Knowledgebase

Core Files: Assessing Their Accuracy & Building a Longitudinal File for Trend Analysis (Technical Note / Method Note)

The attached documents represent the products of an analysis conducted in 1997 examing the accuracy of Core files (created from IRS Return Transaction Files) from 1990-1996 and using them to construct a longitudinal file with complete data for each fiscal year for each "live" organization in each year.

Key Findings:

• Error rates are acceptable for most research purposes.
• Total revenue and its component sources (e.g., investment income, public support), total expense and balance sheet items from the RTFs (the Core files) can be used for the vast majority of observations.
• Our ability to reduce the number of errors, beyond identification of a relatively small number of egregious errors, is limited. Given the size of the data sets, it is impossible as a practical matter to locate the majority of the errors unless thousands of individual organizations are contacted directly.

The first document, a Word file, is the text of the Technical Appendix for a report on the finances of operating public charities. The second document contains all of the accompanying tables.

Researchers interested in the accuracy of the Core financial data or in building a longitudinal file should find these documents useful.

MISSING VARIABLES

The 1990 and 1991 Core files did not include the following variables:

• Total revenue,
• Program service revenue,
• Rental net income and expenses, and
• Special event net income and expenses.

Additional Keywords: financial analysis, data quality