@TomPepinsky Fair enough. Still, usually likelier to hold at sub-nationally than cross-nationally. But it should be shown, not just assumed.
— Fabrizio Gilardi (@fgilardi) November 25, 2013
I disagree that the unit homogeneity assumption is “more likely to hold” within countries than across countries as a general statement. (I also don’t think that we can ever “show” and assumption is true or false, we can only argue that it plausible or useful.) It all comes down to the research question, and in some cases assuming unit homogeneity actually makes a lot more sense across national boundaries. Here are some examples.
- Dan Posner‘s article on Chewas and Tumbukas in Malawi and Zambia. Here, the design rests on the assumption that ethnic groups are comparable across national boundaries, and that the national boundary captures a variable which varies at the national level: the relative sizes of ethnic groups.
- Michael Ross‘s book on rent seizing and tropical forest resources. The design is a focused comparison of four governments: Indonesia, Sabah, Sarawak, and the Philippines, each confronting a similar problem (“resource booms” in tropical hardwoods). Of note here is that Sabah and Sarawak are states within Malaysia compared with national governments in Indonesia and the Philippines. Sabah and Sarawak are not compared with Peninsular Malaysian states because we have no reason to expect that they would respond similarly to timber booms, not least because their tropical forest resources pale in comparison.
You might imagine other examples. I’ve often wondered about the effects of national language regimes on the development of Malay languages; for that, comparing Riau Malay to Johor Malay would make good sense. This map clarifies the matter:
Comparing Johor Malay to Kelantan Malay would not work (they both in the same country, so no variation in the language regime), and comparing Johor Malay to Pattani Malay would not either (they are different countries but starting from very different dialects prior to the language regime’s formation).
The takeaway is that unit homogeneity is always an assumption, one that only can be adjudicated with reference to the research question at hand. Should we assume that Sabah has the features that make it comparable to the Philippines to study forest management? Should we assume that Chewa-ness and Tumbuka-ness are comparable ethnic categories? Should we assume that Johor Malay and Riau Malay were comparable prior to the formation of Bahasa Indonesia and Bahasa Malaysia? All of these seem reasonable to me.