Data. The model is PPML regression with 10,000 randomized permutations for creating the sampling distribution. Control variables not presented in the table include total profession fundraising fees in the past year, state population, state poverty rates, GRP, state average income, and state unemployment rates. The panel presents 5th and 95th percentile confidence intervals from permutations tests and p-values from a one-tailed test. *p < 0.10, **p < 0.05, ***p < 0.01.

7 Discussion and Conclusion

This article examines the impact of Medicaid expansions on nonprofit hospitals. Using a difference-in-differences strategy, the results indicate Medicaid expansions have little or no impact on the number of nonprofit hospitals. The hospital-level analysis implies a negative relationship between Medicaid expansions and nonprofit general hospitals’ contributions; in other words, the nonprofit general hospitals’ revenue from contribution in the expansion states see an annual decline. However, the decline only accounted for a small proportion of the nonprofit hospital revenues since contributions add up to around 2% of total revenues. While such a decline might not be a critical threat to the operation of these hospitals, it could potentially impact certain activities, mainly charitable/community outreach and care provision.

Moreover, starting from 2021, the federal government will reduce the DSH payment by more than $4 billion. In the long-term, the decrease in federal funding may exacerbate nonprofits’ financial situations. Our analysis focuses on the changes in hospitals’ finance prior to changes in Federal DSH payment and finds that nonprofit hospitals neither see an increase in contributions nor an increase in financial margins. Based on these results, it is reasonable to expect that the reduction in DSH payment could negatively affect the financial stability and the operation of nonprofit hospitals, especially those serving more Medicaid patients. Even if private donations increase after the reduction of DSH payment, donations only make up a small proportion of total hospital revenue and may not be able to make up the loss. We might also expect the decline to be a bit steep in the near future and the impact more noticeable.

The results continue to draw attention to previous research which suggests a crowding out effect of government funding on private donations (Andreoni and Payne 2011; Duncan 1999; Heutel 2014; Luksetich 2008; Warr 1982). The public usually donates to a nonprofit as an expression of belief in its mission and to support its work; these donations are much needed when the organization is under financial strain. The increased reliance on government funding might send a signal to the public that the nonprofit has secured a stream of revenue that potentially compensates for private donations; a private donor might then channel their support to another organization that lacks enough resources. At the same time, we should not dismiss the possibility that private donations are sought as much as they are granted. Attracting donations requires an organization to invest resources, time, and efforts; an organization with limited government funding would likely be more willing to make such an investment in comparison to another that has enough funding. It is possible for the increase in government support – through Medicaid reimbursement – to therefore crowd out private donations and other funding opportunities for nonprofit hospitals in the expansion states. We cannot confirm this possibility here due to data limitations and leave it for future examination.

Contrary to Lam and Grasse’s (2019) finding that Medicaid expansions increase excess margins for nonprofit community health centers, we do not find Medicaid expansions have a statistically significant impact on nonprofit hospitals. One plausible explanation that requires further exploration is that nonprofit hospitals are different from nonprofit community health centers in terms of clients, management structure, and are operated under different strategies.

Here, it might be worth considering whether Medicaid expansion reimbursements could also impact the level of volunteering at nonprofit hospitals. Individuals tend to donate their money or time to nonprofits (AbouAssi and An 2017); if they donate less money to nonprofit hospitals in the expansion states due to the increase in government funding, the question is whether they are instead willing to volunteer their time, especially with the increase in the demand for volunteers due to the expansion of medical services at these hospitals. Such a question needs further exploration, focusing on the demand side of volunteering. On the supply side, Sohn and Timmermans (2017) have already established readiness to volunteer. Medicaid expansions were associated with a 38% increase in voluntary activity participation, which is due to improved financial security and a sense of health-related to Medicaid expansions (Sohn and Timmermans 2017).

The results also indicate that Medicaid expansions may not have the same impact on different types of hospitals, underscoring the premise of the population ecology theory. The discussion resonates with Porter’s (1980, 1985 theory of generic competitive strategy; an organization performs based on its competitive scope: narrow or broad. Generalists tend to focus on social performance, while specialists tend to enhance economic performance (Porter 1980, 1985). For example, Guo and Brown (2006) contend that specialist foundations (e.g., city or county foundations), which serve smaller geographic segments, outperform generalist foundations (e.g., regional or national foundations) in terms of fiscal efficiency (i.e., how efficient a foundation is in acquiring and managing financial resources). Such efficiency is one of the causes of dissolution among 41 nonprofit associations in Spain (Fernandez 2008).

