counting organizations can be a tricky thing

A number of recent conversations have revolved around the following issue: how, exactly, does one count organizations? If I wanted to do my own population level study, how do I develop a list of all relevant organizations? For the sake of argument, let’s assume you have a decent definition of the population. You know an organization of type X when you see it. There are number of strategies you can employ to build a census or sample:

  1. Publicly certified organizations: In some cases, organizations might be required to register themselves with the government. Political action committees, for example, must be registered. There may also be accreditation: an organization needs the state or some other group to sanction them. Then some government (or quasi-gov’t) organization will likely have a list of all orgs in your population.
  2. “People friendly” organizations: These are organizations which make it possible for outsiders to contact them. For example, they may be listed in directories. They might also be in the telephone book, or they may have a professional association that lists them in a nice book.
  3. Visible organizations: Some industries have organizations that are easy to spot. For example, it’s not too hard to find every auto firm that has ever produced a car because when you produce cars, you are covered in all kinds of media.

Ok, that was easy. But some populations are harder to sample or enumerate:

  1. Hypernetwork sampling: Consider churches. Some are huge, some are tiny. Some literally exist in garages, others on the internet. And there is no requirement that they register with the government, unless they need tax exempt status. So you take a sample of people and ask them to list their religious affiliations. That’s called “hypernetwork” sampling. Eliminate the repeats, or weight by frequency.
  2. Snowball sampling: Some groups don’t want to be found. Or they only serve specific populations and this won’t show up in a random sample of people. Then you find one or two prominent orgs in the field, and ask leaders to list other org. Go to the leaders of those orgs and repeat. If you do this a few times with different starting points, you’ll get a reasonable sample.

As you can see, these is nothing obvious or easy about the task, especially for organizations that are under no obligation to register with the state or otherwise present themselves to the public. It’ll will also depend on your research question. For example, if I wanted to study highly influential antiwar orgs (a project I am currently working on) then I only need to find some threshold and work from there (e.g., registered as a 501(c)). But if I wanted to study the small army of antiwar organizations that exist in every small town then I will not have the luxury of organizational directories. I will need a more aggressive sampling strategy that may involve hypernetwork sampling, media analysis, and so forth. And of course, if I am studying illegitimate or marginalized orgs (e.g., criminal groups or anarchist synidcates), ethnographically driven snowball sampling is my only choice.

Written by fabiorojas

August 24, 2009 at 12:22 am

Posted in fabio, mere empirics

9 Responses

Subscribe to comments with RSS.

  1. This is an enormous challenge for people studying civil society organizations in China, as only a small percentage are registered as such. Others are registered as businesses or simply lack legal status. Directories may reflect political biases. Hypernetwork can be difficult, as surveys of certain size entail government approval. Snowball seems to be the most common method people use when selecting case studies.

    Another question I have, though this may be covered above by the assumption that “you know an organization of type X when you see it,” is how to deal with the phenomenon of nested organizations. In the PRC, organizations will often exist as subdivisions of larger organizations, or exist under the wing of a university or think tank. The know-it-when-see-it standard can get a bit confusing, especially if the research is concerned with inter-organization coordination or institutional configurations. (Maybe “organization” is not the best unit there.) Does anyone have any recommendations for reading on this sort of question?



    August 24, 2009 at 4:03 am

  2. With regard to nested orgs, perhaps a two stage sampling frame: sample the federation/big org and then the units inside them. Of course, in China, it’s probably a huge problem to get lists of units in orgs, but it’s worth thinking about.



    August 24, 2009 at 4:17 am

  3. The anti-war organization sample is even trickier because recognizing “organization of type X” is not always easy to do. As you well know, not all anti-war organizations look like anti-war organizations – their specialty may be in a completely different social movement industry but they may occasionally ally themselves with anti-war organizations and participate in anti-war rallies. Does that kind of organization fit into the population?



    August 24, 2009 at 1:37 pm

  4. Ok, Brayden, you’ve just triggered the SHAMELESS SELF PROMOTION. Yes, in my antiwar work, we’ve exploited the fundamental fact that being an antiwar organization is ambiguous. In the recent Research in the Sociology of Organizations, I looked at groups that contacted antiwar protesters instead of defining “antiwar orgs.” Then I compared groups that defined themselves as peace groups vs. “outsiders.”

    Fabio Rojas. 2009. “Technology, Structure, and Heterogeneity among American
    Antiwar Organizations.” Research in the Sociology of Organizations 26: 221-247.

    My collaborator and I exploit this ambuguity in a working paper:

    Click to access Hybrid_Politics.pdf

    Both of these are based on hypernetwork samples of protester-org links.



    August 24, 2009 at 6:00 pm

  5. That RSO volume sounds like an excellent book! Maybe I should write a post about it sometime. :)

    Fabio, shameless self-promotion is useful when the topic is so close to home. I like the way you dealt with the population question in those papers.



    August 24, 2009 at 6:25 pm

  6. The interdisciplinary journal Nonprofit and Voluntary Sector Quarterly has several articles on “dark matter” – organizations that are so small or new that they are difficult to identify.

    See articles like the following:
    “Lost in Cyberspace: Shedding Light on the Dark Matter of Grassroots Organizations.” Lori A. Brainard and Jennifer M. Brinkerhoff. Nonprofit and Voluntary Sector Quarterly, Sep 2004; vol. 33: pp. 32S – 53S.

    “The Rest of the Nonprofit Sector: Grassroots Associations as the Dark Matter Ignored in Prevailing “Flat Earth” Maps of the Sector.” David Horton Smith. Nonprofit and Voluntary Sector Quarterly, Jun 1997; vol. 26: pp. 114 – 131.



    August 25, 2009 at 6:27 pm

  7. […] organizations we encounter turn out to be Potemkin Village facades. (Fabio recently addressed the conundrum of sampling organizations.)  I’ve described instances of this in previous posts about OEM medical research, pet food, […]


  8. I’ve been thinking about the challenges of counting organizations thought to be of a kind, and so far, one of the few things I can conclusively say is that it’s a lot harder to explain than prior research would suggest. Even though most org theory uses counts of specific types of organizations as independent variables, I think this is a good argument for doing research that treats counts as dependent variables that need explaining.



    September 1, 2009 at 2:19 am

  9. […] leave a comment » Previously on orgtheory: counting organizations can be a tricky thing. […]


Comments are closed.

%d bloggers like this: