In our conversations and observations, we often hear colleagues bemoaning, “No wonder the public is losing confidence in our sector—I heard the error rate on the Internal Revenue Service Form 990 is 80 percent…” “…50 percent…” “…20 percent… “Because these percentages vary so wildly, we asked the folks of the National Center for Charitable Statistics at the Urban Institute for clarity. Operating on the basis of inaccurate information can be damaging.

So what is the truth?

There is no single meaningful estimate of the error rate for information reported on the IRS Form 990. A typical form, plus the attached Schedule A, requires a charity to submit over 600 lines of information ranging from standard accounting items to executive compensation, major contractors, and seemingly obscure regulatory questions. Although research over the past several years has addressed the strengths and weaknesses of the data reported on the form, “the” error rate depends on what you’re looking for.

The Form 990 has multiple audiences, including federal and state regulators, the public, donors and funders, academic and other researchers, the media, and practitioners.

A major problem in data quality for a researcher may be of little consequence to others. Misspelling a name or forgetting a decimal point can be easily identified on a single return—but imagine running a salary comparison for hundreds of “executive directors” if some entries say “exactive director.” Similarly, if summarizing total revenue, a dropped decimal point turning $1 million into $100 million is a nightmare for researchers (or rather, their beleaguered research assistants) who have to manually track down such errors. On the other hand, forms that do not have an officer’s signature present a problem for regulators but not for researchers.

It’s time to replace the anecdotal horror stories surrounding discussions of data quality with some data. The Urban Institute’s National Center for Charitable Statistics (NCCS) and Guidestar have collaborated to digitize much of the information on the Form 990 for the past several years. This means that NCCS now has a research database that can tell us more about quality issues. Below is a tour through initial findings on some of the 990 sections of greatest interest to researchers and practitioners.

Financial Data—Parts I, II, and IV of the 990: In 1999, researchers analyzed 32 items on the revenue, expenses, and balance sheet sections of almost 33,000 Forms 990 and 12 items on almost 18,000 Form 990-EZ, a shorter form used by smaller organizations.1 The average number of errors was 0.77 for Forms 990 and 0.38 for Forms 990-EZ. Only about 20 percent had two or more errors, well over half of which were errors of omission—for example, a value that was supposed to appear on more than one line that was not copied over. Since they neither misrepresent financial data nor influence analysis, they don’t affect regulators or researchers in most cases.

When researchers compared data from Forms 990 to audited financial statements, the results were encouraging. The most recent study concluded that “…the IRS Form 990 return is a reliable source of information for basic income and balance sheet entries…” and that it was potentially more useful than financial statements for the study of nonprofits because of the level of detail, seldom found on financial statements.2

After reviewing scanned images of the Forms 990 in the last two years, we estimate that about 80 percent were created by professionals using accounting software. As software improves and more forms are prepared by accountants based on audited financials, the data quality should continue to improve.

Program Service Accomplishments—Part III of the 990: The NCCS research database of Forms 990 contains over 450,000 descriptions of what nonprofits do. A review of a sample of forms that do not include program descriptions revealed that much of the problem is because of data entry difficulties. Frequently, missing descriptions were included in an attachment or did not transfer well from paper to the scanned image. Improved data entry will help address this problem.

In the past, nonprofits were only classified by type of organization (arts, environment, human services, etc.), using the National Taxonomy of Exempt Entities (NTEE). For its new research database of Form 990 information, NCCS has created a system to classify the programs, activities, and services of nonprofits, as described in Part III of the 990. An initial analysis of 4,014 organizations with program descriptions found that only 1.3 percent of the descriptions did not contain enough information to be assigned a code under the new Nonprofit Program Classification System.3

Another analysis of program descriptions for a random sample of 400 organizations rated the quality of the descriptions according to whether they included the following three elements: a clear statement of the activity, some indication of quantity of service, and a description of beneficiaries.4 About 98 percent met at least one of our criteria and could be coded using the new system. Over 90 percent had at least two out of the three elements, and one-quarter had all three elements.
As more and more potential donors, volunteers, and funders go to the Internet to read completed 990s located at GuideStar and NCCS, nonprofits are realizing that this section may often represent the first introduction of the organization. Once again, the outlook for improved attention to this section is good.

Executive Compensation—Part V of the 990: This section lists the nonprofit’s officers, directors, trustees, and key employees. An initial analysis on this section using the NCCS research database of over 220,000 filers found about one-fifth of the entries did not fully complete this part of the form.5 Of these, about half had filled out names but did not include compensation information. The vast majority that left the section blank turned out to have very low annual expenses—less than $75,000. It appears that in most cases, these were organizations run by volunteers, and they had simply omitted the zero for salary paid. In this section, then, it seems that many of the omissions would have little impact on analysis.

There are many types of information now available on the Form 990 that we have not addressed in this brief review. As NCCS continues to develop and explore its research database of Form 990 information, we will know more and more about data quality—but trying to come up with a simple error rate would not help assess the usefulness of the data. There are many types of errors and many types of uses.

Although more research is needed on the question of data quality, one conclusion is clear—the number of errors on 990s is falling. Easy Internet access to these public documents and IRS regulations calling for greater disclosure continue to increase the visibility of the information, prompting organizations to pay more attention to their filings and increase the quality even more. Progress on electronic filing will also help, as we work on improved software, more standardized accounting practices, and standardized program descriptions.

1. Chuck McLean and Linda Lampkin. 1999. “Now That We Have the Data, Can We Use Them? Assessing the Quality of IRS Form 990 Data.” ARNOVA Conference, Washington, DC, November.
2. Karen Froelich, Terry Knoepfle, and Thomas Pollak. 2000. “Financial Measures in Nonprofit Research: Comparing IRS Form 990 Return and Audited Financial Statement Data.” Nonprofit and Voluntary Sector Quarterly 29: 232-254.
3. Linda Lampkin, Sheryl Romeo, and Emily Finnin. 2000. “Introducing the Nonprofit Program Classification System: The Taxonomy We’ve Been Waiting For.” ARNOVA Conference, Miami, FL, December.
4. Linda Lampkin. 2001. “Preliminary Analysis of Quality of Form 990 Part III Program Descriptions.” Unpublished, May.
5. Eric Twombly, Marie Gantz, and Nicholas Stengel. 2000. “Introducing the Nonprofit Program Classification System: Adding Clarity to the Debate.” ARNOVA Conference, Miami, FL, December.

Linda M. Lampkin is program director of the National Center for Charitable Statistics (NCCS), a program of the Center on Nonprofits and Philanthropy at the Urban Institute. She works closely with the Internal Revenue Service, key nonprofit groups, and the scholarly community to maximize usage of nonprofit data and to promote and conduct research on the sector. Thomas H. Pollak is assistant director of NCCS. His recent projects include developing a pilot electronic project for the Form 990 and studies using the digitized database.