Journal of Health Economics 22 (2003) 1–22
Competitive spillovers across non-profit and for-profit nursing homes David C. Grabowski a,∗ , Richard A. Hirth b a b
Department of Health Care Organization and Policy, University of Alabama at Birmingham, 330 RPHB 1655 University Boulevard, Birmingham, AL 35294, USA Department of Health Management and Policy, University of Michigan, Ann Arbor, MI, USA Received 1 February 2002; received in revised form 1 July 2002; accepted 21 August 2002
Abstract The importance of non-profit institutions in the health care sector has generated a vast empirical literature examining quality differences between non-profit and for-profit nursing homes. Recent theoretical work has emphasized that much of this empirical literature is flawed in that previous studies rely solely on dummy variables to capture the effects of ownership rather than accounting for the share of non-profit nursing homes in the market. This analysis considers whether competitive spillovers from non-profits lead to higher quality in for-profit nursing homes. Using instrumental variables to account for the potential endogeneity of non-profit market share, this study finds that an increase in non-profit market share improves for-profit and overall nursing home quality. These findings are consistent with the hypothesis that non-profits serve as a quality signal for uninformed nursing home consumers. © 2002 Elsevier Science B.V. All rights reserved. JEL classification: I11; L15; L31 Keywords: Consumer information; Non-profit nursing homes; Quality; Instrumental variables
1. Introduction Kenneth Arrow (1963) was the first to hypothesize that non-profit organizations exist in health care markets to provide quality assurance to poorly informed consumers. The quality of care provided by the US nursing home industry has received a great deal of attention over the last three decades (e.g. US Senate, 1974; Institute of Medicine, 1986; US General Accounting Office, 1998; Institute of Medicine, 2001). A common theme within ∗ Corresponding author. Tel.: +1-205-975-8967; fax: +1-205-934-3347. E-mail address:
[email protected] (D.C. Grabowski).
0167-6296/03/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 6 2 9 6 ( 0 2 ) 0 0 0 9 3 - 0
2
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
these reports is that low quality may result from opportunistic behavior by for-profit nursing homes, which account for over two-thirds of the nearly 17,000 nursing homes in the US. If non-profit homes, often affiliated with religious or charitable organizations are less willing to compromise care for the sake of profit then ownership status may provide a low cost signal that the promised quality (i.e. quality commensurate with the price being charged) will be delivered. Given the favorable tax treatment that non-profit nursing homes receive, there has been a great deal of interest in evaluating whether the non-profit sector serves a socially useful function. Unfortunately, most of the existing empirical work is flawed, because it treats each ownership type as if it existed in isolation without controlling for the ownership of competing firms in the marketplace. Recent theoretical work has emphasized that finding (or not finding) quality differences between the two ownership types does not constitute verification of the benefits (or lack thereof) associated with the non-profit sector (Hirth, 1999). The existence of differences across for-profit and non-profit homes is neither necessary nor sufficient to conclude that non-profit enterprise is socially desirable. Omitting a measure of non-profit market share, which could capture a quality spillover effect, may bias the coefficient on the ownership variables and yield misleading policy implications. Using 1995–1996 data for nearly all US nursing homes, this study offers empirical analyses that include a measure of non-profit market share to analyze the role of the non-profit sector in the provision of quality.
2. Conceptual framework Economic theory implies that for-profit firms will produce the socially efficient product array when consumers can easily evaluate products before purchase, contract over delivery terms, monitor contractual compliance, and obtain redress for violations. When these conditions are not met, however, producers may have some discretion to misrepresent quality (Hansmann, 1980). There are reasons to believe that this may in fact be the case in the nursing home industry. Although care is fairly non-technical in nature, monitoring of care can often be difficult, and the learning period may be non-trivial relative to the length-of-stay in some instances.1 The patient is often neither the decision-maker nor able to easily evaluate quality or communicate concerns to family members and staff. Furthermore, the elderly who seek nursing home care are disproportionately the ones with no informal family support to help them with the decision process (Norton, 2000). Finally, there are relatively few nursing home-to-nursing home transfers. Using 1994–1996 minimum data set assessments, Hirth et al. (2000) document a transfer rate of 4.4% in New York and 8.0% in Maine during the first 1 Some forms of nursing home quality may be easily observable by prospective patients (or their agents), such as size of the room, cleanliness of the facility, and the number of staff. Other dimensions of quality are frequently more difficult for patients to ascertain, for example, the quality of nursing home staff. This type of information may take considerable time to learn. For example, problems such as bedsores or infections may take weeks (or even months) to develop. Even after problems develop, residents or their proxies must still draw difficult inferences about whether the problems were truly attributable to inadequate care and whether alternative care arrangements would have prevented them. Garber and MaCurdy (1992) document an average length-of-stay for chronic care (i.e. non-Medicare) nursing home patients of approximately 125 days with substantial variation around this mean. Thus, the learning period may be relatively long compared to the typical length-of-stay for many nursing home patients.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
3
6 months post-admission, with substantially lower rates later in a stay. Movement among homes may be impeded by tight markets due to supply constraints such as certificate-of-need (CON) and construction moratorium laws and health concerns regarding relocation (termed “transfer trauma” or “transplantation shock”). The following framework considers both a model in which some nursing home consumers are poorly informed regarding quality and a model in which all consumers are well informed. The broad implications of this theoretical framework are reviewed here, and the reader is referred elsewhere for a formal mathematical treatment of these issues (see Hirth, 1999). 2.1. Asymmetric information model We first consider a model of the nursing home market in which some consumers lack information on quality (i.e. the asymmetric information case). In this model, for-profit homes are assumed to maximize profit and will provide less than the promised level of quality if the presence of uninformed consumers makes that strategy profitable. We assume a strictly enforced non-distribution constraint, which prohibits payment of profits to owners or employees, on the non-profit sector.2 The non-distribution constraint motivates honest behavior by non-profits, ensuring that they deliver the promised level of quality and do not simply act as ‘for-profits in disguise’. Because the for-profit sector does not deliver the first best outcome unless the fraction of poorly informed consumers is low, there exists an opportunity for non-profit status to serve as a signal of quality by attracting those poorly informed consumers into non-profit homes (Arrow, 1963; Hansmann, 1980). Because non-profits disproportionately attract those uninformed consumers, consumers remaining in the for-profit sector are better informed than a random draw from the patient population. Thus, the larger the non-profit market share, the higher the likelihood that for-profits deliver the promised quality. This can be thought of as an ‘Inverse Gresham’s Law’ under which the good (non-profits delivering the promised quality) drive out the bad (those for-profits attempting to exploit a poorly informed clientele by delivering less than the promised level of quality). Effectively, the non-profit sector exerts a beneficial, competitive spillover effect on the performance of the for-profit sector. As a result, even if non-profit and for-profit homes are observed to have similar quality, eliminating the non-profit sector could have deleterious welfare consequences by changing the prevailing market equilibrium. Thus, this model generates two predictions. First, an increase in non-profit market share will improve quality in for-profit facilities. And second, by improving quality in the for-profit sector, an increase in non-profit market share will also improve overall market quality. 2.2. Full Information model We next consider an alternative model with well-informed consumers (i.e. the full information case) where non-profit status is no longer a necessary signal of quality. In this case, non-profit homes are still thought to stake out the high quality end of the market due to managerial preferences (for example, managers of non-profits may value their community image 2 As the non-distribution constraint becomes less binding, non-profit status becomes an increasingly imperfect signal of quality (Hirth, 1999).
4
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
as high quality caregivers).3 Effectively, non-profits crowd high quality for-profits out of the market, ceding the low price/low quality portion of the market to their for-profit competitors. In this environment, quality differences between for-profit and non-profit homes may be substantial, but prices reflect these differences and the distribution of ownership types at the market level is irrelevant to consumer well being. As a result, the quality of care in for-profit homes will decrease as the share of non-profit homes increase. However, the level of market quality depends only on the willingness to pay of consumers and public payers and will be unaffected by the share of non-profit homes. An empirical model, which only accounts for the effect of ownership type and does not include price measures (which are typically unobserved in secondary data sets used for nursing home research), would mistakenly attribute this observed quality difference across sectors to the desirability of the non-profit sector. Thus, this model generates two predictions. First, an increase in non-profit market share will decrease quality in for-profit facilities. And second, non-profit market share will have no effect on overall market quality.
3. Previous literature The majority of previous empirical studies of the relationship between ownership and quality measure the effect of profit status by using a dummy variable for type of ownership but omit a measure of the relative prevalence of for-profit and non-profit firms. This omission can bias inferences about the effects of ownership. If beneficial spillovers occur (due to poorly informed consumers), the coefficient on an ownership variable will be biased towards zero, because the performance of for-profits and non-profits will tend to converge in areas with high non-profit shares. Conversely, if provision of high quality care by non-profits crowds out the provision by for-profits (as would occur in the case of full consumer information), the coefficient will likely be biased away from zero. Not surprisingly, previous studies that measure the effect of ownership type on quality have yielded little in the way of consistent findings. Three different reviews of this literature have drawn three different conclusions regarding the relationship between ownership type and quality. O’Brien et al. (1983) suggests that quality is identical across non-profit and for-profit facilities. Hawes and Phillips (1986) conclude that studies generally find quality superior in non-profit facilities. Finally, Davis (1991) argues that the evidence is inconclusive in regards to whether non-profit facilities provide higher quality. More recent work in the literature has continued to find mixed evidence on this issue (Holtmann and Ullman, 1991; Gertler, 1992; Cohen and Spector, 1996; Harrington et al., 2001). There has been only limited empirical work incorporating a measure of non-profit market share into the analysis. Using data from the 1985 National Nursing Home Survey (NNHS), 3 There is some evidence that non-profit nursing homes are more likely to stake out this high price/high quality end of the market. First, data from the 1985 National Nursing Home Survey (NNHS) show that, on average, the private-pay price at a non-profit nursing home is greater than the private-pay price at a for-profit facility (Hirth, 1993). This trend persists across the distribution with non-profits dominating the highest price categories. Second, the data analyzed within this study indicate that non-profits have a higher proportion of private-pay residents relative to their for-profit counterparts. On average, the typical non-profit facility has 31% private-pay residents while a for-profit home has 21% private-pay residents.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
5
Hirth (1993) estimated the effect of ownership on measures of staffing skill mix (i.e. registered nurse hours divided by total hours) and found that non-profit ownership was associated with higher quality. He also found evidence favoring the asymmetric information model in that increased non-profit market share improved both overall and for-profit quality. Using the 1987 National Medical Expenditure Survey (NMES), Spector et al. (1998) estimated the effect of non-profit status on various measures of quality including mortality, infections, bedsores, hospitalizations and functional disabilities and found that non-profit homes provide a higher quality of care than for-profit homes. However, these authors did not find that increased non-profit market share would raise either overall or for-profit quality.4 Thus, the results of this study favor the full information model outlined above. Although both these studies have found higher quality in the non-profit sector, it is unclear whether this higher quality reflects beneficial spillovers (the asymmetric information case) or the crowd-out by non-profits of high quality for-profits (the full information case). This paper offers a series of innovations towards resolving this issue. First, previous empirical analyses that have incorporated non-profit market share have relied on nationally representative samples of nursing facilities (i.e. the NNHS and the NMES). Although we discuss our estimation strategy in detail below, this study proposes to use data for all of the approximately 17,000 Medicaid and Medicare certified nursing homes in the US, which represent over 96% of all facilities nationwide. Furthermore, we will use recent nursing home data (1995–1996) and have access to a range of structural, procedural, and outcome measures of quality. Finally and most importantly, those previous studies that have included a non-profit market share variable have been based on limited econometric specifications. The direct evidence regarding this issue has been based on single equation models that are identified solely by the inclusion of a non-profit market share variable. Unfortunately, these single equation models are observationally equivalent with two very different interpretations. Consider the asymmetric information case. If an increase in non-profit market share improves quality in the for-profit sector, one interpretation is that this result was due to the competitive spillover effect. However, an equally plausible alternative interpretation is that the relationship between non-profit market share and quality reflects unobserved demand for quality and non-profit share at the market-level. For example, those areas with a strong preference for nursing home quality (e.g. increased consumer oversight; stricter regulatory enforcement) may be exactly those areas that adopt tax laws or regulatory procedures favoring non-profits. In order to address this potential issue, we employ an instrumental variables (IVs) estimation method to check the robustness of our single equation results.
4. Empirical specification The standard empirical approach to examining the effect of the non-profit sector on the provision of quality within the nursing home industry has been to estimate a reduced form 4 Spector et al. (1998) only show the overall market results, but from other unreported work, the authors conclude that there is little evidence to support “Hirth’s conjecture that increased non-profit market share would raise quality in the for-profit segment of the market” (p. 649).
6
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
equation that includes dummy variables measuring ownership type. This approach fails to account for non-profit market share, which would capture the spillover effects associated with the non-profit sector. This omission may bias the coefficient on the categorical ownership variable and lead to misleading policy implications. However, in order to benchmark our work to the current literature, we present an initial set of estimates using this “dummy variable” approach. The basic specification for this approach is: Y = Nβ + Xδ + ε
(1)
where Y refers to the quality measure, N a vector of ownership dummy variables, X includes an intercept and a set of exogenous controls, and ε the residual. The variable N represents the three nursing home ownership types: non-profit (28% of all facilities), for-profit (66%) and government owned (6.6%). Of the 4688 non-profit homes found in the 1995–1996 Online Survey, Certification, and Reporting (OSCAR) file, 1114 facilities were church related, 3299 were corporate owned, and 275 were classified as “other non-profit.” Of the 11,174 for-profit homes, 9603 facilities were corporate owned, 1228 were owned by a partnership and 343 were individually owned. The 1116 government facilities were owned and operated by the state (113 facilities), county (558 facilities), city (144 facilities), city/county (105 facilities), hospital district (192 facilities) and federal government (4 facilities). In this study, quality Y was represented by several outcome, process and resource-based measures.5 The outcome-oriented measure of quality was the number of residents with pressure sores (or decubitis ulcers), commonly associated with immobility in the elderly. Pressure sores are areas of the skin and underlying tissues that erode as a result of pressure or friction and/or lack of blood supply. Pressure sores can often be prevented or resolved by frequently repositioning the immobile resident. The proportion of residents with catheters, feeding tubes and physical restraints were also used as procedural measures of quality. Because labor constitutes 60–70% of nursing home costs, these procedures may be employed as labor-saving practices on the part of nursing homes with potential negative consequences for resident health (Zinn, 1993). Finally, two resource-based measures were used as proxies for quality. First, the total number of registered nurses (RNs), licensed practical nurses (LPNs) and nurses’ aides (NAs) per resident days was used to represent quality. In order to account for staffing skill mix, the proportion of registered nurses (RNs) per total nursing staff was used as a second resource-based measure. A series of exogenous variables X were included as controls in this study. The exogenous demand variables were the median income of people living in the nursing home’s county; the population of individuals over age 65 in the county; two measures of the health status (case mix) of the home’s residents; and a Herfindahl index. The first measure of case mix is based on an activities of daily living (ADL) index, which includes bathing, dressing, eating, toileting and walking. A home’s ADL score was calculated by summing 5 Given the fact that nursing homes bundle the board and care functions, overall nursing home quality would likely encompass both aspects of technical quality (e.g. staffing, pressure sore rates) and resident amenities (e.g. home-like atmosphere, organized resident activities). Unfortunately, amenities are not typically included within administrative data files such as the OSCAR system. Thus, we rely solely on technical measures of quality within this study.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
7
the number of ADLs that residents needed assistance with at the time of the survey and dividing by the total number of residents in the home.6 The result is an index of the average need for assistance of the residents in each facility. ADLs form the cornerstone of nursing home resident classification and play a major role in all state-level Medicaid case-mix adjustment payment systems. Fries (1990) argues that ADLs explain case mix for the vast majority of nursing home residents except for those heaviest care individuals. Thus, we include a second case mix variable measuring the proportion of residents requiring skilled nursing care. This variable should capture those heaviest case-mix residents within the facility. A Herfindahl index is a measure that is negatively related to the competitiveness of a market. This index was constructed by summing the squared market shares of all facilities in the county. The index ranges from 0 to 1, with higher values signifying a higher concentration of facilities. As a note, Kessler and McClellan (2000) have argued that the inclusion of a Herfindahl index within hospital quality regressions may be endogenous because hospitals with higher quality will obtain higher market shares. Historically, this issue has not been relevant for the nursing home industry where certificate-of-need laws and construction moratoria prevented expansion, but recent work has argued that these regulations may be less important towards constraining nursing home quality competition (Grabowski, 2001). As a sensitivity check, all of the results presented within this study are robust to replacing the Herfindahl index with a measure of the number of homes within the marketplace. Importantly, the county was used to approximate the market for nursing home care within this study. Most economic studies have used the county as a proxy for the nursing home market (e.g. Nyman, 1985; Cohen and Spector, 1996). As noted by Banaszak-Holl et al. (1996), the county may be a reasonable approximation of the market for nursing home care given patterns of funding and resident origin. For example, federal block grant funds for long-term care services are distributed at the county level. Furthermore, Gertler (1989) found that 75% of residents in New York facilities had previously lived in the county where the home was located. Similarly, Nyman (1994) found 80% of residents in Wisconsin facilities chose a nursing home located in the county in which they resided before entering the home. Importantly, all of the analyses presented in this study are robust to excluding those 720 counties with only one nursing home. An exogenous supply variable included within X was the CMS area hospital wage index. Binary indicators were also included for whether the home was part of a multiple-facility chain and for whether the nursing home was a hospital-based facility. A proxy measure of the facility’s age was also included and will be described in more detail below. Finally, a 6 A potential limitation of the ADL index is that lower quality within a facility may lead to greater ADL dependency. However, the nursing home case-mix literature has generally argued that need-based factors such as ADLs are sufficiently invariant to provider influence (e.g. Fries et al., 1994). The ADL most likely to be influenced by quality of care is toileting. That is, homes providing more extensive bladder and bowel retraining programs are likely to have fewer residents dependent in toileting. Thus, we re-estimated the model with a modified ADL measure that did not include toileting. The results were not sensitive to this change. There also been recent experimental evidence suggesting that nursing home patients exposed to aggressive rehabilitative programs such as weight bearing exercises have achieved significant improvements in the ADLs that have historically been thought of as exogenous like walking and transferring. However, very few facilities have instituted these experimental programs.
8
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
number of variables were included to measure the generosity and method of state Medicaid reimbursement. Medicaid is the dominant payer of chronic nursing home care in the US accounting for approximately 50% of expenditures and 70% of bed-days. Rather than including a facility-level reimbursement rate, which would be endogenous to a facility’s quality level, the analysis utilized the average rate for the state (Grabowski, 2001). If the state deals in aggregates, no individual home can affect the state’s reimbursement rate. Thus, to the individual home, the average state Medicaid rate was exogenous. In addition to the Medicaid reimbursement rate, state-level indicators were also included for payment system type (i.e. retrospective, prospective, flat-rate, and combination), whether the state used a case mix adjusted reimbursement methodology, whether a state reimbursed hospital-based facilities differently and whether the state allowed an upward adjustment during the rate year due to additional cost information. Although the dummy variable approach measures differences in quality across sectors, it does not account for potential quality spillovers. Thus, we next estimate a series of models that include a measure of non-profit market share to account for the role of spillovers across sectors. The basic specification for this revised approach can be expressed as Y = Sγ + Xδ + ε
(2)
where S refers to the non-profit market share. Although previous empirical work has considered a specification that includes nursing home ownership measures at both the home and market levels, this approach may be flawed due to multicollinearity (i.e. home and market level ownership are associated with a statistically significant correlation coefficient of 0.43) and the potential endogeneity of ownership at the home level. Thus, this current study will follow the approach of Kessler and McClellan (2001) in the hospital literature by examining only spillovers at the market-level. With this specification, we test for the effect of non-profit market share on for-profit quality by limiting the analysis to only those for-profit facilities. Then, we test the effect of non-profit market share on overall quality by including all nursing homes in the model. Although this revised specification now accounts for the potential spillovers across sectors, it may suffer from bias due to the suspected endogeneity of non-profit-market share and quality. Assume that non-profit market share S has the following reduced form S = Zλ + Xγ + µ
(3)
where X is the same set of variables that appeared in the quality equation, Z a set of variables correlated with non-profit market share but not the error term in the quality equation, and µ the residual. A key econometric issue is that non-profit share S may be correlated with the error term in the quality equation. Although there is limited within-home and within-market variation over time in ownership, there may be unobservable factors that influence both the demand for quality and non-profits at the market-level. If this is the case, the error terms ε and µ will be correlated, which violates the assumptions underlying the linear regression model. However, we can still generate a consistent estimate of the effect of non-profit market share on quality if we can identify a set of variables Z that are correlated with non-profit share but not ε, the error term in the quality equation. Given Z, we can calculate an IV estimate of the effect of non-profit share on quality.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
9
In this study, we identify a set of variables Z. A first plausible instrument is the growth in the demand for nursing home care. Those areas in which nursing home demand is growing rapidly are likely to have a higher for-profit share because the capital market constraints faced by non-profit firms make rapid expansion difficult. Thus, there may be greater incentive for entry and expansion by for-profit owners than by their counterparts in the non-profit sector. There is evidence from both the nursing home and hospital industries that a percentage change in the aging population influences the proportion of non-profit firms (Hansmann, 1987; Steinwald and Neuhauser, 1970). Furthermore, Sloan et al. (2001) use this instrument in a recent analysis of the effects of ownership type on hospital quality. In this study, we used the percentage change in the population aged 65 and above for the 5-year period 1991–1996 as an instrument. A potential criticism of this instrument is that growth in demand—to the extent that it is correlated with the entry of new nursing homes—may also be correlated with quality if homes improve their quality through learning over time. Some of the growth within the nursing home industry has occurred through the expansion of existing homes rather than the entry of new facilities. Total beds in the US grew from 1.31 to 1.81 million (an increase of 38%) between 1978 and 1998 while the number of facilities grew from 14,264 to 17,458 (an increase of 22%) over this same 20-year period (Harrington et al., 1999). However, in an effort to control for the provision of quality by new firms, we include a dummy variable within the model measuring whether the facility has had fewer than three previous OSCAR surveys under its current provider number.7 Second, the non-profit market share in other health care industries may also serve as a plausible instrument by identifying those areas that are favorable towards non-profit health care production. For example, those areas with a higher hospital non-profit market share are expected to also have a larger non-profit nursing home share. The relative share of non-profits in different parts of the country is rooted in historical factors such as the age of the city and different patterns of voluntarism and charitable provision that have little to do with the advanced technology and prevalence of third party payment that characterize the current health care environment; e.g. see Stevens (1989) for a history of the organizational structure of the US hospital industry. However, hospital non-profit market share may be related to the provision of nursing home quality because certain nursing homes may face competition from hospitals for certain services. As a result, we used a lagged measure, the non-profit share of hospital beds in 1986, as an instrument. A potential criticism of this second instrument is that patients choose non-profits due to an underlying demand for high quality across all types of health providers. If this is the case, then the hospital share instrument may be correlated with the error term in the nursing home quality equation. In support of Stevens’ argument above, the difference in the cross-sectional non-profit hospital share is longstanding. Using multiple editions of the American Hospital Association annual survey of hospitals, we find that hospital non-profit market share is remarkably static over time, which provides evidence of a weak correlation between current consumer tastes for non-profit hospital care and unobserved nursing home quality. 7 This proxy measure will capture newly certified facilities, but it will also detect facilities that changed ownership and thus received a new provider number.
10
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
Thus, the identifying assumption is that high demand growth and lagged hospital non-profit market share are correlated with S, the non-profit nursing home market share, but are not correlated with ε, the error term in the quality equation. In the first stage of IV estimation (Eq. (3)), non-profit market share is regressed on these instruments plus the other regressors X from the second stage of IV estimation. In the second stage, quality is regressed on the instrumented non-profit market share and the other regressors X. The quality of the instruments used within the analysis is discussed below. Efficient estimates of the parameters are given by the weighted least squares (WLS) estimator where the weights are the number of residents per home. Importantly, for the bedsore, catheter, tube feeding and physical restraint measures, Y refers to the proportion of residents in a facility that satisfy a particular definition. Although the WLS approach does not recognize the binary nature of these dependent variables, it does facilitate the tractable estimation of the IV models. As an important sensitivity check, the WLS estimations proved robust to other specifications. Because these measures were reported at the facility-level (e.g. the proportion of residents with a bedsore), it was straightforward to convert the data into binary choices grouped at the facility-level. An alternate set of analyses using these grouped facility-level data for maximum-likelihood estimates of a probit or logistic model generated marginal effects that were similar in magnitude and precision to the single equation WLS results presented here. A final methodological point concerns the “grouped” nature of certain explanatory variables (e.g. non-profit market share), which may have introduced heteroskedasticity and biased the estimates of the parameter standard errors. When the true specification of the residual variance-covariance matrix follows a grouped structure, Moulton (1990) has shown that estimates of the standard errors will be biased downwards. A straightforward and unrestrictive approach to addressing this issue was to adjust the standard errors with the Huber–White robust estimator accounting for intra-county correlation.
5. Data The analyses contained within this study used merged data from five distinct sources (see Table 1 for summary statistics). The primary data source was the Online Survey, Certification, and Reporting system. The OSCAR system contains information from state surveys of all federally certified Medicaid and Medicare homes in the US. Certified homes represent almost 96% of all facilities nationwide (Strahan, 1997). Collected and maintained by the Centers for Medicaid and Medicaid Services (CMS), the OSCAR data are used to determine whether homes are in compliance with federal regulatory requirements. Every facility is required to have an initial survey to verify compliance. Thereafter, states are required to survey each facility no less often than every 15 months, and the average is about 12 months (Harrington et al., 1998a). The data for this analysis were collected within the 15-month interval of October 1995 through December 1996 and contain 16,978 unique nursing home surveys. If a home was surveyed multiple times during this 15-month interval, the most recent survey was included in the dataset. Four other data sources were utilized within this study to supplement the OSCAR data. First, the OSCAR data were merged with aggregate county level demographic,
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
11
Table 1 Descriptive statisticsa Variables
Number of homes
Proportion with bedsores Proportion with catheters Proportion with feeding tubes Proportion with physical restraints Total nursing staff per resident day Registered nurses per total nursing staff Non-profit market share Percentage change in elderly (>65) population (1991–1996) (ARF) Lagged (1986) hospital non-profit share (AHA) For-profit facility Not-for-profit facility Government owned and operated facility Hospital-based Chain facility Facility has fewer than three previous surveys Average number of activities of daily living with which the residents needed help Proportion of skilled nursing residents in the home Median per capita county income (ARF) Population >65 (ARF) Herfindahl index CMS area wage index (CMS) The average Medicaid rate (Harrington et al., 1998b) Hospital facilities reimbursed differently (Harrington et al., 1998b) Prospective reimbursement system (Harrington et al., 1998b) Retrospective reimbursement system (Harrington et al., 1998b) Flat-rate reimbursement system (Harrington et al., 1998b) Combines prospective and retrospective systems (Harrington et al., 1998b) Allows rate adjustment upward during or after a rate period (Harrington et al., 1998b) Employs case-mix reimbursement (Harrington et al., 1998b)
16,978 16,978 16,978 16,978 14,638 14,604 16,978 16,978
0.071 0.078 0.065 0.17 3.73 0.114 0.24 0.057
0.075 0.092 0.086 0.17 2.38 0.088 0.22 0.076
16,303 16,978 16,978 16,978 16,978 16,978 16,978 16,978
0.55 0.66 0.28 0.066 0.13 0.52 0.15 3.66
0.35 0.47 0.45 0.25 0.34 0.50 0.36 0.62
16,978 16,978 16,978 16,978 16,978 16,978 16,978
Mean
0.36 23,017.24 82,578.85 0.20 9,374.01 85.28 0.18
S.D.
0.32 5,594.54 170,176.80 0.23 1,769.94 20.64 0.39
16,978 16,978 16,978 16,978
0.76 0.014 0.17 0.059
0.43 0.12 0.37 0.24
16,978
0.41
0.49
16,978
0.55
0.50
a The data are from the 1995–1996 Online Survey, Certification and Reporting (OSCAR) system unless otherwise noted. The other data sources are the Area Resource File (ARF), the Centers for Medicare and Medicaid Services (CMS), the American Hospital Association (AHA), and the 1996 State Data Book on Long-Term Care Program and Market Characteristics (Harrington et al., 1998b).
socio-economic and health status data from the Bureau of Health Professions’ Area Resource File (ARF). Second, state-level Medicaid reimbursement methods and levels were obtained from the 1996 State Data Book on Long-Term Care Program and Market Characteristics published by Harrington et al. (1998b). Third, the CMS hospital area wage indexes were linked with the data. And finally, we obtained a measure of the share of non-profit hospitals from the 1986 American Hospital Association annual survey of hospitals.
12
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
6. Results 6.1. Specification tests As a model sensitivity check, we employed a Hausman test for endogeneity. Under this test, we generally rejected the null hypothesis of the exogeneity of non-profit market share for the quality equations across both the for-profit and overall market specifications. We also constructed an augmented regression test developed by Davidson and MacKinnon (1993) that also generally rejected the null hypothesis of exogeneity of non-profit market share. Our test of the overidentifying restrictions (Wooldridge, 2002) failed to reject the null hypothesis of appropriate specification in every case presented below except the staffing intensity model. Thus, although we present both the WLS and IV results below, we will primarily focus the discussion on the IV estimates. 6.2. Quality of the instruments The plausibility of the IV estimates presented here hinge on the strength of the instruments used within the analysis. Bound et al. (1995) have argued that the use of instruments that jointly explain little of the variation in the endogenous variables can do more harm than good. If a set of instruments is weakly correlated with the endogenous explanatory variable, then the authors have shown that even a small correlation between the instruments and the error can seriously bias estimates. Their results suggest that the partial R2 and F-statistics on the excluded instruments in the first-stage regression are useful as rough guides to the quality of the IV estimates. Staiger and Stock (1997) argue that 10 is an acceptable value of the F-statistic associated with the hypothesis that the coefficients on the instruments in the first-stage regression are jointly equal to zero. The set of instruments used in this study meets the standard of Staiger and Stock. In the first stage IV estimation, the hypothesis that the coefficients on the instruments are jointly equal to zero is rejected. The first-stage coefficients, F-statistic, and partial R2 associated with the excluded instruments are presented in Table 2 for both the for-profit and overall first-stage results. For the for-profit only model, the instruments have an F-statistic equal to 27.77 and R 2 = 0.02. For the overall model including all ownership types, the instruments have an F-statistic equal to 33.79 and R 2 = 0.05. All of the first-stage coefficients on the instruments were of the expected sign. An increased growth in the aging population over the 1991–1996 period was associated with fewer non-profits, and an increase in the lagged hospital non-profit share was associated with a higher non-profit nursing home share. Both of these instruments were statistically significant at the 1% level in both specifications. In addition to the assumption regarding the instruments being strongly associated with the endogenous variable, there is also the requirement that the instrument must not be correlated with the error term in the second stage of IV estimation. If it is still correlated, then the instrumented variable will still be endogenous. Although it is impossible to confirm the null hypothesis that these instruments are uncorrelated with the error term in the quality equation, a standard practice within the health economics literature is to report whether the instruments are correlated with those observable factors believed to be correlated with the unobservable factors that affect the second-stage error term. Thus, Table 3 takes each instrument and
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
13
Table 2 First stage of IV estimation (dependent variable = non-profit market share) Regressors
For-profit homes
All homes
Percentage change in elderly (1991–1996) Lagged hospital non-profit share Hospital-based Chain New facility Average ADLs Percent skilled care Herfindahl index Per capita income (US$ 1000s) Population age >65 (10,000s) Wage index (1000s) Medicaid rate Hospital-based reimbursed differently Retrospective system Flat rate system Combination system Rate adjustment allowed Case-mix adjusted system Constant R2
R2 F-statistic of instruments Number of observations
−0.22 (−3.49) 0.069 (6.12) 0.0055 (0.51) 0.0040 (0.96) −0.0098 (−1.69) −0.0041 (−0.93) 0.0078 (0.99) −0.23 (−14.63) 0.0036 (2.48) −0.00035 (−1.54) −0.011 (−1.94) 0.0013 (4.34) 0.038 (2.84) 0.051 (1.41) −0.033 (−2.57) 0.018 (1.27) 0.0080 (0.80) 0.027 (2.39) 0.10 (2.63) 0.26 0.02 27.77 10,697
−0.31 (−3.61) 0.10 (7.29) 0.065 (7.15) −0.015 (−3.23) −0.016 (−2.31) −0.0061 (−1.16) −0.0080 (−0.94) −0.038 (−1.61) 0.0037 (2.24) −0.00055 (−1.36) −0.014 (−1.96) 0.0017 (5.09) 0.039 (2.38) 0.076 (1.98) −0.034 (−2.22) −0.015 (−0.83) −0.0063 (−0.52) 0.042 (3.09) 0.13 (2.61) 0.19 0.05 33.79 16,303
Notes: The estimations are weighted by the total number of residents in each facility and the Huber–White adjusted t-statistics (corrected for intra-county correlation) are presented in parentheses.
divides the variables used within this study by those observations that are above the mean of the instrument and those that are below the mean. Table 3 presents the means for the non-profit market share, explanatory and quality measures across these two groups. Means of the groups for the non-profit market share indicate that those markets with low demand growth and lagged hospital non-profit share have higher non-profit market share. These results are consistent with the assumption that the instruments are correlated with the endogenous variable. In comparing the means of the explanatory variables, the two groups are statistically different across most of the variables. However, the large number of nursing homes (N = 16,978) within our study provides a high degree of precision. Despite being statistically significant, the case-mix measures, hospital-based measure, Herfindahl index, Medicaid index and wage rate variables are all fairly similar across the two groups. For example, the two case-mix measures do not differ by more than 4% for either of the two instrument groupings. Not surprisingly, ownership status at the facility-level, which is correlated with non-profit market share, appears to be related to the instruments. Because most chain owned homes are for-profit, chain ownership also appears to be related to the instruments. Furthermore, the elderly population appears to be quite dissimilar across the two groups. This dissimilarity may be due to a heavily skewed distribution of these measures due to their grouped nature.
14
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
Table 3 Comparison of variables by IV groups Variable
Non-profit share Explanatory variables Non-profit owned Government owned Investor owned Chain owned New facility Hospital-based Average ADLs Percent skilled Medicaid rate Wage index Income Herfindahl index Population >65 Quality measures Bedsores Physical restraints Feeding tubes Catheters Total nurse staff RNs per total staff Number of homes
Age change (%) (1991–1996)
Lagged hospital non-profit share
Below-average
Below-average
0.29
Above-average 0.21∗∗
0.20
Above-average 0.29∗∗
0.29 0.084 0.63 0.47 0.11 0.064 3.61 0.33 91.33 9,299 23,305 0.19 79,048
0.21∗∗ 0.065∗∗ 0.72∗∗ 0.58∗∗ 0.09∗∗ 0.056∗ 3.76∗∗ 0.33 86.24∗∗ 10,005∗∗ 24,504∗∗ 0.17∗∗ 108,507∗∗
0.20 0.077 0.72 0.55 0.12 0.070 3.69 0.32 86.10 9,568 22,222 0.21 106,963
0.29∗∗ 0.075 0.63∗∗ 0.49∗∗ 0.09∗∗ 0.054∗∗ 3.67∗∗ 0.33∗∗ 91.11∗∗ 9,637∗ 24,918∗∗ 0.17∗∗ 81,923∗∗
0.062 0.17 0.063 0.066 2.82 0.11
0.070∗∗ 0.19∗∗ 0.073∗∗ 0.070∗∗ 2.87 0.12∗∗
0.068 0.18 0.073 0.070 2.95 0.10
0.065∗∗ 0.18 0.063∗∗ 0.066∗∗ 2.78∗∗ 0.13∗∗
9,437
7,541
7,266
9,712
Notes: The means are weighted by the total number of residents in each facility. ∗ Statistically different at 5% level. ∗∗ Statistically different at 1% level.
As expected, a higher percent age growth was associated with lower quality and a higher lagged hospital non-profit market share was associated with higher quality. These comparisons represent crude (or unconditional) IV estimates of the effect of non-profit market share on quality. For example, a 1 percentage point increase in the non-profit market share was associated with 0.1 percentage point decrease in bedsores (the 1991–1996 percentage change between-group difference in bedsores of 0.8 percentage points divided by between-group difference in non-profit market share of 8 percentage points). 6.3. Effect of ownership on nursing home quality The conceptual framework provided two tests of the asymmetric and full information models. In the asymmetric case, an increase in non-profit market share will improve both for-profit and overall market quality, and in the full information case, an increase in non-profit market share will decrease for-profit quality and have no effect on overall quality. Before turning to the analyses that incorporate non-profit market share, we examine results
Regressors
Non-profit Government R2 N
Dependent variables Pressure ulcers
Physical restraints
Catheters
Feeding tubes
Total nurse staff
RNs per total staff
−0.0084 (6.95) −0.0093 (4.48) 0.11 16,978
0.017 (3.68) 0.041 (4.30) 0.10 16,978
−0.011 (8.17) −0.0056 (1.99) 0.16 16,978
−0.016 (7.95) 0.0016 (0.57) 0.20 16,978
0.14 (2.92) −0.66 (8.02) 0.09 14,638
0.0045 (1.82) 0.0056 (1.56) 0.16 14,604
Notes: All estimations are weighted by the total number of residents in each facility with the absolute value of the Huber–White adjusted t-statistics (corrected for intra-county correlation) presented in parentheses. All models include variables measuring the hospital area wage index, a Herfindahl index, the median per capita income, the number of elderly individuals in the county, the average Medicaid reimbursement rate, Medicaid reimbursement system, an activities of daily score, the proportion of residents requiring skilled care and binary indicators for chain owned, hospital-based and newly surveyed facilities.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
Table 4 Weighted least squares estimates of the effect of ownership type on nursing home quality
15
16
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
from the standard dummy variable approach used within the literature (see Table 4). Next, we incorporate the role of competition across sectors by examining the effect of non-profit market share on for-profit quality (see Table 5) and overall quality (see Table 6). Importantly, bedsores, physical restraints, catheters and tube feedings are negative measures of quality (e.g. more bedsores entails lower quality). Only information associated with non-profit status (for-profit status is the reference category) and non-profit market share are reported in these tables. The full results are available upon request from the authors. The standard dummy variable approach from the literature indicates that non-profit ownership was associated with an increase in nursing home quality based on five of the six quality measures. Non-profit status was associated with 0.84 percentage points (or 11.8% of the mean) fewer bedsores, 1.1 percentage points (or 14.1%) fewer catheters and 1.6 percentage points (or 24.6%) fewer tube feedings. Non-profit homes were associated with 0.14 (or 3.8%) more nursing staff and a 0.005 (or 3.9%) higher staffing skill mix. However, non-profit quality was found to be 1.7 percentage points (or 10%) lower when represented by the physical restraint measure. The coefficient on the non-profit dummy variable was statistically significant at the 10% level for all six of the quality measures. The evidence was inconclusive as to whether quality was higher in government sector relative to the for-profit sector. Although this model captured the effect of non-profit ownership, it may have provided biased estimates of the overall effect because it did not account for the spillover effect of the non-profit sector on nursing home quality. Thus, the results presented in Tables 5 and 6 examine the effect of non-profit market share on nursing home quality. Table 5 provides the first test of the asymmetric and full information models by isolating the effect of non-profit market share on quality in the for-profit sector. For each of the six measures of nursing home quality, Table 5 contains two columns of results. The first column for each quality measure reports the WLS estimation results and the second column reports the IV estimation results. Across both the WLS and IV models, an increase in non-profit market share was associated with higher for-profit quality for all six measures except for the WLS total nursing staff case. These results were statistically significant at the 10% level for the tube feeding, catheter, staffing skill mix, total nurse staff (IV only), bedsore (IV only), and the physical restraint (WLS only) measures. The magnitude of the quality spillovers from the non-profit sector to the for-profit was fairly sizable. Specifically, the non-profit market share elasticity of quality implied by the IV estimate from the bedsores model was −0.19[−0.056(0.24/0.071)], −0.44 for the feeding tube model, −0.12 for the restraint model, −0.13 for the catheter model, 0.09 for the total staff model and 0.32 for the staffing skill mix model. These results imply that a 10% increase in non-profit market share was associated with between 0.9 and 4.4% increase in for-profit nursing home quality. Thus, these first results are consistent with the asymmetric information case where an increase in non-profit market share provides a positive quality spillover within the for-profit sector. Table 6 reports the second test of the asymmetric and full information models by examining the effect of non-profit market share on overall quality. Once again, the first column for each quality measure reports the WLS results and the second column reports the IV estimation results. Similar to the for-profit quality results, an increase in non-profit market share was associated with higher overall quality for all six measures except for the WLS total nursing staff case. The non-profit market share coefficient was statistically significant
Regressors
Dependent variables Bedsores
Non-profit share R2 N
Physical restraints
Catheters
Tube feedings
Total nurse staff
RNs per total staff
WLS
IV
WLS
IV
WLS
IV
WLS
IV
WLS
IV
WLS
IV
−0.0031 (0.72)
−0.056 (2.40)
−0.036 (2.02)
−0.085 (1.05)
−0.0076 (1.84)
−0.041 (1.68)
−0.015 (2.32)
−0.12 (3.10)
−0.016 (0.08)
1.46 (1.79)
0.041 (4.19)
0.23 (3.57)
0.04 9743
0.15 10,171
0.05 9723
0.11 11,174
0.09 10,697
0.11 11,174
0.11 10,697
0.16 11,174
0.15 10,697
0.18 11,174
0.12 10,697
0.06 10,191
Notes: All estimations are weighted by the total number of residents in each facility with the absolute value of the Huber–White adjusted t-statistics (corrected for intra-county correlation) presented in parentheses. All models include variables measuring the hospital area wage index, a Herfindahl index, the median per capita income, the number of elderly individuals in the county, the average Medicaid reimbursement rate, Medicaid reimbursement system, an activities of daily score, the proportion of residents requiring skilled care and binary indicators for chain owned, hospital-based and newly surveyed facilities.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
Table 5 Weighted least squares (WLS) and instrumental variables (IVs) estimates of the effect of non-profit market share on for-profit nursing home quality
17
18
Regressors
Dependent variables Bedsores
Non-profit share R2 N
Physical restraints
Catheters
Tube feedings
Total nurse staff
RNs per total staff
WLS
IV
WLS
IV
WLS
IV
WLS
IV
WLS
IV
WLS
IV
−0.013 (4.78)
−0.031 (2.05)
−0.0069 (0.62)
−0.0035 (0.07)
−0.017 (5.78)
−0.016 (1.03)
−0.025 (5.87)
−0.089 (3.47)
−0.12 (1.09)
0.29 (0.60)
0.018 (3.13)
0.16 (4.13)
0.08 14,017
0.16 14,604
0.05 13,984
0.11 16,978
0.11 16,303
0.10 16,978
0.10 16,303
0.16 16,978
0.16 16,303
0.20 16,978
0.17 16,303
0.08 14,638
Notes: All estimations are weighted by the total number of residents in each facility with the absolute value of the Huber–White adjusted t-statistics (corrected for intra-county correlation) are presented in parentheses. All models include variables measuring the hospital area wage index, a Herfindahl index, the median per capita income, the number of elderly individuals per square mile, the average Medicaid reimbursement rate, Medicaid reimbursement system, an activities of daily score, the proportion of residents requiring skilled care and binary indicators for chain owned, hospital-based and newly surveyed facilities.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
Table 6 Weighted least squares (WLS) and instrumental variables (IVs) estimates of the effect of non-profit market share on overall nursing home quality
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
19
at the 10% level for the bedsore, tube feeding, staffing skill mix and the catheter (WLS only) measures. The results were relatively large in that the non-profit market share elasticity of quality implied by the IV estimate from the bedsores model was −0.10. Similarly, the elasticity implied by the IV estimate from the feeding tube model was −0.33, −0.05 for the catheter model and 0.22 for the staffing skill mix model. Thus, based on these estimates, a 10% increase in non-profit market share was associated with between 0.5 and 2.2% increase in overall nursing home quality. Although these results are not as conclusive in terms of statistical significance as the for-profit estimates, they once again are generally consistent with the asymmetric information case where an increase in non-profit market share improves overall quality. Further, the smaller magnitudes of the overall effects in the pooled models (relative to the spillover effects estimated in the for-profit only models) are reassuring in terms of model specification. Had the overall effects been as large as the spillover effects, it would have implied that the quality of non-profits rises as rapidly as the quality of for-profits when non-profit market share grows. Such a finding would have raised the concern that the non-profit market share coefficient was only capturing a correlation with unobserved variables that simultaneously cause higher quality and higher non-profit market shares.
7. Discussion The existing empirical literature on ownership and quality has generally focused singlemindedly on whether or not quality differences exist across sectors. The literature based on this focus has yielded highly inconsistent findings on whether non-profit nursing homes are “better” than for-profit facilities. We argue that inferring quality differences from the coefficient of a non-profit ownership dummy variable is vulnerable to bias that may be responsible for at least part of the inconsistent findings on quality. The key implications of the approach offered in this analysis are that the existence of differences is neither necessary nor sufficient to conclude that non-profit enterprise is socially desirable. This paper offers a novel instrumental variables approach to incorporating the ownership of a firm’s competitors into the analysis to test the asymmetric and full information models of the nursing home market. The empirical results presented within this paper favor an asymmetric information model. Thus, non-profit ownership may help alleviate inefficiencies associated with poorly informed consumers. Weisbrod (1988) has outlined two goals for public policy towards the non-profit sector: (1) public policy should help encourage non-profits to achieve their social goals, and (2) public policy should help achieve a better balance of institutional responsibilities between non-profits, for-profits and governments. With these two goals in mind, this model has several policy implications. If non-profits have a competitive advantage in “trustworthiness” while for-profits have greater incentives for efficiency, intersectoral competition can yield better outcomes than a market consisting exclusively of one type of firm. If non-profits attract the most poorly informed consumers, the likelihood that for-profits behave honestly (i.e. deliver the promised level of quality) rises with the non-profit market share. Likewise, competition from for-profit firms can limit inefficiency or the exercise of market power by non-profits. Recent work by Kessler and McClellan (2001) found that areas with a stronger presence of for-profit hospitals have 2.4% lower overall hospital expenditures.
20
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
The limited literature on mixed industries has focused on the negative side of competition: for-profits might erode non-profits’ ability to simultaneously achieve social goals and break even. Such an outcome is certainly possible, particularly if the non-distribution constraint is non-binding, which would provide little protection from entry by “for-profits in disguise” that debase non-profit status as a signal of trustworthy behavior. However, this paper demonstrates the beneficial aspects of intersectoral competition by showing that convergence in behavior of for-profit and non-profits may not be bad. Failure to observe large differences is not prima facie evidence that the non-profit sector is not socially beneficial. The recent appearance of more vigorous competition and cost containment pressures within the health care sector have drawn into question whether the cross-subsidization of “good works” that non-profits are presumed to practice should in fact be continued (Salkever and Frank, 1992). For example, published quality rankings and quality monitoring through managed care networks may have already lessened quality spillovers within the hospital sector. Both the federal government and several national services have recently begun to publish quality rankings of nursing facilities. However, quality rankings are likely to be more useful in sorting out the poorest facilities in absolute terms than in verifying that quality of care is commensurate with the prices charged. Additionally, the growth of managed care within the long-term care sector has lagged behind the acute care market. As a result, many of the broader trends in the health care sector may be less relevant for the nursing home market. Thus, the non-profit sector may continue to serve a socially beneficial role within the nursing home industry. Although we find evidence consistent with the type of spillovers predicted by an asymmetric information model, we recognize that this study does not directly observe the sorting of poorly informed consumers into the non-profit sector. Thus, we cannot rule out the possibility that the observed spillover effect (though inconsistent with the alternative, full information model) arises through some other mechanism (e.g. a greater presence of non-profits generates more vigorous non-price competition in quality). If an exclusively for-profit market fails to deliver optimal quality, however, it is worth noting that the ultimate welfare implications of the spillover effect would be similar even if the mechanism through which it operates is not the sorting of poorly informed consumers into the non-profit sector. Finally, there have been a number of explanations put forth for the low level of quality often observed in the nursing home market. Health economists have generally focused on the presence of supply constraints such as CON and construction moratoria that have restricted entry and thereby impeded competition within the nursing home industry; see Grabowski (2001) for a review of this literature. The nursing home industry has argued that low nursing home quality is attributable to inadequate payment levels by state Medicaid programs (e.g. American Health Care Association, 2001). Less attention has been paid to the lack of quality information available to many patients and the implications that asymmetries of information between nursing homes and patients may have for consumer welfare. Due to physical, cognitive and emotional disabilities, many nursing home consumers may fall far short of the homo economicus assumed in most economic models of behavior. This paper has provided evidence consistent with the hypothesis that non-profit status may remain a useful and relatively inexpensive signal that the promised level of quality will be delivered.
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
21
Acknowledgements We would like to thank Jim Burgess, Vivian Ho, Joseph Newhouse, and two anonymous referees, and seminar participants at the AEA annual meetings, the APHA annual meetings and the University of Pennsylvania for helpful comments on an earlier draft of this paper. We are also grateful to Robert Edwards for his excellent research assistance.
References American Health Care Association, 2001. A briefing chartbook on shortfalls in Medicaid funding for nursing home care. Prepared by BDO Seidman, LLP, Accessed 6 October 2001 at: http://www.ahca.org/brief/ seidman/seidmanstudy.pdf. Arrow, K.J., 1963. Uncertainty and the welfare economics of medical care. American Economic Review 53, 941–973. Banaszak-Holl, J., Zinn, J.S., Mor, V., 1996. The impact of market and organizational characteristics on nursing care facility service innovation: a resource dependency perspective. Health Services Research 31 (1), 97–117. Bound, J., Jaeger, D.A., Baker, R.M., 1995. Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variable is weak. Journal of the American Statistical Association 90 (430), 443–450. Cohen, J.W., Spector, W.D., 1996. The effect of Medicaid reimbursement on quality of care in nursing homes. Journal of Health Economics 15, 23–48. Davidson, R., MacKinnon, J.G., 1993. Estimation and Inference in Econometrics, 4th ed. McGraw-Hill, New York. Davis, M.A., 1991. On nursing home quality: a review and analysis. Medical Care Review 48, 129–166. Fries, B.E., 1990. Comparing case-mix systems for nursing home payment. Health Care Financing Review 11 (4), 103–119. Fries, B.E., Schneider, D.P., Foley, W.J., Gavazzi, M., Burke, R., Cornelius, E., 1994. Refining a case-mix measure for nursing homes: resource utilization groups (RUG-III). Medical Care 32 (7), 668–685. Garber, A.M., MaCurdy, T.E., 1992. Payment source and episodes of institutionalization. In: Wise, D.A. (Ed.), Topics in the Economics of Aging. University of Chicago Press, Chicago, pp. 249–271. Gertler, P.J., 1989. Subsidies, quality and the regulation of nursing homes. Journal of Public Economics 38 (1), 33–52. Gertler, P.J., 1992. Medicaid and the cost of improving access to nursing home care. The Review of Economics and Statistics 74, 338–345. Grabowski, D.C., 2001. Medicaid reimbursement and the quality of nursing home care. Journal of Health Economics 20 (4), 549–569. Hansmann, H.A., 1980. The role of non-profit enterprise. Yale Law Journal 89, 835–901. Hansmann, H., 1987. The effect of tax exemption and other factors on the market share of non-profit versus for-profit firms. National Tax Journal 40, 71–82. Harrington, C., Carillo, H., Thollaug, S.C., Summers, P.R., 1998a. Nursing Facilities, Staffing, Residents, and Facility Deficiencies, 1991 Through 1996. Department of Social and Behavioral Sciences, University of California, San Francisco, CA. Harrington, C., Swan, J.H., Griffin, C. Clemena, W., Bedney, B., Carillo, H., Shosak, S., 1998b. 1996 State Data Book on Long-Term Care Program and Market Characteristics. Department of Social and Behavioral Sciences, University of California, San Francisco, CA. Harrington, C., Swan, J.H., Wellin, V., Clemena, W., Carrillo, H.M., 1999. 1998 State Data Book on Long Term Care Program and Market Characteristics. Department of Social and Behavioral Sciences, University of California, San Francisco, CA. Harrington, C., Woolhandler, S., Mullan, J., Carillo, H., Himmelstein, D.U., 2001. Does investor ownership of nursing homes compromise the quality of care? American Journal of Public Health 91 (9), 1452–1455.
22
D.C. Grabowski, R.A. Hirth / Journal of Health Economics 22 (2003) 1–22
Hawes, C., Phillips, C.D., 1986. The changing structure of the nursing home industry and the impact of ownership on quality, cost and access. In: Gray, B.H. (Ed.), For-profit Enterprise in Health Care. National Academy Press, Washington, DC, pp. 492–541. Hirth, R.A., 1993. Consumer Information and Ownership in the Nursing Home Industry. Ph.D. dissertation. University of Pennsylvania, Philadelphia, PA. Hirth, R.A., 1999. Consumer information and competition between non-profit and for-profit nursing homes. Journal of Health Economics 18 (2), 219–240. Hirth, R.A., Banaszak-Holl, J.C., McCarthy, J.F., 2000. Nursing home-to-nursing home transfers: prevalence, time pattern and resident correlates. Medical Care 38 (6), 660–669. Holtmann, A.G., Ullman, S.G., 1991. Transactions costs, uncertainty, and not-for-profit organizations. Annals of Public and Cooperative Economy 62, 641–653. Institute of Medicine, 1986. Improving the Quality of Care in Nursing Homes. Committee on Nursing Home Regulation, National Academy Press, Washington, DC. Institute of Medicine, 2001. Improving the Quality of Long-term Care. Committee on Improving Quality in Long-Term Care, National Academy Press, Washington, DC. Kessler, D.P., McClellan, M.B., 2000. Is hospital competition socially wasteful? Quarterly Journal of Economics 115 (2), 577–615. Kessler, D., McClellan, M., 2001. The effects of hospital ownership on medical productivity, NBER Working Paper No. 8537. Moulton, B.R., 1990. An illustration of a pitfall in estimating the effects of aggregate variables on micro units. The Review of Economics and Statistics 72 (2), 334–338. Norton, E.C., 2000, Long-term care. In: Cuyler, A.J., Newhouse, J.P. (Eds.), Handbook of Health Economics, vol. 1. Elsevier, Amsterdam, pp. 955–994. Nyman, J.A., 1985. Prospective and ‘cost-plus’ Medicaid reimbursement, excess Medicaid demand, and the quality of nursing home care. Journal of Health Economics 4, 237–259. Nyman, J.A., 1994. The effects of market concentration and excess demand on the price of nursing home care. The Journal of Industrial Economics 42 (2), 193–204. O’Brien, J., Saxberg, B.O., Smith, H.L., 1983. For-profit or not-for-profit nursing homes: does it matter? Gerontologist 23, 341–348. Salkever, D.S., Frank, R.G., 1992. Health services. In: Clotfelter, C.T. (Ed.), Who Benefits from the Non-profit Sector? The University of Chicago Press, Chicago, pp. 24–54. Sloan, F.A., Picone, G.A., Tayler, D.H., Chou, S.-Y., 2001. Hospital ownership and cost and quality of care: is there a dime’s worth of difference? Journal of Health Economics 20 (1), 1–21. Spector, W.D., Selden, T.M., Cohen, J.W., 1998. The impact of ownership type on nursing home outcomes. Health Economics 7, 639–653. Staiger, D., Stock, J.H., 1997. Instrumental variables regression with weak instruments. Econometrica 65 (3), 557–586. Steinwald, B., Neuhauser, D., 1970. The role of the proprietary hospital. Law and Contemporary Problems 35 (4), 817–838. Stevens, R., 1989. In Sickness and in Wealth. Basic Books, New York. Strahan, G.W., 1997. An overview of nursing homes and their current residents: data from the 1995 national nursing home survey. Advance Data Number 280, National Center for Health Statistics, Rockville, MD. US General Accounting Office, July 1998, California Nursing Homes: Care Problems Persist Despite Federal and State Oversight. Report to the Special Committee on Aging, US Senate Pub. No. HEHS-98-202. US GAO, Washington DC. US Senate, Subcommittee on Long-term Care, Senate Special Committee on Aging, 1974. Nursing Home Care in the United States: Failure in Public Policy. US Government Printing Office, Washington, DC. Weisbrod, B.A., 1988. The Non-profit Economy. Harvard University Press, Cambridge, MA. Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. The MIT Press, Cambridge, MA. Zinn, J.S., 1993. The influence of nurse wage differentials on nursing home staffing and resident care decisions. The Gerontologist 33, 721–729.