Printer Friendly

Determinants of hospital tax-exempt debt yields: corrections for selection and simultaneous equation bias.

The cost of capital for hospitals is a topic of continuing interest as Medicare's new capital payment policy is implemented. This study examines the determinants of tax-exempt revenue bond yields, the primary source of long-term capital for hospitals. Two important methodological issues are addressed. A probit analysis estimates the probability that a hospital or system will be observed in the tax-exempt market. A selection-corrected two-stage least squares analysis allows for the simultaneous determination of bond yield and bond size. The study is based on a sample of hospitals that issued tax-exempt revenue bonds in 1982-1984, the years immediately surrounding implementation of Medicare's new payment system based on diagnosis-related groups, and an equal number of hospitals not in the market during the study period. Results suggest that hospital systems and hospitals with high occupancy rates are most likely to enter the tax-exempt revenue bond market. The yield equation suggests that hospital-specific variables may not be good predictors of the cost of capital once estimates are corrected for selection.

Capital finance is an area of research inquiry of considerable interest to managers and policymakers in the hospital industry. Investment decisions in hospitals determine the industry's capacity to produce services and thus their availability to various subgroups of the population. Today's investment and financing decisions may affect future access to care. Differential access to capital may characterize the hospital industry in the coming years and may determine which hospitals survive. Access to capital may determine ownership patterns and the availability of services and facilities. The result may be serious inequities that jeopardize access to care for specific patient groups or geographic locations. Issues related to capital investment are particularly compelling now as Medicare implements its new capital payment policy.

The cost of capital for hospitals, as for any economic enterprise, is the weighted average of the cost of equity and the cost of debt (Copeland and Weston 1983). A limited number of studies have addressed issues related to the cost of capital for hospitals. Three studies (Cleverley and Rosegay 1982; Austen, Corman, and LiCalzi 1986; Grossman et al. 1990) examined the determinants of tax-exempt bond yields. The first two of these used data from bonds issued before the time of Medicare's prospective payment system (PPS); the third used more recent data. Sloan, Morrisey, and Valvona (1987) used 1979-1983 data to estimate the cost of debt utilizing all forms of debt. Most recently, Sloan et al. (1988) produced estimates of the cost of both debt and equity capital for hospitals.

This study focuses on the cost of tax-exempt debt. For most hospitals, the primary source of long-term capital and, therefore, the most important component of the cost of capital, is tax-exempt debt. Tax-exempt financing represents 60-70 percent of hospital capital financing (Cohodes and Kinkead 1984).


The analysis presented in this article addresses two potentially important methodological issues -- selection bias and simultaneous equation bias -- that have received little attention in previous hospital studies. To understand these issues, it is necessary to consider the bond-issuing process for a typical, tax-exempt revenue bond.

Most hospital tax-exempt revenue bonds are negotiated financings. Unlike competitive bond issues, negotiated financings allow the issuer to obtain a preliminary estimate of the interest rate (yield) to be paid prior to the actual commitment to sell the bonds (Shields 1983). Based on this preliminary assessment, the hospital may decide to enter the bond market, wait for interest rates to decline, or reduce the principal amount to be borrowed (bond size). Thus, for those hospitals that choose to enter the market, we observe actual values for interest rates and principal. However, these observed values are preceded by a market-testing process that determines the hospitals that enter the market and the amount that they borrow.

The bond-issuing process raises two methodological issues. The first issue relates to the use of bond size as an independent variable in the yield equation. Empirical tests of bond yield determinants have produced conflicting results regarding the relationship between interest rates and bond size. Kessel (1971) found an inverse relationship between size and yield that was attributed to economies of scale in underwriting bond issues. Some have argued that an inverse relationship exists because larger issues are easier to trade in secondary markets, thus reducing marketability risk (Lev 1974). Hendershott and Kidwell (1978) found a positive coefficient for their size variable, which may suggest that the risk of default increases with issue size.

Cleverley and Rosegay (1982) concluded that the insignificant coefficient for their bond size variable suggested that most hospital issues are too small to have a secondary market that could reduce marketability, risk. Austen, Corman, and Licalzi (1986) concluded that size has a positive but diminishing effect on interest costs because the difficulties of marketing large issues outweigh any economies of scale. All of these studies have treated bond size as an exogenous variable.

