Financial Disclosure Management by NPO

Published on January 2017 | Categories: Documents | Downloads: 48 | Comments: 0 | Views: 268
of 46
Download PDF   Embed   Report

Comments

Content

Financial Disclosure Management by Nonprofit Organizations 1
Ranjani Krishnan, Michelle H. Yetman, Robert J. Yetman*
Eli Broad College of Business, Michigan State University, East Lansing, MI 48824. Tippie College of Business, The University of Iowa, Iowa City, IA 52240

______________________________________________________________________________ Abstract This paper examines how nonprofit organizations respond to incentives to manage their publicly available financial information. Prior research identifies two operating ratios donors commonly use to evaluate the efficiency and effectiveness of nonprofits (i.e., the program service ratio, defined as the fraction of total expenses committed to advancing the charitable mission of the organization, and the fundraising ratio, defined as the ratio of fundraising expenses to donations revenue). Nonprofit managers have an incentive to over-report the expenses classified as program services and under-report the expenses classified as administrative and fundraising in order to improve these ratios. We examine whether nonprofits respond to these incentives, and we find evidence consistent with opportunistic cost shifting to improve the program service and fundraising ratios. Additional analysis finds that smaller nonprofits that are more reliant on donations revenue manipulate their operating ratios to a greater extent. JEL classification: M4; L3 Key words: Nonprofit organizations, earnings management, disclosure, hospitals. ______________________________________________________________________________

*Corresponding author. Tel.: (319) 335-0841; fax (319) ; email: [email protected] We thank Ashiq Ali, Ramji Balakrishnan, Leslie Eldenburg, Lil Mills, Shiva Sivaramakrishnan, and workshop participants at the University of Arizona and Texas A&M University for their helpful comments.
1

1. Introduction This paper examines how nonprofit organizations respond to incentives to manage their publicly available financial information through accounting expense allocations. Although nonprofit organizations do not have an observable stock price, there are nonetheless significant contractual incentives for financial disclosure management. In particular, implicit contracts with donors provide nonprofits with an incentive to appear efficient in raising their donations and in allocating their resources to charitable, rather than administrative, outputs. Typically, two ratios are used to measure a nonprofit organization’s donative and operational efficiency. Nonprofit financial statements aggregate all expenses into one of three categories (i.e., program services, fundraising, and administrative), and the ratios are based on these expense categories. The first ratio, known as the program service ratio, is program service expenses divided by total expenses. Because program service expenses are those directly related to the nonprofit’s primary charitable output (as opposed to either administrative or fundraising expenses), this ratio measures a nonprofit’s operating effectiveness. The second ratio, known as the fundraising ratio, is fundraising expenses divided by donations revenues and captures a nonprofit’s fundraising efficiency. Numerous industry watchdog groups suggest that donors use these two ratios to determine which charities are “worthy” of receiving donations, and a growing body of empirical research finds evidence consistent with donors responding to information contained in these ratios (Weisbrod and Dominguez 1986; Posnett and Sandler 1989; Callen 1994; Tinkelman 1999; Okten and Weisbrod 2000; Yetman and Yetman 2002; Baber, Daniel and Roberts 2002). To the extent that nonprofit managers believe that donors respond to these ratios, there is an incentive to allocate expenses out of the fundraising and administrative categories and into the

1

program services category in order to improve reported operating ratios. Although these kinds of accounting expense allocations have long been suspected, they have never been empirically documented (Khalat and Hueslein 1992). For example, an article in Forbes states, “Moreover from a public relations standpoint, charities have every incentive to maximize the charitable commitment figure and minimize reported fundraising expenses. Dubious accounting isn’t rare.” (Barret, 1999). Our paper provides empirical evidence consistent with nonprofits using accounting expense allocations to overstate their program service expenses and understate their fundraising expenses, thereby making themselves more attractive to potential donors. Further analysis estimates the magnitude of these accounting expense allocations, and examines the effects of political costs and reliance on donations as a revenue source on a nonprofit’s propensity to make expense allocations. This research is of potential interest for several reasons. First, nonprofits provide a natural setting (free of price-related incentives) in which to examine the effects of contractual incentives on financial reporting choices. Second, the effects of disclosure management on the quality and decision usefulness of publicly available financial information generally has not been addressed in the nonprofit setting. Nonprofit organizations, which constitute over 10 percent of national gross domestic product, are an important part of the United States economy (Independent Sector 2001). Annual donations received by nonprofits total over $200 billion (American Association of Fundraising Council Trust for Philanthropy, 2002). To the extent that donors make resource allocation decisions based on reported financial information and are unable to disentangle disclosure management, economic resources are potentially misallocated. Third, a small but growing body of research uses these ratios for empirical analysis without considering the implications of managerial manipulation (Baber, Roberts, Visvanathan 2001; Yetman 2001;

2

Baber, Daniel, and Roberts 2002;). By documenting the extent and method of nonprofit disclosure management, we provide valuable information for future research. We conduct our analysis using two databases. The first database contains matched sets of IRS 990s and state regulatory reports for a pooled sample of 719 observations from California hospitals for the years 1994 to 1998. The IRS 990, which is freely available for all nonprofits, is considered to be the principal publicly available source of financial information for nonprofit organizations (Joint Committee on Taxation, 2000). The state regulatory reports are available from the California Office of Statewide Health Planning and Development and are referred to as OSHPD reports. For reasons discussed later, we believe that the OSHPD reports are not likely to be used by donors to evaluate a particular hospital’s operating efficiency or effectiveness and therefore managers have less incentive to manipulate their OSHPD reports. We compare the amounts of program service expenses reported by the hospitals on their OSHPD reports with the amounts reported on their publicly available IRS 990 financial statements. Results of this analysis find that, although nonprofit hospitals in California report the same amounts of total expenses (i.e., the sum of program service, administration, and fundraising) on their OSHPD and IRS 990 forms, they report an average of $13.9 million more program service expenses (and $13.9 million less administrative and fundraising expenses) on their IRS 990 forms than on their OSHPD reports. These expense allocations increased the average program service ratio from 68 percent to 83 percent, causing the average organization to appear to be more operationally effective in that a larger proportion of total expenses are for charitable purposes. This result is consistent with nonprofit managers understating the amounts of program expenses reported on their publicly available financial reports (i.e., the IRS 990s) in order to appear more operationally efficient to potential donors.

3

The advantage of our first empirical test is that it measures disclosure management by comparing data on two sets of matched financial reports as self-reported by the nonprofits (i.e., each organization acts as its own control). The disadvantage of our first test is its limitation to a relatively small sample (i.e., 719) of California hospital observations. To overcome this limitation, we conduct a second empirical analysis using a second database of over 100,000 organization-year observations from 1982 to 1998, which includes educational, medical and charitable nonprofits in all 50 states (plus the District of Columbia). Because the only publicly available data for this larger sample is the IRS 990, we employ an estimation technique to determine expected expenses, and then compare this expected amount with the amounts reported on the IRS 990s, with the difference representing accounting expense allocations. Our estimation technique uses a regression model based on prior research to derive nonprofit-level estimates of fundraising expenses. We examine only fundraising expenses with this larger sample (rather than program services or administrative expenses) because prior research supplies models that can be used to estimate fundraising expenses. Results of our second analysis show that educational, medical, and charitable nonprofits under-report an average of $1,656 thousand, $129 thousand, and $227 thousand, respectively, in fundraising expenses on their IRS 990s. These expense allocations decreased the average fundraising ratios for educational, medical, and charitable nonprofits by 16 percent, 3 percent, and 7 percent, respectively, causing the average organization to appear to be more efficient in raising donations. Our third set of empirical tests is based on our observation that almost 40 percent of all nonprofits that report earning some amount of private donations (i.e., from individuals or corporations) report zero fundraising expenses. We investigated these nonprofits further because

4

it seems unlikely that such a large percentage of nonprofits could earn substantial private donations (average of $1.8 million each) with zero fundraising effort. To conduct this analysis, we randomly selected a sample of 10 percent of nonprofits that report zero fundraising expenses yet earn private donations and searched their Internet web pages for evidence of fundraising activities. Of the 349 organization web pages searched, we found evidence of fundraising activities for 130 organizations (37%). The instructions to the IRS 990 require that any and all expenses related to fundraising activities, such as those we found on the internet, be reported as fundraising. The results of this analysis imply that a substantial portion of nonprofits that report zero fundraising expenses actually do incur some amount fundraising expenses. This analysis corroborates the findings of our two previous tests that nonprofits allocate costs among various accounting categories so as to appear more operationally effective and efficient to potential donors. In the fourth and final portion of our analysis, we test the prediction that accounting expense allocations are negatively related to political costs (as measured by size) and positively related to the relative importance of donations as a revenue source. Results generally support our predictions. Our results have at least two implications. First, as with earnings management by for-profit entities, nonprofit financial disclosure management accomplished through accounting manipulations has potential resource allocation implications. Second, the results suggest that future nonprofit research using the program service or fundraising ratios must account for possible managerial manipulation. Prior research examined earnings (i.e., net income) management by nonprofit hospitals and finds that they smooth their net income and do not avoid reporting losses (Leone and Van Horne

5

2001). We differentiate our research in several ways. First, we focus on the manipulation of operating ratios rather than on net income because, in the nonprofit setting, operating ratios are a more commonly used evaluation metric. Prior research suggests that, since a nonprofit’s value is not reflected in a price metric, nonprofit objective functions generally do not include net income maximization, but rather include maximizing charitable output and meeting donor expectations (Rose-Ackerman 1996). Second, we consider all kinds of nonprofits, not just hospitals. Finally, we develop and test predictions about the characteristics associated with disclosure management; while prior research does not. The next section reviews the nonprofit financial disclosure environment. Section III discusses our data; Section IV presents our empirical analyses; Section V provides robustness tests, and the final section our conclusion.

2. Nonprofit Financial Disclosure Environment 2.1. The Agency Problem in Nonprofit Organizations Public demand for nonprofit financial statement data is at least partially driven by the inherent agency conflict within the nonprofit setting. The economics literature typically assumes that not-for-profit firms have different objectives and behave differently from for-profit firms. Much of the work in this area draws on the seminal research of Arrow (1963), who suggests that nonprofit hospitals arise in response to the asymmetry of information between patients and providers of healthcare. Fama and Jensen (1983) argue that nonprofits arise for particular forms of organizations such as charities because unrestricted donations pose agency problems for any organization with residual claimants. In order for donors to be assured that residual claimants do not consume their resources, the nonprofit form arises, which has no alienable residual claims.

6

Fama and Jensen (1983) conclude that “the absence of residual claims avoids the donor-residual claimant agency problem and explains the dominance of nonprofits in donor-financed activities.” However, the absence of residual claimants does not automatically ensure that nonprofit managers will use donations in the most efficient manner. Managers can consume more resources than efficient production requires for various reasons. For example, managers can consume excessive perquisites and expend resources in activities peripheral to the stated mission of the organization. Also, in some cases efficient production requires additional effort by the manager who would otherwise be unwilling to expend that effort in the absence of high-powered incentive contracts. Donors have an implicit contract with nonprofits in that donors supply net cash flow in exchange for the nonprofit efficiently and effectively providing some charitable output with the provided funds. Accounting information can assist donors in monitoring their implicit contracts by providing a means (i.e., the fundraising and the program services ratios) for donors to evaluate whether the nonprofit is using their donations in the most efficient and effective manner. The following sections discuss the financial disclosure requirements for nonprofit organizations that facilitate financial analysis of nonprofits by donors.

2.2. IRS 990 Financial Disclosures Federal regulations require that nonprofits annually prepare and disclose their IRS 990, which contains typical financial statements, including a statement of revenues and expenses and a balance sheet, as well as a substantial amount of other information related to the nonprofit’s charitable purpose and activities. IRS 990 data on most nonprofits is freely available on the Internet at www.guidstar.org. Congressional reports suggest that the intent of the IRS 990 is to

7

provide the public with the necessary information to evaluate the performance of a nonprofit and that the IRS 990 is the primary source of publicly available nonprofit financial information (Joint Committee on Taxation 2000). Nonprofit financial statements on the IRS 990 group all expenses into one of three accounting categories: program service, administrative, and fundraising expenses. Instructions to the IRS 990 indicate that fundraising expenses are any and all incurred either directly or indirectly to generate, maintain, or keep track of government grants or donations, both from the general public and from “feeder” organizations such as the United Way. Program service expenses are those incurred to further the charitable mission of the nonprofit, such as professors’ salaries at nonprofit universities, direct patient care at nonprofit hospitals, and specimen acquisitions at nonprofit zoos and aquariums. Expenses that are not properly classified as program service or fundraising are considered to be administrative. Overhead expenses, which are shared among the three categories of expenses, have to be allocated using a reasonable method of cost allocation. Although nonprofits can elect to undergo financial statement audits, there is no requirement that these audited financial statements be made public, nor is there any IRS requirement that the IRS 990 conform to audited financial statements. Although most accounting rules for both the IRS 990 and for financial statements presented in accordance with Generally Accepted Accounting Principles coincide, there are some differences. For example, the IRS 990 does not use materiality thresholds, any amount of fundraising must be separately stated. Industry watchdogs use information from the IRS 990s to compute the program service and fundraising ratios and make annual recommendations to the public via printed publications and the World Wide Web. The largest watchdog groups include the American Institute of

8

Philanthropy (AIP), the Philanthropy Advisory Service (PAS) of the Council of Better Business Bureaus (CBBB), and the National Charities Information Bureau (NCIB). In addition to these agencies, the popular press (e.g., Forbes , Money, and The Wall Street Journal) utilize IRS 990 data in news articles and/or in annual lists comparing nonprofits’ relative program service and fundraising ratios. 2, 3

2.3. OSHPD Financial Disclosures In addition to IRS disclosure requirements, hospitals providing medical care in the state of California must provide additional disclosures to state regulatory authorities. The California Office of Statewide Health Planning and Development (OSHPD), a department of the California Health and Human Services Agency, collects, analyzes, and disseminates data on hospitals licensed in California. The information collected by OSHPD includes financial reports, patient statistics, and usage statistics. Expenses on the OSHPD report are provided in the following format: The OSHPD data include totals (and some details) for general service, administrative, and program service expenses, but does not generate a separate total for fundraising expenses. General services include expenses such as printing, non-patient food services, grounds, security, parking, accounting, and data processing. Administrative expenses comprise hospital administration,

2 3

For example see “Giving Smartly: A Guide to Charities 2001” in http://www.forbes.com/charities. For example, Barrett (1999) in a Forbes article titled “Look Before you Give” reports, “The Cancer Research Institute funneled 89% of its budget to scientific inquiry, much greater than the 62% of the far larger American Cancer Society, or the 56% of the smaller Cancer Fund of America. Much of the difference had to do with fundraising. Only 3 cents of every dollar raised by the Cancer Research Institute are spent on generating the donations. At the Cancer Fund, 24 cents go out the door that way, and at the American Cancer Society, 30 cents.”

9

personnel, library, staff administration, and public relations. Program service expenses 4 as per the OSHPD data include daily hospital services such as medical/surgical intensive and acute care, coronary care, pediatric care, obstetrics, and psychiatric care; ambulatory services such as emergency room, medical transportation, and home health; ancillary services such as radiology, occupational therapy, and renal dialysis; research expenses; education costs; insurance and malpractice; and other operating overhead and miscellaneous costs not classified as administrative or general expenses. There are two primary users of the OSHPD reports: First, the Health Planning and Policy (HPP) Division of OSHPD, which conducts research on issues related to healthcare cost containment, access to needed services, and improving the quality of care, demands a great amount of accurate, detailed information. Second, health care researchers and other medical policy makers also rely on the data. Therefore, OSHPD reporting guidelines are very detailed (the instruction manual is about 1,000 pages) and specific for each line item. We believe that hospitals are less likely to manipulate their operating ratios on the OSHPD reports as compared with their IRS 990s for four reasons. First, the OSHPD data have to be purchased at a cost of $375 per year; while the IRS 990s are freely and easily available to potential donors. Second, the OSHPD reports contain large amounts of complex data (hundreds of accounts) that must be aggregated across various categories and cost centers in order to compute a nonprofit’s performance ratios. In contrast, the IRS 990 contains only three categories of expenses which can be easily used to construct the operating ratios (all data necessary to calculate the two ratios are clearly presented on page one of the IRS 990). Third,

4

The OSHPD database does not specifically refer to these expenses as program service expenses. However, consistent with the IRS 990 classification, all expenses other than administrative, fundraising, and general are program service expenses.

10

the popular press and all watchdog agencies use IRS 990 data in forming their recommendations. Finally, the OSHPD data are available only for California hospitals; while the IRS 990 data are available for all types of nonprofit organizations located throughout the country. Donors are likely to be interested in comparing a variety of nonprofits in a number of locations before making decisions. It is also possible that, rather than overstating the amount of program service expenses on the IRS 990, California hospitals are actually understating their program service expenses on their OSHPD reports, although we are not aware of any incentive to manipulate the OHSPD reports in this manner. To the contrary, it is possible that there are incentives, such as expense reimbursement policies or state regulatory scrutiny, for hospitals to overstate, rather than understate, the program expenses on their OSHPD reports. The effect of overstated OSHPD program expenses is to bias our tests against finding results. The differing opportunities and incentives for earnings management between the IRS 990 and the OSHPD reports guide our interpretation that the reported expenses on the OSHPD data are less managed compared with the financial information on the IRS 990. However, to the extent that hospitals do simultaneously overstate their program services on both their OSHPDs and IRS 990s, our estimates of expense allocations will be biased downward, conservatively influencing our results.

11

3. Data 3.1. Sample Selection 3.1.1. Sample of California Nonprofit Hospitals Our pooled cross-sectional sample of California hospitals is based on the intersection of two databases: the OSHPD Hospital Financial data and the IRS 990 data from the “Statistics of Income” (SOI) files. The SOI files contain data on all 501(c)(3) organizations with more than $10 million in assets plus a random sample of approximately 4,000 smaller organizations and include most financial variables on the IRS 990. The SOI data files were obtained from the National Center for Charitable Statistics (NCCS). The NCCS, which is a project of the Center on Nonprofits and Philanthropy (CNP) of the Urban Institute, is the national repository of data on the nonprofit sector in the United States (http://nccs.urban.org). We matched each nonprofit hospital in the OSHPD database with each nonprofit hospital in the IRS 990 database. Since the two databases have different firm identifier numbers (a unique 9-digit hospital number in the OSHPD data and a unique 9-digit Employer Identification Number (EIN) in the IRS 990s), we matched by hospital name and zip code. Because external changes in the hospital industry make the analysis of a large panel of data noisy, we restrict our analyses to the most recent five years for which data are available (1994-1998). The matching process produced 719 hospital-year observations. The number of hospitals per year ranges from 167 in 1994 to 122 in 1998. Although there are about 262 nonprofit hospitals in California, not all hospitals are represented in the IRS 990 database. Our sample is less likely to include very small hospitals (under 25 beds), because the IRS 990 data only contains a random sample for hospitals with less than $10 million in assets.

12

3.1.2. Larger Sample of Educational, Medical, and Charitable Nonprofits The second sample is based on the pooled cross-section of all educational, medical, and charitable nonprofits contained in the IRS SOI sample files for the years 1982 to 1998, for a total of 163,093 organization-year observations. We reduce our sample by 39,383 observations which report zero donations. We further reduce our sample by 2,590 observations by trimming off the top and bottom one percent of the most extreme values of all analysis variables, although results are robust to no trimming and further trimming up to 10% of extreme observations. After these data screens, our sample contains 25,501, 37,602, and 48,017 educational, medical, and charitable nonprofit-year observations, respectively.

3.2. Sample Characteristics 3.2.1. California Nonprofit Hospitals Table 1, Panel A provides descriptive statistics for the hospital sample. The average number of staffed beds is 209; the percentage of Medicare patients is 45.41 percent; the percentage of Medi-Cal (i.e., the California state Medicaid program) patients to total patients is 20.82 percent and the occupancy rate is 59.54 percent. These data suggest that the hospitals represented in the sample are representative of the overall population of nonprofit hospitals in California (average staffed beds = 192, percentage of Medicare patients = 41, percentage of Medi-Cal patients = 21, and percentage occupancy rate=59). Hospitals in the sample provide charity care of about 2.4 percent of revenue. The average net income margin as a percentage of revenue is 3.51 percent. A large percentage of revenue is from program services (93.9 percent). Donations constitute only 2.8 percent of revenue. However, because average net income margins are only 3.51

13

percent, donations are likely to make a difference between the hospital’s breaking even versus incurring a loss.

3.2.2. Educational, Medical, and Charitable Nonprofits Table 1 panel B reports descriptive statistics for our sample of 25,501, 37,602, and 48,017 educational, medical, and charitable observations, respectively. Educational nonprofits receive the largest average dollar amount of donations, followed by charitable nonprofits. Medical nonprofits’ average program service revenues (i.e., the sales of products and services) is over 43 times as large as average donations, suggesting that medical nonprofits rely primarily on program revenues. Charitable nonprofits earn roughly equal amounts of donations (mean of $3 million) and program revenues (mean of $4 million).

4. Empirical Analysis and Results 4.1. California Nonprofit Hospitals Expense Allocations In this section, we compare the amount of program service expenses as reported on a nonprofit’s IRS 990 to those reported on its OSHPD report for a matched set of 719 California hospitals. Note that this analysis can determine whether excess expenses were allocated to program services, but it cannot determine whether those expenses were allocated from fundraising or administrative expenses because the OSHPD database aggregates both fundraising and administrative expenses into a single expense category. The program services ratio as reported on the IRS 990 is program service expenses divided by total expenses. We compute the program service ratio using the OSHPD data by aggregating expenses across the following expense centers: patient care expenses; research expenses;

14

education expenses; insurance, malpractice, and other operating overhead; and miscellaneous costs not classified as administrative or general expenses. Using the means from the sample, we find that the average California hospital reports $13.9 million more program service expenses on their IRS 990 than on their OSHPD report (and correspondingly $13.9 million less fundraising and administrative expenses on their IRS 990 than on their OSHPD report). The top portion of table 2 contains a two-sample matched pair ttest of the difference between reported program service ratios and program service expenses as per the IRS 990 and the OSHPD data for our sample of California Hospitals. The mean differences are highly significant across all the years (t=28.3, p<0.001), and also for each individual year (untabulated), with no discernable pattern observable across the years. These results suggest that nonprofit hospitals overstate their program service ratios on their IRS 990s by overstating their program service expenses. These expense allocations increased the average program service ratio from 67.9 percent to 82.8 percent, making the average organization appear to donors to be more effective at delivering its charitable output.

4.2. Educational, Medical, and Charitable Nonprofits Expense Allocations For the larger sample of nonprofit organizations, there is no publicly available data source (similar to the California OSHPD reports) that we can use as a benchmark of the amount of expenses nonprofits should report as program services, fundraising, or administrative. Therefore, we employ an alternative procedure in which we first estimate, using regression analysis, the fundraising expenses we would expect a nonprofit to report on its IRS 990, and then compare that estimate to the fundraising expenses actually reported on the IRS 990, with the difference representing our estimate of allocations out of fundraising. For this analysis, we focus

15

on fundraising expenses only (and not on program services or administration) because prior research provides us with a means of estimating a particular nonprofit’s fundraising expenses. The model we use to estimate nonprofits’ expected fundraising expenses follows those used by Weisbrod and Dominguez 1986, Posnett and Sandler 1989, and Okten and Weisbrod 2000, as well as by many others. The intent of these models is to estimate the determinants of private donations to nonprofits, and they generally take the form:

Private donationsit = α + β1 Fundraising expensesit-1 + β2 Priceit-1 + β3 Ageit-1 + β4 Program revenuesit-1 + ε i. (1)

Unlike either program service expense or administrative expense, fundraising expense has a direct revenue counterpart (i.e., donations revenues). Because of this relationship, it is possible to use regression analysis to derive estimates of fundraising expenses. Note that a similar procedure could not be used to estimate program service expenses or administrative expenses because they have no direct revenue counterparts. The dependent variable, Private donations, is the dollar amount of donations from individuals and corporations and excludes government grants and feeder donations, such as those from the United Way. Fundraising expenses are those incurred, directly or indirectly, to generate or otherwise manage donations. Fundraising can affect donations in two ways. First, fundraising will increase donations to the extent it increases donor awareness and communicates information about the quality of the nonprofit’s output (Nelson 1974). Second, excessive fundraising will reduce donations if donors are concerned with the portion of their donation that is spent on additional fundraising. The variable Price is defined as 1 / (1 - Fundraising expenses / Private donations) and captures donor sensitivity to

16

nonprofits’ fundraising efficiency. The variable Age is the number of years a nonprofit has existed and is intended to measure a nonprofit’s stock of reputation capital. We include Program revenues to control for possible crowding-in or crowding-out effects of this alternative revenue source. All variables (except for Age) are scaled by assets, although our results are qualitatively robust to alternative specifications that use total expense or total revenue scaled variables as well as using logarithmically transformed variables. All independent variables are lagged, because a donor does not have access to nonprofit financial information until the end of the year. Consistent with prior research, we interpret the coefficient estimate β 1 as the average dollar amount of donations generated by one dollar of fundraising expenses. We derive an organization-specific estimate of the fundraising expenses a nonprofit should report on its IRS 990 by dividing a nonprofit’s private donations by our coefficient estimates (i.e., if the coefficient estimate were four, suggesting that one dollar of fundraising generates four dollars of donations, we would divide the nonprofit’s reported amount of donations by four to derive an estimate of the fundraising expenses the nonprofit incurred). We then compare our organizationlevel estimate of fundraising expenses based on model (1) with those reported on the nonprofit’s IRS 990, with the difference representing our estimate of a nonprofit’s allocation of expenses out of the fundraising category. By using this estimation methodology we make several assumptions. First, we assume that nonprofits’ private donations revenue generation process is constant within nonprofit type (i.e., educational, medical, or charitable). Second, we assume that all fundraising expenses are incurred to generate private donations and not feeder or government grants (i.e., the dependent variable is only private donations; whereas the independent variable Fundraising expenses is all fundraising expenses). Although it is possible that nonprofits make fundraising expenditures to

17

raise grants, we make this assumption for two reasons. First, prior research has not recognized any link between grants and fundraising expenses, leaving us with no prior model of the relationship between fundraising and grants. Second, several institutional factors render the association between fundraising expenses and grants difficult to determine at best and tenuous at worst. 5 To the extent that fundraising expenses are incurred to generate either feeder donations or government grants, our estimates of expense reallocations are conservatively biased. Table 3 contains the results of estimating equation (1). We estimate the model controlling for a first-order auto-regressive process as well as general heteroscedasticity. 6 We find that each dollar of fundraising is associated with $1.94, $4.29, and $5.58 of private donations for educational, medical, and charitable nonprofits, respectively. The magnitude of these results are similar to those found in prior studies using similar models (Weisbrod and Dominguez 1986; Posnett and Sandler 1989; and Okten and Weisbrod 2000). Because a working supposition of our paper is that the reported amount of fundraising expenses are systematically understated, equation (1) contains a regressor with systematic measurement error. Econometrically, this measurement error is reflected in the error term, ε, and if ε is correlated with the measured variable (i.e., Fundraising expenses), then ß1 is biased. The direction of the bias depends on the directions of the covariance between the measurement error and both Fundraising expenses and Private donations; if the direction is equal (unequal), then ß1

5

Our discussions with grant writers at various charities and hospitals around the country suggests that feeder donation and government grant contracts usually span several years, with the majority of the “fundraising expenses” being front-loaded, causing the relationship between grants and fundraising expenses to be difficult to estimate. In addition, grants are usually less costly to acquire than private donations because the fundraising effort is directed at a particular agency; whereas the generating efforts for private donations are comparatively broad. 6 A Durbin-Watson tests suggests that our model error terms exhibit first order serial correlation. A Lagrange multiplier test shows that the errors exhibit non-constant variance. We use a generalized autoregressive procedure (i.e., GARCH) with a long memory process (i.e. all past squared residuals are used to estimate the current error variance). As compared with using ordinary least squares , the GARCH procedure produces higher expense allocation estimates for educational nonprofits, and qualitatively similar allocation estimates for medical and charitable nonprofits.

18

is biased upwards (downwards). Given that fundraising is performed to generate donations, it seems reasonable that the covariance between Fundraising expenses and Private donations is positive, biasing ß1 upwards and our results downwards. Additionally, it seems reasonable that the larger the true fundraising expenses, the more opportunity there will be for unreported fundraising expenses, making an additional case for a positive covariance between reported fundraising and the measurement error which would again conservatively bias our results. In sum, the expected covariance structures predict an upwards bias on ß1 , causing us to derive too high an estimate of “true” fundraising expenses, which when compared with reported fundraising expenses on the IRS 990, would result in an understatement of the allocated fundraising expenses producing a mechanical conservative bias in our results. Although this analysis suggests that the presence of systematic measurement error in reported fundraising expenses of the type we suspect exists would cause a conservative bias in our results, we supplement that analysis with simulation results. To conduct our simulation analysis, we construct plausible models of the measurement error and directly estimate their effects on our equation (1) regression estimates. Our error models take the form:

True fundraising expenses = φ Reported fundraising expenses + γ, True fundraising expenses = Reported fundraising expenses λ + η,

(2) (3)

where φ and λ are greater than 1, and γ and η are random errors with mean zero. We conducted analysis holding φ and λ cross-sectionally constant as well as permitting them to randomly vary across organizations. These models assume that the systematic error component is increasing [linearly in equation (2) and non-linearly in equation (3)] in the amount of true fundraising

19

expenses. 7 We consider these relationships to be plausible representations of the actual systematic error, although it is not possible to determine ex-ante the exact nature of the error. Our simulation results (untabulated), which varied the parameters φ and λ across a wide range of positive numbers, find that the errors represented by equations (2) and (3) produce coefficient estimates of ß1 that are systematically biased upwards, in turn biasing our estimated expense allocations downward. These simulation results support our analytical error analysis above and suggest that, if one assumes that our error models are descriptive, our results contain a conservative bias of unknown magnitude. This inherent conservative bias, in conjunction with the previously discussed conservative bias caused by assuming that no fundraising expenses are incurred to generate grants, suggests that our estimations of fundraising expense allocations are a lower bound and should be interpreted accordingly. The bottom portion of table 2 reports our aggregate estimates of nonprofits’ fundraising expenses and compares those estimates with the fundraising expenses reported on the IRS 990. The average educational nonprofit in the full sample reports a mean of $493 thousand of fundraising expenses. Using our estimation methodology, our expectation of the mean fundraising expenses for educational nonprofits is $2,320 thousand, suggesting that educational nonprofits allocate a mean of $1,827 thousand in expenses out of the fundraising expense category. We estimate that the average medical nonprofit allocates a mean of $168 thousand of expenses out of the fundraising category;

It is well known that random measurement error (i.e., γ and η) biases coefficient estimates towards zero. In an attempt to control for the random portion of the measurement error in fundraising expenses, we replicated both our simulation and primary results using two-stage least-squares where the value of fundraising that enters the estimation model is the fitted value from a regression of fundraising expenses on the exogenous variables in equation (1) as well as lagged values of fundraising expenses (up to three lags were included). The general inferences of the results were not altered by the two-stage procedure.
7

20

while the average charitable nonprofit allocates a mean of $299 thousand of expenses out of the fundraising category. Median values show similar patterns, although with smaller magnitudes. We estimate that educational nonprofits reduced their average fundraising ratios using accounting expense allocations from 21.0 percent down to a reported figure of 2.8 percent. Medical nonprofits used accounting expense allocations to reduce their average fundraising ratios from 5.2 percent down to 1.5 percent; while charitable nonprofits reduced their average fundraising ratios from an estimated 13.3 percent down to a reported 3.3 percent. To the extent that our estimation methodologies are reasonable, our results are consistent with a behavior that has long been suspected by the popular press. If donors rely on the reported ratios to make their donations decisions, and are unable to disentangle managerial expense allocations, then financial disclosure management by nonprofit organizations likely results in economic misallocation of resources.

4.3. Analysis of Nonprofits that Report Zero Fundraising Expenses This analysis examines the portion of our samples that, despite earning some amount of private donations, report zero fundraising expenses. We focus on private donations because, as compared to feeder or government grants, they are the kind most likely to be associated with continuous fundraising efforts. Table 4 shows that 95 of 122 California hospitals in the 1998 sample report earning private donations but record zero fundraising expenses. For the larger sample, we find that 25 percent, 46 percent, and 33 percent of educational, medical, and charitable nonprofits, respectively, that earn private donations report zero fundraising expenses. The average dollar amount of private donations received by a nonprofit that reports zero fundraising expenses is over $1.8 million.

21

Although it is possible that a nonprofit could earn private donations without incurring any fundraising effort, it seems unreasonable that such a large proportion of nonprofits could earn such significant amounts of private donations with zero effort. In an attempt to determine if any of these nonprofits actually do incur some amount of fundraising expenses, we conducted a search of the nonprofits’ Internet web pages for evidence of fundraising activities. We examined web pages for all of the 95 California hospitals from the year 1998 (most recent year) that reported earning private donations but reported zero fundraising expenses. Rather than examining all 5,000 web pages of the larger set of educational, medical, and charitable nonprofits for the year 1998, we randomly selected a sample of 10 percent of the observations in the sample year 1998 (most recent year) for each of the three nonprofit types. To perform our investigation we first accessed the nonprofit’s most recent IRS 990 at www.guidestar.org to make sure that it continues to earn private donations in its most recent year (i.e., a positive amount on page 1 line 1a of the IRS 990 for the year 2001) and still reports zero fundraising expenses (i.e., zero on line 15 of page 1 on the IRS 990 for the year 2001). For those nonprofits that report earning private donations and yet report zero fundraising expenses in their most recent year’s IRS 990, we searched the Internet (using the Yahoo search engine) for their web pages and examined their web pages for evidence of fundraising activities. We consider evidence of fundraising activities to be either direct solicitations for donations or providing instructions on how a potential donor could contribute. Table 4 reports the results of our web page search. For the 95 California hospitals in 1998 that reported earning private donations yet reported zero fundraising expenses we found evidence of fundraising for 19 hospitals (20 percent of web sites searched). For the larger sample of nonprofits we found evidence of fundraising activities for 26 of 38 educational nonprofits, 13 of

22

91 medical nonprofits, and 72 of 125 charitable nonprofits. If our samples are representative, these results suggest that over one-third of nonprofits that report zero fundraising expenses on their IRS 990s actually do incur some amount of fundraising. A list of the nonprofits that we found evidence of fundraising for is included in Appendix A. For many of the nonprofits we found evidence of substantial fundraising activities beyond simply soliciting donations. Some web sites had links to separate offices for fundraising development while many others contained the names of individuals whose job titles included “Director of Foundation and Government Grants” and “Director of Individual Giving”. It is possible that organizations engage in other types of fundraising efforts that are not revealed by an internet web page search (such as mailings or radio announcements). If true, our analysis provides a conservative estimate of fundraising expense under reporting. These results provide additional evidence that nonprofits under-report the true extent of their fundraising activities on their publicly available financial reports.

4.4. Characteristics Associated with Expense Allocations The purpose of this analysis is to use the estimated expense allocations from our prior analyses as dependent variables to test the hypotheses that accounting expense allocations are decreasing in political costs and increasing in a nonprofit’s reliance on donations as a revenue source. Prior research suggests that larger for-profit firms make more conservative accounting choices so as to avoid drawing unwanted attention to themselves (Watts and Zimmerman 1978), and we test this hypothesis in the nonprofit setting. The incentive to report more favorable fundraising and program service ratios is likely to be higher for those nonprofits that rely more heavily on donations as a funding source. We first conduct this analysis using the expense

23

allocation estimates for the sample of 719 California hospitals, and then using the allocation estimates we derived for the larger set of educational, medical, and charitable nonprofits. Although both analyses test the same research hypotheses, they include somewhat different control variables because the OSHPD dataset permits us to include an institutionally richer set of controls. Results are not qualitatively sensitive to the choice of control variables. We empirically test our predictions using the combined OSHPD and IRS 990 data on a sample of 719 California hospitals by estimating the following regression:

Program allocations it = α + β 1 Sizeit + β2 Marginit + β 3 Donations it + β4 LOSit + β5 Medicareit + β6 Medi-Calit + β 7Occupancyit + β 8Charityit + ε it, (4)

where Program allocations is a nonprofit’s reported program service ratio (i.e., program service expenses / total expenses) calculated using IRS 990 data less its program service ratio calculated using the OSHPD data; Size is total assets in $millions; Margin is net income divided by net revenue; Donations is donations revenue divided by total revenue; LOS is the average length of stay from admission to discharge; Medicare is the proportion of Medicare patients to total patients; Medi-Cal is the proportion of Medi-Cal patients to total patients; Occupancy is the actual patient days divided by available patient days and Charity is charity care expense divided by net revenue. Our primary variables of interest are Size and Donations, and we predict that the sign of the coefficient estimate will be negative for Size and positive for Donations. Medicare, Medi-Cal, LOS, and Occupancy are control variables. Prior health care studies suggest that these variables have an influence on hospital outcomes (e.g., for Size: Alexander and Lee, 1996; French, 1996; Mick and Wise, 1996; Robinson and Phibbs, 1989; for Medicare and Medi-Cal:

24

Dranove, 1988, Lynk, 1995; for LOS: Dranove, Shanley, and White, 1993; Lynk, 1995). The model includes, but does not tabulate, yearly fixed-effects. 8 To empirically test our predictions for the larger sample of nonprofits, we estimate the following regression:

Fundraising allocations it = α + β1 Sizeit + β 2 Marginit + β3 Donationsit + ε it.

(5)

Fundraising allocations is our estimate of a nonprofit’s fundraising expense allocations scaled by total revenues (to be consistent with equation (4) above). Size is total assets in $ millions; Margin is net income divided by revenue and Donations is private donations revenue divided by total revenue. Once again our primary variables of interest are Size and Donations, and we predict that the sign of the coefficient estimate will be negative on Size and positive on Donations. Margin is included as a control variable. Because the IRS 990 data do not include as much institutional detail as the OSHPD reports, we are not able to include as many control variables for our sample of medical nonprofits in equation (5) as for equation (4). We estimate equation (5) separately by nonprofit type and include (untabulated) yearly fixed-effects. Table 5 contains the results from estimating equation (4); while table 6 contains the results of estimating equation (5). Results in table 5 support our prediction that hospitals with a greater reliance on donations as a revenue source have larger expense allocations into program services, but they do not support our hypothesis that expense allocations are sensitive to political costs as measured by total assets. Although not a part of our hypothesis, the results for the charity care

8

Because we use time series data, we cannot eliminate the possibility of auto-correlation. To control for this, we use a linear auto-regression model whose error term is assumed to be an autoregressive process. Our models also control for conditional heteroscedasticity (ARCH) and generalized heteroscedasticity (GARCH).

25

variable are of interest. Hospitals that perform a greater proportion of charity care have larger differences in reported program expenses. Charity care is a measure of community service used by regulators to examine whether a nonprofit is deserving of its tax exemption (Bryce 2001, Potter and Longest 1994) 9 . Nonprofit hospitals can also manage their expenses in various ways, such as pricing the charity services at gross prices and including other community services under this category. These results show that hospitals having greater reported charity expenses also have greater reported program service expenses as per the SOI. With respect to the remaining control variables, the results demonstrate that hospitals with higher occupancy rates had greater differences in reported program service expense. Hospitals with a greater proportion of Medicare and Medi-Cal patients over-report program service expenses to a lesser extent. Table 6 contains the results of estimating equation (5) and, consistent with the above analysis, shows that fundraising expense allocations are increasing in the relative importance of donations as a revenue source across all nonprofit types. The results in table 6 also support our hypothesis that expense allocations are decreasing in political costs as measured by assets. In comparing the results of our analyses across samples, we find the magnitudes and signs of the donations coefficient using the larger sample of medical nonprofits (0.24) is qualitatively similar to that obtained from the model using only California hospitals (0.21).

9

The criteria set by the IRS for tax-exempt status by a hospital under Section 501(c)(3) as stipulated in its 1969 Ruling 69-545 and subsequent 1983 revision (83-157) require that: (1) the hospital should provide care to all insured patients, including those sponsored by government programs such as Medicare and Medicaid, (2) the hospital should provide fulltime availability of emergency room services to anyone needing them, including the indigent, (3) the hospital should select its board of trustees from the community, (4) the hospital should provide medical staff privileges to all eligible physicians, and (5) the operating surplus of the hospital should be applied to capital replacement, expansion, debt amortization, improvements in patient care, medical training, education, and research (United States Internal Revenue Service Ruling 69-454 (1969) and Ruling 83-157 (1983). Also see Bryce (2001) for a discussion.

26

5. Robustness Tests We conducted a series of analyses intended to test the robustness of our results to various assumptions as well as to provide cross-checks of our results across the various modeling procedures. In our first robustness test, we replicate equations (4) and (5) using alternative deflators including assets, total revenues, and total expenses for each independent and dependent variable in various combinations. Results were not qualitatively altered, and inferences were unchanged by using these alternative deflators. In our second robustness test, we replicated the fundraising expense allocation procedure contained in part 4.2, which was originally conducted on the larger sample of nonprofits, on the smaller sample of 719 California hospitals. Results of this analysis find that the smaller California hospital sample allocated an average of $331 thousand out of fundraising expenses. This compares with the results for the larger sample of medical nonprofits (primarily hospitals), which table 2 reports allocated an average of $168 thousand out of fundraising expenses. We then estimated equation (5) on this sample of 719 California hospitals and find that, similar to the results in table 5, the ratio of fundraising expense allocations to total expenses is increasing in a nonprofit’s reliance on donations as a revenue source but is not related to size. In our third robustness test, we replicate equations (4) and (5) using only those variables that were common to both samples (i.e., the IRS 990 and the OSHPD databases). These variables included total assets, total donations, and net income margin. Again, the general inferences of our results were not altered. As our fourth robustness analysis, we correlated the expense allocation measures derived using the California hospitals and those derived using the larger sample of medical nonprofits for the 719 hospitals that are common to both datasets. Recall that the allocation estimates

27

generated using the California hospital database (i.e., OSHPD) are estimates of expenses allocated into program services; whereas the expense allocation estimates generated from the larger sample analysis were estimates of expenses allocated out of fundraising. Although the two measures are based on different expense categories, they should nonetheless capture the same economic construct of financial ratio management using accounting expense allocations and, to the extent that expenses allocated out of fundraising were allocated into program services, the two measures should be statistically correlated. We find that the Pearson (Spearman) correlation between our two measures of expense allocations is 22 (15) percent and is significant at the 0.0001 (0.0016) level. This result provides some comfort that our estimation procedures capture common elements of nonprofit financial reporting behavior. In our next robustness test, we use the sample of 719 California hospitals to replicate the results of equations (4) and (5), but we switch the dependent variables. Equation (4) originally used the dependent variable constructed by comparing the OSHPD database with the IRS 990 database. We replicate equation (4) using as the dependent variable our fundraising expense allocations generated by the regression equation (1). In a similar manner, we re-estimate equation (5), which originally used the fundraising expense allocations generated by the regression equation (1) as a dependent variable, with the amount of expense allocations constructed by comparing the OSHPD database with the IRS 990 database. This procedure provides an integrity test of our two alternative measures of expense allocations. If the two measures are each independently capturing financial statement management, then they should produce similar results when swapped out for each other in subsequent empirical analysis. Results (untabulated) produce similar statistical inferences (although somewhat different coefficient magnitudes) as does the primary analysis in the paper.

28

As our final robustness test, we reclassify the hospital expenses on the OSHPD report based on what is required, under all circumstances, to be only administrative expenses rather than taking the reported amounts at face value. The purpose of this test is to address concerns that some of the expenses reported on the OSHPD dataset as administrative were due to the differences in reporting format between the OSHPD and the IRS 990s, and should have more appropriately been reported as program services. To conduct the reclassification of possible differences due to the differences in reporting format between the OSHPD and the IRS 990 databases, we reclassified all those expense items that could be possibly included as program services. While performing this reclassification, we used a donor perspective, i.e., we examined each expense item line-by-line and determined whether donors would prefer their money to be used in that expense rather than in direct program service expenses. We reclassified the following items out of administrative and general expenses and into program services from the OSHPD data: dietary, laundry and linen, social work services, central patient transportation, central services and supplies, pharmacy, auxiliary groups (hospital volunteer groups), chaplaincy services, medical records, nursing float administration, utilization management, and community health education. 10 Even after this strict reclassification, the reported administrative and general expenses as per the OSHPD data was greater (21.7%) than the reported administrative expenses as per the IRS 990s (17.2%). We then re-estimated all subsequent empirical analysis using these re-classified expense allocation estimates with inferences generally unchanged, although the magnitudes of the estimated expense allocations were smaller.

10

After the reclassification, the following expense items remained as a part of administrative and general expenses as per the OSHPD data: education administration office, student housing, printing and duplicating (non-patient), kitchen (non-patient), non-patient food services, purchasing and stores (excluding medical purchases), grounds, security, parking, housekeeping, plant operations (non-medical), plant maintenance, communications, data processing, accounting, credit and collection, admitting, registration, hospital administration, governing board expense, public relations, management engineering, personnel (non-medical), employee health, library, staff administration, insurance (excluding medical and malpractice), and employee benefits (non-payroll related).

29

6. Conclusions This paper examines how nonprofit organizations respond to incentives to manage their publicly available financial information as disclosed on their IRS 990s. The IRS 990 is considered to be the primary source of nonprofit financial information, and prior literature suggests that donors evaluate whether a nonprofit is “worthy” of their donations by examining two operating ratios based on IRS 990 data. We examine the extent to which nonprofits allocate expenses between various accounting expense categories so as to improve these ratios, thereby appearing more efficient and effective to potential donors. Because nonprofits do not have a stock price, they provide a natural setting (free of price related incentives) in which to examine the effects of contractual incentives on financial reporting choices. Our empirical analyses uses two different data sets: a matched set of IRS 990s and hospital regulatory financial data (OSHPD reports) for 719 observations from nonprofit hospitals located in California, and a larger set of over 100,000 IRS 990 for educational, medical, and charitable nonprofits. Our analyses suggest that nonprofit organizations over-report expenses in the program services category and under report their fundraising expenses, which have the effect of improving the ratios used by donors when making their resource allocation decisions. The results are robust to the data source (i.e., the IRS 990 and OSHPD financial statements) and a variety of specification tests. Additional results suggest that nonprofits that are more reliant on donations as a revenue source and face lower political costs manipulate their operating ratios to a greater extent. Some of the limitations of our study should be noted when drawing inferences from our results. First, our empirical analysis on California nonprofit hospitals, which combines the IRS 990 and the OSHPD financial database, assumes that donors are more likely to use the IRS 990.

30

Although this assumption is plausible given the intended purposes of the databases, it is possible that nonprofit hospitals manage their operating ratios on both the IRS 990s and the OSHPD reports. If true, our results based on the sample of California hospitals are conservatively biased. Additionally, in our empirical analysis of educational, medical, and charitable nonprofits we rely on a model to estimate fundraising expenses. The accuracy of the estimates is a function of the model, its related assumptions, and independent variable measurement issues. Although plausible, our estimates undoubtedly capture the true underlying expense allocations with error and likely contain several conservative biases. This study makes several important contributions to accounting research. First it demonstrates that expense allocation choices affect the quality and usefulness of financial disclosures in a nonprofit setting. To the extent that donors base their contribution decisions on the ratios reported in the nonprofits’ financial statements, opportunistic management of these ratios can reduce social welfare because of misallocated resources. Second, prior accounting research that empirically examines the financial performance of nonprofit organizations has not considered the extent to which the incentives to manage reported operating efficiency affect the quality of accounting disclosures in nonprofits’ financial statements. Future research could examine the extent to which the ability to manage expense ratios affects the nonprofit’s growth and survival. It would also be interesting to examine disclosure management in nonprofits that have converted from non-profit to for-profit status and vice-versa. Of late, such conversions have been frequently appearing in the hospital industry. For example, between 1991 and 1997, 185 hospitals comprising approximately 6.5% of total hospitals in the country converted status. Of these, 127 hospitals converted from non-profit to for-profit and 58 hospitals converted from for-profit to nonprofit status (Thorpe, Florence, and Seiber 2000).

31

Because such conversions affect revenue sources as well as the objective function of hospitals, they are likely to have an impact on disclosure management. Another area for future research is to explore disclosure management by various types of nonprofits within the same industry such as religious nonprofits vis a vis private nonprofits. Because managerial ideology is an important factor affecting nonprofit behavior (Rose-Ackerman 1987), there are likely to be differences in disclosure management across the types of nonprofits.

32

REFERENCES Alexander, J.A. and S.Y. Lee. 1996. The effects of CEO succession and tenure on failure of rural community hospitals. Journal of Applied Behavioral Science 32, 70-88. American Association of Fundraising Council Trust for Philanthropy. 2002. Giving USA, Sewickley, PA. Arrow, K. 1963. Uncertainty and the welfare economics of medical care. American Economic Review, 53,941-973. Baber, W.R., A.A. Roberts, and G. Visvanathan. 2001. Charitable organizations’ strategies and program-spending ratios. Accounting Horizons 15, 329-343. Baber, W.R., P.L. Daniel and A.A. Roberts. 2002. Compensation to Managers of Charitable Organizations, An Empirical Study of the Role of Accounting Measures of Program Activities". Accounting Review, forthcoming. Barrett, W.P. 1999. Look before you give. Forbes. Vol. 164 (14), December 27,208-216. Bryce, H.J. 2001. Capacity considerations and community benefit expenditures of nonprofit hospital. Health Care Management Review 63, 24-39. Callen, J. 1994. Money donations, volunteering, and organizational efficiency. The Journal of Productivity Analysis, 5, 215-228. Dranove, D. 1988. Pricing by nonprofit institutions, The case of hospital cost shifting. Journal of Health Economics 7, 47-57. Dranove, D., M. Shanley, and W.D. White. 1993. Price and concentration in hospital markets, The switch from patient-driven to payer-driven competition. Journal of Law and Economics 36, 179-204. Fama, E.F., and M.C. Jensen. 1983. Agency problems and residual claims. Journal of Law and Economics. 24, 327-349. French, H.E. III. 1996. Competition and Monopoly in Medical Care, Washington, DC, The AEI Press. Independent Sector. 2001. The New Nonprofit Almanac. Joint Committee on Taxation. 2000. Study of Present-Law Taxpayer Confidentiality and Disclosure Provisions as Required by Section 3802 of the Internal Revenue Service Restructuring and Reform Act of 1998. Volume I. (January).

33

Khalat, R. and W. Hueslein. 1992. The accounting games charities play; evaluating fundraising efficiency. Forbes 150(10), 252-256. Leone, A.J. and R.L Van Horne. 2001. Earnings management in not-for-profit institutions, Evidence from hospitals. Working paper, University of Rochester, Rochester, NY. Lynk, W.J. 1995. Nonprofit mergers and the exercise of market power. Journal of Law and Economics 38, 437-461. Mick, S.S. and C.G. Wise. 1996. Downsizing and financial performance in rural hospitals, Health Care Management Review 21, 16-25. National Center for Charitable Statistics. 1999. Internal Revenue Service Statistics of Income IRS 990 sample files. Center on nonprofits and Philanthropy / Urban Institute online database, www.nccs.org. Nelson, P. 1974. Advertising as Information. Journal of Political Economy 82 No. 4, 729–754. Okten, C. and B.A. Weisbrod. 2000. Determinants of donations in private nonprofit markets. Journal of Public Economics 75, 255-272. Potter, M.A. and B.B. Longest. 1994. The divergence of federal and state policies on the charitable tax exemption of nonprofit hospitals. Journal of Health Politics, Policy and Law 19, 393-419. Posnett, J. and T. Sandler. 1989. Demand for charity donations in private non-profit markets. Journal of Public Economics 40, 187-200. Robinson, J.C. and C.S. Phibbs. 1989. An evaluation of Medicaid selective contracting in California. Journal of Health Economics 8, 437-455. Rose-Ackerman, S. 1987. Ideals versus dollars, Donors, charity managers, and government grants. Journal of Political Economy, 95, 810-823. Rose-Ackerman, S. 1996. Altruism, nonprofits, and economic theory. Journal of Economic Literature, 34, 701-728. Tinkelman, D. 1999. Factors affecting the relation between donations to not-for-profit organizations and an efficiency ratio. Research in Governmental and Nonprofit Accounting 10. United States Internal Revenue Service Ruling 69-545 (1969) and Ruling 83-157 (1983). Weisbrod, B. and N. Dominguez. 1986. Demand for collective goods in private nonprofit markets, Can fund-raising expenditures help overcome free-rider behavior? Journal of Public Economics 30, 83-95.

34

Yetman, R.J. and M.H. Yetman. 2001. The effect of nonprofits’ taxable activities on the supply of private donations. Working paper, The University of Iowa, Iowa City, IA. Yetman, R.J. 2001. Tax-Motivated Expense Allocations by Nonprofit Organizations. The Accounting Review 76, 297– 311.

35

TABLE 1 Descriptive Statistics (dollar amounts in $thousands) _____________________________________________________________________________________________ Panel A: California Nonprofit Hospitals Description Financial variables: Private donations Program revenues Assets Net income margin Liabilities 1,251 87,116 208,209 0.04 101,498 114 69,582 91,954 0.04 46,646 4,075 77,784 643,638 0.10 294,312 Mean Median Std. Dev.

Other variables: Beds Medicare Medi-Cal Occupancy Charity LOS 209 45.41% 20.82% 59.54% 2.39% 6.24 190 46.63% 15.76% 58.05% 1.49% 4.47 136 14.78% 16.56% 18.74% 3.28% 7.95

Panel B: Larger Sample of Educational, Medical, and Charitable Nonprofits Description Educational organizations (n =25,501) Private donations Program revenues Fundraising expenses Price Age Assets Net income margin Medical organizations (n= 37,602) Private donations Program revenues Fundraising expenses Price Age Assets Net income margin 1,106 47,588 90 1.08 32.10 62,056 0.11 170 20,715 0 1.00 34.92 30,011 0.05 5,319 75,193 798 0.26 18.95 86,711 0.21 4,501 17,612 493 1.23 36.52 71,711 0.18 1,746 6,008 183 1.11 38.94 24,722 0.13 10,045 49,873 927 0.40 17.40 149,498 0.20 Mean Median Std. Dev.

36

TABLE 1 Descriptive Statistics Panel B (Continued): Larger Sample of Educational, Medical, and Charitable Nonprofits _____________________________________________________________________________________________

Charitable organizations (n = 48,017) Private donations Program revenues Fundraising expenses Price Age Assets Net income margin 3,114 4,198 259 1.12 30.89 21,544 0.16 427 243 0 1.00 26.98 11,284 0.08 9,757 13,724 976 0.28 21.37 35,032 0.29

________________________________ Notes: Private donations are those not received from governmental or other granting agencies. The primary source of private donations is from individuals and corporations. Program revenues are from the sales of products and services. Fundraising expenses are any expense incurred to generate or maintain donations. Net income margin is net income scaled by total revenues. Beds is the number of staffed beds. Medicare (Medi-Cal) is the proportion of medicare (Medi-Cal) patients to total patients. Occupancy is the actual patient days divided by available patient days. Charity is charity care expenses divided by total revenues. LOS is the average length of stay in days from admission to discharge. Price is 1 / (1 - Fundraising expenses / Private donations). Age is the age of the nonprofit in years. All amounts are from the OSHPD report or the IRS 990. _____________________________________________________________________________________________

37

TABLE 2 Paired Two-Sample t-test of Expense Allocations (Dollar amounts in $thousands) ____________________________________________________________________________________________________________________________________ Sample of California Hospitals: Program service expenses Per IRS 990 California hospitals (n = 719) 71,104 Per OSHPD 58,211 Difference 13,893* Program service ratio Per IRS 990 82.8% Per OSHPD 67.9% Difference 14.9%*

Larger Sample of Educational, Medical, and Charitable Nonprofits: Fundraising expenses Per IRS 990 Educational (n = 25,501 ) Medical (n = 37,602 ) Charitable (n = 48,017 ) ______________________ 493 90 259 As estimated 2,320 258 558 Difference 1,827* 168* 299* Fundraising ratio Per IRS 990 2.8% 1.5% 3.3% As estimated 21.0% 5.2% 13.3% Difference 18.0%* 3.7%* 10.0%*

Notes: The IRS 990 is a nonprofits’ publicly available annual information return. The OSHPD report is a state-level regulatory report for California hospitals only. Program service expenses are those devoted to accomplishing the nonprofit’s primary charitable purpose. The amounts of estimated fundraising expenses are from Table 3. The Program service ratio is the amount of program service expenses scaled by total expenses (i.e., the sum of program service, administrative, and fundraising expenses). The Fundraising ratio is the amount of fundraising expenses scaled by total expenses (i.e., the sum of program service, administrative, and fundraising expenses). * Significant at the one-percent level ____________________________________________________________________________________________________________________________________

38

TABLE 3 Regression Analysis Used to Estimate Nonprofit Fundraising Expenses _____________________________________________________________________________________________ Private donationsit = α + β1 Fundraising expensesit-1 + β2 Priceit-1 + β3 Ageit-1 + β4 Program revenues it-1 + εi.. _____________________________________________________________________________________________ α β1 β2 β3 β4

Adj. R2

n

Educational nonprofits

0.12 (88.5)

1.94 (73.1)

-0.01 (-13.3)

-0.00 (-32.3)

-0.05 (-44.8)

0.41

22,708

Medical nonprofits

0.07 (118.9)

4.29 (433.5)

-0.01 (-24.3)

-0.00 (-190.4)

-0.01 (-60.6)

0.49

31,891

Charitable nonprofits

0.33 (119.6)

5.58 (378.3)

-0.09 (-41.7)

-0.00 (-85.9)

-0.02 (-20.7)

0.43

38,776

_______________________ Notes: Private donations are those not received from governmental or other granting agencies. The primary source of private donations is from individuals and corporations. Program revenues are from the sales of products and services. Fundraising expenses are any expense incurred to generate or maintain donations. Price is 1 / (1 Fundraising expenses / Private donations). Age is the age of the nonprofit in years. All data are from the IRS 990. All models use asset-scaled variables and control for first-order autocorrelation and general heteroscedasticity ( i.e., AR(1) GARCH (1,1) model). Influential observations that had a Cook’s D statistic greater than 1 were deleted. t statistics are in parentheses. * All coefficients are significant at the five percent level. _____________________________________________________________________________________________

39

TABLE 4 Analysis of Nonprofits that Report Zero Fundraising Expenses and Report Earning Private donations ____________________________________________________________________________________________________________________________________

Full sample California Hospitals Educational Nonprofits 122

Number reporting zero fundraising and donations 95

Percent reporting zero fundraising and donations 57%

Average private donations ($) 1,251,000

Web sites Searched 95

Web sites found 95

Web sites with evidence of fundraising 19

Percent of web sites with evidence of fundraising 20%

2,706

663

25%

3,650,000

66

38

26

68%

Medical Nonprofits Charitable Nonprofits ____________________

4,555

2,090

46%

1,385,000

209

91

13

14%

6,693

2,242

33%

1,751,000

224

125

72

58%

Notes: The full samples examined included all unique nonprofits from the most recently available year of data (i.e., 1998). The number of web sites searched was the full sample of California hospitals and a randomly chosen sample of 10 percent of the other nonprofit organizations. We considered evidence of fundraising activities to be requests for donations from individuals or providing instructions to individual donors on how to give. A list of web sites searched is attached in Appendix A. ____________________________________________________________________________________________________________________________________

40

TABLE 5 Regression Analysis of Characteristics Associated With Program Service Expense Allocations California Nonprofit Hospitals _____________________________________________________________________________________________ Program allocationsit = α + β1 Size it + β2 Margin it + β3 Donationsit + β4 LOSit + β5 Medicare it + β6 Medi-Calit + β7 Occupancyit + β8 Charityit + εit _____________________________________________________________________________________________ Variable Size Margin Donations LOS Medicare Medi-Cal Occupancy Charity Intercept Adjusted R2 N Coefficient 0.00001 -0.029 0.212 -0.00005 -0.116 -0.071 0.096 0.365 0.167 0.07 601 t-statistic 0.20 -0.50 2.61*** -0.06 -2.64*** -1.76* 3.20*** 2.01** 4.97***

________________________________________ Notes: Program allocations are the estimated amount of expenses that a nonprofit has allocated to the program services category. Size is total assets in $ millions. Margin is net income scaled by total revenues. Donations are private donations that are those not received from governmental or other granting agencies, scaled by total revenues. The primary source of private donations is from individuals and corporations. LOS is the average length of stay from admission to discharge. Medicare (Medi-Cal) is the proportion of medicare (Medi-Cal) patients to total patients. Occupancy is the actual patient days divided by available patient days. Charity is charity care expenses divided by net revenues. All amounts are from the OSHPD report or the IRS 990. The model controls for first-order autocorrelation and general heteroscedasticity ( i.e., AR(1) GARCH (1,1) model). *, **, *** Significant at the ten percent, five percent, and one-percent level, respectively (two-tail). _____________________________________________________________________________________________

