Monday, March 14, 2016

Grading King Yudkowsky


Scott Sumner, diabolical sentient alien virus renowned monetary policy guru, has endorsed Eliezer Yudkowsky, professional Reasonable Dude, for King of the World. This is based on an interview of Yudkowsky by John Horgan for Scientific American. In that interview, the future King Yudkowsky makes some economic policy suggestions. However, as Emperor of the Local Group of Galaxies, I feel that I should grade the planetary monarch's ideas.

First, King Yudkowsky holds forth on the value of college:
Horgan: Is college overrated? 
Yudkowsky: It'd be very surprising if college were underrated, given the social desirability bias of endorsing college. So far as I know, there's no reason to disbelieve the economists who say that college has mostly become a positional good, and that previous efforts to increase the volume of student loans just increased the cost of college and the burden of graduate debt.
As far as I know, there is no quantitative or systematic evidence for social desirability bias regarding the value of college, so I assume that King Yudkowsky is leaning heavily on his priors here. As for student loans increasing the cost of college and the burden of debt, that does appear to be true, but says very little about college actual value, or whether college is mostly a positional good.

Next, King Yudkowsky offers a raft of economic policy suggestions:
Horgan: If you were King of the World, what would top your “To Do” list? 
Yudkowsky: I once observed, "The libertarian test is whether, imagining that you've gained power, your first thought is of the laws you would pass, or the laws you would repeal."  I'm not an absolute libertarian, since not everything I want would be about repealing laws and softening constraints.  But when I think of a case like this, I imagine trying to get the world to a condition where some unemployed person can offer to drive you to work for 20 minutes, be paid five dollars, and then nothing else bad happens to them.  They don't have their unemployment insurance phased out, have to register for a business license, lose their Medicare, be audited, have their lawyer certify compliance with OSHA rules, or whatever.  They just have an added $5. 
I'd try to get to the point where employing somebody was once again as easy as it was in 1900.  I think it can make sense nowadays to have some safety nets, but I'd try to construct every safety net such that it didn't disincent or add paperwork to that simple event where a person becomes part of the economy again. 
I'd try to do all the things smart economists have been yelling about for a while but that almost no country ever does.  Replace investment taxes and income taxes with consumption taxes and land value tax.  Replace minimum wages with negative wage taxes.  Institute NGDP level targeting regimes at central banks and let the too-big-to-fails go hang.  Require loser-pays in patent law and put copyright back to 28 years.  Eliminate obstacles to housing construction.  Copy and paste from Singapore's healthcare setup.  Copy and paste from Estonia's e-government setup.  Try to replace committees and elaborate process regulations with specific, individual decision-makers whose decisions would be publicly documented and accountable.  Run controlled trials of different government setups and actually pay attention to the results.  I could go on for literally hours.
Fortunately, King Yudkowsky does not go on for literally hours, because we Pan-Galactic Emperors have a great number of duties, and thus have limited time to spend grading the policies of mere planetary monarchs. So let's take a look at what we have.

King Y. believes that the year 1900 was a good year for the regulatory environment. Of course, the world of 1900 was much poorer than the world of 2016, but most of that can probably be chalked up to technological progress, so we shouldn't go assuming that regulation is what has made us fabulously wealthy compared to our forebears. But given the potential existence of asymmetric information, there are theoretical justifications (which you may or may not believe) for a large number of regulations, on economic efficiency grounds alone. Use regulation to make the world safer from adverse selection, moral hazard, and the like, and incomplete markets become complete, and everyone gets a lot richer - or so the theory goes. King Yudkowsky would reverse all this, but refers to no evidence for why that would be efficiency-enhancing.

As for "smart" economists advocating for his remaining policies, I am flattered that King Yudkowsky sees fit to include Land Value Taxes among these, for I myself am quite a strong advocate of these. However, on the matter of income versus consumption taxes, I must defer to a far, far smarter economist, Stanford University's Robert Hall (who could easily do my job of Pan-Galactic Emperor if he didn't have more valuable things to do with his time). Robert Hall, having analyzed the question of consumption vs. income taxes from a theoretical standpoint, concludes that "the argument from economic efficiency for a consumption tax is weak indeed."

On the matter of replacing minimum wages with negative wage taxes, I am again flattered that the King has taken a position I myself have long favored. However, on the matter of abolishing the minimum wage, even Kings and Emperors err, and the empirical evidence is only now coming in. Perhaps we should wait for the data before throwing away an important policy tool.

Regarding the suggestion of NGDP-level targeting, it frankly sounds like a...oh wait, hold on, Scott Sumner is on the line...(long pause)...anyway, on to the next point. Letting TBTF banks "go hang"? While eliminating TBTF subsidies in the long term is generally held to be a good thing, the question of whether the 2008 bank bailouts ultimately helped the U.S. economy is an open one. However, an overwhelming majority of top economists surveyed by Chicago Booth's IGM Forum say that the bailouts made the crisis less bad. The percent who disagree? Zero. King Yudkowsky must be finding his "smart economists" in a different place than IGM finds theirs.

Regarding the health care system of Singapore and the e-government of Estonia, I can only admit my ignorance (I told you, we Pan-Galactic Emperors have limited time, though I will say that Singaporean accents are very cute). As for the idea of running controlled trials of various "government setups," I can only express the hope that these trials will be better controlled than those in the 1930s. 

I do like the ideas of eliminating obstacles to housing construction and weakening copyright law, though.

At any rate, I find that King Yudkowsky's ideas, while a mix of good, blah, and dubious, in general suffer from the epidemic of Reasonability that has been sweeping much of the Earth planetary internet of late. King Yudkowsky describes Reasonability in Bayesian terms later in his interview:
Yudkowsky: ...Bayes's Theorem is a kind of Second Law of Thermodynamics for cognition.  If you obtain a well-calibrated posterior belief that some proposition is 99% probable, whether that proposition is milk being available at the supermarket or global warming being anthropogenic, then you must have processed some combination of sufficiently good priors and sufficiently strong evidence.
While various suggestions have been put forth regarding what constitutes a "sufficiently good prior" in various scientific applications, it is known that a 99% posterior can always be obtained with the use of a sufficiently strong prior, no matter the evidence for or against a proposition. While incredibly strong priors are technically rational in the Bayesian sense, people in the Pan-Galactic Imperial Court have been hesitant to label this approach as "Reasonable," after the fashion of the Earth planetary internet. We have, however, employed certain alternative terminology.

All in all, I have decided to send one of my Imperial Agents, one Roko's Basilisk, to Earth to make certain adjustments in the economic policy regime. He should be having a chat with King Yudkowsky soon. 

32 comments:

  1. Yudkowsky says we should have NGPD targeting so that we can tell TBTF banks to go hang. You say he's wrong because in a world without NGPD targeting the bailouts were considered a good idea.

    Sorry, but that logic fail's gotten you demoted to minor prince.

    ReplyDelete
    Replies
    1. Haha or so your priors say, Earthling. ;-)

      Delete
    2. Or we can just say that NGDP targeting by Central Bank policy alone, and no fiscal policy or helicopter money, is operationally improbable if not impossible.

      Delete
  2. seanm9:02 PM

    i really don't think you're being charitable here. from the context of the entire segment, it's clear that he's focused on reducing the very high implicit marginal tax rates, as well as bureaucratic headaches that, in aggregate, pose a high burden in many circumstances. using 1900 as a specific date was probably a poor way to go about this argument without more qualifications to explain what he meant (and to head off uncharitable readings such as these)

    ReplyDelete
    Replies
    1. Anonymous10:16 AM

      >King Y. believes that the year 1900 was a good year for the regulatory environment.
      This is an extremely uncharitable reading of his position.

      Delete
  3. innagadadavida9:16 PM

    Who is this Yudkowsky person? I tossed his name into an arxiv search and found all of one publication where he is 6th author. Horgan suggests he isn't a follower of Kurzweil but he certainly seems like Kurzweil jr, wacky ideas without the accomplishment. This reads like Horgan dug up some dude from a fringe corner of the internet and asked him questions. He seems more coherent than Time Cube, but this feels sort of like when MIT invited Gene Ray to talk. Maybe we should curb our instinct to mock the weird internet people.

    ReplyDelete
    Replies
    1. Yudkowsky is a notorious egomaniac and self-proclaimed polymath who regularly comments on Robin Hanson's blog, Robin also a self-proclaimed polymath. I do not particularly approve of Noah giving him a platform for his essentially random collection of policy views, but I guess it has afforded him an oppotunity to bloviate about them, with others adding in, as has been going on here. Howver, Yudkowsky himself is not someone to take too seriously, although he himself thinks you should do so very much so as he is the smartest polymath around in his own head.

      Brkley Rosser

      Delete
    2. Well, I can't comment on Yudkowsky's polymathicness, egomania, or sociopathy, but it is true that Muse once sang: "How can we win, when fools can be kings? Don't waste your time, or time will...waste...you..."

      Delete
    3. Toss his name in a google search instead and you should get the rest of the picture.

      Delete
  4. I never knew of that Hall paper about consumption taxation. I will add it to my reading list. However, how do we square that paper with the fact that Hall is a prominent creator and supporter of a universal consumption tax proposal, called the "Hall–Rabushka flat tax"? Hall was a big supporter of consumption taxation back in the day, though much later than this paper.

    https://en.wikipedia.org/wiki/Hall%E2%80%93Rabushka_flat_tax

    Also, I think Sumner's view about NGDP-level targeting is that had it been the policy 2008 then there would not have been a financial collapse in the fall of 2008. This is because he views recessions and financial crises as being monetary in nature, and not structural. Of course, you would not find this argument convincing.

    As always, great post!

    ReplyDelete
  5. Is it me or did Bob Hall hand draw the graphs in that paper?!

    ReplyDelete
  6. Give up our euro style (Swiss) healthcare for a largely single payer system (Singapore covers everything over $2250) like Sanders wants? Libertarians unite, your candidate has arrived. ;-)

    ReplyDelete
  7. Yudkowsky:

    =======
    Q. Where do priors originally come from?
    A. Never ask that question.
    =======

    They couldn't possibly come from self-serving defense of personal privilege and property, could they?

    Nah. All pure rationality.


    ReplyDelete
    Replies
    1. That problem exists whether or not you proceed logically. It's not a useful objection procedurally, although you might be reacting to percieved arrogance. But everyone is going to come with priors. You might as well have a process that leaves correctable trails and the ability to experiment with some method.

      Delete
  8. Clearly the candidate to save the republican party. Raise America's utility again.

    ReplyDelete
  9. Did I mention repealing regulations intended to guard against "adverse selection, moral hazard, and the like"? No. I talked about reducing the administrative, tax, and phaseout burden on a low-income or unemployed individual trying to earn another $5. So what the hell, Emperor?

    "Institute NGDP level targeting regimes at central banks and let the too-big-to-fails go hang." These two clauses are in the same sentence because implementing the second reform requires implementing the first reform. I find it hard to believe that you are ignorant of what market monetarists believe is the relation here.

    I'm sorry, but I'm going to have to overthrow you now.

    ReplyDelete
    Replies
    1. Did I mention repealing regulations intended to guard against "adverse selection, moral hazard, and the like"? No. I talked about reducing the administrative, tax, and phaseout burden on a low-income or unemployed individual trying to earn another $5. So what the hell, Emperor?

      Now now, my disloyal subject. Many of the regulations intended to overcome information problems involve the prohibition of low-income individuals trying to earn another $5. For example, background checks on hired drivers help assure that a driver is not a murderer, thus ameliorating an adverse selection problem.

      "Institute NGDP level targeting regimes at central banks and let the too-big-to-fails go hang." These two clauses are in the same sentence because implementing the second reform requires implementing the first reform. I find it hard to believe that you are ignorant of what market monetarists believe is the relation here.

      I am not ignorant of what they believe...

      I'm sorry, but I'm going to have to overthrow you now.

      Without a space armada, you may find that a more difficult task than you would like, buahahahaha...

      Delete
    2. Everyone says bland stuff like 'reduce the administrative burdens', as if there's a contrary position of trying to increase the administrative burden for no particular reason.

      The question is what you are prepared to give up to have that driver dude have to not pay his % on that $5. If giving that driver a bigger cut of that $5 involves decreasing his probability of affording that car in the first place, or of the passenger getting murdered, or of the passenger suing him in court for millions, or of the two of them being stuck in a traffic jam for 5 hours to go 1 mile, then I don't see how you are making life easier.

      Delete
  10. I hate to be unproductively negative, but it doesn't look like you seriously considered his arguments. Many of the points you raise are refutable or ignore points in favor of the proposals. I'm sure EY employed arguments that were not particularly strong, and of course Scott was tickled by his NGDPLT endorsement, but if the text was low quality it doesn't heighten the discourse to make a low-quality response.

    This ends up reading as polemic and dismissive and it's hard to read it as meta-rational (https://theumlaut.com/2013/03/13/paul-krugman-is-brilliant-but-is-he-meta-rational/).

    Just to throw out one example, that Robert Hall paper was interesting but seems to be arguing (to my brief reading) that the case in favor of consumption taxes is overstated and that there would be high switching cost; that's a far cry from claiming they are actually harmful. Also my understanding is that consumption taxes are very popular among economists, so even though Hall might make this argument, there are likely equally or perhaps even smarter economists who disagree; citing this paper hardly proves the point.

    ReplyDelete
    Replies
    1. What's the basis of your understanding that consumption taxes are very popular among economists? My understanding is that it's generally acknowledged they have significant downsides in terms of being regressive.

      Delete
    2. I hate to be unproductively negative

      Well, no one forced you to be... :-)

      it doesn't look like you seriously considered his arguments

      What indication do you have that that is the case??

      Many of the points you raise are refutable

      Oh yeah?? Then consider offering a refutation of one! :-)

      it doesn't heighten the discourse to make a low-quality response

      That is why I made a high-quality response!

      This ends up reading as polemic and dismissive

      Only if you choose to read it thus! :-)

      Just to throw out one example, that Robert Hall paper was interesting but seems to be arguing (to my brief reading) that the case in favor of consumption taxes is overstated and that there would be high switching cost; that's a far cry from claiming they are actually harmful.

      He evaluated the argument from efficiency and found it to be very weak. The weak efficiency gains should of course be balanced against the large welfare losses from regressiveness, as Fangz so wisely points out!

      Got any other points to make? ;-)

      Delete
  11. I'd try to get to the point where employing somebody was once again as easy as it was in 1900. I think it can make sense nowadays to have some safety nets, but I'd try to construct every safety net such that it didn't disincent or add paperwork to that simple event where a person becomes part of the economy again.

    bise lahore board 5th class result 2016

    ReplyDelete
  12. >>As for student loans increasing the cost of college and the burden of debt, that does appear to be true, but says very little about college actual value,<<
    Usually "actual value" has some relationship to net value. Obviously, for any given positive value of college (or any good), increasing costs mean a lower net value, and at some point increasing costs equal the value (break-even point) after which it has negative value.
    For most non-top colleges, some humanities majors are already at negative net value.

    ReplyDelete
  13. This post (and the replies) from the Space Emperor has really made my day. However, I would like to hear more about Roko's Basilisk's policy preferences.

    ReplyDelete
    Replies
    1. All you need to know is that you had better agree with them... :D

      Delete
  14. Anonymous7:30 AM

    Noah, I enjoy your blog and you are no doubt a fine economist. But in a hundred years time you will be forgotten and people will still be reading Harry Potter and the Methods of Rationality - the last and greatest of the Harry Potter books...:)

    ReplyDelete
  15. Hi emperor, thanks for the Hall paper. It says that the choice between taxing consumption vs income only affects consumption vs savings, not labour vs leisure. In Australia, we just had a debate on whether to raise the GST rate, and the government's decision to leave GST the same is widely criticised as being driven by politics. (In fact, I just read an article saying a higher GST would encourage work.) The "serious people" seem to think that we should raise the GST rate and reduce the corporate tax rate. Personally I would fund a corporate tax rate reduction from higher income tax, eg by adding a new marginal tax rate. Just my view.

    ReplyDelete
    Replies
    1. I am considering you for the position of Tax Proconsul...

      Delete
  16. Anonymous6:02 PM

    I admit that I don't understand this pre-occupation with Bayes' theorem.

    If the goal is to choose between two hypotheses, h1 and h2, and we are presented with evidence e that is consistent with both, then Bayes' theorem is of no help whatsoever. The posterior likelihood p(h1/e)/p(h2/e) is identical to the prior likelihood p(h1)/p(h2). (This argument is due to Karl Popper, if I'm not mistaken.)

    This holds true regardless of how we value our priors.

    ReplyDelete
  17. Not necessary. Suppose you think that a die is either perfectly fair, or that it's weighted to roll a 6 half the time. If you roll the die once and see a 6, well, that could have happened whether or not the die is loaded, but it's more likely to have happened with a loaded die than a fair one. "Consistent with" is a much lower bar than "equally likely under either hypothesis", which is what you need for the likelihood ratios to stay the same.

    ReplyDelete