Expense preference behavior and management “outsourcing”: a comparison of adopters and non-adopters of contract management in U.S. hospitals

Abstract

Specialized managerial expertise, coupled with the threat of non-renewal should improve efficiency in firms that opt for contract management arrangements. To examine this we apply a generalized version of tests for expense preference behavior to U.S. hospitals in the 1990s. Extending prior literature, we create a quasi-experimental design for a comparison of adopters and non-adopters of contracts using propensity score methods. We generate the distribution of ‘expense preference’ parameters for all contract adopters in both the pre- and post-adoption states, and for a matched control group of non-adopters over the same period. Our results show that contract adoption leads to reduced expense preference behavior, but that this result depends critically on the input being examined.

This is a preview of subscription content, access via your institution.

Fig. 1

Notes

  1. 1.

    Hospital care may be the only industry where information on contract management is collected on a regular basis; however, trade reports suggest that its use is widespread in service industries, government contracting, and among non-profit organizations. A recent government report surveys the use of comprehensive contract-management services as an intermediate step towards privatization, especially in construction projects (CETS 2000). The closest analogy to hospitals occurs in education where 15–20% of all charter schools use for-profit EMOs (Education Management Companies—the equivalent terminology for contract-management firms) for comprehensive management (Molnar et al. 2001).

  2. 2.

    Furthermore, the trade/practitioner literature views contract management as means by which boards can weaken the community oversight over the management function by contracting with an outside corporate entity not bound by the hospital’s charter, a matter which has been tested in the courts (Kimberly and Rosenzweig 1998).

  3. 3.

    U L and U K are realizations of the state-dependent utility, under labor-preference and capital preference, respectively. State dependent postscripts are suppressed for notational convenience. U E is independent of state, while E i  =  W i . Adding probabilities p L and p K = 1 −  p L to each state will not affect the basic result.

  4. 4.

    The banking literature has similarly treated labor as the more abundant input which can be more easily manipulated by the expense-preferring manager; see Hannan and Mavinga (1980) and Awh and Primeax (1985).

  5. 5.

    A full theory of incentives under contract-management remains to be developed. However there is substantial and relevant literature on managerial incentives in nonprofit organizations. For instance, Roomkin and Weisbrod (1999) and Brickley and Van Horn (2002) discuss limitations placed on hospital boards by community-oriented charters and member organizations. This implies weakened board oversight, leading to management entrenchment, a problem shared by other industries when boards are said to be “weak” (e.g., Denis et al. 1997). The principle-agent literature has dealt extensively with the issue of pay-for-performance contracts when executive compensation is tied to stock options, dividends and the like. This tends to less relevant to an industry dominated by non-profits and community oversight.

  6. 6.

    An alternative strain in the literature considers the problem of overall firm inefficiency along the firm’s production function or cost function for given level of inputs. Therein, the stochastic frontier (SF) specifies a truncated or half-normal error term for observations that lie below the least-cost frontier, in addition to the usual (normal) regression error. (e.g., Fare et al. 1994). Skinner (1994) notes that SF is sensitive to the distributional assumptions made, and may lead to biased inefficiency scores if errors are skewed. Moreover, SF does not provide a direct measure of allocative inefficiency due to varying input proportions, which is of main interest to us. In a related review, Dor (1994b) notes that panel data yield more robust estimates of cost differences among nursing homes or hospitals. In this analysis we track longitudinal changes in inefficiency, as well as input-specific behavior.

  7. 7.

    Dor et al. (1997) and Mester (1989) constrain the expense-preference for the non-preferred input to be positive (z k  > 0). Here we demonstrate that an increase in the preferred input requires a concomitant decrease in the non-preferred input. Although constraints on z k can be imposed in our estimation framework using a penalty function, in practice this was not necessary since convergence of the unconstrained nonlinear system of equations occurred at the permissible range.

  8. 8.

    This is akin to multiple cause-multiple indicator (MIMIC) models. See Van Vliet and Van Praag (1987).

  9. 9.

    Note that by emphasizing preferences for a given input, the expense preference model essentially assigns behavioral motivation for allocative inefficiency. Other models, such as Data Envelopment Analysis (DEA) focus on technical efficiency, namely maximizing output given the level of inputs employed. However, DEA is essentially a linear programming algorithm that does not take into account random error and statistical noise, thus producing inefficient estimates. While the non-parametric estimation in DEA is an attractive property in some applications where explicit functional form is not of interest and hence restrictive, structure is required for the problem at hand, namely identifying changes in unobservable preferences due to adoption of contracts.

  10. 10.

    Labor costs are defined as the sum of total facility payroll expenses and total facility employee benefits. Unfortunately the survey does not provide a breakdown of expenses by type of labor; this category covers registered nurses, licensed practical nurses, and administrative staff. Depreciation plus interest accounted for 8% of total expenses in 1993, the latest year in this study for which the AHA data reported these items specifically.

  11. 11.

    The AHA conversion is based on adjusted patient days = actual inpatient days · (total patient revenue/inpatient revenue). (AHA Annual Survey Database).

  12. 12.

    Another strain in the literature has focused on estimating the multi-output cost function of the hospital, and related marginal costs (e.g., Vita 1990; Dor and Farley 1996; Carey 1997). In this paper, we are not interested in output-specific costs per se, but it may be the case that inefficiency varies by output category or that preferences for inputs vary. As far as we can tell, the problem of output-specific inefficiency has not yet been treated in the econometrics literature. In our case, obtaining estimates of output-specific z’s would require having data on the share of each input within each output category. Such data are unavailable.

  13. 13.

    The issue of monopsonistic labor markets for nurses has been the subject of some debate in the empirical literature, yet evidence on this point is inconclusive. For instance Staiger et al. (1999) find some evidence in favor of monopsony, in the form of inelastic supply for nurses. Conversely, Hirsch and Schumacher (1995) found no support for the presence of monopsony power in the nursing labor market. Due to this ambiguity we implemented a Hausman test for endogeneity of the wages in each of the four samples. The test resulted in our failure to obtain a rejection of the null hypothesis of no endogeneity in any of the four cases.

  14. 14.

    For instance see, Mobley and Magnussen (2002). Recently the trade literature has begun to focus on the problem of nurse ‘shortage’ in hospitals, suggesting that hospitals tend to under-employ nurses (Green and Nordhaus-Bike 1998). However this concern is limited to certain high-end specialties of registered nurses, and does not seem to apply to licensed practical nurses, nurse-aides and the like. Moreover, even for registered nurses as a whole the national trend has been that of increased employment in hospitals during most of the period observed in our data (Buerhaus and Staiger 1999).

  15. 15.

    This can also be referred to as a ‘difference-in-difference’ estimator. See Angrist and Krueger (1999).

  16. 16.

    At the Federal level, special bonus payments are available to physicians located in Health Professional Shortage Areas (HPSAs). A variety of similar state-level programs are also available. See Shugarman and Farley (2003).

References

  1. Alexander JA, Rundall TG (1985) Public hospitals under contract management. Med Care 23(3):209–219

    Article  Google Scholar 

  2. Angrist JD, Krueger A (1999) Empirical strategies in labor economics. In: Handbook of labor economics V3A, Chapter 23. Elsevier, New York

  3. Awh RY, Primeaux WJ (1985) Managerial discretion and expense preference behavior. Rev Econ Stat LXVII(2):224–231

    Article  Google Scholar 

  4. Berndt E, Christensen L (1973) The translog function and the substitution of equipment, structures, and labor in U.S. manufacturing 1929–1968. J Economet 1(1):81–113

    Article  Google Scholar 

  5. Blair DW, Placone DL (1988) Expense-preference behavior, agency costs, and firm organization: the savings and loan industry. J Econ Bus 40(1):1–15

    Article  Google Scholar 

  6. Brickley J, Van Horn RL (2002) Managerial incentives in nonprofit organizations: evidence from hospitals. J Law Econ 45(1):227–249

    Article  Google Scholar 

  7. Buerhaus PI, Staiger DO (1999) Trouble in the nurse labor market? Recent trends and future outlook. Health Affairs 18(1):214–222

    Article  Google Scholar 

  8. Carey K (1997) A panel data design for estimation of hospital cost functions. Rev Econ Stat LXXIX:443–453

    Google Scholar 

  9. Carey K, Dor A (2004) Trends in contract management: the hidden evolution in hospital organization. Health Affairs 23(6):192–199

    Article  Google Scholar 

  10. Commission on Engineering and Technical Systems (CETS) (2000) Outsourcing management functions for the acquisition of federal facilities. Washington DC: National Academies Press

  11. Dehejia RH, Wahba S (1999) Causal effects in non-experimental data: re-evaluating the evaluation of training programs. J Am Stat Assoc 94(448):1053–1062

    Article  Google Scholar 

  12. Dehejia RH, Wahba S (2002) Propensity score matching methods for nonexperimental causal studies. Rev Econ Stat 84(1):151–161

    Article  Google Scholar 

  13. Denis D, Denis D, Sarin A (1997) Ownership structure and top management turnover. J Finan Econ 45(2):193–223

    Article  Google Scholar 

  14. Dewatripont M, Maskin E (1995) Contractual contingencies and renegotiations. RAND J Econ 26(4):704–719

    Article  Google Scholar 

  15. Dor A (1994a) Are contract-managed hospitals more efficient? Rockville, MD: U.S Department of Health and Human Services, AHCPR Publication No. 94-0004

  16. Dor A (1994b) On stochastic frontiers and non-minimum cost functions for health care providers. J Health Econ 13(3):329–334

    Article  Google Scholar 

  17. Dor A, Farley DE (1996) Payment source and the cost of hospital care: inferences on quality from a multiproduct cost function. J Health Econ 15(1):1–21

    Article  Google Scholar 

  18. Dor A, Duffy S, Wong H (1997) Expense preference behavior and contract-management: evidence from U.S. hospitals. Southern Econ J 64(2):542–554

    Article  Google Scholar 

  19. Edwards FR (1977) Managerial objectives in regulated industries: expense-preference behavior in banking. J Polit Econ 85(1):147–162

    Article  Google Scholar 

  20. Fama E, Jensen M (1983) Separation of ownership and control. J Law Econ 26(2):301–325

    Article  Google Scholar 

  21. Fare R, Groskoff S, Lovell CAK (1994) Production frontiers. Cambridge University Press, Cambridge, UK

    Google Scholar 

  22. Gallant AR (1987) Nonlinear statistical models. New York: Wiley

    Google Scholar 

  23. Greene J, Nordhaus-Bike AM (1998) Nurse shortage, where have all the RNs gone? Hosp Health Netw 72:78–80

    Google Scholar 

  24. Hannan TH, Mavinga F (1980) Expense preference and managerial control: the case of the banking firm. Bell J Econ 11(2):671–682

    Article  Google Scholar 

  25. Hirsch BT, Schumacher EJ (1995) Monopsony and relative wages in the labor market for nurses. J Heath Econ 14:443–476

    Article  Google Scholar 

  26. Imbens GW (2000) The role of the propensity score in estimating dose-response functions. Biometrika 87(3):706–710

    Article  Google Scholar 

  27. Kimberly JR, Rosenzweig PM (1988) Hospitals boards and the decision to renew the full service management contract. Hosp Health Serv Adm 33(4):449–465

    Google Scholar 

  28. Kralewski JE et al (1984) Effects of contract management on hospital performance. Health Serv Res 19(4):479–497

    Google Scholar 

  29. Lalonde R (1986) Evaluating the econometric evaluations of training programs. Am Econ Rev 76:604–620

    Google Scholar 

  30. Mester LJ (1989) Testing for expense preference behavior: mutual versus stock savings and loans. RAND J Econ 20(4):483–498

    Article  Google Scholar 

  31. Mobley L, Magnussen J (2002) The impact of managed care penetration and hospital quality on efficiency in hospital staffing. J Healthcare Finan 28(4):24–42

    Google Scholar 

  32. Molnar A, Morales J, Wyst AV (2001) Profiles of for-profit education management companies: 2000–2001. Center for Education Research, Analysis, and Innovation, University of Wisconsin-Milwaukee, Milwaukee, WI

    Google Scholar 

  33. Preyra C, Pink G (2001) Balancing incentives in the compensation contracts of nonprofit hospital CEOs. J Health Econ 20(4):509–525

    Article  Google Scholar 

  34. Roomkin M, Weisbrod B (1999) Managerial compensation and incentives in for-profit and nonprofit hospitals. J Law Econ Organ 15(3):750–781

    Article  Google Scholar 

  35. Rosenbaum PR (1987) The role of a second control group in an observational study. Stat Sci 2:292–316

    Article  Google Scholar 

  36. Rosenbaum PR, Rubin DB (1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70:41–55

    Article  Google Scholar 

  37. Rosenbaum PR, Rubin DB (1984) Reducing bias in observational studies using the subclassification on the propensity score. J Am Stat Assoc 79:516–524

    Article  Google Scholar 

  38. Rosenbaum PR, Rubin DB (1985) Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 39:33–38

    Article  Google Scholar 

  39. Rundall TG, Lambert WK (1984) The private management of public hospitals. Health Serv Res 19(4):519–544

    Google Scholar 

  40. Scott L (1994) Firms see pressure, not profits. Modern Healthcare Press, January 31

  41. Shugarman LR, Farley DO (2003) Shortcomings in medicare bonus payments for physicians in underserved areas. Health Affairs 22(4):73–187

    Article  Google Scholar 

  42. Skinner J (1994) What do stochastic frontier cost functions tell us about inefficiency? J Health Econ 13(3):323–328

    Article  Google Scholar 

  43. Staiger D, Spetz J, Phibbs C (1999) Is there monopsony in the labor market? Evidence from a natural experiment. NBER Working paper 7258

  44. Van Vliet R, Van-Praag B (1987) Health status estimation on the basis of mimic-health care models. J Health Econ 6(1):27–42

    Article  Google Scholar 

  45. Vita M (1990) Exploring hospital production relationships with flexible functional forms. J Health Econ 9:1–21

    Article  Google Scholar 

  46. Wheeler RC, Zuckerman HS (1984) Hospital management contracts: institutional and community perspectives. Health Serv Res 19(4):499–517

    Google Scholar 

Download references

Author information

Affiliations

Authors

Corresponding author

Correspondence to Kathleen Carey.

Additional information

Earlier versions were presented at the NBER Summer Institute, the BU/Harvard/MIT health economics workshop 2002, and the American Economic Association annual meetings, 2003. The authors are grateful for comment received. Avi Dor acknowledges financial support from the Presidential Research Initiative grant at Case Western Reserve University. The views expressed here are those of the authors and do not necessarily represent the Department of Veterans Affairs or NBER.

Appendix: Propensity scores calculation

Appendix: Propensity scores calculation

Propensity score methods are commonly used in observational studies to replicate the benefits of a controlled experimental trial in which two similar groups are compared, but only one group receives a treatment. A propensity score is defined as the probability of assignment to a treatment, given the distribution of known covariates, just prior to intervention. It is used to balance the distribution of covariates between a treatment group and a control group in cases where a random draw from the general population results in a poor match (Rosenbaum and Rubin 1983, 1984; Imbens 2000) thereby alleviating the bias due to systematic differences between the treated and comparison groups (Dehejia and Wahba 2002). The most common algorithm for converting propensity scores into non-experimental matching criteria is to stratify the unit scalar into quintiles or similar groupings (Rosenbaum and Rubin 1985; Rosenbaum 1987). Dehejia and Wahba (1999) demonstrate that results are robust with respect to percentile definitions within the propensity score.

A random drawing of non-contract hospitals failed to create a reliable control for hospitals identified as contract-adopters, because the sampling strategy involves drawing all hospitals that adopted contract management across years, and because those hospitals differ in profile from internally managed hospitals. More specifically, a much higher proportion of contract-managed hospitals are rural (see Table 2). These hospitals are also more likely to be government-affiliated hospitals, less likely to be not-for-profit, are lower in case-mix index, and have longer lengths of stay. To account for these differences we used propensity scores to reduce the bias in drawing the comparison group.

Propensity scores, or conditional probabilities of adopting contract management, were estimated for each hospital using logistic regression, pooling contract-adopters with all non-contract hospitals in the database. We performed two regressions corresponding to the ‘pre’ and ‘post’ periods. The first regression included all available observations for the adoption years 1993–1998 and the second included those for adoption years 1992–1996. We incorporated all variables used to explain variation in cost as covariates in the logit regression. Table A1 shows the results of these regressions. Although results of these regressions are tangential to our main question, a number of results are worth noting: What matters in the distinction of contract adopters is ownership form and locality. Contract management is on the order of two times as likely to be found among hospitals with government control, non-profit status, or rural location.

Table A1 Propensity score model (logit)

Stratification proceeded by dividing the propensity scores from the combined groups into quintiles. Stratification also took into account the distribution of contract-managed hospitals over time. Table A2 reports stratum boundaries for propensity scores and sampling distributions by quintile-year cells. Counts pertain to hospitals in the contract-managed group. The control groups were finally drawn randomly within propensity score quintile-year cells (without replacement). For purposes of cost function estimation we produced two control groups each containing three times the number of contract adopters that matched the adopter groups’ distributions. Cost function estimation pertained to two years prior and two year after the year of adoption.

Table A2 Propensity score distributions by quintile and year

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Carey, K., Dor, A. Expense preference behavior and management “outsourcing”: a comparison of adopters and non-adopters of contract management in U.S. hospitals. J Prod Anal 29, 61–75 (2008). https://doi.org/10.1007/s11123-007-0065-3

Download citation

Keywords

  • Expense preference
  • Hospitals
  • Contract management

JEL Classifications

  • D24
  • L24