Ben Okonkwo: Marcus, hey — I have a question before we even start. Do you think railroads caused American industrialization?
Marcus Vale: I mean — yes? Obviously? That's like asking if the internet caused e-commerce.
Ben Okonkwo: Robert Fogel won a Nobel Prize arguing you're wrong. His answer — remove railroads from nineteenth-century America entirely, the economy grows almost identically. A century of assumed causation, gone.
Marcus Vale: Hold on. That's the actual finding?
Ben Okonkwo: That's the actual finding. And the method he used — counterfactual reasoning — is exactly what today is about. The core idea in one sentence: you imaginatively remove a suspected cause, hold other conditions fixed, and check if the outcome still follows. If it does, the thing wasn't a difference maker.
Marcus Vale: It's like — okay, the alarm clock thing. Did your alarm wake you up, or were you waking up anyway? You run the experiment in your head, alarm absent, and see what changes.
Ben Okonkwo: Exactly that structure. And Max Weber was formalizing this over a hundred years ago — the 'what-if' as a disciplined analytical move, not just speculation.
Marcus Vale: So Fogel isn't even the origin. He's applying a very old tool.
Ben Okonkwo: Right — rediscovering it, really. Historians have always done this. The July 1914 countdown — Austria-Serbia, Britain's decision to enter — that's a canonical test case. Change one diplomatic choice, does the war still happen? Same logic.
Marcus Vale: And that's the part that starts to feel slippery to me. Because the moment you say 'hold other variables constant' — in a world where railroads are reshaping capital markets and settlement patterns and political incentives — what does 'constant' even mean?
Ben Okonkwo: That's the load-bearing assumption the whole method rests on. And it's... hm. It's not obviously solvable. Which is what makes this interesting rather than just settled.
Marcus Vale: So we've got a real method, serious intellectual history behind it, and a genuine crack in the foundation. That's the episode.
Ben Okonkwo: And that crack — that's actually where the method gets interesting, not where it falls apart. Because the point of the exercise isn't to perfectly freeze history. It's to correct a specific kind of error. Three of them, actually.
Marcus Vale: Three errors. Okay, what are they?
Ben Okonkwo: Right — so imagine a pharmaceutical researcher. She's staring at a failed trial on a Wednesday afternoon. The patient recovered. The drug was administered. And the instinct — the very human instinct — is to say the drug worked. But her whole job is to ask: would the patient have recovered anyway? That's the counterfactual move. And the three errors it catches are — one, assuming the factor was indispensable. That it had to be there for anything to happen. Two, assuming it was causally decisive — that it must have produced the outcome. And three, assuming the outcome was inevitable regardless. Fogel's railroad work is correcting all three simultaneously.
Marcus Vale: So the error isn't just 'railroads mattered' — it's 'railroads were the only possible path to that economy.' That's the claim being punctured.
Ben Okonkwo: Exactly. And the tool for puncturing it is ceteris paribus — hold everything else fixed, remove only railroads, and see if growth still follows. If it does, railroads were not a difference maker at that scale. Now — Jack Levy, political scientist, he actually formalized best practices for this. Precisely because the plausibility constraint — the rule that you can only alter one historically credible condition — is incredibly easy to claim and almost never rigorously enforced.
Marcus Vale: Wait — Levy's point is that most people say 'I held variables constant' without actually doing it?
Ben Okonkwo: That's exactly it. The plausibility constraint sounds like a rule. It functions more like... hm. More like an honor system. You can't remove railroads and then also quietly assume canals stayed underdeveloped, that capital didn't reroute, that settlement patterns held. Those things were already moving in response to railroads before railroads were fully built. So when Fogel says 'remove railroads,' the tractability of that — the reason it's a good counterfactual — is that he bounded it tightly. One condition altered, not the entire era.
Marcus Vale: That's the thing though — if historians can't run a second trial, how do they actually know they held the right variables constant? Like, the researcher can run the trial again with a placebo. Historians get one timeline.
Ben Okonkwo: That's — yeah. That's the question the whole method has to answer. And the honest answer is: they use plausibility and generalization from adjacent evidence as proxies for experimental control. Which is real, but it's not the same thing.
