There have been a lot of such manifestos, so it will be pretty hard to isolate the effects of any particular one.
Such manifestos can be divided into two classes: those which set p-values in opposition to confidence intervals and those that describe them as the same. The first class has a positive suggestion: use confidence intervals. It seems to me that positive manifestos are more likely to be adopted than those which merely condemn p-values, but offer no alternative. This manifesto seems to condemn confidence intervals, but makes the very vague suggestion to pay attention to effect sizes. In the end, maybe that amounts to the same thing.
I reject your examples. Sex and drugs are almost always hypocrisy, not one community trying to impose its standard on another.
Do we want a War on Hypocrisy?
There are lots of examples where the optimal state has some kind of consistency property (eg, lack of hypocrisy). It’s probably always possible to use the failure of consistency of the current state to improve, but I think that there are lots of examples where naively trying to improve consistency makes things worse, not better.
I argue in the last piece that it is common even now for people to engineer blackmail material against others and often also against themselves, to allow it to be used as collateral and leverage. That a large part of job interviews is proving that you are vulnerable in these ways.
I don’t see anything about existing practices for job interviews in the previous piece.
This is all so abstract. Does it match how the world works?
Why are we talking about this? Because the National Enquirer knows how to get away with blackmail? What do you think of the hypothesis that blackmail is legal if lawyers get a cut? That sounds to me like the worst of both worlds.
I recently learned that UK politics is openly described in terms of blackmail. There is some equivocation and I’m not sure it’s actually true, but it isn’t a norm violation.
Usually this comes up in the context of financing education, where they are sometimes called human capital contracts. This was previously discussed on LW. I left a comment there giving several of education (England, France, Yale) and another example more like this post: “360 deals” in the music industry, where instead of touring existing to promote the album or vice versa, the record label owns a proportion of everything musician does, so that their interests are more aligned. I think that the most famous example is Lady Gaga, who signed such a deal fairly early in her career. (The wikipedia article I link gives as examples even more famous musicians Madonna and Jay-Z, but those were late in their careers and, I guess, maybe in the opposite direction, where the tour promoter buys a chunk of the record.)
Yes, you should draw calibration curves without binning.
The calibration curve lives in a plane where the x-axis is the probability of the prediction and the y-axis is the actual proportion of outcomes. We can make each prediction live in this plane. A scored prediction is a pair of an outcome, b, either 0 or 1, and the earlier prediction p. The value p belongs on the x-axis. The value b is sort of like a value on the y-axis. Thus (p,b) makes sense on the x-y-plane. It is valuable to plot the scatterplot of this representation of the predictions. The calibration curve is a curve attempting to approximate this scatterplot. A technique for turning a scatter plot into the graph of a function is called a smoother. Every smoother yields a different notion of calibration curve. The most popular general-purpose smoother is loess and it is also the most popular smoother for the specific task of calibration curves without bins. Frank Harrell (2-28) suggests tweaking the algorithm, setting α=1.
I’ve asserted a lot of things. Should you believe me? You don’t know that I’m disinterested (though you should damn well know that 300 pages of advertising copy is very, very interested). But I can provide a different perspective; you can’t imagine that anything good is going on, but I can try to widen your imagination.
More broadly and aside from allegations of specific wrongdoing, the claim is that HFT is just shaving the margins of everyone who makes trades more slowly.
The main difference in perspective I want to promote is how you carve up the market. You’re carving it into HFT and Everyone. I say that you should carve it into market makers and investors. HFT makes money. It makes money by shaving the margins off of someone. But why think that it’s shaving “everyone”? It’s competing with market makers and shaving their margins. The market makers are mad about that. That’s enough to explain everything that is observed. Maybe something bad is going on beyond that, but no one says what, no one except liars.
I could say more, but I think it would distract from that one point. (Indeed, I think my prior comments made that mistake.)
I haven’t read it, but Vanguard wrote about how it loves HFT. A lot of what I know is from Matt Levine, but he is such a fragmentary blogger that he probably doesn’t say much in one place and it’s hard to find any particular thing he’s written.
Your abstractions like “benefit” seem confused to me. Where is money flowing? How?