However, if we assume that a hospital's demand for loanable funds is elastic and that a hospital can adjust the size of a bond in response to initial assessments of interest rates, then the size of the bond cannot be considered exogenous. The endogenous variable that is used as a regressor will be contemporaneously correlated with the disturbance term in the equation, and this will result in biased and inconsistent estimates of ordinary least squares (OLS) coefficients (Kennedy 1984). Grossman et al. (1990) made a similar argument in their analysis of hospital tax-exempt yields.

The second methodological issue relates to the potential for selection bias. Yield and bond size are observed only for those hospitals that actually enter the bond market. If there are variables that influence the probability of entering the bond market and influence the observed yield, and if some of these variables are omitted from the structural equation, then estimates of the regression coefficients in the yield equation will be biased (Heckman 1977). This issue has not been addressed in previous hospital capital studies.

The potential problem of simultaneity can be addressed by using two-stage least squares to produce an instrumental variable that can serve as a substitute for bond size in the yield equation (Kmenta 1986). Selection bias can be addressed by first estimating a probit equation for the probability of observing a hospital in the tax-exempt market. A selection term from this model is used as an explanatory variable in the two-stage least squares equations (Lee, Maddala, and Trost 1980; Maddala 1983).

This study employs the commonly used assumption that the price of tax-exempt debt (the yield to maturity) is a function of factors both internal and external to the debt-issuing hospital (Fisher 1959; Kessel 1971; Tanner 1975; Hendershott and Kidwell 1978). Specifically, we assume [P.sub.i] = f (X, B, F), where [P.sub.i] = the yield to maturity on the date of issue for the ith bond, X = characteristics of the bond-issuing organization that affect the risk of investment in the ith bond, B = bond instrument characteristics that affect the risk of investment in the ith bond, and F = financial market characteristics that affect the risk of investment in tax-exempt debt in general. The study utilizes data from hospital tax-exempt issues in the years immediately surrounding implementation of Medicare's prospective payment system, 1982-1984.

Table 1 presents operational definitions for the study's variables. Independent variables included in the yield equation are:

1. Hospital-specific variables that relate to the risk of default. Descriptive statistics of bond ratings and empirical tests of their determinants suggest that the following hospital characteristics may be associated with the risk of default: ownership, teaching status, location, rate regulation, system affiliation, occupancy rate, operating margin, debt-to-assets ratio, and percent of revenue from Medicare and Medicaid (Cleverley and Nutt 1984; Cohodes and Kinkead 1984; Sloan, Morrisey, and Valvona 1987).

2. Financial market variables that relate to purchasing power risk and marketability risk: outstanding state debt per capita, Visible Bond Supply, state tax rate, and Treasury rate. The first two of these are measures of bond supply, at the state and national levels, respectively. One unique characteristic of tax-exempt securities is the market segmentation on the demand side. Tax-exempt bonds traditionally have been purchased by commercial banks, insurance companies, and high-income individuals (Shields 1983). Individuals who purchase tax-exempt bonds have been viewed as the residual investors in this market. It has been argued that if institutional demand for tax-exempt issues is low, then higher yields must be offered to attract more individual investors (Van Horne 1984). However, empirical tests of the effects of bond supply or institutional demand have produced mixed results. Some results suggest a positive relationship between bond supply and bond yields; others suggest no significant relationship between these two variables (Hendershott and Kidwell 1978; Campbell 1980; Buser and Hess 1985).


The marginal state income tax rate affects investors in states with dual exemption. In some states interest earned on municipal securities issued within the state is exempt from both federal and state income taxes. The tax benefit of investing in in-state tax-exempt securities increases with the tax rate. The Treasury rate variable serves as a control for general market conditions at the time of sale.

3. Debt instrument characteristics that relate to default risk and interest rate risk: term to maturity (duration), refunding status, and credit enhancement such as bond insurance or a bank letter of credit.

The size of the bond is presumed to be a function of the factors that determine hospital investment in physical capital. The neoclassical investment model presumes that investment is a function of factor prices and output prices (Jorgenson 1963). In this study these would be interest rates (the endogenous variable), average employee wage, and net charge per patient day. Modifications of that model have included measures of capacity utilization and profits (Eisner 1962; Kuh 1963). In this study these factors are represented by occupancy rate and operating margin.

Variables unique to hospital investment behavior include payer-related variables such as the Medicare and Medicaid share of revenue and state rate regulation. Payer methods that reimburse on a cost basis for interest and depreciation may create perverse incentives to invest in physical capital and to finance those investments with debt (Sloan et al 1988; Wedig et al. 1988; Wedig, Hassan, and Sloan 1989). Other hospital-specific variables in this equation include hospital size and average age of assets. This system of equations is estimated using data from hospitals in the tax-exempt market. Values for price and quantity are those that can be observed for the in-market group. The probit equation estimates the probability that a hospital will be observed in the market. This equation is estimated using the in-market hospitals from the yield and size equations and an equal number of hospitals not in the market.

Explanatory variables in the probit equation include those that can be observed for both the in-market and out-of-market hospitals and that affect both yield and bond size. In essence, the probit equation is a reduced-form equation for yield and size. Estimates from the probit equation can be used to construct a selection term that becomes an independent variable in the reduced form and structural equation in the two-stage least squares analysis.


This study utilizes a random sample of 150 hospital bond issues in the years 1982-1984. Official statements for each bond issue in the sample were requested from the underwriters and served as the data source for bond-specific variables. The data source for the hospital-specific variables was the American Hospital Association's (AHA) Annual Survey of Hospitals.

To estimate the probit equation, a sample of hospitals not in the tax-exempt market was drawn. The population for this sample was all acute care hospitals and systems that met the AHA definition of a community hospital and were eligible to issue tax-exempt revenue bonds (American Hospital Association 1983). A random sample of the out-of-market group was drawn for each of the three years of the study.(4)

The unit of analysis in this study is the bond issue. Tax-exempt revenue bonds may be issued by hospital systems, by individual hospitals affiliated with systems, or by independent hospitals. All three types of issuers were included in this study. For the bonds issued by systems, data for the independent variables were aggregated across all hospitals in the system.(5)


Table 2 presents the results of the probit analysis. Three significant independent variables predict the probability of observing a hospital/ system in the market-systems, occupancy rate, and debt-to-asset ratio.
Table 2: Probit Analysis-In-Market versus out-of-Market
Variable Coefficient t-test
Constant -3.850 2.550(***)
Ownership 0.057 0.181
Year 83 -0.160 -0.464
Year 84 0.119 0.342
Teaching status 0.205 0.501
Location 0.188 0.595
Rate regulation -0.388 -1.516
System 0.928 2.009(**)
System affiliate -0.267 -0.862
Beds -0.0003 -0.594
Debt per capita -0.00005 -0.294
Occupancy rate 0.020 1.689(*)
Asset age 0.041 1.023
Operating margin 0.011 0.572
Debt-to-assets 0.014 2.466(**)
Percent Medicare 0.005 0.340
Percent Medicaid -0.025 -0.839
Average wage 0.00003 0.423
Charge per day 0.0009 0.942
Tax rate 0.001 -0.046
Log-likelihood -107.34
Restricted (slopes = 0) -207.94
Chi-squared (df = 19) 201.21, p < .0001
(*)p < 10.
(**)p < .05
(***)p < .01.

Systems are more likely to be observed in the market compared to independent hospitals (the reference group). This is not surprising given that system bonds are typically issued for more than one hospital. Hospitals in the market also have higher occupancy rates. Hospitals nearing capacity may have a greater need to expand and therefore will be more likely to be observed in the market.

Occupancy rate may also be a good measure of future creditworthiness so that hospitals with high occupancy rates are offered lower yields. A recent analysis of hospital credit downgrades demonstrated the importance of occupancy rate as a measure of creditworthiness (McCue, Renn, and Pillari 1990). If high-occupancy hospitals find it easier to gain access to the bond market because of lower yields on their debt, then hospitals in overbedded areas may find it more difficult to expand. This implies that the bond market creates an incentive for the efficient use of capital. This incentive would be reinforced by Medicare's decision to fold capital costs into their DRG rates. Hospitals in the market are more highly leveraged. This result seems counterintuitive, since one would expect those with less debt to be more likely to enter the market. This result may reflect the large number of hospitals that took advantage of declining interest rates over this period and advance-refunded their outstanding debt.

