One Sector, Several Labor Markets
The U.S. nonprofit sector is not one labor market with one pay rule. The useful question now is where nonprofit and for-profit employers actually meet, what kinds of workers and jobs are concentrated in those places, and how the comparison changes across industries. This matters because a worker–firm model can separate a worker component from an employer component only where the data contain meaningful links and comparable employment. Health care, education, publishing, and the arts show why the finding that selection explains little of the average penalty must be read as a conditional result rather than a universal description of nonprofit work.
The comparison gets stronger where employers overlap
A comparison between nonprofit and for-profit pay needs a plausible counterfactual. If nonprofit organizations and for-profit organizations employ people in the same industry, recruit similar skills, and provide related services, a difference in pay can be examined as a relationship between employer types. If the sectors occupy entirely different activities, the difference may mostly describe the activities and workers each sector contains. No statistical adjustment can create a real-world comparison where the labor markets do not overlap.
The Census evidence shows that overlap is uneven. Nonprofit and for-profit firms coexist in some industries, but many sectors contain very few nonprofit employers. Education is a prominent exception in the study's industry breakdown: nonprofits make up a larger share of firms there than for-profits, roughly 40 percent compared with just over 15 percent. Manufacturing, mining, and wholesale trade have very little nonprofit presence. A national nonprofit average therefore is not a balanced sample of every part of the economy. It is weighted toward the activities in which nonprofit organizations actually operate.
The BLS data make that concentration concrete from the worker side. Nonprofit employment is heavily represented in hospitals, religious organizations, elementary and secondary schools, colleges and universities, social services, and other professional services. The sector also contains a large share of professional and related occupations, alongside service and administrative work. These are not just labels in a table. They describe different credential requirements, revenue sources, working conditions, and forms of service. A nonprofit nurse, teacher, social worker, clergy member, or administrator may face a labor market with little resemblance to that of a worker in a nonprofit arts organization.
Concentration changes the meaning of the overall average in two ways. First, a large industry can dominate the sector estimate even if its nonprofit pay relationship is unusual. Second, an industry with a high share of nonprofit employment may still be a poor comparison if there are few for-profit firms doing similar work. The fact that religious organizations are overwhelmingly nonprofit, for example, tells us that they are important to the composition of nonprofit employment but does not by itself reveal a nonprofit employer effect. A comparison needs both a meaningful number of nonprofit workers and a plausible alternative employer.
This is why “within industry” is useful but not automatically sufficient. Holding broad industry constant removes one major source of composition, but a broad category can contain distinct services and occupations.