By the biggest players, I mean investment firms. I thought that the biggest investors are bigger than the exchanges, but maybe they’re only equally big. For example, BATS and Fidelity both have a market cap of $60 billion (NASDAQ $14B, NYSE $6B).* HFT firm Virtu has a market cap of $5B. That is, the net present value of the profits Fidelity extracts from its retirement accounts is $60B, while the NPV of the profits BATS extracts from all retirement accounts in the US is about the same. BATS allows (and subsidizes) HFT because it thinks that Fidelity wants it. That’s what BATS says and that’s what Fidelity says. Maybe they’re lying and actually BATS has market power to extract money from Fidelity, but I’ll get to that later. [And why would Fidelity go along with such a lie?] [* looking up these numbers again, I find completely different numbers. But they’re roughly right if we swap NYSE and BATS.]
Transaction costs are way down. This is easily and objectively measured in terms of bid-ask spreads and trading fees. This is money that is not going to the middlemen, neither exchanges nor market makers, but is saved by the investment firms, which is why they love HFT. There is a more subtle argument that market liquidity is an illusion that will go away “when it matters” and produce flash crashes. I think that this is also false, both painting too rosy a view of the past and exaggerating the damage caused by flash crashes, but it is much harder to argue about rare events.
HFT and running exchanges are not terribly lucrative businesses, not by the standards of Wall Street. HFT makes orders of magnitude less money than market makers made (in aggregate) even 20 years ago. Individual HFT firms make a lot of money, but there used to be a huge number of market makers who specialized in very small numbers of stocks. When HFT first appeared and drove out these people, they reduced the aggregate money going to market makers and the small number of HFT firms have continued to compete away their own profit. This is a pretty simple metric that is exactly opposed to many common stories. It’s not that simple because many of the market makers were vertically integrated into firms that did other things, so it’s not that easy to aggregate the market makers. In particular, I claim that’s what Brad K’s old job was (understanding market structure and making money off of the difference between markets, allowing the rest of the firm to think in terms of stocks, not markets). If you buy that, it makes the old market makers look even more bloated and thus the new HFT look even more efficient, but I don’t claim that it is obvious.
But are you saying that Lewis is saying that the exchanges are sucking all the profits out of the HFT? I don’t think that exchanges are very lucrative (see numbers in beginning). Does he give any numbers? I think that there are only about 4 companies running exchanges in the US (including IEX), which doesn’t sound very competitive. But that’s because they keep buying out new exchanges, so it can’t be that hard to enter the market. And they keep the exchanges around, so they do see value in diversity. IEX being the 12th exchange and the 4th company does make the market more diverse and competitive, but they’re probably only fill a small niche. Whereas the dozens of dark pools are already very competitive.
Everything I’ve ever heard attributed to Michael Lewis on this topic was false. Good ideas shouldn’t require lies to sell them. And I can tell it’s false because I carefully read the quoted passages. In particular, you conclude from his claim that “the richest people on Wall Street” are angry that they employ HFT algorithms. But that’s not at all true. HFT is tiny. If the biggest players on Wall Street are angry, it’s because they’d rather trade with HFT than with Brad Katsuyama.*
Is IEX listing a stock a significant milestone? I doubt it. IEX has been selling other stocks for 4 years, first as a dark pool and then for 2 years as an exchange. (But aren’t “dark pools” evil incarnate?) IEX claims to be 2.7% of the total volume. I never looked at that number before today. I’ve previously claimed that IEX was a failure, and I was surprised to see that the number was so high. I welcome experimentation and I’m happy that they’ve found a niche. But if you thought it was a much better product that would quickly win in the marketplace, maybe you should reconsider this, 2-4 years on.** But the future is long. Maybe IEX listing individual stocks will matter, though you should be suspicious if no one can explain why. And it’s hard to rule out the possibility that’s it’s a much better product that will take a long time to win.
* It was probably a bad idea to use Brad K as metonymy, because he plays two roles. I meant the kind of trader he was at the beginning of the book, who was outcompeted by HFT. I don’t mean IEX, the market he now runs. In as much as IEX exists as a place to trade with people like him, it seems like most people wouldn’t want to trade there, either, but it has a lot of room to evolve.
** To put the 2.7% in context, there are 12 exchanges, so I think IEX is the smallest, but I don’t know. There are dozens of dark pools, so the 1.5+% market share IEX had when it transitioned to exchange was pretty big for a dark pool.
Is there no middle ground? You say that Kegan paints it as a binary (“nothing inside the range of what we think of as a normal workplace”). But you suggest that kaizen is an intermediate. Your summary as two phrases suggests that they are separable (“everyone talks about mistakes and improvements, and where the personal/professional boundaries are broken down”).
Also, the negative book is about how things actually work, while the positive book is about the system working as promised. But this could be cherry-picking successes. Is there any reason to believe that DDO is self-correcting? Why shouldn’t we expect the worst of both worlds, implementations with the face of a DDO that actually work as describe in Moral Mazes? (which is probably what people insinuate with the word “cult”) Does the book make an argument, or does it just profile success stories?
Added: Indeed, “The Western Elite from a Chinese Perspective” talks about the discussion of emotion at business school and implies that people are faking it.
To add a little on different terminologies being a feature: if you ultimately want to apply linear algebra, you’ll have to build a bridge from the theory to the particular application that is probably even more difficult than the bridge between two presentations of the theory. So it’s probably good to practice building bridges.
(I’m also suspicious of people who say that the last book that they read on a subject was the best book, because that’s when it clicked. How much did the first books prepare them?)
I don’t know. My claim was based on reasoning from first principles. It was intended as an illustrative example that there could be positive externalities, not to measure them. If you have to triage nets, it’s probably the way to go, but if you’re triaging nets, you’ve probably made a bad decision. I can think of so many reasons to concentrate nets in one village, rather than spreading them out and micro-managing the deployments in the villages. One reason is habit formation. Another is the cost of distribution, which is probably low for marginal nets and high for a new village. A third is that there positive externalities compound, at least if you cross over the threshold of locally wiping out malaria. (Under that threshold, I’m not sure.)
Parasites in general and malaria in particular are pretty specific. For example, humans developed immunity shortly after speciation from chimps and malaria only jumped back 30kya (but probably did so multiple times to produce the several species of malaria). It’s pretty clear that it doesn’t have other hosts in the New World because the strategy of treating all humans in an area for 3 weeks wipes it out. But it’s hard to rule out the possibility that it has other hosts in Africa.
Ewald has written lots of great papers. Here (ungated) is a paper summarizing his career. Mostly it’s about explaining the past, but he goes on to say that we should design interventions to shape the evolution of infectious agents. His main claim about the past is that malaria is debilitating because it can be passed on from someone who can’t move. Thus if we keep the mosquitoes out of beds or out of homes, then malaria will evolve to be less debilitating. But I’m not sure where he says this. Scientific American? TED?
Or you could just look at the weather report, now that you know what to look for.
Between rational and irrational is PD: individually rational, collectively irrational. Lanrian gave two reasons that nets benefit people other than they person paying. One is that they are most valuable for children. The other is that they protect people not using the nets. Probably the most valuable nets are those deployed on people who already have malaria, to prevent it from spreading to mosquitoes, and thus to more people. (See also Paul Ewald.)
If you regularly get sunburn, you should use sunscreen, because sunburn is unpleasant. Death isn’t the only thing that matters. If you haven’t noticed that you get sunburn or haven’t thought about the possibility of using sunscreen or making simple interventions, like buying sunscreen, or putting it near sports gear as a reminder, there are great opportunities for improvement.
Whether to use sunscreen in situations where you won’t get sunburn is controversial. But these situations are intermediate and it’s probably less important to get them right than the extremes.
Added: maybe this sounds like trivial advice, but it’s important to figure out what the typical advice actually means in terms of who it’s aimed at (confer).
SSC’s argument that the dermatologists’ factual claims are wrong are the least of the problems. Even if the dermatologists really are experts at skin cancer, they aren’t expert at the trade-offs.
On the other hand, if SSC is correct, that only eliminates one option that shminux gave. It doesn’t necessarily reject the claim that you should stay out of certain sun conditions.
By “really ancient” you mean bronze age, right?
Classical antiquity definitely has plagues in the modern sense, like the Antonine plague. Indeed, in your paper you endorse the fairly standard claim that it was smallpox. That seems to me worth mentioning here, more than the negative claim about Hippocrates.
Yes, if the game has many opportunities for betting, you should focus on the instrumental use of the money, which is via compounding, thus the instrumental value is geometric, and so you should use the Kelly criterion. In particular, if your edge is small (but can be repeated), the only way you can make a lot of money is by compounding, so you should use the Kelly criterion.