New Eyes, New Wool: Sunstein on History

Cass Sunstein is a marvelous salesman but an intellectual prestigitator. Well, what else can you call a man who is half of the team that half a decade ago had a lot of prominent libertarians very nearly convinced that manipulating people’s choices was not only morally acceptable but a way to maintain individual freedom? With a recent essay weighing in on the latest iteration of the ongoing controversy about the place of counterfactuals in the study of history, I think we’re in a position to identify his MO:

(1) Give due attention to a subject such that you can rightfully claim to know enough about it to comment
(2) Posit that whatever the controversial activity is is unavoidable, part of the “architecture” of the field
(3) By failing to say what the limits on acceptable indulgence in this activity are or should be, leave the reader with the impression that it is more acceptable than it is

That’s how Thaler and Sunstein sold “libertarian paternalism,” and it’s how Sunstein is selling historical counterfactuals now.

The essay is a book review in the New Republic of Richard J. Evan’s recent (and controversial) attack on the use of counterfactuals in history. Of course, what Evans has in mind are fanciful counterfactuals – of the kind that imagine what would have happened but for something that observably did happen. I haven’t read Evans’ book, so I can’t really say anything about whether he covered his bases here. The most common complaint I see about his book on the internet is philosophical: namely that to say a thing A caused another thing B, it is necessary to posit that without that thing A, B would have happened differently, or not at all. I’m not sure that’s entirely true – not strictly speaking, anyway – but it’s a cogent rebuttal, and I’ll take it on assumption that Evans didn’t sufficiently address that point in his book. This gives Sunstein the opening he needs:

Yet the most fundamental problem is that Evans does not grapple sufficiently with the fact that historians do not only offer narratives; they also offer explanations. They say that some event—the rise of Nazism, the Vietnam war, the election of Ronald Reagan, the attacks of 9 / 11—had particular causes. It is not possible to take a stand on the existence of causes, or on their relative importance, without thinking about what the world would be like if one or another were removed.

That’s true, but I think it’s the limits of such thinking that’s at issue, not whether such thinking can take place. So when Sunstein follows that passage with this one:

If we say that the Vietnam war or the passage of the Civil Rights Act of 1964 was “caused” by the Kennedy assassination, we must be imagining a world in which Kennedy was not assassinated, and hence making a claim about what would have happened in that alternative and historically unrealized world.

he’s maybe going a step too far.

When we say that the Civil Rights Act of 1964 was “caused” by the Kennedy assassination, I think we’re explicitly NOT saying two things Sunstein seems to assume we’re saying: (1) That the Civil Rights Act would not have happened without the Kennedy assassination and (2) that the Kennedy assassination was the sole or most important cause of the Civil Rights Act. Or, to use a legal term of art that Sunstein, as a legal scholar, would be comfortable with, we’re not necessarily asserting that the Kennedy Assasination is a but-for cause. It’s obvious what that means, eh? “But for X, Y would not have happened.”

But-for causation shows up in law as a component of proving culpability. The recognized problem with it as a standard for legal blame is that it is (a) unacceptably easy to prove and (b) does not in and of itself assign culpability. For example, “but for my decision to drive in the rain, the accident where I slid into that playground and killed that child would not have happened.” True, but legally irrelevant, because most drivers manage to drive in the rain regularly without killing or even injuring or even scaring anyone. A but for cause is merely a necessary precondition for proving culpability. After that, you have to show that the action in question was also a proximate cause, which means a lot of things, including but not limited to:

(1) The harm could reasonably have been predicted (the “extraordinary in hindsight” rule)
(2) There are no (significant) intervening causes (the “direct causality” test)
(3) If the action were repeated, the risk of harm would increase (the “foreseeable risk” rule)
(4) The action is risky in the way that would have predicted the harm (the “risk rule”)

The Kennedy Assassination seems to fail all of these tests as a cause of the Civil Rights Act – if you’ll allow me to substitute “result” for “harm” of course. There was no way to predict that asassinating Kennedy would lead to the passage of the Act, so (1) is out, there are plenty of intervening causes (Johnson is not subsequently assassinated, Johnson fails to bring the bill to vote before the winter recess – as it was during the winter recess that the opinion polls showing support for it in the North came out, probably shifting some votes – Kennedy had decided to make passing it a legislative priority immediately prior to his assassination so that it was fresh in the public mind, all of the Congressional elections are what they were, previous civil rights bills had been watered down, etc.), so (2) is out, (3) is inapplicable to historical counterfactuals, so it either passes or fails on a technicality if you insist, and (4) is certainly out, since it isn’t hard to find examples from history where assassinations result in signature legislation being defeated rather than approved.