Table 3 presents the results of the reduced-form equation for bond size with the selection term from the probit equation. There are two significant variables, refunding and duration. The estimated coefficient suggests that refunding issues are $10.104 million larger than new-issue bonds, ceteris paribus. Refunding issues include bonds that are only issued to advance-refund existing debt and bonds that include a mixture of refunding and new debt. Thus, the principal amounts for bonds that were issued entirely for refunding were determined by previous borrowing; for bonds that were a mixture of refunding and new debt, the principal amount included both fixed and variable components. The incentive to advance-refund existing debt should increase with issue size. As interest rates declined over the study period, the hospitals with large outstanding debt issues would have been more likely to opt for advance refunding than those with small amounts of outstanding debt.
Table 3: Reduced-Form Equation for Bond Size
with Selection Term
Variable Coefficient t - Test
Constant 174.201 0.454
Ownership -2.626 -0.150
Year 83 2.266 0.108
Year 84 -8.132 -0.402
Teaching status -3.015 -0.116
Location -7.900 -0.346
Rate regulation 22.014 0.709
System bond -31.877 -0.459
System affiliate 11.276 0.433
Credit enhancement 7.762 1.432
Refunding 10.104 1.823(*)
Beds 0.030 1.155
Debt per capita 0.005 0.520
Occupancy rate -0.648 -0.409
Age of assets -1.419 -0.384
Operating margin -0.737 -0.420
Debt-to-assets -0.625 0.544
Percent Medicare -0.140 -0.174
Percent Medicaid 0.823 -0.350
Average wage -0.0002 -0.054
Bond supply -1.385 -0.628
Charge per day -0.036 -0.385
Tax rate -0.065 -0.049
Treasury rate -2.481 -0.519
Duration 3.497 2.544(**)
Selection term -56.865 -0.560
Log-likelihood -614.19
Restricted (slopes = 0) -689.64
Chi-squared (df = 25) 150.90, p < .0001
(*)p < .10.
(**)p < .05.
(***)p < .01.

The other significant variable in the reduced-form equation for bond size is duration. Holding other variables constant, the amount borrowed increased by $3.497 million for each unit increase in duration.(6) If financial managers typically match the maturity of the debt instrument to the life of the assets being financed, then long-term bonds are used to finance investments in long-term assets. Long-term assets are typically larger, more expensive assets, which would account for the increase in bond size with duration.

When this reduced-form equation for bond size is estimated without the selection term, some additional variables become significant. Table 4 shows that the credit enhancement and hospital size coefficients are positive and significant, suggesting that larger hospitals and those that purchase bond insurance borrow more money.
Table 4: Reduced-form Equation for Bond Size
without Selection Term
 Variable Coefficient t-Test
Constant -40.610 -1.324
Ownership -0.331 -0.075
Year 83 -4.640 -0.054
Year 84 -3.065 -0.520
Teaching status 6.162 1.105
Location 0.128 0.028
Rate regulation 5.941 1.715(*)
System bond 6.018 1.098
System affiliate -0.324 -0.075
Credit enhancement 7.704 2.167(**)
Refunding 10.148 2.747(***)
Beds 0.019 3.747(***)
Debt per capita 0.003 1.328
Occupancy 0.206 1.278
Age of assets 0.300 0.491
Operating margin 0.189 0.577
Debt-to-assets 0.007 0.083
Percent Medicare 0.089 0.466
Percent Medicaid -0.254 -0.719
Average wage 0.0009 1.117
Charge per day -1.155 -0.773
Bond supply 0.005 0.228
Tax rate 0.137 -0.412
Treasury rate -2.108 0.639
Duration 3.580 3.789(***)
Log-likelihood -628.98
Restricted (slopes = 0) -689.64
Chi-squared (df = 24) 121.31, p < .0001
 (*)p < .10.
 (**)p < .05.
(***)p < .01.

Table 5 presents the results for the yield equation. Independent variables include the instrumental variable for bond size derived from the reduced-form equation and the selection term derived from the probit equation. There are nine significant variables.
Table 5: Structural Equation for Yield with Selection Term
 Variable Coefficient t-Test
Constant 1.751 0.697
Ownership -0.240 -0.051
Year 83 -2.117 -7.995(***)
Year 84 -2.697 -8.817(***)
Teaching status 0.204 0.539
Location 0.215 0.711
Rate regulation 0.479 1.782(*)
System bond 0.749 1.424
System affiliate -0.032 0.111
Credit enhancement -0.508 -2.534(**)
Refunding 0.400 1.751(**)
Debt per capita -0.0002 -1.249
Occupancy rate -0.013 -1.035
Operating margin 0.018 -0.841
Debt-to-assets 0.004 0.496
Percent Medicare 0.010 0.877
Percent Medicaid 0.003 0.124
Bond supply -0.145 -1.969(**)
Tax rate -0.006 -0.284
Treasury rate 1.102 6.748(***)
Duration 0.547 8.395(***)
Bond size instrument -0.036 -2.778(***)
Selection term 0.532 0.762
Log-likelihood -177.10
Restricted (slopes = 0) -274.75
Chi-squared (df = 22) 195.31, p < .0001
 (*)p < .10.
 (**)p < .05.
(***)p < 01.

