Methods: We use the National Center for Charitable Statistics (NCCS) datasets to explore the overall trends related to organizational survival or closure across various nonprofit sectors. Machine learning decision trees were utilized to explore what types of organizations remain open under a variety of socioeconomic shifts. A cluster analysis was applied to explore the relationship between the number of non-profits establishments and poverty rate across zip codes.
Data: The NCCS data archive contains information from tax-exempt organizations in the US from 1988-2014. This study uses data from all years, and included information from each non-profit organization that filed a 990 form. The final sample was n = 7,462,907.
Measures: We use Employer Identification Numbers (EINs) as unique identifiers across the dataset. Each organization’s established year was the year that the EIN first appeared in the dataset. Its closure year is the last year the EIN was observed, before 2014. The type of the organization was pulled from categories created by the IRS. The most frequent organizational category assigned by the IRS to each EIN was used to define non-profit type. Revenue was the average revenue for each EIN across all the years it appeared in the dataset. Poverty rate for each zip code was estimated by the U.S. Census Bureau.
Results: Across the data span, human service related organizations were the most common non-profit organizations, while the proportion of Arts/Humanities, as well as Education organizations has increased. The pattern of Community Improvement organizations and Philanthropy/Grantmaking foundations were similar, with the data reflecting slight increases over all 27 years. While the number of religious organizations increased annually, the overall proportion decreased. Moreover, an organization formed after 2012 and before 2014 with an average revenue of $61,000 was more likely to survive when compared to all others. Organizations started before 2008 with lower average revenues were the most likely to fail. The cluster analysis revealed that areas with high poverty are associated with a lower overall number of non-profit organizations.
Conclusions and Implications: The non-profit sector has changed dramatically in the past two decades. Organizational type matters regarding longevity, as each category displayed different change patterns. These findings suggest that important information is contained within 990 forms which could help schools and students of social work better understand the environment of their future employment. Lastly, high poverty areas should remain a focus of investment, to ensure that at-risk population have access to important social services.