Inferences regarding type II errors also vary across prior research. Dechow et al. (1995, p. 194) found that power is low even for the models with the lowest standard error (the Jones and modi®ed Jones models had the lowest standard errors, with mean standard errors of about 9% of total assests for forecast accruals over their random sample of 1,000 ®rms). Earnings management would have to exceed 18% of total assets for the average ®rm (or 1% of TA for each ®rm in a sample of 300 ®rms) to generate signi®cant statistics that would reject the null hypothesis of no earnings management. Kang and Sivarama- krishnan (1995, pp. 361) paint a more optimistic picture, because they move from the individual ®rm level to samples of 100 ®rms each. When they seeded their ®rms with a random accrual that averaged 2% of total assests (it actually ranged between 0.064% and 30.469%), their rate of rejection of the null hy- pothesis at the 5% level increased from 5% (before seeding) to 23% for the Jones model (and 33% and 47% for their IV and GMM models).
Given the above trade-os, the researcher's goal is to identify contexts where the incentives to manage earnings are of interest and reliable and powerful measures of earnings management can be identi®ed. Several studies taking a speci®c accruals approach have attempted to do so by focusing on a speci®c industry or set of industries. For example, McNichols and Wilson (1988) choose industries with the highest ratios of receivables to total assets and the largest number of ®rms in the industry: publishers, business services and nondurable wholesalers. The ®rst criterion serves to identify ®rms for which the allowance for uncollectibles is likely a material account and the second serves to identify industries with a suciently large number of ®rms to permit reliable model estimation. A number of studies focus on discretion in loan loss pro- visions in the banking industry. These studies include Beaver et al. (1989), Moyer (1990), Scholes et al. (1990), Wahlen (1994), Beatty et al. (1995), Collins et al. (1995), and Beaver and Engel (1996). A distinctive feature of each of the above studies is the use of generally accepted accounting principles to specify what the nondiscretionary component of an accrual should be.
Economic consequences of regulatory accounting in the nuclear power in- dustry: market reaction to plant abandonments, The, 161±187. Eects of changes in cost allocations on the assessment of cost containment regulation in hospitals, The, 97±112. Eect of relative performance evalua- tion on earnings management: a game- theoretic approach, The, 377±397. Empirical analysis of auditor report timing by large municipalities, An, 263±281.
Daniel Beneish has been Special Associate Editor for the next four articles in this issue of the JournalofAccountingandPublicPolicy (Vol. 19, Nos. 4/5). Those four articles were submitted as a result of his request for manuscripts pertaining to the management of earnings (see Beneish, 1998). We appreciate his work.
techniques to a service setting. The increasing magnitude of health care costs, however, necessitates that hospitals adopt costing systems to provide better dataand greater insights for cost control and cost management. The revenue constraints imposed by Medicare preclude hospitals from maximizing pro®ts today as they did in the past by simply increasing the number of services. Hospital revenues were severely constrained in the 1980s and continue to be so through the 1990s. In response, one would suspect that hospitals would react to this by changing their cost accounting systems since their control over the revenues provided by third-party payers, particularly Medicare, has decreased signi®cantly. When prices are dictated by external parties, pro®t maximization can only come through managing and controlling costs. Yet, as of 1990, a surprising number of hospitals still lacked a costing system able to provide data on the costs of procedures. The ability to monitor the cost of procedures, rather than merely aggregating costs, is critical because it allows hospitals to track the costs by patient. This is important since Medicare Õ s change in reim- bursement methodology means that hospitals are reimbursed per patient rather than per service.
Empirical studies that examine the eects of regulation on cost containment fre- quently ignore the impact of changes in accounting practices. This results in a potential bias of research ®ndings. For example, some studies found evidence of cost containment for inpatient services after a change in Medicare reimbursement in 1983. However, Eldenburg and Kallapur (1997, p. 33) found that more costs were allocated to outpa- tients and correspondingly less to inpatients after 1983, which could bias the cost comparisons. In this paper we examine changes in inpatient costs relative to outpatient costs, after controlling for allocations, to determine whether the magnitude of allocation changes was large enough to bias the ®ndings of studies that ignored these accounting practices. As in previous health-care studies (noted in our paper), we found that in- patient full costs (i.e., direct cost plus allocated costs) decreased relative to outpatient full costs after 1983. However, when cost allocations were excluded, inpatient direct costs increased relative to outpatient direct costs, thus providing no evidence of cost- containment. When regulation provides incentives that have the potential to aect ac- counting practices andpublicpolicy researchers do not consider the implications of these accounting practices, analyses of the success ofpublic regulation may reach im- proper conclusions. Accordingly, subsequent policy based on such research ®ndings may be incorrectly motivated. Ó 2000 Elsevier Science Ltd. All rights reserved.
