Friday, August 31, 2012
Signing Off -- And More on the Nirvana Fallacy
This will be my last post during this guest-blogging stint. Thanks to Dan and the others for having me. It has been fun.
Over at Truth on the Market, my colleague Adam Mossoff has a response to my post about his paper. I argued that he misconceives the Nirvana fallacy, because to me it pertained to a situation where one was comparing against a idealized and fictional alternative. Adam’s response is that nobody—at least nobody in economics or law—compares against a fictional alternative. He challenges me to name a single legal or economics scholar who does. A much longer post--where I think I answer his challenge--will follow after the break.
If the only question is an economist comparing against a fictional baseline, then every economist does that. Every economics student learns on their first day to compare everything against the fictional alternative of a perfectly competitive market involving perfectly rational actors, which nobody thinks is descriptive of real life. The blog he is blogging at, “truth on the market,” is named after a theory that says that a perfectly competitive capital market will instantaneously incorporate all publicly available information.
To be sure, economists spend much of their time relaxing the assumptions of the models and also testing the models to see if they approximate reality. So we don’t think that the market is perfectly competitive, but we think it is highly competitive, and thus the truth on the market theory works well enough. But the analysis begins with a fictional and idealized conceptual baseline and works backwards. And Demsetz’s whole point with the Nirvana fallacy is that often economists don’t do enough work on the back end to ensure that their fictional and idealized conceptual baseline actually matches reality. To that extent, Adam is correct that “the question is whether such an idealized counterfactual is a valid empirical metric or not.”
But the emphasis is not on the words “empirical validity” but “idealized counterfactual.” And my point is that there is no “idealized counterfactual” in the scholarship that Adam criticizes if we correctly understand that term. Adam seems to define an idealized counterfactual as any time someone is wrong on a empirical question, with his example being someone who predicted that Nixon would lose the election in 1972. If that is the definition of an idealized counterfactual, then every public policy debate will degrade into mutual accusations of the Nirvana fallacy. Patent scholars have strongly differing intuitions about whether the patent system fosters of hampers innovation on net. Nobody currently has rigorous evidence about this. By Adam’s logic, at least one side must be living in Nirvana.
Rather, the idealized counterfactual is what economists do when they begin their analysis first with an assumption of perfectly rational actors in perfectly competitive markets and then work backwards. The analogue in the patents-to-real-property context would be to (a) first theorize a perfect real property system with zero transaction costs, (b) have an intuition that (a) is a valid baseline because the actual real property system has low transaction costs, and (c) therefore compare the actual patent system to (a). There would be two possible critiques of this practice: (1) the empirical intuition (b) is not correct, and (2) one shouldn’t be comparing real life systems to hypothetical fictional baselines in the first place. My point is that people in the patent debate are not doing this. Precisely because we take Demsetz’s critique to heart, we are avoiding positing a comparison to an idealized alternative of a real property system of zero transaction costs, and positing the actual real property system with non-zero and empirically-verifiable transaction costs as the baseline. We are comparing the patent system to the actual real property system, and arguing that the patent system has higher transaction costs than we understand the actual real property system to have. There is only one possible criticism, which is that our empirics are wrong. But we have taken Demsetz’s critique to heart precisely because we are engaging with the question at an empirical level: the question is framed conceptually as “what are the actual transaction costs of the actual real property system?”
Adam can of course say that our empirics are wrong, but there is no longer a conceptual problem with the conceptual baseline, which I understand the Nirvana fallacy--i.e. conceptual mistake--to be about. I just don’t see where Adam is pointing to a false conceptual premise, except by reference to the fact that he thinks the critics are empirically wrong (I am putting aside here his perfectly valid point about “land” rather than “estates”; that is a conceptual mistake, but that can be fixed by moving to a comparison between invention boundaries—i.e. claims—and land boundaries). If Adam wants to say that the actual real property system has high transaction costs, then he should just say that and either put forward empirical evidence or his own intuitions about it. Even saying that the critics’ empirical intuitions are unsupported and that is itself a problem is fine. But I don’t think he has made the case for saying there is a logical fallacy here. To paraphrase Adam’s post’s title, the Nirvana fallacy is not the “you are empirically wrong” fallacy.
Finally, in his blog post, Adam also says that “comparing an actual institutional system to a ‘loose intuition’ about another institutional system” is exactly what Harold Demsetz defined as the Nirvana fallacy. In one sense he is correct about this, so I must tease this out a little. In every comparison between two different systems, there is necessarily an embedded hypothetical fictionalized alternative. When I compare patents to real property, at some conceptual level it cannot be done. What I am really comparing is (a) the actual patent system, and (b) a hypothetical patent system that works like the actual real property system. So there is a fictional baseline there, and there is a valid conceptual critique. So I take back everything above if that is Adam's point.
