The impact of medicare capital prospective
payment regulation on hospital capital
expenditures
Ran Barniv
a,*, Kreag Danvers
b, Joanne Healy
aaDepartment of Accounting, Graduate School of Management, Kent State University, P.O. Box 5190, Kent, OH 44242-0001, USA
bDepartment of Accounting, Eberly College of Business, Indiana University of Pennsylvania, Pennsylvania, USA
Abstract
Our study examines the impact of the capital prospective payment system (CPPS), implemented by Medicare in 1991, on capital expenditures and cost-eective behavior of non-proprietary hospitals. As noted in the paper, we use audited ®nancial statement data for a large national sample of hospitals. Univariate analyses demonstrate a sta-tistically signi®cant decline in capital expenditures in the years following the CPPS regulation without signi®cant changes in relative aggregate operating expenses. These preliminary ®ndings suggest that CPPS induces some cost-eective behavior by hospital managers. Ordinary least-squares (OLS) regressions indicate that capital expenditures before and after CPPS are dierently aected by the changes in most explanatory variables. Further OLS regressions indicate that high-cost (low-cost) hospitals decrease (increase) capital expenditures following CPPS, once other factors are controlled for. Managerial accounting implications for hospitals include the eect of the regulation on capital budgeting decisions. Greater accounting disclosure may be necessary so that alternative modes of coping with the regulation can be discerned. Policymakers and regulators should also be aware that although reductions in capital expenditures may have favorable short-term eects of reducing health care costs, a potentially negative public health impact may result if capital expenditures continue to decrease. Ó 2000
Published by Elsevier Science Ltd. All rights reserved.
*Corresponding author. Tel.: +1-330-672-2545-x379; fax: +1-330-672-2548. E-mail address:Rbarniv@bsa3.kent.edu (R. Barniv).
1. Introduction
The 1983 amendments to the Social Security Act introduced the Medicare prospective payment system (PPS) and modi®ed the mechanism through which payments are made to hospitals for providing care to Medicare patients (see e.g., Cotterill, 1991, p. 79; Soderstrom, 1993, pp. 155±156; Eldenburg and Kallapur, 1997, p. 32). Although the 1983 amendments altered payments of hospital inpatient operating costs, implementation of the capital cost pro-spective payment system (CPPS) was deferred and Medicare continued pay-ments based on actual capital-related expenses (Wedig et al., 1989, p. 518). The debate on when and how to implement the new regulation continued for al-most eight years.1Finally, CPPS was implemented with a 10 year gradual and proportional transition period on October, 1 1991, and became eective for ®scal years beginning on this date (Burke, 1991, p. 36; Grimaldi, 1991, p. 72). Articles published during the early 1990s expressed concerns about unfavor-able consequences of the capital regulation, including a potential reduction in future capital expenditures by hospitals (Luggiero, 1990, p. 3; Burke, 1991, pp. 34, 35; Anderson, 1992, p. 38).
The purpose of our study is to explicitly investigate the impact of the ®nal CPPS regulation on capital expenditure decisions in the hospital industry, and to assess whether it induces overall cost-eective behavior by managers. We hypothesize that hospitals may reduce capital expenditures as a result of the new regulation. Furthermore, we attempt to determine how hospitals have adapted to the post-CPPS environment. Previous articles have not empirically scrutinized these issues.
Using data from audited ®nancial statements provided by the Merritt Sys-temâ
database, we examine capital expenditures for a national sample of non-proprietary general service hospitals prior and subsequent to the ®nal CPPS regulation.2The database includes ®nancial accounting data for 2048 hospi-tals from 1988 to 1996 of which 1949 are useful for our study.
We found a statistically signi®cant decline in capital expenditures in the years following the implementation of the ®nal Medicare capital regulation. Another important issue examined in this study is whether CPPS had the de-sired eect in promoting overall cost-eective behavior. Although a reduction in capital expenditures was observed, operating expenses such as salaries and
1
In 1986, the proposed regulation for incorporating capital payments into PPS was issued by the United States Department of Health and Human Services (Cotterill, 1991, p. 79). However, the ®nal capital PPS regulation was delayed on several other occasions and actual cost-based payments continued until 1991 (Grimaldi, 1991, p. 72).
2We use the term proprietary hospitals throughout the manuscript. These hospitals are
other expenses (i.e., the aggregate of drugs, supplies, operating leases, and rental expenses) as percentages of revenues remained unchanged across the two subperiods. Furthermore, pro®tability improved perhaps due to declining in-terest rates. The CPPS regulation seems to motivate managers to reduce capital expenditures and to change other variables such as ®nancial leverage, but not to change operating expenses.
We examined the impact of hospital characteristics on capital expenditures across subperiods and found that the impacts of certain explanatory variables were statistically dierent before and after CPPS. In response to CPPS, hospital managers have changed several ®nancial and operational characteristics, which could explain some aspects of the reduction of capital expenditures following CPPS. For example, the other-expense variable, which includes drugs, supplies, leases and rentals, was signi®cantly negative in the period subsequent to CPPS. This result indicates that operating leases and rentals may potentially substitute for capital expenditures. Other empirical results suggest that high-cost (low-cost) hospitals decrease (increase) capital expenditures following CPPS, once other factors are controlled for. This ®nding supports Medicare's expectations regarding the redistribution eect of the new regulation (Federal Register, 1991, pp. 43427±43428). An important public policy issue is that although reductions in capital expenditures may reduce health care cost, policymakers need to be cognizant of the potentially negative long-term impact on public health.
2. Some regulatory background
Prior to the Social Security Amendments (United States Statutes at Large, 1983, Public Law 98-21, Section 601, pp. 149±163), Medicare reimbursed hospitals on a retrospective cost basis. Under this cost-based reimbursement scheme, Medicare-related hospital revenue was tied directly to actual operating expenses incurred, providing no incentives for hospitals to control either risk or cost. In response to in¯ationary concerns and the solvency of the Medicare program, the United States Congress directed the Department of Health and Human Services to develop a new reimbursement method that would incor-porate incentives for hospital eciency (Guterman and Dobson, 1986, p. 97). PPS was implemented in 1983, but certain hospital costs, including capital-related expenses, were speci®cally excluded and continued to be reimbursed on a retrospective cost basis (Wedig et al., 1989, p. 518).
Prior to CPPS, Medicare reimbursed hospitals for actual capital-related expenses, primarily depreciation and interest expenses.3 In addition, some
3Medicare reimburses for annual capital-related expenses, such as depreciation and interest, and
other less substantial expenses such as insurance were reimbursed. Following the implementation of CPPS, Medicare began to reimburse these expenses based on predetermined national averages. Medicare promoted two objec-tives of CPPS which are relevant for our study. First, CPPS was expected to stimulate eciency in capital spending and induce hospitals to become more cost-eective, thereby reducing total Medicare cost. Second, CPPS might provide redistribution of payments and encourage hospitals with lower capital-related expenses, for example hospitals in rural areas, to increase capital expenditures (Federal Register, 1991, 30 August, pp. 43428± 43517).
Through this regulation, Medicare fundamentally changed the way it pays hospitals for capital expenses attributable to Medicare inpatients. Rather than basing reimbursements on actual expenses, hospitals would receive a payment per case based on factors similar to the existing PPS system (Burke, 1991, p. 36). The CPPS regulation aects all PPS hospitals and provides for a 10 year gradual and proportional transition period that precedes full imple-mentation of the new system (Grimaldi, 1991, pp. 72±76). During this tran-sition period, Medicare established two tiers of hospitals, based upon hospital-speci®c capital-related expenses relative to an adjusted national av-erage. Hospitals were classi®ed as low-(high-) cost if their capital-related expenses were lower (higher) than the national average. Low-cost hospitals are reimbursed based on a fully prospective method, while high-cost hospitals are reimbursed based on a hold-harmless method (Burke, 1991, p. 36).4We expect this new payment scheme to impact capital budgeting decisions by hospital management.
