Public Policy Sources #35:
Evidence

[Previous] [Contents] [Next]

The theoretical models which reflect the central feature of the Alberta plan - the opportunity for private, potentially for-profit firms to compete with government hospitals - are interesting but not determinative in their predictions; consequently, examination of the evidence regarding the implications of these changes is desirable. As always in economic analysis of policy, the proof is in the empirical pudding.

Enterprise comparisons

Non-profit versus for-profit

The theoretical discussion of non-profits and for-profits suggests no compelling reasons to expect significant differences in firm behaviour, particularly if physicians can effectively control hospital operations (Pauly and Redisch, 1973). This expectation is borne out by numerous empirical studies. Thus, the verdict rendered by economist Mark Pauly in 1987 is still relevant: "there is little ownership-related difference in hospital care given quality, or in quality given bed size, teaching status, and other proxies for type of output" (Pauly, 1987).

Of the 24 articles comparing for-profit and non-profit hospital performance reviewed in this study, 7 found that for-profits performed better, 5 found that non-profits performed better, and 12 found no difference in performance. Consider first the studies detecting better performance by for-profits.

Early among these was Clarkson (1972), who discerned a number of differences in the features of the two hospital types in US data. Clarkson discovered that for-profit administrators devoted significantly more time to supervisory control of employees, and were significantly less likely to grant automatic salary increases than their non-profit counterparts. In addition, non-profit hospitals were far less homogeneous in their input choices (e.g., personnel per bed) than for-profits were, implying that non-profit managers are less driven by market forces to choose the input mix that minimizes costs (Clarkson, 1972).

Average cost per admission in US for-profit and non-profit hospitals was compared by Bays (1979). After controlling for a variety of differences among his sample of hospitals, including hospital size, bed turnover, and case-mix, he determined that for-profits had lower average costs than non-profits.  In addition, he found that among for-profits, those belonging to regional or national chains had even lower costs than those which were independently operated.

Wilson and Jadlow (1982) compared US non-profit and for-profit provision of nuclear medicine services (diagnostic and therapeutic radiation services) in order to avoid the problem of general hospital-level comparisons, in which interhospital differences in service mix complicate comparison. The mode of evaluation employed by Wilson and Jadlow was to measure the disparity in production efficiency between for-profit and non-profit nuclear medicine departments. This analysis revealed how close each type of hospital came to producing the maximum possible amount of output, given its choice of input mix. They found that for-profit nuclear medicine services were significantly closer to maximum production than non-profits were.

Robinson and Luft (1985) examined over 5000 US hospitals to ascertain if cost differences existed between for-profits and non-profits. Their analysis controlled for a wide range of differences among hospitals, including number of compe- titors, case-mix, size, occupancy, and labour costs. They discovered that for-profits had significantly lower cost per admission and per day than non-profits.

A more subtle test of non-profit inefficiency was conducted by Hoerger (1991), who compared the variability in profits over time between US non-profits and for-profits. Hoerger reasoned that if non-profits were not simply maximizing profits, as for-profits were assumed to do, then non-profits would maintain a certain amount of slack, i.e., resources devoted to goals which reward managers but reduce profits. Consequently, in response to adverse economic shocks (such as the reduction in government reimbursement studied by Hoerger), the slack-maintaining non-profit could reduce slack and maintain profits at roughly the same level. Conversely, the for-profit, which has no slack, would suffer a reduction in profits. The same logic would apply to a favourable shock. Hoerger found that for-profits did indeed experience greater variability in profits, implicitly confirming his theory that non-profits maintain more slack.

Ferrier and Valdmanis (1996) focused their evaluation of for-profit and non-profit performance on US rural hospitals, which comprised approximately 50 percent of all short-term general hospitals and 30 percent of all beds in the year they examined (1989). They compared four dimensions of efficiency: technical (are inputs like labour and capital wasted in production?), allocative (is the right combination of inputs being used?), cost (technical and allocative efficiency combined), and scale (is the hospital producing output in its lowest-cost range?).  They discovered that - controlling for other differences among hospitals, including quality (defined as a hospital's excess mortality), hospital size and occupancy rate, and the proportion of services provided to outpatients and in intensive care - for-profits were efficient compared to non-profits by three of the four measures described above (the exception was allocative efficiency).

The final study detecting better performance by for-profits, reported by Altman and Shactman (1997), is based on analysis of US hospital profits earned from treating Medicare (elderly government-financed) patients. This study revealed that since 1991, profits from Medicare patients at for-profit hospitals have exceeded the comparable "profits" at non-profit hospitals. Because reimbursement for such patients is identical regardless of hospital type, Altman and Schactman conclude that the higher profits among for-profits must have resulted from lower costs at those hospitals relative to non-profits.

Other studies find better performance by non-profits than by for-profits. Pattison and Katz (1983) compared profit margins, use of ancillary services (e.g., pharmacy), service to government-funded and charity patients, and costs for US non-profits and for-profits. In all cases, meaningful conclusions are precluded by the failure of Pattison and Katz to determine if differences between for-profits and non-profits were statistically significant. Nevertheless, they find that for-profits: generated larger profits (smaller losses) in general and on ancillary services, provided more services per patient-day for those ancillary services on which they earned a profit, exhibited no difference in revenues from government-funded or charity cases, and incurred operating expenses 2 percent more per admission and 6 percent more per patient day. Besides neglecting the question of whether these differences were statistically significant, Pattison and Katz did not control for potential quality differences between the two types of hospital ownership, although they do note that the average age of assets held by national for-profit chains (4 years) was below the overall average in the sample (6 years).

Mark (1996) examined US psychiatric hospitals to determine if differences existed between for-profits and non-profits. She compared the two types along two dimensions: quality and cost. Quality was defined as the per-hospital number of regulator-identified violations and patient-initiated substantiated complaints. Mark found that for-profits exhibited higher frequency of violations and complaints, controlling for various relevant factors including volume of patients. She also compared efficiency for the two firm types, and discovered that there was no difference in operating costs related to ownership type.

Koop, Osiewalski, and Steel (1997) applied atypical but sophisticated statistical methodology to evaluate whether there are production efficiency differences between US non-profit and for-profit hospitals. Based on particular prior assumptions about hospital efficiency, they estimated that for-profits are less efficient than non-profits. But because their analysis does not control for any measurable differences between the two hospital types besides ownership form, they acknowledged that the for-profits' apparent "low efficiency might actually be capturing higher quality" (Koop, Osiewalski, and Steel, 1997).

An analysis of US hospitals by Woolhandler and Himmelstein (1997) juxtaposed for-profit and non-profit costs. Specifically, they found that after adjusting for differences in case-mix and local wage rates, for-profit hospitals had higher costs both per day and per discharge compared to non-profits. They did not, however, report whether these differences were statistically significant.

Most recently, Silverman, Skinner, and Fisher (1999) compared per capita spending by US Medicare (elderly government-funded) patients in non-profit and for-profit hospitals. Their analysis adjusted for differences in demographics and some basic hospital characteristics. After this adjustment, they found that per capita spending was significantly higher in for-profit hospitals. Their results do not directly address the efficiency question, however, for two reasons. First, they do not distinguish whether their spending difference arises from greater cost per service, greater extent of service provision, or both. Second, their "study could not address possible differences in the quality of care or amenities provided to Medicare beneficiaries" (Silverman, Skinner, and Fisher, 1997).

A number of other articles discern no efficiency difference between non-profit and for-profit hospitals. Psychiatric inpatient departments at US for-profit and non-profit hospitals were compared by Hrebiniak and Alutto (1973). They evaluated performance by these departments in three ways: discharge rate, cost per discharge, and cost per patient-day. On each of these three measures, there was no significant difference between the for-profit and non-profit hospitals studied.

In their study of US hospitals, Lewin, Derzon, and Margulies (1981) made a number of comparisons between non- and for-profits. The comparison of greatest relevance for efficiency purposes is operating cost, both per day and per admission. They found no significant difference in either type of cost between the two categories of hospitals. Moreover, for-profits had significantly lower full-time-equivalent staff per daily patient, a "crude measure ... of productivity" (Lewin, Derzon, and Margulies, 1981).

Becker and Sloan (1985) examined non-profit and for-profit hospital costs in the US, and attempted to control for measurable differences in addition to ownership type, including teaching status, case-mix, hospital size, and crude measures of quality (local per capita income, hospital employee wages). Given these controls, they found that while for-profits had significantly higher costs per patient day, they also had significantly shorter stays, and thus no significant difference in cost per admission. Furthermore, even the significantly higher cost per patient day was only one-tenth of one percent of daily patient costs.

In a large nationwide sample, Freund, Shacht- man, Ruffin, and Quade (1985) assessed the relative efficiencies of US non-profits and for-profits. The efficiency measure they considered was average length of hospitalization, controlling for differences in case-mix as well as geographical region, insurance, and hospital size. With these adjustments, they found that hospital ownership type (non-profit or for-profit) was statistically insignificant in determining average length of hospitalization.

Another method to evaluate the efficiency of non-profit relative to for-profit hospitals was used by Register, Sharp, and Bivin (1985). They tested whether US non- and for-profits operated according to the same production process, reasoning that if there was no significant difference in the manner in which labour and capital were combined in production, then there could be no difference in efficiency. Their analysis indicated that no significant difference existed in the production technologies of non- and for-profits, implying that no efficiency differences could exist, either. It is worth noting that while their test effectively controls for the quality of labour, it does not, however, do so for capital, which was simply measured as "staffed beds."

Renn, Schramm, Watt, and Derzon (1985) examined total patient-care costs per admission among US for-profit and non-profit hospitals. After adjusting this data for cross-hospital differences in case-mix and labour costs, they assessed whether this adjusted cost per admission was significantly higher in either hospital type. They found no significant difference in cost between for-profits and non-profits.

A separate study by many of the same authors (Watt, Derzon, Renn, Schramm, Hahn, and Pillari, 1986) considered a sample of 80 US hospital pairs (one for-profit and one non-profit in each pair) matched in terms of various operational similar- ities such as location, patient load, length of stay, and use of inpatient and intensive care. To further facilitate meaningful comparison, they adjusted their cost data (which included capital and medical-education costs) for differences in case-mix. They discovered that adjusted costs per admission and per day were not significantly different between for-profit and non-profit hospitals.

The technical efficiency of non-profits relative to for-profits was examined by Register and Bruning (1987). To endeavour to control for case-mix, they limited their sample of US hospitals to those within the 100-to-250-bed range. Controlling for bed size and market concentration, they discovered that for-profit firms exhibited no difference in technical efficiency; for-profits produced no more and no less than non-profits, given the same quantity of inputs. Differences in quality were not included among the control variables.

A similar exercise was conducted by Bruning and Register (1989). For a large sample of over 1,200 US hospitals, they estimated whether a relationship existed between technical efficiency and ownership form. Their analysis controlled for hospital size, case-mix, and geographical region. In none of their linear regressions did they find a significant relationship between ownership form and technical efficiency.

Besides the nine studies just described which found no significant efficiency difference between for-profit and non-profit hospitals, there are three other studies which offer mixed evidence. First among these is Robinson and Luft (1988), which compared per-admission costs among US for-profit and non-profit hospitals in two separate years, 1982 and 1986. They detected that these costs, which were adjusted for wage and case-mix differences among other factors, were significantly lower among for-profits in 1982 but significantly lower for non-profits in 1986.

The second ambiguous study, Burgess and Wilson (1995), calculated two different measures of technical efficiency for a sample of US for-profit and non-profit hospitals for each year of the period 1985-1988. For 1985, both measures indicated the efficiency of for-profits relative to non-profits, as did one of the two measures for 1986. All of the other efficiency comparisons revealed no significant difference between the two types of firms.

Finally, Mark (1999) indirectly compared the two forms. She examined conversions of hospitals from one ownership form to the other to determine the effect of such conversions on costs and profits. Her results indicated that both types of conversion - non-profit to for-profit, and for- profit to non-profit - led to increased operating costs as well as increased profits. This finding is attributable, in her view, to "The similar ability of for-profit and nonprofit management to turn around ailing hospitals" (Mark, 1999).

Public versus private

As with the comparison of non-profits and for-profits, the evidence relating the performance of public and private hospitals falls into three categories: studies finding that private hospitals apparently perform better, that public hospitals apparently perform better, and that there is no significant difference in performance. Of the 15 studies comparing public and for-profit hospital performance reviewed in this article, 8 found that private hospitals performed better, 3 found that public hospitals performed better, and 4 found no difference in performance. Later discussion will suggest, however, that even the apparent cost advantages of public hospitals are illusory.

Among the 8 studies finding superior performance by private hospitals was Clarkson (1972), discussed earlier, who contrasted a variety of aspects of performance for US public and private hospitals: administrator monitoring of employees, granting of automatic salary increases, and homogeneity of input choice. In each case, he determined that government hospitals performed worse. Specifically, private hospital administrators devoted significantly more time to employee supervision, were significantly less likely to grant automatic salary increases, and were significantly more likely to choose the same input mix as their competitors.

Psychiatric departments in US hospitals were the unit of analysis for Hrebiniak and Alutto (1973), discussed earlier. Besides comparing for-profit and non-profits (between which there was no significant difference in performance), they also evaluated public vis-à-vis private psychiatry departments. They found that departments in public hospitals performed worse than those in private hospitals; public departments had a significantly lower discharge rate, and signi- ficantly higher cost per discharge and per day.

Lindsay (1976) compared US government and for-profit hospitals. Because public hospital administrators are only rewarded for providing services that are visible to their bureaucratic masters, Lindsay predicted that there would be a distortion in service provision towards certain measurable (and thus visible) activities. For example, a public hospital could increase its visible output of patient-days by keeping patients longer than is medically desirable. In fact, Lindsay found evidence that for 13 common procedures, public hospital average stay lengths were substantially greater than comparable private hospital stays (although he did not calculate whether these differences were statistically significant). But because public hospitals correspondingly skimp on other, unobservable activities (e.g., bedside manner, clean floors), Lindsay found that their average costs per patient day were lower, suggesting that government hospitals will provide less of these bureaucratically unmeasurable attributes than is efficient.

An analysis of US nuclear medicine services was made by Wilson and Jadlow (1982), described earlier. In addition to comparing non-profit and for-profit entities, they also evaluated public facilities. They discovered that in terms of technical efficiency (maximum output for given inputs), government providers performed worse than private non-profits, which performed worse than for-profits.

The comparison of variation in profits which Hoerger (1991) made for non-profits and for-profits (discussed earlier) was also extended to government hospitals. The logic of the comparison was that firms with greater operating slack (i.e., where resources are devoted to non-profit-maximizing purposes) would exhibit smaller variation in profits in response to external shocks. Hoerger ascertained that government hospitals experienced even less variation in profits than did non-profits, which themselves had less profit variation than for-profit firms. Consequently, this provides indirect evidence that government hospitals maintain more slack than private hospitals, either non-profit or for-profit, and are thus less efficient.

Ferrier and Valdmanis (1996), discussed earlier, examined rural US hospitals. Their calculations of four different efficiency measures were made to compare public and private hospitals, as well as for-profits and non-profits. On each of the four efficiency measures, they determined that, in the rural setting, public hospitals were inefficient compared to private ones.

Profits from Medicare (elderly government- financed) patients at different hospital types were compared by Altman and Shactman (1997), discussed earlier. By tracking profits, they were able to form inferences as to cost differences between public and private hospitals. Specifi- cally, because Medicare reimbursement was identical across the different hospital types, and because profits were higher among private (for-profit) hospitals than among public hospitals from 1991 through 1996, Altman and Shachtman concluded that costs must have been lower at private hospitals.

A different test of public-private differences was undertaken by Coles and Hesterly (1998), who analyzed the decision by US hospitals either to contract out services or perform those services using internal staff. Among the 15 services compared were laboratory, housekeeping, computing, food, maintenance, billing, supplies, emergency room, and pharmacy. Coles and Hesterly generated two forms of evidence distinguishing public and private hospitals in terms of the decision to contract out (or not). One test indicated that public hospitals were significantly less likely to contract out these services. The other test determined that public hospitals were not significantly influenced in their contracting decisions by the transactions costs associated with those decisions. In other words, public hospitals were not motivated to contract out even when the costs of doing so were low, unlike private hospitals. Coles and Hesterly interpret this as reflecting "the fact that public institutions ... are likely subject to less severe efficiency pressures ... than private organizations operating in the same market" (Coles and Hesterly 1998).

Three articles detected superior performance by public hospitals. In addition to their comparison of US non-profit and for-profit hospitals, Pattison and Katz (1983), discussed earlier, examined government hospitals relative to private alternatives. Most important among their findings was the higher cost per patient-day and per admission exhibited by for-profits relative to government hospitals. These differences, however, were not tested for statistical significance.

Robinson and Luft (1988), discussed earlier, compared US public and private hospital costs in 1982 and 1986. They adjusted cost per admission for interhospital differences in case-mix and labour cost. They found that in both years, public hospital cost per admission was significantly below that in private hospitals.

Although focused on comparisons of admini- strative cost, Woolhandler and Himmelstein (1997), previously discussed, also addressed the subject of total cost comparisons between public and private US hospitals. They reported that for-profit firms had higher costs per day and per discharge than public ones did, even after adjusting for differences among hospitals in case-mix and local wage rates. Whether these differences were statistically significant was not reported, however.

Four studies discerned no significant difference between US government and for-profit hospital costs, or found mixed results. Becker and Sloan (1985) compared total costs per patient day and per admission, controlling for other potential sources of influence, including case-mix, hospital size, and crude quality measures (local per capita income, hospital employee wages). They found that while cost per patient-day was higher in for-profit hospitals (a very small difference in relative magnitude), for the more appropriate measure, cost per admission, there was no significant difference between government and for-profit hospitals.

As well as the comparison of for-profit and non-profit US hospitals found in Renn, Schramm, Watt, and Derzon (1985), previously described, that study also contained a comparison of public and private firms. The authors compared cost per admission, adjusted for differences across hospitals in case-mix and labour costs. They detected no significant difference in cost between public and private entities.

Mixed results were reported in two papers. As described earlier, Robinson and Luft (1985) examined a large data set of US hospitals in order to compare average costs per day and per admission. When they controlled for differences among hospitals in a variety of relevant areas, including number of competitors, size, occupancy, and labour costs, they found that public hospitals had significantly higher costs per admission and per day. When they were able, in addition, to control for differences in case-mix for a subset of their data, however, they detected no significant cost differences between public and private entities.

Burgess and Wilson (1995), discussed previously, calculated two efficiency measures for four different hospital types (private non-profit, for- profit, federal government, and local govern- ment) for each year in the period 1985-1988. For each year, one can compare each type of public hospital (federal, local) with for-profit hospitals, using each of the two efficiency measures. Of the 8 instances in which federal government hospitals are compared with for-profit entities, public hospitals were found to be superior in efficiency terms in six cases. However, when the comparison is made between local government hospitals and their for-profit counterparts, no difference in efficiency is found in any of those eight calculations.

Hospital competition

The final area of evidence bearing on the impact of the proposed Alberta reform is that regarding the effect of competition in hospital markets. Besides allowing activity by private for-profit firms, the Alberta proposal also might confer benefits by engendering competition. But, as noted in the theoretical discussion, the predicted effects of competition are ambiguous. Consequently, it is vital to consult the evidence on this topic.

While the effects of hospital competition have been frequently studied, we are fortunate that this research has been ably surveyed and summarized in recent years by several prominent health economists. Prior to the mid-1980s, increases in competition among US hospitals were found to increase prices and costs, contrary to the effect of competition in most markets, due to a "medical arms race" from burgeoning quality-based competition (Kessler and McClellan, 1999). But from the mid-80s onward, intensified US competition was found to lower prices and costs (Dranove and White, 1994). This orthodox price-reducing effect of competition has been detected in US non-profit markets as well (Gaynor and Haas-Wilson, 1999). The desirable effects of competition in both for-profit and non-profit in US markets have been more recently confirmed by Keeler, Melnick, and Zwanziger (1999).

Nevertheless, three limitations in the existing empirical literature have been identified by Kessler and McClellan (1999). First, the existing literature does not permit definitive conclusions regarding the beneficial or adverse effects of competition because these studies do not fully account for the financial or health consequences of enhanced competition. Second, inappropriate definitions of markets are used, creating bias in the estimates of competition's effect. Finally, most studies have not adequately controlled for other factors, again potentially biasing estimates of the effects of competition.

Having targeted these problems, Kessler and McClellan (1999) endeavour to remedy them. While the details of their modifications are beyond the scope of this overview to describe, their analysis of the recent effects of competition among US hospitals stands as a leading example of meticulous scholarship. Based on their statistical modifications, Kessler and McClellan discovered that in the 1980s, US hospital competition did increase costs but also improved some health outcomes. Thus, "Whether competition increased welfare depends on an assessment of whether the additional health associated with competition was worth the higher associated cost of care" (Kessler and McClellan, 1999). But this ambiguity was removed beginning in 1991. From that point forward, increased competition not only improved health outcomes, it also lowered costs.

[Previous] [Contents] [Next]

info@fraserinstitute.ca You can contact us at the above email address for any comments or information requests. Please report any dead links or technical problems.

If you know someone who would be interested in this web page, please enter their email address below, and we will forward this URL to them: Email Address:

Last Modified: Thursday, August 5, 1999.