Marcus Vale: Proxies for control. That's doing a lot of work.
Ben Okonkwo: It is. And this is where the difference-maker framing becomes load-bearing rather than just philosophical. Historians aren't claiming strict causation — they're claiming that railroads were or weren't a difference maker. That if you'd removed them, the outcome shifts. Now, difference making is neither necessary nor sufficient for causation in a strict metaphysical sense. The discipline basically accepts that gap because — well, because the alternative is silence. You can't say nothing about cause and effect in history.
Marcus Vale: So they're naming a limitation and calling it a methodology. Which — I mean, I get why. But that's the crack we need to go deeper on.
Ben Okonkwo: But naming it doesn't make it manageable. And this is where the crack goes deeper than just 'proxies for control.' Think about — okay, picture an intelligence analyst. It's July 28, 1914. Austria-Hungary has just declared war on Serbia. She's sitting with a cable in her hand asking: what if Britain stays out? Clean question. Except — Britain's decision to enter wasn't a standalone variable. It was woven into German strategic calculations about whether a two-front war was survivable, French morale about whether their eastern flank held, and global credit markets already pricing in collapse. You can't remove Britain's entry and hold those constant. They were moving *because* of the anticipated British decision, not after it.
Marcus Vale: So the cascade starts before the event you're testing.
Ben Okonkwo: Right — which is the actual objection, stated precisely. Changing one event plausibly alters the social, political, and ideological systems that explain *why* that event mattered. So you haven't frozen the context. You've just stopped noticing it moving.
Marcus Vale: That's objection two of the three. What are the other ones? Because I want to name them explicitly.
Ben Okonkwo: So — one: counterfactual pasts leave no empirical traces. You can't go check. The July 1914 world where Britain stays out doesn't exist in any archive. Two — the cascade problem we just walked through: ceteris paribus breaks because the system is non-modular. Three, and this is the one that I think is most dangerous: counterfactual narratives are especially prone to wishful thinking. Motivated reasoning. You build the alternative history that confirms what you already believed.
Marcus Vale: That's the Niall Ferguson problem.
Ben Okonkwo: Exactly. Ferguson's World War I counterfactuals — serious historian, real method — but critics read them as counterfactuals of convenience. Built backward from a preferred interpretation of British foreign policy. Not genuinely testing causation. Confirming it.
Marcus Vale: And the defense against that is — what? 'We have epistemic norms. Plausibility constraints.' That sounds like honor system again.
Ben Okonkwo: That's exactly what defenders argue — that plausibility constraints and generalization norms can answer all three objections. And I mean, I don't think that's nothing. Those norms keep you from saying 'what if the Roman Empire had nuclear weapons.' They constrain the space. But constraining the space isn't the same as disciplining what you pick *within* that space. A contingent world, one where outcomes really did hinge on specific decisions, is also a more interconnected one. More contingency means more cascade risk, not less.
Marcus Vale: No, I don't buy that the norms solve it. Because Ferguson passes every plausibility test and still gets accused of convenience. So the constraint didn't catch him — or didn't catch the bias, anyway.
Ben Okonkwo: Which is the honest place to leave objection three. There's no mechanical fix for motivated reasoning. What experimental design would tell you, before you publish, that you rigged your own counterfactual? That's — hm. That's not solved. And actually, the gap between 'difference maker' and 'cause' is going to make that problem worse when we get to what formal frameworks like Pearl's structural causal models actually expose — because they don't rescue historians from this, they just make the assumption visible.
Marcus Vale: So the method corrects three specific errors and creates a fourth one that's harder to see. That's the 'but' moment of this whole conversation.
Ben Okonkwo: And the fourth problem is actually embedded in the language historians use. 'Difference maker.' That phrase sounds causal. It isn't. Here's the concrete version — picture a wildfire investigator, February morning, she's standing in a burned hillside in Portugal, and she's trying to explain why this fire consumed forty thousand acres when the last one didn't. She finds the match. Specific match, specific hand, specific moment. Remove that match — no fire. So the match is a difference maker. But the cause? The cause is a drought that had been building for eleven months. A different match, a different hiker, same day — same fire.
Marcus Vale: The match was substitutable. The drought wasn't.
Ben Okonkwo: Right. And that's the distinction that historical counterfactuals almost never make explicit. A factor can be a difference maker without being the cause, because a substitute was available. And a genuine cause can fail to be a difference maker if the outcome would have happened anyway through another path. Those are different failure modes and the method — the standard but-for test — doesn't separate them.
Marcus Vale: So Fogel proves railroads weren't the match. But could he prove whether they were the drought?
Ben Okonkwo: No. And I don't think he claimed to. That's — hm. That's actually the honest version of what he did. The problem is the field often reads difference-maker findings as causal bedrock. Which is where Judea Pearl becomes relevant, because his structural causal models are trying to formalize exactly this separation.
Marcus Vale: Pearl's the directed graphs guy.
Ben Okonkwo: Directed acyclic graphs, structural equations — you map which variables causally influence which others before you run the counterfactual. So you're not just removing the match and watching. You've specified in advance: drought feeds fire-risk, fire-risk plus ignition source equals fire. The graph tells you where the match sits in the causal chain and where the drought sits. Now you can tell them apart.
Marcus Vale: That sounds like it solves it. What's the catch?
Ben Okonkwo: The catch is — actually, it's a large catch — you need the full causal graph specified before you start. Every variable, every arrow. Historians almost never have that. For July 1914, do you know in advance whether German strategic calculation causally precedes British entry, or whether British entry causally precedes German calculation? They're simultaneous, they're co-constituting. The graph Pearl's framework needs doesn't exist. So the SCM makes the assumption visible. It doesn't supply the assumption.
Marcus Vale: It's a flashlight, not a map.
Ben Okonkwo: That's exactly it. Which is why Li Qian's weak-necessity framework is interesting as an alternative — instead of demanding the full graph, it relaxes the strict but-for test. You're not asking 'was this factor strictly necessary?' You're asking 'does removing it make the outcome substantially less likely under a plausible model?' It's designed for exactly the cases where strict necessity is epistemically too demanding.
Marcus Vale: Does that fix the difference-maker problem though? Or does it just lower the bar for claiming you've found a cause?
Ben Okonkwo: It — no. It doesn't fix it. And I think that's the honest landing place. What Li Qian's framework does is give historians permission to say 'this was a plausible difference maker under a specified model' without overclaiming. Which is actually more honest than what the field usually says. The practical implication is this: when historians run a counterfactual and find the outcome shifts, they've established that one factor was a difference maker in that specified model. Not causal bedrock. That framing should be in the claim itself.
Marcus Vale: So the method isn't broken — the language around it is. Historians should say 'we identified a plausible difference maker' and stop there, not slide into 'we found the cause.'
Ben Okonkwo: That's exactly it. And I think that's actually — hm. That's the honest product of the whole enterprise. You ran a disciplined counterfactual, you found a plausible difference maker, and you stopped there. That's not a consolation prize. That's what the method can actually deliver.
Marcus Vale: Which brings us back to Fogel. Because he didn't prove what caused American industrial growth. He proved railroads weren't indispensable to it. Those are different claims.
Ben Okonkwo: Right — and that distinction is actually the point, not a limitation of it. Correcting a deterministic error — the assumption that railroads were causally decisive, that the outcome was inevitable through them — that's real progress. It's just quieter progress than 'we found the cause.'
Marcus Vale: And the alternative to that quiet progress is — what? Pretending history is more deterministic than it actually is. Which is worse.
Ben Okonkwo: Much worse. That's where you get the real methodological damage — treating every correlation as a cause, every outcome as inevitable. Weber saw this, Pearl formalized a version of it, Levy tried to discipline it with rules. The whole lineage is just people trying to hold that particular error at bay.
Marcus Vale: You asked me at the top whether railroads caused American industrialization. I said yes, obviously. Like it was settled.
Marcus Vale: Now I'd say — railroads probably weren't the difference maker Fogel's contemporaries assumed. Which is a lot less satisfying than 'yes, obviously.' But it's probably closer to true.