3. Previous literature and incremental contribution
Prior accounting studies examined hospital behavior in response to federal regulatory changes. Several of these studies analyzed issues associated with the 1983 Medicare PPS. For example, Soderstrom (1993) scrutinized hospital management's tendency to increase income through changing admission or reporting policies, whereas Eldenburg and Kallapur (1997) examined the ten-dency of hospitals to maximize cash ¯ows by changing patient mix and cost
4
allocations.5In addition, studies that examined the impact of the 1983 PPS on hospital ®nancial and operating characteristics found that hospitals coped fairly well by undertaking cost-cutting behavior (Guterman et al., 1988, p. 68; Feder et al., 1987, pp. 869±872; Friedman and Shortell, 1988, p. 264).6
Other studies focused on state level regulatory eects. Blanchard et al. (1986, pp. 11±14) studied hospitals in the State of Washington that were subject to revenue regulation and suggested that those hospitals report biased budget information to increase revenue. Eldenburg and Soderstrom (1996, pp. 23±24) re-examined Blanchard et al.'s (1986, pp. 1±2, 14) hypotheses and concluded that hospitals bias budget information and shift costs across payor types to increase reimbursement.7
Research that deals speci®cally with the issue of hospital capital costs within the prospective payment system is very limited. For example, studies discuss the potential impacts of a CPPS-type payment structure on the cost of capital and capital structure decisions (Boles, 1986; Wedig et al., 1988, 1989). Prior to, and concomitant with the ®nal capital regulation, some conjectures were made regarding its impact on the hospital industry's ®nancial viability, future levels of capital expenditures and access to capital (Burke, 1991; Anderson, 1992; Health Industry Today, 1992).8Cotterill (1991) and Kauer (1995) examined issues related to the ®nal capital regulation. Recently, Lynch (1998, pp. 1±3)
5
Eldenburg (1994) investigated management information sharing with physician groups as a hospital cost control mechanism in response to the 1983 PPS. Lambert and Larcker (1995, pp. 1±3) found that hospitals more vulnerable to the 1983 PPS regulation tend to implement more bonus-based compensation contracts as motivational devices. Carey (1994, p. 275) found that during the years following PPS implementation small, rural hospitals tended to allocate more costs to outpatient departments than did large, urban hospitals.
6Studies also investigated the market response of for-pro®t hospitals related to PPS events
(Folland and Kleiman, 1990) and the eect of a diagnosis related group (DRG) payment system on the eciency of New Jersey general acute-care hospitals (Borden, 1988). These studies found little eect of a DRG payment system on either the market value of for-pro®t hospitals (Folland and Kleiman, 1990, pp. 61±66) or the eciency of hospitals in New Jersey (Borden, 1988, pp. 87±93).
7Hospital accounting studies also provide evidence that disclosing quasi-dividend payments to
physicians may enhance the informativeness of hospital operating statements (Mensah and Chiang, 1996), classi®cation of hospital ®nancial ratios (Zeller et al., 1996) and the relative eciency of proprietary versus non-proprietary hospitals (Carter et al., 1997). Mensah and Li (1993) extended a translog budget model within a not-for pro®t setting. Mensah et al. (1994a) examined accounting conservation and earnings management by HMOs, whereas Mensah et al. (1994b) examined earnings management by HMOs that may be subject to political cost.
8
found that hospitals in California marginally change capital investment be-havior and reduce the use of long-term debt in response to changes in Medicare regulation.
Our study diers from the prior literature in several aspects. First, we ex-amine empirically the impact of CPPS on capital expenditures, while prior articles do not provide empirical evidence on this issue. Second, we scrutinize how hospitals have adapted to the post-CPPS environment by studying po-tential changes in their cost-eective behavior. Third, we examine the impact of other hospital characteristics on capital expenditures before and after the im-plementation of CPPS. Fourth, we examine hospital response to regulation utilizing audited ®nancial statement information for a large, national sample of PPS hospitals, whereas many previous studies use a single-state sample. The results of our study are relevant for policymakers, regulators, hospital man-agement, creditors, accounting standards setters and scholars as well as other users of ®nancial and managerial accounting information.
4. Research methods
4.1. Hypotheses
Non-pro®t hospitals use standard capital budgeting techniques to allocate scarce capital resources (Cleverley and Felkner, 1984, p. 45). The decoupling of Medicare payments from actual costs requires more rigorous project evalua-tion since hospitals are forced to assume a higher level of risk for investment decisions. Also, increasing unpredictability in the update factors of capital payments (Burke, 1991, p. 34) leads to a higher risk adjusted cost of capital. Therefore, it may be expected that increasing uncertainty and lowering ex-pected future cash ¯ow under CPPS decreases the number of acceptable in-vestment projects. Our primary objective is to identify whether changes in the level of hospital expenditures for property, plant and equipment (PPE) are associated with the issuance of the ®nal CPPS regulation. This leads to the ®rst hypothesis tested in our study:
H1: Non-proprietary hospitals subject to CPPS will exhibit no change in capital expenditures across the pre-regulation and post-regulation periods. The alternative hypothesis is that hospitals sub-ject to CPPS will decrease capital expenditures in the post-CPPS period.
oper-ating expenses may be observed. The regulation may result in overall cost-eective behavior if capital expenditures decline by more than the increase in these operating expenses. This is the logic behind the second alternative hypothesis. Thus, the second hypothesis examines behavior by managers for operational activities other than capital expenditures, across the two subperiods.
H2: Non-proprietary hospitals subject to CPPS do not change cer-tain operating expenses across the pre-regulation and post-regula-tion periods. The alternative hypothesis is that hospitals increase certain operating expenses during the second subperiod.
The third hypothesis examines how hospitals subject to Medicare's two-tier classi®cation may dierently change capital expenditures. While all PPS hos-pitals are expected to be aected by the capital regulation, particular hoshos-pitals may be less adversely impacted than others. For example, Medicare expected that payments to high-cost hospitals (primarily large, urban hospitals) would decrease, and therefore, these hospitals might reduce capital expenditures. However, payments to low-cost hospitals (primarily small, rural hospitals) would increase subsequent to CPPS (Pallarito, 1991, p. 4) and the higher ex-pected payments to these hospitals may increase capital expenditures (Federal Register, 1991, pp. 43427±43428). This leads to the statement of our third hypothesis:
H3: Non-proprietary hospitals de®ned by Medicare as low-cost and those de®ned as high-cost will exhibit no change in capital expendi-tures across the pre-regulation and post-regulation periods. The alternative hypothesis is that high-cost hospitals will decrease their capital expenditures, while low-cost hospitals will increase their capital expenditures.
Both univariate and multivariate analyses were used to test the three hypotheses. In particular, we assessed the impacts of hospital characteristics on capital expenditures through the use of ordinary least-squares (OLS) multivariate regression models. Furthermore, we examined the eects of changes in these characteristics on capital expenditures across the two sub-periods.
4.2. Variables
4.2.1. Dependent variable
The variable used to measure capital expenditures represents the change in net PPE, adjusted for depreciation and in¯ation. This variable is used in both univariate and multivariate analyses.
CEXGFA (capital expenditures/CIPI)t/gross ®xed assetstÿ1. Capital
ex-penditures are calculated as the change in net ®xed assets from year tÿ1 to yeart, plus depreciation expense in yeart. Thus the numerator of CEXGFA includes purchases of ®xed assets in period t minus the book value of the disposed assets during the period. Construction-in-progress is included in the numerator and the denominator, but it is not separately available in the data-base. Gross ®xed assets at time tÿ1 include all depreciable and non-depre-ciable ®xed assets at historical cost before accumulated depreciation. Except for the CIPI adjustment, CEXGFA corresponds to the Zeller et al. (1996, p. 170) capital expenditure growth rate variable.9
4.2.2. Independent variables
We used the following eight independent variables in our ®rst regression model and for the univariate analyses.10These variables include capital related characteristics along with pro®tability, liquidity, eciency and ®nancial le-verage measures. The ®rst two variables include depreciation in the numerators since it is the primary capital-related expense reimbursed by Medicare.11 Similar capital expense variables are presented by Deloitte and Touche (1996, p. 235) and Zeller et al. (1996, p. 170). Although we hypothesize the directions of various independent variables on capital expenditures, the complexity of the analyses makes interpretations of several coecients rather dicult. All independent variables, except the denominator of the ®rst, are measured at periodt.