But, if that is what Adam is saying, then the logical implication is to stop all comparisons between patents and real property, period, because the conceptual differences between patents and real property are unbridgeable. That is, to my understanding, not Adam’s argument, because he makes the conceptual comparison between patents and real property in his own work as well.
Posted by Tun-Jen Chiang on August 31, 2012 at 11:33 AM | Permalink
TrackBack URL for this entry:
Listed below are links to weblogs that reference Signing Off -- And More on the Nirvana Fallacy:
Nice post, TJ. This reminds me of one of my favorite economist jokes (told to me by an economist).
A physicist, a chemist, and an economist are trapped on a desert island. A case of tuna cans washes ashore, and the three start to argue how to open the cans.
The chemist says, "Well, let's take some sea water and spread it on top of the can. The metal will corrode and then open."
The physicist says, "That's too dangerous and will take too long. Let's take that rock over there, and with the proper force and leverage, we can pop the can open."
The economist then says, "You both have got it all wrong. FIRST, assume you have a can opener...."
Posted by: Michael Risch | Aug 31, 2012 12:54:07 PM
On a more serious note, it seems like the real difference between you and Adam is just how "unrealistic" the alternative has to be to qualify as "Nirvana." Perhaps you both agree that perfect competition fits that bill, as does my can opener. Also fitting the bill must be the assumption that regulation will always fix things (the Demsetz example), or perhaps that there is NO traffic at 4AM (thanks Wikipedia). And so the question becomes how "loose" the view of the alternative has to be.
I gather from Adam that the traditional comparative view of trespass law is too idealized - too far from real, too idealized. I gather from TJ that so long as you can provably decide one way or the other, then it's not really the Nirvana fallacy, it's just an error of fact.
I feel like Adam has to be right, or there is a line drawing problem. The idealized nirvana of perfect competition is empirically false no less than the empiric falsity of easy trespass (assuming Adam is right about that - I'm not a property guy). So, once you've got an idealized empiric falsity, I don't see where you draw the line. One can certainly be wrong empirically and then use that wrong scenario in an idealized way. Politicians do it all the time. It is obviously a factual error, but I think it also leads to a conceptual error. I think it is why philosopher types are very, very careful when they write.
That said, I don't get the hubbub about it. Whether you call it a conceptual error or a factual one, you're (figuratively, of course) still wrong, and you can't be conceptually wrong without first being factually wrong. The only question is how much work you put into being factually wrong - whether your idealized world is based on false data, a misinterpretation of real data, or a loose interpretation of what the data is.
Posted by: Michael Risch | Aug 31, 2012 2:39:59 PM
Michael, no, you are fundamentally misunderstanding my point, and under your test patent scholars will charge each other with mutual accusations of the Nirvana fallacy because each side thinks that the other is not only empirically wrong, but grossly empirically wrong. My point is that the Nirvana fallacy points to a conceptual mistake that occurs regardless of the empirics. Even in a situation where the idealized theory (e.g. perfect competition) is empirically pretty close (e.g. a capital market that is highly efficient), there is a conceptual mistake in comparing some other system to it. The conceptual mistake is to compare a real life system to a hypothetical and fictional perfectly competitive market, which is still an idealization even if real life is empirically close. The correct conceptual baseline in such a case is to compare the real life system to the real life highly-but-imperfectly efficient capital market. The point of the Nirvana fallacy, to me, is that one should not compare ANYTHING to the theoretical ideal. Our conceptual baseline should always be defined as the actual manifestation, and then we can empirically try and discern what level of efficiency it has.
Second, on the point of "does it matter." Well, it matters quite a lot in scholarly discourse and a situation of empirical uncertainty. Saying that an argument is empirically wrong invites the response that it takes empirical evidence to refute an empirical claim. Saying that an argument contains a logical fallacy is a stronger statement and one that requires no empirical backing. I am fighting Adam on this move in particular because it transforms every empirical dispute into a charge of Nirvana fallacy.
Posted by: TJ | Aug 31, 2012 4:07:58 PM
And in case it is not clear, the corollary to my point is that, so long as someone employs a conceptual baseline of actual reality, they are not committing the Nirvana fallacy no matter how empirically wrong they turn out to be about the content of actual reality. So Ms. Tael who thinks that Nixon would lose the 1972 election based on a sample of her friends is not committing the Nirvana fallacy, because her conceptual baseline is still (presumably) the actual votes of actual voters in the United States. She has an horribly unrepresentative sample and this led to a grossly wrong empirical result, but those are simply empirical mistakes. Being really really wrong empirically might be an "idealization" in colloquial language, and we may well say that she is living in her own bubble world, but it is not a logical fallacy in the same way as theorizing a perfectly competitive market or perfectly motivated public servants or a can opener, where the point from the start is that they are fictional.