Hospitals are different in their specialization and organizational structure. For example, a specialist hospital focuses on a specific target group and provides specialized and tailored services, while a general hospital usually provides multiple services to a larger base of customers with diverse demands. These generalist organizations are typically large in size and capacity (AbouAssi, Makhlouf, and Tran 2019; Archibald 2007). While they are equipped to deal with any change, an increase in demands from heterogeneous groups that they serve places them at a disadvantaged position to adjust to policy environment changes. Besides, small and specialized organizations could adjust to these changes more quickly. Lam and Grasse (2019) note that nonprofit community health centers are usually smaller than hospitals and focus on local community residents. The unique features allow nonprofit health centers to adjust to uncertainties and raise their profit margins noticeably after Medicaid expansions. Kanter et al. (2020) further argue that although these hospitals are expected to sufficiently contribute to the communities they serve when their financial burden is eased, their direct community spending rather decreased.

Regardless of how Medicaid expansions affect the different types of hospitals, these hospitals are nonprofit entities; their financial status and dependence on government support have implications on their autonomy, legitimacy, and participation in policy advocacy (Jung and Moon 2007). As Frumkin and Andre-Clark (2000) recommend, the focus of nonprofits is not just on enhancing efficiency but, more importantly, on developing the distinctive niches – based on their unique values – to compete against for-profits in the increasingly competitive government-contracting market (Haslam, Nesbit, and Christensen 2019). One way that hospitals can better position themselves to compete—or to survive—is through acquisitions and mergers; an increase in hospital acquisitions was reported after the enactment of the ACA (Young 2017). Such a trend reshapes the landscape possibly resulting in more hospitals under health systems. This trend should be closely examined in future research.

Situating the results back in the policy domain, Mosley (2010) contends that while many human service nonprofits do not engage in policy advocacy, among those that do are organizations with the ability to secure high levels of government funding; that is because they “may pay more attention to proposed policy changes or have better contact with administrators” (Mosley 2010, p. 71). It is possible then for nonprofit hospitals that receive more Medicaid expansions reimbursements to get more involved in advocacy work. However, the effects of Medicaid expansions are not uniform across the different types of hospitals; we should then expect some variations in the degree of involvement in advocating for and shaping public policies among these entities.

In addition, existing research indicates a variation in advocacy involvement based on the organizational niche; for example, charitable nonprofits serving minorities – a narrower customer base – are more likely to engage in policy advocacy than their generalist counterparts (MacIndoe 2014, MacIndoe and Whalen 2013). This means advocacy involvement could then depend on the type of hospital. The concern here is that end goals of each of the three types of nonprofit hospitals are not necessarily aligned; it is possible to witness a certain level of competition in this subfield of the nonprofit sector. MacIndoe and Whalen (2013) caution that such competition could pose a serious risk to specialist nonprofits; specialists are expected to falter in a turbulent resource environment which is where policy advocacy typically takes place.

Several limitations are important to acknowledge. First, the IRS’s 990 and BMF data has its limitations. As suggested by Grønbjerg (2002), some nonprofit organizations may not register under the same category every year. In the analysis of financial data, we dropped those organizations that switch their NTEE codes and states, which might bias the estimations. In addition, we use a difference-in-differences strategy, which enables controlling for time-invariant heterogeneities. However, the identification will fail if there are time-variant unobserved variables, such as the shift of political attitudes among the public or the changes in organizational structures and personnel with each organization. Related to that, the results provide only short-run effects of Medicaid expansion on hospital numbers; the data limitation confine the ability to estimate the long-run trends. Lastly, we identify each organization by the employer identification numbers (EIN); therefore, the analysis would not capture the trends in hospital acquisitions; this is because, unlike hospital mergers, hospital acquisitions do not necessarily lead to changes in hospitals’ EINs.

Nevertheless, this analysis provides sufficient interpretations of the impact of Medicaid expansions on nonprofit hospitals’ financial status. It accounts for endogeneity in the providing of government support by exploring the different timing of the implementations of Medicaid expansions using a difference-in-differences design. The article also further expands the understanding of the functions of different types of nonprofit hospitals at the state-level and opens the door for future research.