Charitable objectives or donor benefits? What sponsor language reveals about donor-advised fund priorities and resource flows
Data files
Apr 09, 2025 version files 525.72 KB
-
DAF_Concentration.csv
59.76 KB
-
DAF_Financial_Performance.csv
325.99 KB
-
Data_Description_for_DAF_Concentration.csv
7.32 KB
-
Data_Description_for_DAF_Financial_Performance.csv
7.38 KB
-
Data_Description_for_EM_Comparisons.csv
517 B
-
Data_Description_for_EM_Scraping.csv
4.94 KB
-
EM_Comparisons.csv
43.12 KB
-
EM_Scraping.csv
71.19 KB
-
README.md
5.51 KB
Abstract
Recent years have seen a dramatic rise in donor-advised funds (DAFs). Though housed in public charities, DAFs are often characterized as de facto private foundations due to the deference sponsors typically give to donors’ wishes. The consequence has been frequent calls to institute DAF grant disbursement requirements and other restrictions akin to those on foundations. Despite their growing importance, we know little about what distinguishes different DAF sponsoring organizations beyond a commonly used three-type split between community foundations, national sponsors, and single-issue sponsors. To better understand variation in behavior across DAF sponsoring organizations – which may, in turn, be driven by the donors they attract – we develop a proxy measure of the priorities they display in the language they use on their websites. The measure seeks to identify the extent to which a sponsor emphasizes achieving charitable objectives versus providing extrinsic benefits to donors. In addition to presenting a new method of classifying DAF sponsors, we also show how this measure complements existing sponsor type classifications, with national sponsors emphasizing donor benefits more on average but also exhibiting the most meaningful within-type variation in their emphasis. Most notably, among national sponsors, greater emphasis on donor benefits is highly predictive of greater DAF assets, and this feature is largely attributable not to greater contribution receipts but rather to lower payout rates. Our results suggest that variation in the language used by DAF sponsors can help inform which organizations would be affected by regulatory proposals targeting DAFs, and what effects such proposals would have on charitable activity.
https://doi.org/10.5061/dryad.6wwpzgn8r
Description of the data and file structure
Each file uploaded to Dryad has a corresponding Data Description file which lists and describes the variables.
-————————————————————————————-
DATA FILE PRODUCTION
Files in bold text represent the publicly provided final data files and Stata code.
-————————————————————————————-
1. DOWNLOAD RAW NONPROFIT RETURNS FROM IRS
Process: Manual process
URL: https://www.irs.gov/charities-non-profits/form-990-series-downloads
Outputs: Individual XML files of nonprofit annual 990 returns
2. DOWNLOAD EXEMPT ORGANIZATIONS BUSINESS MASTER FILE FROM NCCS
Process: Manual process
URL: https://urbaninstitute.github.io/nccs/catalogs/catalog-bmf.html
Outputs: CSV business master index file for all exempt organizations
3. CREATE LISTS OF CHARITY EINS
Inputs: Exempt Organizations Business Master File downloaded in step 1; raw XML returns downloaded in step 2
Process: Run R script to output CSV files of unique EINs for:
Donor-advised fund sponsors (2020-2022)
Non-DAF operating 501(c)(3) charities (2020-2022)
Outputs: CSV files of EINs for DAFs and operating charities (2020-2022)
4. CREATE JSON FILES OF IRS DATA
Inputs: Raw XML returns downloaded in step 1; CSV EIN files output in step 3
Process: Run Python scripts to pull required IRS XML Parts and Schedules
into JSON files for analysis; at its core, this script uses the open-source Python package IRSx for parsing the IRS components
Outputs: JSON files for DAFs & operating charities (2020-2022)
5. CREATE LISTS OF CHARITY URLS
Inputs: CSV files output in step 3; JSON files output in step 4
Process: In Tableau, join charity EINs from CSVs to charity URLs from JSONs
Outputs: CSV search lists of EINs and URLs of DAFs & operating charities (2020-2022)
6. SEARCH WEBSITES
Inputs: XLSX dictionary of search terms, created manually; CSV URL lists output in step 5
Preprocessing steps: manually create list of EINs of operating charities size-matched to DAFs and for investment advisors
Process: Run Python script to search websites for dictionary terms
(for DAFs, size-matched operating charities, and investment advisors)
Outputs: CSV file (EM Comparisons.csv) of terms found, with counts per website, for DAFs, size-matched operating charities, and investment advisors
7. CREATE COMBIMED EMPHASIS MEASURE SUMMARY FILE
Inputs: JSON files output in step 4; CSV web search files output in step 6
Process: In Tableau, join input files above to create simple tables with rows for charities and columns for financial info and web search results, using the most recent year of financial info for each charity
Outputs: CSV file (EM Scraping.csv) of most recent year financial information and web search results for DAFs
8. CREATE SUMMARY XLSX FILES FOR DAF FINANCIAL PERFORMANCE SAMPLE
Inputs: JSON files output in step 4; CSV web search files output in step 6
Process: In Tableau, join input files above to create simple tables with rows for charity-years and columns for financial info and web search results
Outputs: XLSX file of financial information and web search results for all DAFs in the EM sample for all years (2020-2022)
9. CREATE CSV FILES for STATA ANALYSIS
Inputs: XLSX file from step 8
Process: Follow cleaning steps (i)-(iii) outlined in the *Sample Description *section of the paper; add columns that calculate relevant data points; repeat process for organization years for which DAF contributions represent at least 90% of all contributions
Outputs: CSV files of financial information and web search results for DAF Financial Performance Sample (DAF Financial Performance.csv) and DAF Concentration Subsample (DAF Concentration.csv).
Note: empty cells reflect entries in 990 filings left blank by filing organizations and are treated as reflecting a value of zero.
10. PERFORM ANALYSIS IN STATA
Inputs: CSV files output in steps 6, 7, and 9; Stata do-file (FM NVSQ Paper Analysis.do)
Process: Open each CSV file, in turn, and run associated do-file steps in Stata
Outputs: Data for statistical analysis discussed in text, data for figures, and data for tables as labeled in FM NVSQ Paper Analysis.do
Code/software
Python (January 2025 release v. 1.97)
R (v. 4.3.3 Angel Food Cake)
StataNow/MP for Mac (v. 18.5)
Tableau (v. 2023.3)
Microsoft Excel (v. 2501 Build 18429.20158)
Access information
Data was derived from the following sources:
- RAW NONPROFIT RETURNS FROM IRS (https://www.irs.gov/charities-non-profits/form-990-series-downloads)
- EXEMPT ORGANIZATIONS BUSINESS MASTER FILE FROM NCCS (https://urbaninstitute.github.io/nccs/catalogs/catalog-bmf.html)
- SEARCH OF ORGANIZATION URLS AS PROVIDED TO IRS
Data collected consists of elements from: IRS Nonprofit 990 Database; NCCS Exempt Organizations Business Master File; website search
Data processing and analysis described in the README file