Although national trends indicate that investors could expect to bear some portion of cancellation losses, rate setting is a decentralized process with considerable variability between rulings by public service commissions at the state level (DOE/EIA, 1983, pp. ix±xxii). Therefore, market reaction to abandonment announcements may be aected by the state-speci®c regulatory environment in which utilities operate, as well as overall national trends. Based on our analysis of the cash ¯ow consequences of regulatory accounting policies noted earlier in our paper, we can derive the following expectations regarding investor behavior. Investors anticipating a full recovery ruling from a state regulatory commission would not be expected to react negatively to aban- donment announcements since the cash ¯ow consequences of abandonment and non-abandonment are identical. Investors anticipating a state-speci®c ruling of no recovery or partial recovery, however, would be expected to react negatively to abandonment announcements since both these rulings have negative cash ¯ow consequences in comparison to the alternative of bringing the project to completion. We use prior regulatory rulings on abandonment costs as a proxy variable for the state-speci®c regulatory environment in order to examine the association between the state regulatory environment and the stock market's reaction to abandonment decisions.
mental regulations (e.g., Wallace et al., 1988, p. 80; Shane, 1995, p. 496; Blacconiere and Northcutt, 1997, pp. 162±163) or following events that were expected to increase regulatory activity (e.g., Bowen et al., 1983, p. 98; Hill and Schneeweis, 1983, p. 1290; Blacconiere and Patten, 1994, p. 368; Patten and Nance, 1998, p. 410). 3 One factor that may have contributed to the decline in market value for ®rms in aected industries was the expectation that new en- vironmental regulations would require the implementation of costly pollution abatement controls. For example, Pashigian (1984, p. 6) suggested that envi- ronmental regulation would favor capital intensive methods of reducing emissions. 4 Following the Bhopal accident, industry observers predicted that the event would indeed lead to higher capital expenditures (Starr, 1985, p. 28). 5
Case Mix Adjusted Admissions per Bed in service (CMAAD/BED) and Case Mix Adjusted Equivalent Admissions per Full-Time-Equivalents (CMAEAD/FTEs) loaded consistently onto a separate factor (Table 3). While these variables do not conform nicely to any one SEA category provided by the GASB, they do seem to ®t the general description of service eorts and ac- complishments. Perhaps they are more appropriately described as representing ``capacity productivity'' and ``manpower productivity'', 13 respectively. Ca- pacity productivity is a measure which correlates inpatient activity produced by each bed with productivity across hospitals. The capacity productivity ratio is produced by ®rst determining the relative level of total inpatient activity ± CMAAD. This is achieved by multiplying admissions by the case mix index. CMAAD is then divided by the number of beds in service or capacity to yield a measure of bed turnover. Manpower productivity is measured by the rela- tionship of case mix adjusted equivalent admissions to FTEs. CMAEAD is a measure of total hospital output, taking into account inpatient turnover, case mix intensity, and outpatient production. FTEs are a good indication of total hospital input. The ratio is indicative of the number of CMAEADs serviced by each FTE. Because manpower productivity is a combination of total utilization and stang, the ratio provides a measure of eciency anda comparison of one hospital relative to another according to FTEs (Van Kampen Merritt Invest- ment Advisory Corporation, 1990, p. 7.26). 14
In general, the results in Table 4 reinforce prior evidence. First, the earnings variable is positive and signi®cant; its coecient of 2.13 (1.95 in model 2) suggests that a change in de¯ated earnings by 1% leads to an analogous change in median-adjusted stock returns by 2.13% (1.95%). In contrast to Hypothesis 1, the outsider representation±earnings interactive term is insigni®cantly dif- ferent from zero, suggesting that board composition is not related to earnings informativeness, unconditionally. This result holds regardless of the de®nition of outsider representation. Importantly, there remain signi®cant dierences in earnings usefulness across board-size categories even after controlling for nu- merous ®rm characteristics such as size, growth, board composition, ownership structure, systematic risk, and default risk. In fact, the interactive term coef- ®cient suggests that the dierence in the earnings±returns relation between ®rms with large boards (with 12 or more members) and ®rms with small boards remains substantially unchanged compared to the dierence suggested by model 3 in Table 3 which excludes the control variables (ÿ1:28 vs ÿ1:23, respectively). The persistence of this ®nding is interesting and remains in line with the increased monitoring eectiveness of smaller boards.
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 reduction 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).
The ®nancial accounting measure of pro®tability, the accounting rate of return, has both critics and defenders. Critics (Atkinson et al., 1997, p. 475; Brealey et al., 1999, p. 161; Fisher and McGowan, 1983, p. 86) point to two theoretical de®ciencies. The ®rst is that, for a single completed investment project, the annual accounting rates of return vary and do not equal the eco- nomic or internal rate of return (Atkinson et al., 1997, p. 475; Brealey et al., 1999, p. 161). The second is that, for an in®nite horizon reinvestment model, the annual accounting rate of return can converge to a number quite dierent from the common economic rate of return of the individual investments (Fisher and McGowan, 1983, p. 86). The critics ®nd these theoretical de®ciencies telling: if the accounting rate of return can fail in these relatively simple settings what con®dence can one have for an actual ®rm which is a complex aggre- gation of projects (Fisher and McGowan, 1983, p. 83)? The defenders of the accounting rate of return refer to the desirable theoretical properties ofa weighted average ofaccounting rates of return (e.g., Edwards et al., 1987, pp. 65±68; Brief and Lawson, 1992, pp. 416±419) and empirical results relating the accounting rate of return to various stock market measures (e.g., Board and Walker, 1990, pp. 186±187; Landsman and Shapiro, 1995, pp. 111±115).