Both of the year variables were highly significant. Yields for 1983 were 211.7 basis points7 lower than 1982 yields; those for 1984 were 269.7 points lower than those for 1982. This is consistent with the general decline in market interest rates over the three-year study period. However, the yield equation also included a Treasury rate variable that reflects market conditions at the time of issue for each observation. The Treasury rates were matched with the bonds in the sample by month of issue and duration. The significant coefficients for the year variables, therefore, reflect additional factors within each year that affected bond yields.

One relevant factor that may have been captured by the year variables was any change in health care policy over this time period. The most significant change was the implementation of Medicare's prospective pricing system in 1983. The year before implementation, 1982, could be considered a time of considerable uncertainty about the impact of Medicare's new system. That uncertainty could have been reflected in higher yields. By 1983, initial payment rates were known; by 1984, hospital margins for Medicare patients were known to be fairly high (Broyles and Rosko 1990). Henkel (1984) has argued that regulatory uncertainty has a more serious impact on the cost of capital for hospitals than regulation itself. Once a new form of regulation is understood, the market can adapt its expectations and the price of

capital to reflect any constraints represented by regulation. He argues that the uncertainty that precedes changes in regulation is more apt to increase investor concerns and resulting yields. The coefficients for these variables may support that contention.

The Treasury rate variable in this function was significant. The marginal effect of a one percentage point increase in Treasury rates was to increase hospital yields by 110.2 basis points. Hospital yields were only slightly more volatile than general market interest rates at the time of this study.

The other significant market variable was the Visible Bond Supply. The negative coefficient implies that a $1 million increase in the supply of tax-exempt securities results in a 14.5 basis point decrease in offering yields. Although an inverse relationship between supply and price would normally be expected, this finding is not consistent with other studies of tax-exempt bonds. It may reflect a problem with the Visible Bond Supply index as a measure of bond supply. Utilization of a measure of bank demand for municipal bonds as well as a measure of bond supply might have been more effective.

Four other significant variables are bond instrument characteristics: credit enhancement, refunding, duration, and bond size. The coefficient for credit enhancement is negative. The purchase of a bank letter of credit or bond insurance reduces yields 50.8 basis points. A hospital purchases bond insurance to reduce the risk premium on the issue's yield. The 50 basis point savings in this study is consistent with estimates given by investment bankers of the interest rate savings that typically accrue from bond insurance and with empirical findings from an earlier study (Carpenter 1991).

The refunding variable in this equation is also significant. The positive coefficient seems to contradict the result in the reduced-form equation for bond size. This may reflect the fact that the issue size is fixed for those bonds that are 100 percent refunding, that is, determined by the initial bond issue and not current interest rates.

The duration variable is significant with a positive coefficient, which is consistent with the liquidity preference theory that investors must be rewarded with higher yields for holding long-term debt and with investor expectations that interest rates will rise in the future. A unit increase yields by 54.7 basis points.

Finally, the bond size instrument is significant and negative. A $1 million increase in the size of the bond will reduce yields by 3.6 basis points. A downward sloping curve can be expected if there are economies of scale in issuing tax-exempt debt. A negative coefficient is also consistent with the concept of marketability risk, that is, that larger issues are more easily traded in secondary markets after the initial bond purchase.

The only variable unique to hospital bonds that is even marginally significant is rate regulation (p < . 10). Bonds issued in states with some form of rate regulation have yields 47.9 basis points higher than bonds issued in states with no rate regulation. This last finding is of particular interest to health care policy analysts. The constraint on revenues represented by rate regulation may increase the risk of investment in rate-regulated hospitals and thus increase their bond yields. While rate regulation programs are instituted to constrain the growth in hospital costs, the findings of this study suggest that rate regulation may increase one component of cost, the cost of long-term debt.

It may also be true that the retrospective cost-based method used by many state rate-setting programs to reimburse capital costs creates little incentive to seek out lower interest rates (Rosko 1989). This interpretation would support the decision to fold capital costs into the Medicare PPS rates. Elimination of the capital cost pass-through may create an incentive to make more prudent investment and financing decisions.

The Medicare and Medicaid variables are also of particular interest because of the changes in payment methods during the study period. Medicare and Medicaid have paid for capital-related expenses on an actual-cost basis. The insignificant coefficients in this study contradict earlier findings that suggest some payer impact on the cost of capital (Cleverley and Nutt 1984; Sloan, Morrisey, and Valvona 1987; Sloan et al. 1988). The study results presented here may suggest offsetting effects of Medicare payment policies. The change in payment for operating expenses created uncertainty regarding future cash flows from Medicare. This should have increased observed yields; however, Medicare continued to pay for capital, including interest, on a cost basis. This could have reduced investor concerns about payment of debt obligations and decreased yields. But the relevant payer variable may no longer be the Medicare or Medicaid share of hospital revenue. A recently developed index of PPS fiscal pressure that combines the Medicare share of discharges with the margin earned on Medicare patients (Hadley, Zuckerman, and Feder 1989) may be a more appropriate measure of bondholders' risk (Grossman et al. 1990).

It is interesting to note that financial ratio variables also were not important predictors in the yield or selection equations. Sloan, Morrisey, and Valvona (1987) had similar findings. Investors may believe that occupancy rate, rather than current financial ratios, is more predictive of future cash flows in hospitals.

When the structural equation for bond yield is estimated without the selection term, some additional variables are significant. In Table 6, occupancy rate has a negative and significant coefficient, implying that yields decrease with an increase in occupancy. Rate regulation, which was only marginally significant in the selection-corrected equation, is now significant at the I percent level.


This study has confirmed previous findings that yields from hospital tax-exempt revenue bonds are determined by many of the same factors that determine yields on other types of bonds: market interest rates, duration, and credit enhancement. These preliminary findings also suggest that future analyses of bond yields need to consider the impact of simultaneous-equation and selection bias.

The negative and significant coefficient for the instrumental variable for bond size in the yield equation suggests that there may be simultaneous determination of yields and bond size. In a single-equation OLS analysis of yield determinants that utilized the same data as this simultaneous-equation model, the variable for bond size
Table 6: Structural Equation for Yield without Selection
 Variable Coefficient t- Test
Constant 3.267 2.003(**)
Ownership -0.241 -0.950
Year 83 -2.052 -7.817(***)
Year 84 -2.775 -8.845(***)
Teaching status 0.135 0.362
Location 0.106 0.411
Rate regulation 0.618 2.859(***)
System bond 0.473 1.312
System affiliate 0.076 0.299
Credit enhancement -0.524 -2.251(**)
Refunding 0.408 1.539
Debt per capita -0.0002 -1.111
Occupancy -0.020 -2.046(**)
Operating margin -0.024 -1.245
Debt-to-assets -0.001 -0.270
Percent Medicare 0.007 0.640
Percent Medicaid 0.009 0.440
Bond supply -0.145 -1.627
Tax rate -0.006 -0.305
Treasury rate 1.109 5.591(***)
Duration 0.546 7.385(***)
Bond size instrument -0.037 -2.674(***)
Log-likelihood -207.25
Restricted (slopes = 0) -274.75
Chi-squared (df = 21) 135.02, p < .0001
 (*)p < .10.
 (**)P < .05.
(***)p < .01.

was not significant (Carpenter 1991). The contradictory findings regarding bond size in previous work may reflect the fact that these estimates did not correct for simultaneous-equation bias.

The probit equation results demonstrate that factors exist that predict those hospitals that enter the tax-exempt market. The factors that proved significant in this model were hospital characteristics such as occupancy rate, system affiliation, and debt-to-asset ratio. In contrast, the selection-corrected equations for bond size and bond yield did not include any significant hospital variables with the exception of a weak relationship between yield and rate regulation. Hospital-specific variables were significant when the selection term was excluded from these equations. This implies that hospital variables previously identified as determinants of bond size or yield may reflect a selection effect that is eliminated when the yield equation is corrected for selection bias.

Future research in this area could expand on the findings in this study. The specification of a selection equation, for example, needs refinement. The intercept term in the probit model was highly significant, suggesting that other factors may be more predictive of which hospitals enter the market.

The model could be expanded to include general obligation and taxable bonds. This study was limited to the tax-exempt revenue bond market. Investor-owned hospitals and public hospitals were included in the population of potential issuers but may be underrepresented in the observed group. At the time of this study, for-profit hospitals could sell small issues of tax-exempt debt. However, few proprietary hospitals exercised this option because of the high cost of issuance and the restrictive covenants associated with tax-exempt financing. Thus, few were observed in the market during the study period. Public hospitals can issue either revenue bonds or general obligation bonds. Thus, it is likely they are underrepresented in the observed group of revenue bond issuers in this study. Since many rural hospitals are organized as public hospital districts and, therefore, can issue general obligation debt, underrepresentation of government hospitals may also affect the impact of the location variables. A model that incorporates more debt financing options may be more effective in identifying any disparities in access to capital that can be associated with ownership or location. This would be particularly helpful at a time when health care managers and policymakers are attempting to evaluate the impact of a change in capital payment policy on various types of hospitals.