41

TABLE 6 Regression Analysis of Characteristics Associated With Fundraising Expense Allocations Educational, Medical, and Charitable Nonprofits _____________________________________________________________________________________________ Fundraising allocationsit = α + β1 Size it + β2 Margin it + β3 Donationsit + β4 Liabilities it + εit. _____________________________________________________________________________________________ Educational Nonprofits Variable Size Margin Donations Intercept Adjusted R2 N Coefficient -.00002 0.341 0.664 -0.064 0.87 24,912 t-statistic -8.05*** 130.22*** 290.68*** -92.63*** Medical Nonprofits Coefficient -0.00001 0.138 0.239 -0.009 0.62 36,968 t-statistic -3.76*** 101.09*** 157.14*** -26.86*** Charitable Nonprofits Coefficient -0.0002 0.225 0.192 -0.025 0.48 46,952 t-statistic -14.18*** 131.95*** 128.33*** -35.69***

________________________________________ Notes: Fundraising allocations are the estimated amount of expenses that a nonprofit has allocated out of the fundraising category. Size is total assets in $ millions. The primary source of private donations is from individuals and corporations. Margin is net income scaled by total revenues. Donations are private donations that are not received from governmental or other granting agencies, scaled by total revenues. All amounts are from the IRS 990. All models control for first-order autocorrelation and general heteroscedasticity ( i.e., AR(1) GARCH (1,1) model). *, **, *** Significant at the ten percent, five percent, and one-percent level, respectively (two-tail). _____________________________________________________________________________________________

42

Appendix A Nonprofit Organization Web Pages Searched for Evidence of Fundraising Expenses _____________________________________________________________________________________________ California hospital web pages with evidence of fundraising expenses:
Nonprofit name

ANAHEIM HOSPITAL BARTON MEMORIAL HOSPITAL CHILDREN’S HOSPITAL OF ORANGE COUNTY CITY OF HOPE NATIONAL MEDICAL CENTER DAMERON HOSPITAL ASSOCIATION EMANUEL MEDICAL CENTER FOOTHILL PRESBYTERIAN HOSPITAL GOLETA VALLEY COTTAGE HOSPITAL LONG BEACH MEMORIAL MEDICAL CENTER MARIN GENERAL HOSPITAL NOVATO COMMUNITY HOSPITAL PRESBYTERIAN INTERCOMMUNITY HOSPITAL SANTA BARBARA COTTAGE HOSPITAL SEQUIOA HEALTH SERVICES ST. JOSEPH HOSPITAL ST. JUDE HOSPITAL SUTTER MERCED MEDICAL CENTER TORRANCE MEMORIAL MEDICAL CENTER WHITE MEMORIAL MEDICAL CENTER

Sample of educational nonprofit web pages with evidence of fundraising expenses:
Nonprofit name ASHLAND UNIVERSITY WILMINGTON COLLEGE CLEMSON ADVANCEMENT FDN FOR DESIGN & BUILDING KRISHNAMURTI FOUNDATION OF AMERICA HARCUM COLLEGE ROWAN COLLEGE FOUNDATION ELISABETH MORROW SCHOOL INC DWIGHT-ENGLEWOOD SCHOOL PRINCETON DAY SCHOOL CHAPIN SCHOOL GEORGIAN COURT COLLEGE UNIVERSITY AT BUFFALO FDN INC ELMIRA COLLEGE D'YOUVILLE COLLEGE ALBANY MEDICAL COLLEGE JAPAN INTERNATIONAL CHRISTIAN UNIVERSITY FOUNDATION INC MITCHELL COLLEGE SALEM STATE COLLEGE FOUNDATION INC BECKER COLLEGE ASSUMPTION COLLEGE CARDINAL STRITCH UNIVERSITY INC AVE MARIA COLLEGE MIAMI UNIV FDN INC BISHOPS UNIVERSITY FDN BROWARD COMMUNITY COLLEGE FDN INC MOUNT SINAI SCHOOL OF MEDICINE OF CITY UNIV OF NEW YORK

43

Sample of medical nonprofit web pages with evidence of fundraising expenses:
Nonprofit name

ST LUKES EPISCOPAL HOSPITAL FRED HUTCHINSON CANCER RESEARCH CENTER RHODE ISLAND HOSPITAL ISLAMIC MEDICAL ASSN OF NORTH AMERICA INC ST JOHNS RIVERSIDE HOSPITAL GARDEN CITY HOSPITAL OSTEOPATHIC VNA CARE NETWORK INC MAINE COAST REGIONAL HEALTH FACILITIES KIN ON HEALTH CARE CENTER SKAGGS COMMUNITY HOSPITAL ASSN BAPTIST HEALTHCARE SYSTEM INC MOUNT SINAI HOSPITAL ENH RESEARCH INSTITUTE

Sample of charitable nonprofit web pages with evidence of fundraising expenses:
Nonprofit name NATIONAL BOY SCOUTS OF AMERICA FDN NEW YORK STATE HISTORICAL ASSOCIATION BOYS HOME OF NORTH CAROLINA INC IQRA INTL EDUCATIONAL FDN COMMUNITY FDN OF SARASOTA CO INC ARIZONA COMMUNITY FDN INC COASTAL BEND COMMUNITY FOUNDATION CENTRAL NEW YORK COMMUNITY FOUNDATION INC NATIONAL HUMANITIES CENTER COMMUNITY FOUNDATION OF BROWARD ST LOUIS COMMUNITY FDN NEW YORK TIMES NEEDIEST CASES FUND INDIANAPOLIS NEIGHBORHOOD HOUSING PARTNERSHIP TIDEWATER JEWISH FDN INC ELKS NATIONAL FDN TROY FOUNDATION HUMBOLDT AREA FOUNDATION FREMONT AREA FOUNDATION ARKANSAS ARTS CENTER FOUNDATION OREGON JEWISH COMMUNITY FOUNDATION MILWAUKEE SYMPHONY ORCHESTRA FOUNDATION TRUST ALLEY THEATRE AURORA FDN MADISON COUNTY COMMUNITY FDN

44

Sample of charitable nonprofit web pages with evidence of fundraising expenses (continued):
Nonprofit name LONG BEACH AQUARIUM OF THE PACIFIC GOOD SHEPHERD FUND INDIANA SPORTS CORP CHICAGO SYMPHONY & LYRIC OPERA FACILITIES FUND LAMBS FARM PARKS AND WILDLIFE FOUNDATION OF TEXAS INC CAL FARLEYS BOYS RANCH FOUNDATION NEBRASKA CHILDREN'S HOME SOCIETY NATIONAL CHURCH RESIDENCES NATIONAL COUNCIL OF TEACHERS OF MATHEMATICS SOUTHWEST MINNESOTA FOUNDATION GOODWIN HOUSE FDN INC ALABAMA SHAKESPEARE FESTIVAL ENDOWMENT TRUST NEW HOPE CHARITIES INC WILDLANDS TRUST OF SE MASSACHUSETTS SOCIETY FOR ASIAN ART IOWA STATE FAIR FOUNDATION NEW YORK STATE ASSN FOR RETARDED CHILDREN INC\SUFOLK ATLANTIC 10 CONFERENCE INDIANAPOLIS SYMPHONY ORCHESTRA FDN NATIONAL COUNCIL OF STATE GARDEN CLUBS INC BGTM INC CINEMA CHICAGO BUSINESS COMMITTEE FOR THE ARTS INC US LACROSSE FOUNDATION INC CATHEDRAL HOME FOR CHILDREN AMERICAN SOCIETY OF CIVIL ENGINEERS HUMANE ANIMAL WELFARE SOCIETY OF WAUKESHA LA HABRA BOYS & GIRLS CLUB ABILENE BOYS RANCH INC HOPE FOR THE CHILDREN NATURAL AREA PRESERVATION ASSOC MENORAH PARK FOUNDATION LEGAL AID BUREAU, INC JAYCEES INTERNATIONAL (JCI) FOUNDATION, INC ROGER HOUTSMA WORLD OUTREACH UNITED CEREBRAL PALSY ASSN OF WESTCHESTER CNTY INC ALBRIGHT CARE SERVICES FLORIDA CHILDERN'S FORUM INC UNITED STATES GOLF ASSOCIATION FDN VENICE FOUNDATION, INC RANCHO SANTA ANA BOTANIC GARDEN MUSTARD SEEDS & MOUNTAINS INC CANTERBURY PLACE PLEASANT RUN CHILDRENS HOMES FDN INC ROSS RAGLAND THEATRE INTERNATIONAL BUREAU FOR EPILEPSY GREYSTONE PROGRAMS INC

45

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close