DNFAdepreciation expenset/net ®xed assetstÿ1. Net ®xed assets are
computed as gross ®xed assets less accumulated depreciation. Examining only the numerator suggests that, as purchases increase depreciation potentially increases, and as equipment is sold depreciation may decrease, depending on
9To adjust capital expenditures for in¯ationary eects we use Medicare's capital input price index
(CIPI), which re¯ects changes in the price levels of a market basket of capital resources used by hospitals for inpatient Medicare services. The CIPI measures the input price change in capital-related expenses and is used by Medicare in determining increases in capital prospective payment rates (Federal Register, 1996, pp. 46196±46203).
10
Several factors have been promoted to in¯uence hospital capital expenditure decisions, including size, patient mix and capital structure (Pallarito, 1990, p. 33). Cotterill (1991, p. 84) found variables that are associated with hospital capital costs include location, bed size, age of plant and ®nancing structure. Several of these variables are used in our study.
11The Pearson correlation coecient between the ®rst two independent variables is 0.065 which is
the age of equipment. These interactions would lead to a positive coecient, not considering the eect of the denominator. Therefore, capital expenditures are expected to be positively associated with DNFA and this relationship may continue following CPPS.12
DTAdepreciation expense/total assets. The denominator of this variable includes nondepreciable assets, which are not represented in the denominator of DNFA. Similar to DNFA, we may expect a positive relation between DTA and CEXGFA and the positive relation may continue during the second pe-riod. However, a negative relation may be anticipated if hospitals reduce the growth rate in capital expenditures along with decreasing other non-®xed as-sets. This negative relation may continue following CPPS. Given the com-plexity of the relation and the potentially dierent impacts of the numerator and the denominator, the ®nal impact is not clear and interpretation of the estimated coecient should be done with caution.
ROAexcess of revenues over expenses/total assets. Hospitals that have a higher numerator may have more funds available for capital expenditures. Therefore, there may be a positive relationship between ROA and CEXGFA. The relationship could weaken following CPPS since hospitals may not invest immediately in capital expenditures due to uncertainty in the update factors used in the DRG-based formula to calculate reimbursement. This and other measures of ROA are used by Soderstrom (1993, p. 179) and Mensah and Chiang (1996, p. 230).
FBTAfund balance/total assets. FBTA may be negatively related to capital expenditures since hospitals with relatively higher equity in their capital structure probably use more costly ®nancing.13However, because higher eq-uity may enable hospitals to more easily ®nance capital expenditures we may also expect a positive relation between FBTA and CEXGFA. Therefore, the ®nal impact of FBTA on CEXFGA and the change in the relationship are complicated and may be positive or negative. Our capital structure variable is identical to the equity ®nancing variable used in Zeller et al. (1996, p. 169). Other measures of ®nancial leverage are used by Soderstrom (1993, p. 179) and Mensah et al. (1994a, p. 80).
LNTAthe natural log of total assets. This variable is used to control for size eect. Since large hospitals tend to be more capital intensive they are
12
Note that prior to 1991, actual depreciation expense was reimbursed by Medicare (Cotterill, 1991, p. 80). Following CPPS depreciation is reimbursed with a declining proportion of 10% a year. After 2001 depreciation will re¯ect only an accrual (Grimaldi, 1991, pp. 72±73).
13Although prior studies argue that tax-exempt debt is a cheaper source of ®nancing than a
expected to invest more in new PPE, but to be more adversely aected by CPPS. Therefore, hospital's size is anticipated to be positively related to capital expenditure levels, and this relationship may become negative subsequent to CPPS.
PPEAGEaccumulated depreciation/depreciation expense. This variable is used as a proxy for the age of the facility. We anticipate a negative relationship with CEXGFA because as facilities become older accumulated depreciated is expected to increase relative to annual depreciation expense. This relationship may be less negative following CPPS because as ®xed assets continue to age some replacement of these assets is necessary. Similar variables are used by Cotterill (1991, p. 82), Soderstrom (1993, p. 179), and Zeller et al. (1996, p. 170).
TURNOVERnet patient revenue/total assets. TURNOVER is used as a measure of asset eciency. We anticipate a negative relationship between TURNOVER and CEXGFA because ecient hospitals should be more successful at controlling capital expenditures, and this negative relationship may strengthen following CPPS. This variable is used by Deloitte and Touche (1996, p. 238) and a similar variable is used by Zeller et al. (1996, p. 170).
LBDFassets designated for capital acquisition/net ®xed assets (see AICPA, 1997, p. 131). The liquid board-designated fund ratio could be expected to increase over time because hospitals may require more internally generated funds to guard against uncertain and declining reimbursement for capital-related expenses. We anticipate a negative relationship between LBDF and CEXGFA, which is expected to become more negative following CPPS because hospitals may be inclined to accumulate these funds for future capital expenditures.
Four additional variables are included to examine other means of adapta-tion to CPPS by managers. These variables are useful to assess the relative level of operating expenses, which may help to determine adaptation of hospital managers to the post-CPPS environment. Missing values for these variables reduce the 10 227 available observations (hospital years) by more than 2300. The following two variables represent certain expenses that potentially sub-stitute for capital expenditures.
OTHREV(medical supplies and drugs + insurance + other expenses, in-cluding operating leases)/net patient revenue. There are several components in the numerator, and therefore, it is not clear if OTHREV will change following CPPS.14 To the extent that operating leases and rentals are material compo-nents of other expenses, we may expect a negative relationship between
14A separation of the numerator into components such as operating leases and rentals is
OTHREV and CEXGFA, which is anticipated to become more negative following CPPS.
SALREVsalary and bene®t expense/net patient revenue. To the extent that a reduction in capital expenditures induces more utilization of labor, we may expect this variable to increase and to have a negative impact on capital expenditures.
The next two variables re¯ect the impact of interest expense on capital expenditures.
INTREVinterest expense/net patient revenue. Due to the decline in mu-nicipal bond interest rates across the subperiods examined, we may expect this ratio to decline. Lower interest expense relative to net patient revenue com-bined with higher capital expenditures in the prior period could generate a negative relationship between this variable and CEXGFA. Subsequent to CPPS, the change in the impact is expected to be positive because low interest rates may induce management to ®nance capital expenditures with debt. This may result in increasing interest expense following CPPS and provide a positive change in the relation between INTREV and CEXGFA. INTREV may also be aected by reimbursement changes and revenue trends across the subperiods which could increase the denominator and along with decreasing capital ex-penditures further result in a positive change in the relation between INTREV and CEXGFA.
INTTDinterest expense/total liabilities. Since this variable represents an eective interest rate for hospitals, we expect a decline in INTTD across sub-periods. Prior to CPPS, this variable may have a negative impact on capital expenditures. For the reasons described for the previous variable, we expect a positive change in the estimated coecient following CPPS.
In addition, we use four control variables that re¯ect inpatient utilization, the proportion of revenue generated by Medicare, the competitive environment and outpatient service mix. Data for these variables are available for less than 3000 hospital-years. These or similar variables are used by Cotterill (1991), Eldenburg (1994), Eldenburg and Soderstrom (1996), Deloitte and Touche (1996).
OCCPCYinpatient days/(beds in service365). The inpatient occupancy percentage measures the utilization of inpatient capacity (see Renn, 1991, p. 25). We expected this variable to continue to decline across subperiods due to ongoing shifts toward outpatient activity and reduced average length of hos-pital stay. To the extent that this variable has an impact on cahos-pital expendi-tures, we anticipated a positive relationship between OCCPCY and CEXGFA, which may increase with the implementation of CPPS because higher occu-pancy rates may begin to result in higher levels of reimbursement for capital expenditures.
due to increasing population age. To the extent that this variable has an impact on capital expenditures, we anticipated a positive relationship between MEDICR and CEXGFA, which is expected to decrease following CPPS due to the unpredictability of update factors utilized in the new DRG-based reim-bursement formulas.
MATHtotal number of hospitals in market area. We expected this mea-sure of market concentration to decrease due to merger and acquisition activity within the industry. To the extent that this variable has an impact on capital expenditures, we anticipated a positive relationship between MATH and CEXGFA, which is expected to decrease following CPPS because consolida-tion may reduce duplicaconsolida-tion of capital expenditures.
OPPCTgross outpatient revenue/gross patient revenue. Due to the on-going shift toward outpatient utilization, we expected this variable to increase across subperiods. To the extent that outpatient services are less costly, we anticipate a negative relationship between OPPCT and CEXGFA, which is expected to become more negative following CPPS.
4.3. Testing the hypotheses
We tested the ®rst hypothesis using univariate analysis and more impor-tantly using multivariate analyses. For the second hypothesis we concentrated on four variables: OTHREV, SALREV, INTREV and INTTD. In addition to capital expenditures, these are the most useful measures in the database to examine other types of cost-eective behavior by hospital managers following CPPS. For testing the third hypothesis, we partitioned the sample into high-and low-cost hospitals, using the hospital-speci®c classi®cation provided to the authors by Medicare and an urban and non-urban partition. We used these partitions to examine changes in CEXGFA by hospital type and how these changes are aected by the explanatory variables across the two subperiods.
4.4. Dummy variables and estimating changes in slope coecients
to space limitations, only the eight-variable restricted and unrestricted models are presented in this section:
CEXGFAjb0b1DNFAjb2DTAjb3ROAjb4FBTAj
where the dummy variable BFAF 0 prior to CPPS and BFAF 1 sub-sequent to CPPS. Each interaction element consists of the value of the re-spective independent variable for each observation subsequent to CPPS, and zero otherwise. The coecientb0u re¯ects the dierence between the expected CEXGFA before and after CPPS, given that the other variables are included in the model.
Eq. (1) assumes that the restricted estimated coecients are the same prior and subsequent to the implementation of CPPS. For Eq. (2) the expected model prior to CPPS is
b0ub1uDNFAjb2uDTAjb3uROAjb4uFBTAjb5uLNTAj
b6uPPEAGEb7uTURNOVERb8uLBDFj;
and after CPPS the expected model is
5. Data and sample selection
This study empirically examines the impact of CPPS on hospital invest-ment levels for a large sample of non-proprietary hospitals from 1988 to 1996. The data source is the Merritt Systemâ
, a private credit and investment analysis database system, which contains audited annual ®nancial statement information, socio-economic information and supplementary operational statistics for 2048 hospitals, representing all 50 states and the District of Columbia.15 Since about 4100 general acute care hospitals participate in Medicare and ®le cost reports (Deloitte & Touche LLP, 1996, p. 1) our sample comprises roughly 50% of this population. This relatively large sample permits inferences to be comfortably made to the non-proprietary population. With some exceptions, the Merritt Systemâ
tends to exclude hospitals that are either speciality service providers, federal government controlled, or propri-etary.
Control and service codes from the AHA (1996) are identi®ed for all hos-pitals in the database and used to exclude any remaining proprietary and special service hospitals from the sample. As shown in Table 1, 21 hospitals represent a proprietary ownership type, while 78 hospitals have speciality service codes. The remaining 1949 hospitals used in the study represent non-proprietary, non-federal, general service types.
A total of 17 541 hospital-years are obtained for the nine-year period, 1988± 96, of which 7796 belong to the subperiod 1988±91 (prior to CPPS) and 9745 pertain to the subperiod 1992±96 (subsequent to CPPS). Missing data elements for CEXGFA and other variables reduce the number of hospital-years. To test the third hypothesis, we use a high-cost/low-cost partition of hospitals ob-tained by the authors from Medicare, which results in 4356 and 4634 useful hospital-years, respectively.
6. Results
6.1. Univariate tests for the ®rst hypothesis
Table 2 shows descriptive statistics, t-tests, and Wilcoxon-Zs for the de-pendent variable prior and subsequent to CPPS. Subsequent to CPPS, CEXGFA shows a statistically signi®cant decrease. For example, the mean (median) CEXGFA is reduced from 9.1% (7.7%) prior to CPPS to 7.8% (6.6%)
15The Merritt Systemâ(now Merritt Millennium) is the product of Van Kampen Management
subsequent to CPPS. Thus, the univariate analysis suggests a preliminary re-jection of the ®rst hypothesis.16
6.2. Univariate tests for the second hypothesis
Table 3 shows the descriptive statistics and statistical tests for the inde-pendent variables across subperiods, including the eight variables used in the ®rst regression (Panel A). Additional operating expense and cost of ®nancing variables (Panel B) and control variables (Panel C) are used in 12-variable and 16-variable regression models, respectively. As shown in Panel A, both DNFA and DTA signi®cantly increase across the subperiods. Also, it appears that hospitals in the sample operate at signi®cantly higher levels of pro®tability in the period subsequent to CPPS. For instance, the mean (median) of ROA
Table 1
Sample selection criteria and sample size
Panel A: Total number of hospitals and hospital-years
Hospitals Hospital-years
Number of hospitals in databasea 2,048 18,432
Less proprietary hospitals (21) (189) Less specialty service hospitals 78 702
Non-proprietary general service hospitals used in study
1,949 17,541
Panel B: Number of hospital-years prior and subsequent to capital prospective payment system (CPPS) with available data for the dependent variable, capital expenditures to gross ®xed assets (CEXGFA)
Prior to CPPS Subsequent to CPPS Total 1988±1991 1992±1996 1988±1996
Hospital-years 7,796 9,745 17,541 Less 1988 (The base year) (1,949) (1,949)
5,847 9,745 15,592 Less hospital-years with missing data 1;614 3;561 5;175
Net hospital-years used 4,233 6,184 10,417
a
Source: The Merritt Systemâ
, now Merritt Millennium (the following is that database's web site: www.merrittmillennium.net), Van Kampen Management Inc., Oakbrook Terrace, Illinois.
16In addition, we examined capital expenditures for a small sample of speciality hospitals not
signi®cantly increases from 3.6% (3.7%) during 1988±1991 to 4.4% (4.4%) during 1992±1996. This improving pro®tability may stem from a statistically signi®cant decline in the cost of ®nancing as discussed below. The proxy for average plant age (PPEAGE) increases signi®cantly from 7.54 (7.36) to 8.30 (8.08) across the two subperiods. We also found a signi®cant increase in TURNOVER, representing increased eciency in utilization of total assets. Finally, LBDF signi®cantly increases, which may represent accumulation of funds designated for future replacement of property and equipment to protect against anticipated declines in Medicare reimbursement. Since it is dicult to assign reasons for variable changes based on univariate tests, this discussion should be interpreted with caution.
The ®rst operating expense variable, OTHREV, has not changed signi®-cantly across the two subperiods, which may indicate that any potential in-creases in operating leases and rental expenses, relative to revenues, are oset by decreases in drugs and supplies. Panel B of Table 3 also shows that SAL-REV remains unchanged across subperiods, which indicates that managers may not substitute labor for reduced capital spending. This part of the uni-variate analyses provides an indication that hospitals have not signi®cantly changed their operating expenses subsequent to CPPS. In addition INTREV and INTTD signi®cantly decline subsequent to CPPS. One possible explana-tion for the declines in both variables may be the reducexplana-tion in interest rates during the period. For example, the average interest rate on municipal bonds decreased from 7.3% in the ®rst subperiod to 6.0% in the second subperiod (Standard and Poor's Bond Guide).
Panel C presents four control variables. OCCPCY signi®cantly decreases, which may re¯ect ongoing incentives to substitute outpatient for inpatient care.