So Kennedy’s 1963 assassination certainly fails to be a proximate cause of the bill’s passage, and in all honesty it probably fails to be a but for cause as well, since it’s entirely conceivable that Kennedy could have gotten the bill passed had he survived the attempt, or had he even failed to go to Dallas and be assassinated on that date under those circumstances by that individual (pace Oliver Stone).

All of which is to say that I don’t know that saying that Kennedy’s assassination led to the passage of the Civil Rights Act really is speculating that the Act wouldn’t have passed without it. And it’s only on the thinnest of possible interpretations that it’s speculating on what the world would have been like had Kennedy not been assassinated. And I think that is Evans’ (possibly inadequately expressed) point: that it’s one thing to say that something caused another thing, but quite a different, though still acceptable, thing to say that absent that cause, the thing would not have happened. What Evans considers impermissible is saying that if a thing caused another thing then but for that thing happening, the world would probably be different in the following specific ways. It’s getting into too many specifics about what would have been different that Evans objects to – and that I object to as well.

Now, you’ll have noticed my weasel words – what’s “too many” specifics? Like with all such things, it’s impossible to draw a clear line. But the fact of there being no clear line doesn’t mean we can’t meaningfully use the phrase, and a useful thing for someone to do would be to provide some guidelines about when we’ve gone too far. This is what Sunstein expressly fails to do, and why I think the whole review as a whole is deceptive.

Evans himself is no exception. In responding to Ferguson’s argument about the likely consequences of British neutrality in 1914, he offers some counterfactual history of his own. He suggests that if Britain had stayed out of the war, Germany would not have scaled back its war aims. And in responding to Charmley’s arguments about the potentially beneficial effects of appeasing Hitler, he suggests that Germany would have attacked Britain in any case, with a higher probability of victory.

That’s all true – but of course Evans was offering those scenarios not to demonstrate that his own counterfactuals are superior to theirs, but rather to demonstrate that counterfactuals are cheap to manufacture and thus inconclusive. If we can freely speculate on both sides of an issue at all times, the method itself is useless, no?

True, his statements on these counts are qualified, but in explaining the rise of Nazism (an area in which he has great expertise), Evans writes more firmly, saying that the “key factor” was “the Nazi’s storm troopers’ escalating use of violence” — which is an unambiguous suggestion that in the absence of that violence, the Nazis might not have come to power.

Alright, fair enough, but that’s not the same thing as speculating about non-violent Nazis, because non-violent Nazis are not Nazis. The party that came to power came to power in this way and ruled with these and those characteristics, but what, exactly, would it mean to say that if the Nazis had not used violence they would not have come to power? I think we’re not speculating so much about a world where Nazis were different as a world where Nazis never existed. We’d be using the word “Nazi” to refer to a historical fantasy: a fascist, nationalist, racist/Voelkish party that was unwilling (or less willing) to use violence to achieve its ends. Could such a thing have existed in Weimar Germany? Well, it did – it was the DNVP – and it did form some governments (including Hitler’s original government in coalition). But you see, that’s the point – the DNVP is the DNVP, it is not the NSDAP. So what exactly is Sunstein accusing Evans of doing? Of saying that the NSDAP could have been the DNVP? Of saying that there was still an NSDAP that differed only on the dimension of “violence” from the existing NSDAP and still managed not to get absorbed into the DNVP? That without the use of violence, the NSDAP would have continued on in irrelevance, and the DNVP would have … what? What would have happened, exactly? The DNVP would have won some elections? And done what? What about the SPD? The KPD? Did they form a Soviet Republic of Germany together? Did the SPD come to power and ban the KPD? WHAT? And this, again, is the point. Evans doesn’t mind saying that this, that or the other is a contributing factor to something – what he’s skeptical about is anyone’s ability to imagine in any detail what would have happened had x been different. It’s the getting into specifics about the alternate world that’s at issue, not the attributing causality to something.

Evans adds that even when historians call a “cause necessary rather than possible or contributory, they almost never speculate about the alternative course events might have taken had it not been operative.” This misses the point.

No, it doesn’t. It IS the point. Because:

Whenever historians call a cause necessary, they are, in the same breath and by virtue of that very statement, speculating about an alternative course.

No they’re not. Or if they are, they’re speculating about it as a completely abstract thing – a formless thing with no details. They’re not drawing any conclusions from it about how the world would have been “but for” cause X, they’re simply noting that it would have been different.

Sunstein consistently pretends to miss this point, but I don’t think he really does. Take this, for example:

Here is another way to make the point. Social scientists test hypotheses. They might hypothesize, for example, that if people have to pay a small tax for plastic bags at convenience stores, they will use fewer plastic bags. To test hypotheses, social scientists usually like to conduct randomized controlled trials, allowing them to isolate the effects of the tax. Such trials create parallel worlds and hence alternative histories—one with the tax and one without it. Historians cannot conduct randomized controlled trials, because history is run only once. Yet they nonetheless develop hypotheses, and they attempt to evaluate them by reference to the evidence

And you see the problem immediately: the analogy doesn’t actually “make the point,” it rather makes the opposite point. Or, actually, it obscures the truth about the difference in scope between historical and scientific hypotheses. Scientists only conclude what they do because they have the ability to “create parallel worlds and hence alternative histories.” Historians do not have this ability, because, as Sunstein says, “history is run only once.” So, historical speculation lacks the critical ingredient that allows people to say with any confidence and in any detail how the world would be different if a particular variable had a different setting. Which is why people shouldn’t indulge in too much historical speculation too seriously: because it can’t be tested. Like, at all.

Sunstein is using a tired old trick here – hoping that by saying something true and obvious and taking the contrary position you’ll mistake what he’s said for evidence for his position, just because he’s the one saying it. Fortunately, we’re too smart for that. This analogy with scientific experimentation is an argument against historical counterfactuals being informative, in fact, no matter who’s offering it. Even if it’s Cass Sunstein.

Evans appreciates the entertainment 
value of the most imaginative counterfactual narratives, but he doesn’t want them to be taken seriously, or to be seen as what historians do. With Thompson and Oakeshott (and countless others), he thinks that historians should explain what did happen, not what didn’t happen. The problem is that, to offer an explanation of what happened, historians have to identify causes, and whenever they identify causes they immediately conjure up a counterfactual history, a parallel world.

The sublte little error in that sentence is “a.” In saying “a parallel world” Sunstein is implying that they have a particular one in mind. But they don’t, and Sunstein knows it. “They immediately conjure up a plethora of counterfactual histories, a plethora of parallel worlds,” would have been the right way to put it.

So thanks for playing, Cass, but we know your game now.

(1) Give due attention to a subject such that you can rightfully claim to know enough about it to comment
(2) Posit that whatever the controversial activity is is unavoidable, part of the “architecture” of the field
(3) By failing to say what the limits on acceptable indulgence in this activity are or should be, leave the reader with the impression that it is more acceptable than it is

It’s critical that (1) be done in all sincerity, and there’s no reason to doubt that Sunstein actually read and digested Evans’ book. That’s actually what makes the response so galling. As for (2), check! For the broadest possible use of the term “counterfactual,” we do indeed have to have them to attribute anything in history to causes. The misdirection happens in (3), when Sunstein leaves you with the impression, without actually constructing an argument for it, that a great deal more counterfactual reasoning is acceptable in historical studies than actually is.

And as I said, we’ve seen this before with “libertarian paternalism.” First, Sunstein and Thaler make reference to a lot of real research on how people behave and so establish their credentials to talk about psychological manipulation. Then they – convincingly (this is crucial) – argue that in most situations it is impossible to structure offerings in a way that fails to influence people’s choices. The problem is with the conclusion: therefore, people should feel free to manipulate people’s choices. But it’s a canard. It’s in failing to identify the limits to that license that they give up their right to the label “libertarian,” because anything “libertarian” is concerned, first and foremost, with drawing the limits around the kinds of powers we allow ourselves over others.

Enough. Historical counterfactuals are dangerous in the obvious way: because they are unconstrained and tend to foster false impressions of what historians do. If Sunstein would care to demonstrate a way that a historian can use a fanciful counterfactual responsibly, that would be a useful thing to do. Until then, his attack on Evans on the basis of an extemely thin technicality is not only unconvincing, it’s harmful

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>