Data for this study were provided by the American Hospital Association, Chicago. Thanks to Lloyd Wackerling at AHA for his assistance with this project. The author is grateful to the following people for their advice in the completion of this study: Bryan Dowd, Roger Feldman, Vernon Weckwerth, Timothy Nantell, and Robert Vigeland of the University of Minnesota, and Michael Rosko of Widener University. Comments from the anonymous referees also were very helpful.


(1.) Hospitals classified as teaching hospitals are those that are members of the Council of Teaching Hospitals. (2.) The Visible Bond Supply index is published weekly by the Bond Buyer in New York. It represents the volume of tax-exempt securities that are expected to be offered in the next 30 calendar days. (3.) The duration of a security is a weighted average of the times in the future when interest and principal payments are to be received. The yield to maturity of the bond is used as the discount rate for future cash flows. (4.) In 1982, there were 5,801 hospitals that met the AHA definition of a community hospital. There were also 249 systems. Thus, there were 6,050 potential issuers of tax-exempt bonds. Of these, 907 issued bonds and were in the in-market group, leaving 5,143 in the out-of-market group. Theoretically, a hospital/system could have had more than one bond issue over the three-year period. In fact, about 5 of the 907 were duplicates. This number was sufficiently small to allow the assumption that the number of bonds was equal to the number of issuers in the market.

The sample for the probit analysis contained equal numbers from the in market and out-of-market groups. Thus, the in-market group is overrepresented in the sample and the out-of-market group is underrepresented. A method proposed by Manski and McFadden (1981) was used to weight the observations to correct for disproportionality. The weight for the in-market group is the true population proportion (907/6050) divided by the sample proportion (150/300) or 0.30. The weight for the out-of-market group is the true population proportion (5143/6050) divided by the sample proportion (1 50/300) or 1. 7. By using these weights, the dependent variable in the probit equation represents the probability of observing a hospital in the market over the three-year study period. (5.) Ratio variables such as operating margin were calculated using aggregate data in the numerators and denominators. Variables such as location were measured as continuous vari beds in a metropolitan statistical area (MSA), and later reclassified as dichotomous variables. (6.) The duration for a bond is not necessarily proportional to years. A 30-year bond, for example, can have a duration from 7 to 10, depending on the pattern of cash flows over the life of the bond and the yield used to weight those cash flows. (7.) A basis point is one-hundredth of a percent; 100 basis points equals 1 percent.


American Hospital Association. AHA Guide. 1983 ed. Chicago: AHA, 1983. Austen, E., H. Corman, and G. LiCalzi. "The Determinants of Net Interest Cost in Negotiated Underwriting of Tax-Exempt Hospital Bonds." American Business Review 4, no. 1 (January 1986): 16-19. Broyles, R. W., and M. D. Rosko. Fiscal Management of Healthcare Institutions. Owings Mills, MD: National Health Publishing, 1990. Buser, S. A., and P. J. Hess. "What Explains the Ratio of Tax-exempt and Taxable Yields." Unpublished manuscript, 1985. Campbell, T S. "On the Extent of Segmentation in the Municipal Securities Market." Journal of Money, Credit and Banking 12, no. 1 (February 1980): 71-83. Carpenter, C. E. "The Marginal Effect of Bond Insurance on Hospital, Tax-exempt Bond Yields." Inquiry 28 (Spring 1991): 67-73. Cleverley, W. O., and P. C. Nutt. "The Decision Process Used for Hospital Bond Rating-and Its Implications." Health Services Research 19, no. 5 (December 1984): 615-37. Cleverley, W. O., and W. H. Rosegay. "Factors Affecting the Cost of Hospital Tax-exempt Revenue Bonds." Inquiry 19, no. 4 (Winter 1982): 317-26. Cohodes, D. R., and B. M. Kinkead. Hospital Capital Formation in the 1980s. Baltimore, MD: Johns Hopkins University Press, 1984. Copeland, T. E., and J. F. Weston. Financial Theory and Corporate Policy. Reading, MA: Addison-Wesley Publishing Company, 1983. Eisner, R. "Investment: Fact and Fancy." American Economic Review 53, no. 2 (May 1963): 237-46. Fisher, L. "Determinants of Risk Premiums on Corporate Bonds." Journal of Political Economy 67, no. 3 (June 1959): 217-37. Grossman, M., F. Goldman, S. W. Nesbitt, and P. Mobilia. "Determinants of Interest Rates on Tax-exempt Hospital Bonds." Presentation at 118th Annual Meeting of the American Public Health Association, New York, September 1990. Hadley, J., S. Zuckerman, and J. Feder. "Profits and Fiscal Pressure in the Prospective Payment System: Their Impacts on Hospitals." Inquiry 26 (Fall 1989): 354-65. Heckman, J. "Sample Selection Bias as a Specification Error." Econometrica 47, no. 1 (January 1977): 153-61. Hendershott, P. H., and D. S. Kidwell. "The Impact of Relative Security Supplies." Journal of Money, Credit and Banking 10, no. 3 (August 1978): 337-47. Henkel, A. Payment Policy Effects on Investor Behavior and Access to Capital. Proceedings of the Fourth Annual American Health Planning Association Health Policy Conference. Washington, DC: American Health Planning Association, March 7, 1984. Jorgenson, D. W. "Capital Theory and Investment Behavior." American Economic Review 53, no. 2 (May 1963): 247-59. Kennedy, P. A Guide to Econometrics. Cambridge, MA: MIT Press, 1984. Kessel, R. H. "A Study of the Effects of Competition in the Tax-exempt Bond Market." Journal of Political Economy 79, no. 4 (1971): 706-38. Kmenta, J. Elements of Econometrics. 2d ed. New York: Macmillan Publishing Co., 1986. Kuh, E. "Theory and Institutions in the Study of Investment Behavior." American Economic Review 53, no. 2 (May 1963): 260-68. Lee, L. F., G. S. Maddala, and R. P. Trost. "Asymptotic Covariance Matrices of Two-Stage Probit and Two-Stage Tobit Methods for Simultaneous Equation Models with Selectivity." Econometrica 48, no. 2 (March 1980): 491-503. Lev, B. Financial Statement Analysis. Englewood Cliffs, NJ: Prentice Hall, Inc., 1974. Maddala, G. S. Limited-Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press, 1983. Manski, C., and D. McFadden, eds. Structural Analysis of Discrete Data with Econometrics Applications. Cambridge, MA: MIT Press, 1981. McCue, M. J., S. C. Renn, and G. D. Pillari. "Factors Affecting Credit Rating Downgrades of Hospital Revenue Bonds." Inquiry 27 (Fall 1990): 242-54. Rosko, M. D. "A Comparison of Hospital Performance under the Partial-Payer Medicare PPS and State All-Payer Rate Setting Systems." Inquiry 26 (Spring 1989): 48-61. Shields, G. Hospital Debt Financing and Capital Formation. Germantown, MD: Aspen Systems, 1983. Sloan, F. A., M. A. Morrisey, and J. Valvona. "Capital Markets and the Growth of Multihospital Systems." Advances in Health Economics and Health Services Research 7 (1987): 83-109. Sloan, F. A., J. Valvona, M. Hassan, and M. A. Morrisey. "Cost of Capital to the Hospital Sector." Journal of Health Economics 7 (1988): 24-45. Tanner, E. J. "The Determinants of Interest Cost on New Municipal Bonds: A Reevaluation." Journal of Business 48, no. 1 (1975): 74-80. Van Horne, J. C. Financial Market Rates and Flows. Englewood Cliffs, NJ: Prentice-Hall, Inc., 1984. Wedig, G. J., M. Hassan, and F. A. Sloan. "Hospital Investment Decisions and the Cost of Capital." Journal of Business 62, no. 4 (1989): 517-537. Wedig, G., F. A. Sloan, M. Hassan and M. A. Morrisey. "Capital Structure, Ownership, and Capital Payment Policy: The Case of Hospitals." The Journal of Finance 43 (March 1988): 21-40.

Address correspondence and requests for reprints to Caryl E. Carpenter, Ph.D., Assistant Professor, Health and Medical Services Administration, 115 Kapelski, Widener University, Chester, PA 19013. This article, submitted to Health Services Research on June 8, 1990, was revised and accepted for publication on January 13, 1992.
COPYRIGHT 1992 Health Research and Educational Trust
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Carpenter, Caryl E.
Publication:Health Services Research
Date:Dec 1, 1992
Previous Article:The effect of the Illness Episode Approach on Medicare beneficiaries' health insurance decisions.
Next Article:The use of formal and informal home care by the disabled elderly.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters