Observational Equivalence in Regional Conflicts
Non-Arab States in the Middle East’s Cold War
Explaining the Israeli-Iranian-Turkish “Axis”
I’m currently reading Trita Parsi’s Treacherous Alliance (Parsi 2007), and while there are many many things in it that blow my mind, one that I think is relevant to social science is the “observational equivalence” which emerges from the following passage:
On July 24, 1958, [David Ben-Gurion] sent a personal letter to U.S. President Dwight Eisenhower in which […] he wrote that “with the purpose of erecting a high dam against the Nasserist-Soviet tidal wave, we have begun tightening our links with several states on the outside of the perimeter of the Middle East: Iran, Turkey and Ethiopia. Our goal is to organize a group of countries, not necessarily an official alliance, that will be able to stand strong against Soviet expansion by proxy through [Egyptian President] Nasser, and which might save Lebanon’s freedom and, maybe in time, Syria’s.” Eisenhower heeded Israel’s call and offered America’s backing to the periphery alliance.
So Israeli intelligence provided Iran—whose military was constantly preparing for potential Iraqi or Egyptian attacks—with extensive intelligence on Egyptian military movements and planning. Together with Turkey, the Iranian and Israeli intelligence services constantly monitored Soviet-Egyptian-Iraqi military cooperation. The three non-Arab countries observed Soviet military shipments destined for Egypt and Iraq as they made their way from the Black Sea to the Persian Gulf through the Suez Canal.
What’s mind-blowing about this, for social-science brains like mine infected with the DAG/causal-inference bug, is how the formation of this pre-Iranian Revolution alliance can be “explained” via two totally different yet plausible narratives:
Hypothesis A: From the rise of Nasser until the Iranian Revolution, countries in the Middle East formed alliances on the basis of Cold War security considerations, with US-aligned / NATO countries therefore polarized against Soviet-aligned countries.
Hypothesis B: From the rise of Nasser until the Iranian Revolution, countries in the Middle East formed alliances on the basis of “ethnic affinities”, with non-Arab countries therefore polarized against pan-Arab / Nasser-aligned countries.
The point being that, despite these two hypotheses representing common competing explanations of global geopoltical patterns (in International Relations/Comparative Politics, at least), they imply the exact same resulting (observed) alliances:
If Hypothesis A is true, then we would expect the following “camps” to emerge:
| US-Aligned | Soviet-Aligned |
|---|---|
| Israel | Egypt |
| Iran | Iraq |
| Turkey | Syria |
If Hypothesis B is true, then we would expect the following “camps” to emerge:
| Non-Arab | Arab |
|---|---|
| Israel | Egypt |
| Iran | Iraq |
| Turkey | Syria |
So… Which one better explains the fact that we did indeed observe the formation of the alliances shown in both tables? The answer, given only these hypotheses and “datapoints”, is that they are exactly observationally equivalent: the explanatory power of the “security hypothesis” and the “ethnic hypothesis” are literally (mathematically?) equal…
Understanding Present-Day Inequality in the US
My point here is not really to make an argument about Cold War history, or to come down on one side or the other, especially since there’s a ton of really great work in social science studying exactly these specific hypotheses: see, e.g., Fearon and Laitin (2003) on the “ethnic hypothesis” in general, and then maybe Kerr (1967) as an OG study and Halliday (2005) as a more up-to-date study of the “Arab vs. non-Arab” hypothesis.
Rather, my point is just that reading Parsi’s book suddenly “clicked” this thought into place in my head, when it had been vaguely bothering me for a long time!
The same issue emerges, for example, in discussions about the role of race vs. class in structuring disadvantage in the US:
Hypothesis A: Material and/or psychological disadvantage in the US is structured around race, with whites therefore obtaining vastly disproportionate privilege
Hypothesis B: Material and/or psychological disadvantage in the US is structured around class, with the rich therefore obtaining vastly disproportionate privilege
Once again, given the extreme correlation between race and class in the US (like there was a correlation between non-Arab-ness and US-alignment in the Cold War), we have that:
If Hypothesis A is true, then we would expect the following pattern to emerge:
| Whites (Disproportionately Wealthy) | Non-Whites |
|---|---|
| Advantages | Fewer Advantages |
If Hypothesis B is true, then we would expect the following pattern to emerge:
| Wealthy (Disproportionately White) | Non-Wealthy |
|---|---|
| Advantages | Disadvantages |
And since the outcome in both tables is precisely what we observe (among millions of concrete examples, see e.g. any of the research from the Opportunity Insights lab at MIT), these starkly-presented hypotheses end up being observationally equivalent, unless we add further context!
This latter example actually does give us a way forward, however, which is why I’m writing this rather than just letting it swirl around in my head: we can adjudicate between these, in a way, as long as we take into account and work to methodologically disentangle the correlations that give rise to this observational equivalence.
Here I have in mind, for example, the articles/books I got to read in my American Politics PhD seminars, such as Gilens (2012). The thing that “clicked” with me from that book in particular was the notion that:
- The relative “responsiveness” of Democrats vs. Republicans to rich vs. poor citizens is complicated by the fact that public opinion of the rich and the poor in the US is actually highly correlated on most issues…
- The only way to disentangle this is to focus in on those topics which exhibit the greatest disagreement between rich and poor.
- And, only once those topics have been identified and isolated, and the analysis has been conducted on Democratic vs. Republican response, do we find that indeed the Democratic Party is more responsive to “middle-income” voters while the Republican Party is more responsive to “high-income” voters (both exhibit literal zero responsiveness to low-income voters).
If it wasn’t nearing 6am I would try to do a better job relating this point back to the two examples, but the gist is that I truly believe social scientists can use this disentangling-through-careful-causal-reasoning process to
- Successfully ascertain the relative impacts of race and class on life advantages in the US, in a concrete, precise, quantitative sense, which in turn can
- Help inform political action/policy in terms of suggesting the most effective “pressure points” on which to push, to reduce and someday eliminate these inequalities.
This final point, that understanding and applying social science methodologies in this way can help inform political action—i.e., help inform which policies or movements people should put their limited energy into—is something I think first “clicked” in my head when I read Reed (2000) back in the day… probably no other book (imo) has done a better job of foregrounding the point that, despite the tug-of-war between race-first vs. class-first explanations of oppression and exploitation in the US (in my paraphrase of his arguments throughout):
- Political developments which make life less shitty for poor people in the US also help make life less shitty for non-whites, and
- Political developments which make life less shitty for non-whites in the US also help make life less shitty for poor people.