Posted by: TJ | Aug 31, 2012 4:36:17 PM
Yeah, I guess I'm not buying your view then. I think you can compare to a conceptualized ideal. And indeed you should if that ideal can be measured at all. I don't think that's what the Nirvana fallacy is all about in my short time reading about it (very short, I admit).
When I compare leaving now to leaving at 4AM because there is NEVER any traffic at 4AM I commit the fallacy because my ideal is not really real, and thus the comparison is not really fair, since I'll surely hit SOME traffic at 4AM - this is the empirics point. But if I compare leaving now vs. leaving at 4AM vs. the crow flying, and measure the likely differences at different times to the mythical crow, I'm not committing the Nirvana fallacy - I'm setting up a real benchmark to compare real alternatives against a measurable --if unattainable--ideal. And yet you seem to be saying that's the ONLY time we commit the fallacy. I guess I just don't understand that.
As for patent scholars, I don't know that we should reject the definition because it would mean that what patent scholars are doing is a conceptual fallacy. On the contrary, I think a whole lot of that goes on (on both sides): comparing a very complex set of real world facts to a simplified and idealized intuition of how things ought to work (or do work in some comparable field). Just because I'm guilty of it sometimes doesn't mean the definition is wrong.
Posted by: Michael Risch | Aug 31, 2012 6:27:14 PM
No, if you compare you driving home (whether now or at 4am) to a crow flying, then you are committing the Nirvana fallacy in the sense that I explained in the last two paragraphs of my post. At a deep conceptual level, you are not comparing driving to flying--the comparison makes no sense. You are comparing real driving against a hypothetical idealized driving that is as fast as flying. And the logical implication is to not compare driving to flying. But Adam is not, to my understanding, making that version of the argument. If he was, I'd agree with him [at least as to there being a conceptual problem, though I would still defend the comparison between real property and patents as being useful, even if it has conceptual problems].
And to clarify, of course I am not saying that a definition of the Nirvana fallacy must be wrong if patent scholars--i.e. me--are committing it. That would be opportunistic rationalization at its worst. I am saying that a label of "Nirvana fallacy" that reduces virtually EVERY SINGLE PUBLIC POLICY DEBATE to mutual accusations of it is wrong, or if it is right (since of course at some level we can always choose our own definitions of terms) becomes unhelpful to the debate. Does the death penalty deter crime or not? Does same-sex marriage produce positive or negative child welfare effects? Does widespread gun possession increase or decrease crime? Each side in those debates has passionate empirical intuitions and believes the other side is not only empirically wrong, but grossly empirically wrong.
Posted by: TJ | Aug 31, 2012 6:41:02 PM
Michael, having re-read your comment, let me add one more thing. Your statement that "[w]hen I compare leaving now to leaving at 4AM because there is NEVER any traffic at 4AM I commit the fallacy because my ideal is not really real," has a crucial ambiguity. The "not really real" part can mean one of two things:
1. You could mean that something is not really real if it is empirically wrong. So if you actually don't hit any traffic at 4am (which I am sure happens on some rural road somewhere), then in this view you would say that the comparison is fully justified.
2. Quite differently, there is a sense that saying you will never hit traffic at 4am is a conceptual mistake, even if it turns out that you actually don't hit any traffic. This happens if you make the statement *without considering* whether you will actually hit traffic. The assumption that you make that you won't hit traffic is "not real" in the sense that it is posited as a matter of theory. Even if it turns out that you were right--by sheer happenstance and luck--there is still a conceptual mistake, albeit harmless in such a case. And it is this conceptual mistake that I think the Nirvana fallacy pertains to.
The not-thinking-about-it phenomenon is the kind of thing that Demsetz was encountering when he coined the phrase. At the time (and even now), you'll see someone show such-and-such problem and then propose that government (e.g. bureaucrats or judges) should solve the problem by, say, regulating the industry and then engaging in cost-benefit balancing. But the proponent will never explain how the actual incentives of government bureaucrats and the information problems of the process would not result in a huge public choice nightmare. And it is surely not that these proponents of regulation actually believe that government has great information or perfect bureaucrats. It is that they never stopped to think about those problems, because they implicitly had all-knowing judges and perfectly-motivated bureaucrats in their mind. Getting people to stop to think about those problems was the practical contribution of the Nirvana fallacy label.
So if you really stopped to think about whether you will hit traffic at 4am and then concluded that you wouldn't hit any traffic at all, then I would say that you are likely to be empirically wrong, but you are not committing the Nirvana fallacy. If you didn't stop to think about it and just assumed it, then I would say you are committing a conceptual mistake even if turns out you didn't hit any traffic at all.
Posted by: TJ | Aug 31, 2012 9:58:34 PM