book: Thinking Fast and Slow, review part 2

…and this is Part 2, in which I have to bring up a serious problem with this otherwise highly worthy book.

Let’s dive right in, but first let’s agree to be fair about this. Very few have accepted the daunting challenge of sorting out people’s various stupidities, or merely vagaries, or even their minor idiosyncrasies. Even when limited to the supposedly well-understood arena of economic behavior, ten career-long attempts fail for every career-long success. Let us now posit that Kahneman is solidly in the success category, that this book is a highly successful survey of his life’s work (so far!), and so let us accept one mistake, even a major one, in the midst of 400 pages of wisdom.

But it’s still a mistake, slander even. In the middle third of this book, Kahneman seems perpetually confused between description and prescription. That is: was Bernoulli’s logarithmic scale of utility (vs wealth) intended to say how people do behave, or how people should behave? And in either case: so what if Bernoulli didn’t get it exactly right? Bernoulli’s idea was that a person values each percent increase in wealth about equally, and this deceptively simple arithmetic leads directly to three profound and absolutely correct human consequences: (1) the rich care about any small unit of wealth (say, $1) less then do the poor; (2) people tend to be risk-averse rather than risk-seeking (since all logarithmic plots of utility vs wealth curve downward); and (3) all people’s risk aversion will get much, much stronger as the size of any gamble becomes a sizeable fraction of their accessible wealth. This is exactly how people behave, and Bernoulli pointed it out well ahead of his time. And to this point Kahneman agrees.

Yes, you can construct cases where logarithmic utility does not apply, but however interesting, most are artificial–for example the guy who desperately gambles his $100 stake all night because his homicidal loan shark demands $200 in the morning.

  • As things are, he will be extremely risk-seeking than logarithmic because $199 doesn’t do him any more good than $1. He’ll gamble like crazy.
  • If he does win $200 or a little more, then he suddenly goes extremely risk-averse–his life depends on keeping the $200. He won’t just stop gambling–he’ll hide in a dark corner until morning.
  • If he should hit the jackpot for say $1,000,000, he’ll go normal: a normal utility curve with normal risk-aversion. He’s past the loan-shark problem and will adjust to the rest of his life after it. Now in reality, such a person may start gambling again, but that would be against Bernoulli’s utility curve, that is…against his own interests. But some people do (and neither Bernoulli nor Kahneman covers these cases).

So over the past 50 years, Kahneman’s and others’ experiments have shown that the logarithmic curve doesn’t describe people’s actual behavior to the left (money-losing) side of the utility vs wealth curve. He calls this Prospect Theory, and if you want to know how normal people behave, his approximation to double the (logarithmic) utility lost is a better approximation than Bernoulli’s logarithmic description. This is an advance, it largely won Kahneman the Nobel Prize, and deservedly so.

But this does not give Kahneman the right to call Bernoulli wrong, or Bernoulli’s logarithmic utility curve a Mistake. Kahneman gratuitously started this fight, but let’s end it.

  • First of all: Bernoulli’s advance was momentous; Kahneman just tweaked the math. So there.
  • Second: for Kahneman to call Bernoulli’s advance a Mistake, Bernoullli would have to have claimed that the logarithmic curve is always correct, that it always describes everyone’s real behavior, and that it could never allow for tweaks like Kahneman’s. But Bernoulli didn’t say that, and Kahneman surely knows that.
  • Third: Even if Bernoulli never explicitly called his logarithmic curve an approximation, even if he never imagined that someone like Kahneman might someday improve on it, that hardly makes his work a mistake! Do we jeer at Newton just because Relativity came along? Do we call Einstein’s Special Relativity a Mistake because Einstein’s own General Relativity, um, generalized on it? Is the idea “DNA carries the genetic code” a Mistake because epigenetics improves on it? Everyone: can’t we please just celebrate all these brilliant advances from intellectuals on whose shoulders we stand, rather than slander their life’s work for book sales?
  • Finally, and for those who would apply Kahneman’s work to make money, by far the most important: Did Bernoulli intend his logarithmic utility curve to be descriptive of real people or prescriptive of optimal economic behavior (absent loan sharks and the like). Kahneman waffles on this over and over, but the distinction is absolutely crucial. Here’s why: Let’s say you follow Kahneman’s double-the-risk-aversion utility curve, which means simply that you will take gambles that maximize Kahneman’s utility curve. And now by contrast, I follow Bernoulli’s curve, taking gambles that maximize the expected logarithm of my wealth. We are presented with a series of gambles with variable odds payoff. And let’s say that you and I can even choose the size of each gamble defined as the maximum wealth we offer to risk. After a long series of these unforced gambles, here’s what happens: I rip your face off. Your artificially timid utility curve, grace à Kahneman, has put you at a slow, grinding disadvantage. Consider amateurs’ record in stock and commodities trading: QED.

Kahneman is brilliant, certainly, this is too much–Bernoulli is not here to defend himself. Someday imperfect Kahneman won’t be here either, and it will be only fair if a 2046 Nobel Laureate in Economics writes a chapter titled “Kahneman’s Mistake”.

And that’s enough about that. This is a great book, marred only by hubris in its middle chapters. In the later chapters I detect a too-hurried brain dump of useful ideas–especially the deadly and somehow invisible error of “Sunk Costs” which seems to reappear in the world like avian flu. If he hadn’t taken such long time (377 pages!) to distinguish between emotional utility and rational utility, indeed if this distinction had been clearer in his own mind as he wrote, he would had a clearer book (and not made so many mistakes himself). And he never quite gets to the greatest importance of his own System 1 vs System 2 distinction, namely: it’s your job, everyone’s job, to develop your thoughtful System 2 to keep your fast but very error-prone System 1 in check.

It’s called intelligence, and Kahneman’s right, in the short term it almost always amounts knowing that your first thought is not usually your best one. That is, always try to hold back your System 1 until System 2 has had a chance to censor it. Yes, that’s it. If you practice it, you will absolutely get much smarter (and far less neurotic in the bargain). In the long term, relentlessly exercise your slow-thinking System 2 to get faster and sharper. Never trust your reflexes (System 1) unless that’s all you’ve got–which would be almost never if you had used your System 2 in advance.

Always give your System 2 time to act. The harder it feels to do, the more important to do it. And until System 2 has responded: keep your mouth shut and your finger off that trigger.

“Thinking Fast and Slow”, Daniel Kahneman, 2011, Farrar, Straus, and Giroux.

Comments are closed.