Before discussing the results of my study, four limitations related to the use of an experimental approach should be noted. As part of an experimental approach participants responded to hypothetical scenarios about earnings management activities. This approach has previously been used in studies to measure ethically related judgments (Becker and Fritzsche, 1987; Flory et al., 1992; Singer, 1996; Singer and Singer, 1997; Singer et al., 1998). The strength of this approach is that it allows for greater control and manipulation of variables than provided by non-experimental approaches. A concern, however, with this approach is that it is not possible for the stimulus materials to contain all relevant information in order for the task to be completed in a timely manner. Second, the case unambiguously manipulated earnings management intent. In general, intent is unobservable and must be inferred from behavior and the surrounding context (Kelley, 1973, p. 107). An unambiguous manipulation of intent was needed, however, in order to appropriately test the hypotheses in- volving intent. Examining ethical judgments when intent is inferred with greater uncertainty represents a topic for further research.
including rerunning the analysis using 2SLS rather than 3SLS, replacing the dependent variable in Eq. (1), cost per patient in revenue departments, with cost per patient in all departments, and replacing the total weighted procedures per patient with the corresponding raw total procedures in Eqs. (1) and (2). Our results are generally robust to these changes except that in the case of the last change, LOS and OCCUP are no longer statistically signi®cant, consistent with our overall result that the reduction in LOS did not produce a reduction in cost. We also reran the Section 4 analysis using a reduced form estimation to avoid the loss of degrees of freedom associated with the simultaneous equation approach. Again, the ordinary least squares (OLS) regression results are consistent with the pro®ling policy having no statistically signi®cant eect on the cost per patient day. Finally, to check the simultaneity speci®cation, we ran the Hausman test (1978, pp. 1264± 1269) of the null hypothesis that all of the variables are exogenous and that the coecients obtained by using OLS and 2SLS are equal. The Hausman test results for Eqs. (1)±(3) indicate rejection of the null hypothesis that the variables of concern are exogenous for Eqs. (2) and (3).
mid-1990s is remarkable ± from the high of 15.1% in the 1970±1980 period to a low of 3.2% in 1997 (Smith et al., 1999, Exhibit 1 p. 87). Unfortunately, this downward trend may be already beginning to reverse, with the premium rate increases projected for 1999 and 2001 at 7.7% and 8.4%, respectively (Smith et al. 1999, Exhibit 1 p. 87). The reasons for the expected upward spike in health insurance costs over the coming years vary. Arguably, managed-care programs may have matured, may be operating on thin margins and, thus, may need to raise prices (see the discussion in Armour, 1998; Hilzenrath, 1998). It is also possible that employers may have achieved most of the potential savings by changing from traditional insurance to managed-care plans (see the comments in McGinley, 1998).
As he notes in his Guest Editorial in this issue (Mensah, 2000), Professor Yaw M. Mensah was Special Associate Editor for this special theme issue. In his guest editorial Mensah (2000) suggests that there is need for further re- search relating to healthcare andaccounting (see Mensah, 2000; also see Mensah, 1996). We agree with his comment. Accordingly, researchers are en- couraged to submit manuscripts in that general area for editorial consideration at the JournalofAccountingandPublicPolicy.
Our paper diers from previous studies in two ways. First, in addition to the aggregate dierence between LAS and IAS earnings, we analyzed the value relevance of individual reconciling items of local GAAP to IAS earnings. More importantly, we conducted our analysis separately for two investor groups, foreign and domestic investors, who hold equity shares in these companies. During the period 1984±1992, the HeSE (Helsinki Stock Exchange) was a partially segmented stock market as unrestricted shares (available to both foreign and domestic investors) and restricted shares (available only to do- mestic investors) of the same set of companies were listed separately. At the same time, a number of ®rms listed on the HeSE provided non-mandatory local (Finnish) GAAP earnings and their reconciliations to IAS (see Kasanen et al., 1996, p. 286). This institutional setting is unique, and it allowed us to compare directly dierences in the value relevance of the two earnings mea- sures across the two dierent investor groups.
Our study uses data from Consolidated Reports of Condition and Income (call reports) provided by the Federal Deposit Insurance Corporation (FDIC) for each bank insured by the FDIC. For more than 11,000 banks, we examined four accountingand ®nancing choices that allow bank managers substantial discretion over the amount recorded. The four choices are the amount of the loan loss provision, loan charge-os, securities gains and losses, and dividends. For each choice, we ®t an ordinary least squares (OLS) regression model with variables to control for non-discretionary behavior and examine whether the remaining (discretionary) variation is associated with capital level, earnings, and marginal tax rate. The OLS regression equations were estimated as a system of simultaneous equations using seemingly unrelated regression (SUR) for each sample year 1987, 1988, and on a pooled basis. During this period, capital adequacy regulations did not vary (Koch, 1995, p. 389; Collins et al., 1995, p. 266).