Table 2
Summary statistics for the dependent variable, capital expendituresa to gross ®xed assets
(CEXGFA),bprior and subsequent to CPPS
Prior to CPPS Subsequent to CPPS
Sample size 4,233 6,184
First quartile 0.051 0.044
Mean 0.091 0.078
Median 0.077 0.066
Third quartile 0.114 0.098 Standard deviation 0.066 0.058
t-test 10.72
Wilcoxon-Z 11.77
a
Capital expenditures are measured by the change in net ®xed assets adjusted for depreciation. Construction in progress is included in capital expenditures and gross ®xed assets. See Table 1 for source.
b
CEXGFA (Capital expenditures/CIPI)t/gross ®xed assetstÿ1. *
MEDICR signi®cantly increases across subperiods, which seems to be con-sistent with demographic trends. MATH signi®cantly declines, which may represent increased merger and acquisition activities during the second sub-period. Finally, OPPCT signi®cantly increases probably as a result of increasing outpatient utilization.
In sum, beyond the reduction in capital expenditures, we do not ®nd major changes in operating expenses representing cost-eective behavior. Improving pro®tability seems to be due to declining interest rates and improvement in eciency (TURNOVER). Some items relevant for further analysis, such as
Table 3
Summary statistics for independent variablesa
Prior to CPPS Subsequent to CPPS Dierences
Mean Median S.D. Mean Median S.D. Meanb Medianc Panel A: Primary ®nancial statements variables
DNFAd 0.115 0.113 0.026 0.124 0.122 0.028 0.009 0.009
DTA 0.050 0.049 0.012 0.051 0.050 0.013 0.001 0.001
ROA 0.036 0.037 0.047 0.044 0.044 0.046 0.008 0.007
FBTA 0.450 0.454 0.195 0.456 0.470 0.201 0.006 0.016
LNTA 11.013 11.018 1.090 11.367 11.331 1.146 0.354 0.313
PPEAGE 7.539 7.363 1.998 8.300 8.084 2.412 0.761 0.721
TURNOVERd 0.888 0.853 0.258 0.904 0.869 0.277 0.016 0.016
LBDF 0.261 0.168 0.301 0.358 0.253 0.385 0.097 0.085
Panel B: Expense and cost of ®nancing variables
OTHREV 0.396 0.392 0.476 0.400 0.391 0.084 0.004 )0.001
SALREV 0.531 0.517 0.475 0.522 0.517 0.105 )0.009 0.000
INTREV 0.036 0.033 0.024 0.028 0.026 0.018 )0.008 )0.007
INTTD 0.074 0.073 0.047 0.065 0.061 0.168 )0.009 )0.012
Panel C: Control variables
OCCPCY 0.669 0.675 0.189 0.624 0.631 0.162 )0.045 )0.044
MEDICR 0.436 0.439 0.094 0.449 0.451 0.098 0.013 0.012
MATH 6.500 3.000 9.290 3.734 2.000 5.869 )2.766 )1.000
OPPCT 0.262 0.250 0.138 0.340 0.332 0.129 0.078 0.082 aIndependent variables: DNFA, DTA, ROA, LNTA, PPEAGE, TURNOVER, LBDF,
OTHREV, SALREV, INTREV, INTTD, OCCPCY, MEDICR, MATH, OPPCT (see text for explanations).
bSigni®cance based ont-test. cSigni®cance based on Wilcoxon-Z.
dThe distributions of DNFA and TURNOVER have extreme observations that in¯uence
de-scriptive and test statistics. The summary statistics presented re¯ect trimming of the two extreme upper and lower observations.
operating leases and rentals, are unavailable.17One accounting implication of these results is that increased accounting disclosure may be required so that alternative modes of coping with changes in regulation can be discerned.
6.3. Multivariate regression analyses for the ®rst and second hypotheses
Table 4 presents a matrix of Pearson correlations among CEXGFA and the 16 independent variables used in the OLS regression models. Most bivariate correlations between the independent variables are small, but at least nine pair-correlations are slightly higher (e.g., the correlation between LBDF and FBTA is 0.297). Further analyses discussed below indicates that multicollinearity is not a material problem in the following regression analyses.
Table 5 presents results for the OLS regressions with slope-dummies for the eight-variable and 12-variable models using CEXGFA as the dependent vari-able. For the eight-variable model, the restricted OLS regression is highly signi®cant (F-test of 243.5) and has moderate explanatory power (adjusted-R2
of 0.159). All estimated coecients for the independent variables, except TURNOVER, are statistically signi®cant.
The unrestricted model includes the eight hospital-speci®c variables, the dummy variable (BFAF) and the eight interaction variables. The adjustedR2
substantially increases to 0.266. The unrestricted regression tests the hypothesis of no change in coecients across the two subperiods. The reported Chow-Fis statistically signi®cant, suggesting rejection of the joint hypothesis that at least one coecient has changed.
Since no violation of OLS regression assumptions was observed for the eight-variable model, we used standard t-tests for testing the estimated indi-vidual coecients. The Durbin±Watson and White-v2 tests indicated,
re-spectively, that assumptions of uncorrelated residuals and homoscedasticity of the residuals are not violated.18 The insigni®cant estimated coecient for
17Furthermore, hospital managers may reduce operating leverage in response to increased risk
associated with CPPS. Although managers have incentives to increase the proportion of variable costs relative to ®xed cost, relevant data are not available.
18
Pearson correlations among the variables used in the regression analysesa;
CEX-GFA
DNFA DTA ROA FBTA LNTA PPE-AGE ROA 0.113 0.014 )0.132 FBTA )0.069 )0.002 )0.016 0.289
LBDF )0.035 0.019 )0.304 0.246 0.297 0.146 0.094 )0.006 OTHREV )0.002 0.002 )0.003 0.002 )0.013 )0.026 )0.014 0.018 )0.013
aCEXGFA, DNFA, DTA, ROA, FBTA, LNTA, PPEAGE, TURNOVER, LBDF, OTHREV, SALREV, INTREV, INTTD, OCCPCY, MEDICR,
MATH, and OPPCT (see text for explanations).
*All correlations at or above 0.019 (0.028) are statistically signi®cant at a probability level of 0.05 (0.01) two-tailed test.
Results of OLS regression estimates of capital expenditure (CEXGFA) on the 8 and 12 determinants
Variablea;b Expected
sign
Intercept 0.2827 24.47 0.2704 15.61 0.3518 21.37 0.4182 16.08
BFAF ) )0.0020 )0.09 )0.1434 )4.40
BFAFSALREV ) 0.0048 0.28
F-value 243.5 219.0 129.2 128.3
AdjustedR2 0.159 0.266 0.163 0.288
Durbin±Watson 1.974 1.970 1.964 1.976
Whitev2for 33.011 49.1 109.9 76.9
Heteroscedasticity (p-value)
(0.8879) (0.9998) (0.0757) (0.9999)
Chow-F 1491.7 1466.8
aBFAF is a dummy variable where BFAF
0 prior to CPPS and BFAF 1 subsequent to CPPS; DNFA, DTA, ROA, FBTA, LNTA, PPEAGE, TURNOVER, LBDF, OTHREV, SALREV, INTREV, and INTTD (see text for explanations).
bTests for the eects of collinear variables which include in¯uential observations, eigenvalues, outliers, and variance in¯ation factors indicate no
multicollinearity. Furthermore, no violation of normality of residuals is detected.
*
Signi®cant at a probability of less than 0.01 (two-tailed test).
**
Signi®cant at a probability of less thafn 0.05 (two-tailed test).
R.
Barniv
et
al.
/
Journal
of
Accountin
g
and
Public
Policy
19
(2000)
9±40
BFAF suggests that the intercept remains unchanged before and after CPPS, given that the other eight hospital characteristics are included and controlled for in the model. Signi®cant estimated coecients for the interactions of DNFA, DTA, FBTA, LNTA, PPEAGE TURNOVER, and LBDF with BFAF indicate changes in the slopes for these variable across the two sub-periods.
The statistically signi®cant estimated coecient for the interaction of DNFA with BFAF suggests that the positive relationship between DNFA and CEXGFA is strengthened subsequent to CPPS. The negative estimated coef-®cient for the product of DTA and BFAF indicates the relationship between DTA and CEXGFA becomes more negative during the second period. The reduction in growth rate of non-®xed assets may explain this ®nding. The es-timated coecient for the product of FBTA and BFAF is negative and sta-tistically signi®cant, given that the impact of other variables are controlled for, indicating that equity ®nancing for capital expenditures has a less positive impact on CEXGFA following CPPS.
The negative statistically signi®cant estimated coecient for the product of LNTA and BFAF suggests that larger hospitals tend to reduce capital ex-penditures more than smaller hospitals following CPPS. The estimated in-teraction of PPEAGE and BFAF is positive and statistically signi®cant, as expected, indicating that the relationship with CEXGFA is less negative during the second subperiod. This ®nding may indicate that managers began to replace some older assets. Finally, the negative estimated coecients for the products of TURNOVER and LBDF with BFAF are consistent with expectations indicating that hospitals with relatively higher asset eciency and greater liquidity further reduce capital expenditures. Thus, the impacts of depreciation, size, asset age, asset eciency and liquidity designated for future capital expenditures have changed in the expected directions following CPPS.
The results for the 12-variable model are also presented in Table 5 for the restricted and unrestricted regressions. The adjusted-R2 increases from 0.163 for the restricted model to 0.288 for the unrestricted model. The Durbin± Watson statistics and the White-v2 indicate no violations of independent
The negative estimated products of OTHREV and BFAF suggests that in-creasing other expenses reduces capital expenditures subsequent to CPPS. This result is according to expectation and may be attributable to increasing oper-ating leases and rental expenses, which are not separately available in the database. SALREV has no impact on CEXGFA across both subperiods. Fi-nally, the positive estimated products of INTREV and INTTD with BFAF suggest a potential change toward increasing interest expenses which may re-sult from increased debt due to less expensive ®nancing, this may induce new capital expenditures subsequent to CPPS.19 Thus, given declining interest rates, the results may indicate that hospitals are borrowing money to further ®nance capital expenditures.
The results for the 16-variable model are presented in Table 6 for the re-stricted and unrere-stricted regressions. The adjusted-R2increases from 0.324 for the restricted model to 0.356 for the unrestricted model. The Durbin±Watson statistics and the White-v2indicate no violations of independent residuals and
homoscedasticity, respectively, for only the restricted model. Similar to the results for the eight-variable and 12-variable models, other tests for multicol-linearity and normality of residuals indicate no violation of these assumptions for both the restricted and unrestricted models. Due to missing values only about 2700 hospital-years remain in the analysis.
The Chow-F test is statistically signi®cant, therefore we rejected the joint hypothesis and found that at least one coecient has changed. Since the
White-v2 test for heteroscedasticity is statistically signi®cant for the unrestricted
model, we used t-statistics based on White (1980) standard errors from the heteroscedasticity consistent-covariance-matrix. The estimated coecient for BFAF is negative and signi®cant, which indicates that CEXGFA signi®cantly decreases subsequent to CPPS, once the impact of the 16 hospital character-istics is controlled for in the model. This result, along with the ®ndings from the 12-variable model, provide further support for rejecting the ®rst null hypoth-esis. While the results resemble those for the eight- and 12-variable models, all estimated coecients for the four control variables are not statistically sig-ni®cant. The estimated coecients for the products of PPEAGE, OTHREV, SALREV and INTTD with BFAF indicate that the impacts of these variables have changed across the two subperiods. In both the 12- and 16-variable models the estimated coecients for INTTD are signi®cantly negative prior to CPPS and the change in the estimated coecients are signi®cantly positive following CPPS.
In sum, the results indicate that capital expenditures are aected by man-agers changing hospital-speci®c characteristics, including OTHREV and
19Additional univariate analyses of revenues and expenses suggest that total interest expense
Table 6
Results of OLS regression estimates of capital expenditure (CEXGFA) on the sixteen determinants
Variablea;b Expected
Intercept 0.3397 9.40 0.4103 5.59
BFAF ) )0.1855 )1.99
DNFA + 1.6127 24.94 1.5877 9.01
DTA ? )4.2375 )26.12 )4.7093 )12.66
ROA + )0.0366 )0.93 0.0092 0.14
FBTA ? 0.0170 1.35 0.0150 0.62 LNTA + )0.0046 )2.19 )0.0051 )1.49
BFAFINTREV + 0.4448 1.64
BFAFINTTD + 0.2657 2.11
BFAFOCCPCY + 0.0006 1.41
BFAFMEDICR ) 0.0006 1.76
BFAFMATH ) )0.0002 )0.64
BFAFOPPCT ) 0.0001 1.23
Sample size 2,685 2,685
F-value 81.40 45.9
AdjustedR2 0.324 0.356
Durbin±Watson 1.954 1.970
Whitev2for 94.9 456.7
Heteroscedasticity (p-value) (0.9990) (0.0000)
SALREV, subsequent to CPPS. However, the control variables seem to have no impact on capital expenditures across the two subperiods.20
6.4. Univariate and multivariate tests for the third hypothesis
We tested the third hypothesis using Medicare's high- and low-cost par-tition. Panel A of Table 7 shows that capital expenditures (CEXGFA) decline for both subsamples. It appears from the univariate analyses that the decline across subperiods is more signi®cant for high-cost than low-cost hospitals.21 For example, the mean (median) CEXGFA for high-cost hospitals declines from 0.085 (0.068) prior to CPPS to 0.070 (0.060) subsequent to CPPS with a t-test (Wilcoxon) of 8.18 (7.28). Similarly, the mean (median) CEXGFA for low-cost hospitals declines from 0.094 (0.081) prior to CPPS to 0.083 (0.071) subsequent to CPPS with a t-test (Wilcoxon) of 5.63 (7.04). These prelimi-nary results somewhat support rejecting the third null hypothesis. However, low-cost hospitals reduce their capital expenditures, which seems to not support the alternative hypothesis that these hospitals will increase capital expenditures.
We further examined how high-cost urban and low-cost non-urban hospi-tals modify their capital expenditures to the post-CPPS environment. First, we partitioned the sample into urban and non-urban hospitals using Standard Metropolitan Statistical Area codes. The results (not reported) suggest that
Table 6 (Continued)
a
BFAF is a dummy variable where BFAF0 prior to CPPS and BFAF1 subsequent to CPPS, DNFA, DTA, ROA, FBTA, LNTA, PPEAGE, TURNOVER, LBDF, OTHREV, SALREV, INTREV, INTTD, OCCPCY, MEDICR, MATH, and OPPCT (see text for explanations).
b
Tests for the eects of collinear variables which include in¯uential observations, eigenvalues, outliers, and variance in¯ation factors indicate no multicollinearity. Furthermore, no violation of normality of residuals is detected.
cThe
t-tests are calculated using White's (1980) consistency covariance matrix, because thev2test
rejects the homoscedasticity assumption.
*Signi®cant at a probability of less than 0.01 (two-tailed test). **Signi®cant at a probability of less than 0.05 (two-tailed test).
20
In further analyses, we examined possible state eects and found no signi®cant coecient for any of the 50 state dummy variables. In addition, the estimated coecients for the independent variables, with the state dummies included, are very similar to those without the state dummies reported in Tables 5 and 6.
21A potential puzzle in the two-tier classi®cation of hospitals provided to the authors by
capital expenditures signi®cantly decline for both subgroups following CPPS.22
Panel B of Table 7 provides a comparison between the two subperiods for high-cost urban hospitals and low-cost non-urban hospitals. The results sug-gest no statistically signi®cant decline of capital expenditures across subperiods for low-cost non-urban hospitals, while high-cost urban hospitals signi®cantly reduce their capital expenditures. For example, the mean (median) CEXGFA
Table 7
Summary statistics of capital expendituresato gross ®xed assets (CEXGFA) for dierent types of
hospitals prior and subsequent to CPPS
Prior to CPPS
Subsequent to CPPS
Prior to CPPS
Subsequent to CPPS
Panel A:High-cost and low-cost
High-cost Low-cost
Sample size 1,800 2,556 1,886 2,748 First quartile 0.046 0.040 0.055 0.047
Mean 0.085 0.070 0.094 0.083
Median 0.068 0.060 0.081 0.071 Third quartile 0.104 0.090 0.118 0.105 Standard deviation 0.065 0.053 0.060 0.062
t-test 8.18 5.63
Wilcoxon-Z 7.28 7.04
Panel B:High-cost urban and low-cost non-urbanb
High-cost urban Low-cost non-urban
Sample size 1221 1714 600 937 First quartile 0.046 0.040 0.048 0.045
Mean 0.084 0.070 0.088 0.087
Median 0.070 0.061 0.071 0.071 Third quartile 0.104 0.091 0.111 0.109 Standard deviation 0.062 0.053 0.070 0.064
t-test 6.61 0.16
Wilcoxon-Z 5.95 0.32
aCapital expenditures are measured by the change in net ®xed assets adjusted for depreciation.
Construction in progress is included in capital expenditures and gross ®xed assets. CEXGFA (Capital expenditurest/CIPI)t/gross ®xed assetstÿ1.
bThe high-cost, low-cost partition is provided by Medicare. The partition for urban and non-urban
hospitals is obtained using Standard Metropolitan Statistical Area Codes.
*Signi®cant at a probability of less than 0.01 (one-tailed test).
22For example, the mean (median) CEXGFA for non-urban hospitals decreases from 0.088
for low-cost non-urban hospitals remains unchanged at 0.088 (0.071) prior to CPPS and 0.087 (0.071) subsequent to CPPS with at-test (Wilcoxon) of 0.16 (0.32). Whereas, the mean (median) CEXGFA for high-cost urban hospitals declines from 0.084 (0.070) prior to CPPS to 0.070 (0.061) subsequent to CPPS with at-test (Wilcoxon) of 6.61 (5.95). These univariate results suggest that we cannot reject the third null hypothesis for low-cost non-urban hospitals, but somewhat indicate rejection of the null hypothesis for high-cost urban hospi-tals. The univariate tests provide only preliminary indications and multivariate analyses are needed to provide further reasons for variable impacts on CEXGFA.
OLS regression results for comparing the impacts of hospital characteristics on capital expenditures across the two subperiods for high-cost vs low-cost hospitals are presented in Table 8.23 The unrestricted regressions for both types of hospitals are highly signi®cant and have reasonable explanatory powers (adjusted-R2s of 0.397 and 0.314 for the high-cost and the low-cost hospitals, respectively). Most estimated coecients for the independent vari-ables are statistically signi®cant and the signs of the estimated coecients tend to be as expected. Subsequent to CPPS, the change in estimated coecients for LNTA and TURNOVER are negative and statistically signi®cant for low-cost hospitals, while they are not signi®cant for high-cost hospitals. Contrary to the aggregate ®nding, for the low-cost hospitals this result indicates that larger and more ecient hospitals within this group tend to reduce their capital expen-ditures subsequent to CPPS. In addition, the changes in estimated coecients are statistically signi®cant and negative for DNFA and FBTA while the changes are positive for DTA and PPEAGE.
An interesting result is the signi®cant negative estimated coecient for BFAF for high-cost hospitals versus the signi®cant positive estimated coef-®cient for BFAF for low-cost hospitals. These signi®cant estimated coe-cients for BFAF suggest that the intercept signi®cantly declines subsequent to CPPS for high-cost hospitals and the intercept signi®cantly increases subse-quent to CPPS for low-cost hospitals, given that the impacts of the other variables are controlled for in the model. These ®ndings provide more insight into the impact of CPPS on capital expenditures for high-cost vs low-cost hospitals, which is not perceived by the univariate analyses presented in Table 7. Thus based on this multivariate analysis, it seems that high-cost hospitals reduce their relative capital expenditures, while low-cost hospitals
23
increase their relative capital expenditures following CPPS, once other factors are controlled for. This ®nding supports rejecting the third null hypothesis and is consistent with Medicare's expectations (Federal Register, 1991, pp. 43427±43428).
The right-hand side column in Table 8 presentst-tests for the dierences in estimated coecients between high- and low-cost hospitals. The dierences in
Table 8
Results of OLS regression estimates of capital expenditure (CEXGFA) on the eight determinants for the high-cost and the low-cost hospitals
Variablea;b High-cost Low-cost t-test for
dierencesc
Intercept 0.2907 11.69 0.2235 8.03 1.80
BFAF )0.0795 )2.38 0.1411 3.94 )4.50
BFAF*ROA 0.0274 0.52 )0.0049 )0.10 0.44 BFAF*FBTA )0.0288 )2.33 )0.0538 )3.92 1.35
BFAF*LNTA 0.0009 0.38 )0.0104 )4.44 3.44
BFAF*PPEAGE 0.0047 4.15 0.0028 2.70 1.22
BFAF*TURNOVER )0.0049 )0.39 )0.0419 )3.90 2.25
BFAF*LBDF 0.0499 6.27 0.0122 1.64 3.47
Sample size 4,300 4,539
F-value 167.63 123.2
AdjustedR2 0.397 0.314
Durbin±Watson 1.996 1.957
Whitev2for 45.1 53.6
Heteroscedasticity (p-value) (0.9999) (0.9989)
Chow-F 128.6 69.6
aBFAF is a dummy variable where BFAF
0 prior to CPPS and BFAF1 subsequent to CPPS, DNFA, DTA, ROA, FBTA, LNTA, PPEAGE, TURNOVER, and LBDF (see text for explana-tions).
bTests for the eects of collinear variables which include in¯uential observations, eigenvalues,
outliers, and variance in¯ation factors indicate no multicollinearity. Furthermore, no violation of normality of residuals is detected.
c
t-test for dierence in estimated coecients for high-cost vs low-cost hospitals.
*
Signi®cant at a probability of less than 0.01 (two-tailed test).
**
the estimated coecients for BFAF are statistically signi®cant. Prior to CPPS the dierences in the estimated coecients for DNFA, DTA, ROA, PPEAGE and LBDF are statistically signi®cant. Subsequent to CPPS, the dierences in the changes for the estimated coecients of DNFA, DTA, LNTA, TURN-OVER, and LBDF are statistically signi®cant. Thus, following CPPS the positive (negative) eect of DNFA (DTA) on CEXGFA is decreasing more rapidly for high-cost hospitals than for the low-cost hospitals. Furthermore, following CPPS, LNTA and TURNOVER more negatively impact CEXGFA while LBDF less negatively impacts CEXGFA, for low-cost hospitals relative to high-cost hospitals.
7. Summary and conclusions
Our study explicitly examined the impact of the CPPS, implemented by Medicare in 1991 (e.g., Cotterill, 1991, p. 79) on capital expenditures and overall cost-eective behavior of non-proprietary hospitals. Prior articles (e.g., Luggiero, 1990, p. 3; Burke, 1991, pp. 34±35) do not empirically examine these issues, but express concerns about unfavorable consequences of the new reg-ulation. We hypothesized that hospitals reduce capital expenditures subsequent to CPPS and examine if the post-CPPS environment aects managers' cost-eective behavior. In addition, we tested if high-cost hospitals reduce capital expenditures while low-cost hospitals increase capital expenditures.
As noted earlier, using a national sample of hospitals from 1988 to 1996, we found a signi®cant decline in capital expenditures subsequent to CPPS. While a concurrent reduction in interest rates in the second subperiod may induce an increase in capital expenditures, we found that capital expenditures decrease. This further suggests that CPPS may have a major impact on the declining capital expenditures. Although an interest rate reduction underlies the im-provement in pro®tability, preliminary results indicates that hospitals have not changed salaries or other aggregate operating expenses relative to revenues during the second subperiod. Thus, univariate analysis indicate that CPPS may not in¯uence other cost behavior by management. It appears that CPPS may stimulate cost-eective behavior in terms of reductions in capital expenditures coupled with no change in relative operating expenses.
The empirical results indicate that hospital managers change several ®nan-cial and operational ratios, which can explain some aspects of the reduction of capital expenditures following CPPS. However, hospital-speci®c control vari-ables re¯ecting service mix and competition seem to have no impact on capital expenditures. Results also suggest that perhaps operating leases and rentals potentially substitute for these expenditures.
should be aware of the potentially negative long-term impact on the public health, if capital expenditures continue to decrease. Finally, an anticipated redistribution of capital payments from high-cost to low-cost hospitals may provide enough incentive for low-cost hospitals to increase capital expendi-tures. The results indicate that low-cost and low-cost non-urban hospitals in-crease their capital expenditures relative to high-cost and high-cost urban hospitals, which reduce capital expenditures following CPPS, given that the impact of other variables is controlled for.
Managerial accounting implications of our study include the eect of the regulation on capital budgeting decisions by hospital management. CPPS may require hospital managers to assume increasing risk for capital expenditures and to explicitly incorporate this risk in the discount rate used for capital budgeting purposes. Finally, it appears that greater accounting disclosure may be necessary to provide external users and regulators with relevant information for ascertaining alternative modes of coping with the new regulation.
Acknowledgements
We acknowledge the helpful comments of anonymous reviewers. We thank Van Kampen Management Inc. for providing the data. The data is from their Merritt systemâ
now Merritt Millennium (the following is that database's web site: www.merritt millennium.net).
References
American Institute of Certi®ed Public Accountants (AICPA), 1997. Health Care Organizations: AICPA Audit and Accounting Guide. American Institute of Certi®ed Public Accountants, New York.
American Hospital Association, 1996. Guide to the Health Care Field. American Hospital Association, Chicago.
Anderson, H.J., 1992. Survey: equipment budgets up; use in outpatient areas growing. Hospitals 66 (18), 38±44.
Blanchard, G., Chow, C.W., Noreen, E., 1986. Information asymmetry, incentive schemes and information biasing: The case of hospital budgeting under rate regulation. The Accounting Review 61 (1), 1±15.
Boles, K.E., 1986. Implications of the method of capital cost payment on the weighted average cost of capital. Health Services Research 21 (2), 189±212.
Borden, J.P., 1988. An assessment of the impact of diagnosis-related group (DRG)-based reimbursement on the technical eciency of New Jersey hospitals using data envelopment analysis. Journal of Accounting and Public Policy 7 (2), 77±96.
Carey, K., 1994. Cost allocation patterns between hospital inpatient and outpatient departments. Health Services Research 29 (3), 275±292.
Carter, R.B., Massa, L.J., Power, M.L., 1997. An examination of the eciency of proprietary hospital versus non-proprietary hospital ownership structures. Journal of Accounting and Public Policy 16 (1), 63±87.
Cleverley, W.O., Felkner, J.G., 1984. The association of capital budgeting techniques with hospital ®nancial performance. Health Care Management Review 9 (3), 45±55.
Cotterill, P.G. 1991. Prospective payment for medicare hospital capital: implications of the research. Health Care Financing Review (Annual Supplement) 79±86.
Deloitte and Touche, 1996. The Comparative Performance of US Hospitals: The Source Book. Deloitte and Touche LLP, Chicago.
Eldenburg, L., 1994. The use of information in total cost management. The Accounting Review 69 (1), 96±121.
Eldenburg, L., Kallapur, S., 1997. Changes in hospital service mix and cost allocations in response to changes in medicare reimbursement schemes. Journal of Accounting and Economics 23 (1), 31±51.
Eldenburg, L., Soderstrom, N., 1996. Accounting system management by hospitals operating in a changing regulatory environment. The Accounting Review 71 (1), 23±42.
Federal Register, 1991. Rules and Regulations 56 (170), 43428±43517. Federal Register, 1996. Rules and Regulations 61 (170), 46196±46203.
Feder, J., Hadley, J., Zuckerman, S., 1987. How did medicare's prospective payment system aect hospitals? The New England Journal of Medicine 317 (14), 867±873.
Folland, S., Kleiman, R., 1990. The eect of prospective payment under DRGs on the market value of hospitals. Quarterly Review of Economics and Business 30 (2), 50±68.
Friedman, B., Shortell, S., 1988. The ®nancial performance of selected investor-owned and not-for-pro®t system hospitals before and after medicare prospective payment. Health Services Research 23 (2), 237±267.
Grimaldi, P., 1991. Capital PPS: trekking through the labryinth. Healthcare Financial Manage-ment 45 (11), 72±87.
Guterman, S., Dobson, A., 1986. Impact of the Medicare prospective payment system for hospitals. Health Care Financing Review 7 (3), 97±114.
Guterman, S., Eggers, P.W., Riley, G., Greene, T.F., Terrell, S.A., 1988. The ®rst 3 years of medicare prospective payment: an overview. Health Care Financing Review 9 (3), 67±77. Johnston, J., 1984. Econometrics Methods, Third ed. McGraw-Hill, New York.
Kauer, R.T., 1995. The eect of ®xed payment on hospital costs. Journal of Health, Politics Policy and Law 20 (2), 303±327.
Kennedy, P., 1992. Guide to Econometrics. MIT Press, Cambridge.
Lambert, R.A., Larcker, D.F., 1995. The prospective payment system, hospital eciency, and compensation contracts for senior-level hospital administrators. Journal of Accounting and Public Policy 14 (1), 1±31.
Luggiero, G., 1990. Capital payments and PPS: Impact on hospital ®nance. Trustee 43 (6), 3±19. Lynch, L.J., 1998. The eect of Medicare reimbursement on hospital balance sheet composition.
Working paper, The University of North Carolina, Chapel Hill.
Mensah, Y.M., Li, S-H., 1993. Measuring production eciency in a not-for-pro®t setting: An extension. The Accounting Review 68 (1), 66±88.
Mensah, Y.M., Considine, J.M., Oakes, L., 1994a. Statutory insolvency regulations and earnings management in the prepaid health-care industry. The Accounting Review 69 (1), 70±95.
Mensah, Y.M., Chiang, C.C., 1996. The informativeness of operating statements for commercial nonpro®t enterprises. Journal of Accounting and Public Policy 15 (3), 219±246.
Nation's CFOs foresee cautious purchase practices, 1992. Health Industry Today 55 (12), 10. Neter, J., Wasserman, W., Kunter, M.H., 1986. Applied Statistical Models. Irwin, Homewood. Pallarito, K., 1990. Shaping hospitals' capital spending decisions. Modern Healthcare 20 (15),
33±48.
Pallarito, K., 1991. Analysis pegs capital winners, losers. Modern Healthcare 21 (15), 4. Payne, C.T., 1995. Strategic Capital Planning for Healthcare Organizations. Probus, Chicago. Renn, S.C., 1991. Executives need to be aware of turns of fast-approaching new medicare capital
rules. Modern Healthcare 21 (16), 25.
Soderstrom, N.S., 1993. Hospital behavior under medicare incentives. Journal of Accounting and Public Policy 12 (2), 155±185.
United States Statutes at Large. 1983. Public-Law 98-21, Section 601: Medicare Payments for Inpatient Hospital Services on the Basis of Prospective Rates. Washington DC, United States Government Printing Oce, pp. 149±163.
Wedig, G.J., Hassan, M., Sloan, F.A., 1989. Hospital investment decisions and the cost of capital. Journal of Business 62 (4), 517±537.
Wedig, G.J., Sloan, F.A., Hassan, M., Morrisey, M.A., 1988. Capital structure, ownership, and capital payment policy: The case of hospitals. The Journal of Finance 63 (1), 21±40. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for
heteroskedasticity. Econometrica 48 (5), 817±838.