tag:blogger.com,1999:blog-17232051.post8280140972505493940..comments2024-03-18T22:32:52.802-04:00Comments on Noahpinion: Economists who don't do it with modelsNoah Smithhttp://www.blogger.com/profile/09093917601641588575noreply@blogger.comBlogger47125tag:blogger.com,1999:blog-17232051.post-55188167643355511772012-06-11T13:00:04.860-04:002012-06-11T13:00:04.860-04:00Sumner:
I think at this late date most macroecono...Sumner:<br /><br /><i>I think at this late date most macroeconomists have basically made up their minds about the "do nominal shocks have real effects" debate, which has been going on for 100s of years. We've gone over the data exhaustively. Most have concluded nominal shocks do have real effect. I'm speaking to that group.</i><br /><br />If one were to conclude that money does affect the real economy, and no economist should deny this, it doesn't imply that central bankers can know what the "correct" supply of money or the "correct" volume of spending should be. The fact that money and spending is controlled by the state, isn't sufficient evidence that they can be expected to know such things through being "advised" by like-minded central planners.<br /><br />You are ignoring, as usual, how central bank inflation, commercial bank credit expansion, and the resulting change in interest rates <i>themselves</i> affect the real economy, in such a way as to make spending and interest rates lose their information content, thus leading investors astray (see Goodhart's Law).<br /><br /><i>I also think that most famous macroeconomists have a "I know it when I see it" attitude toward demand shocks. Thus most look at cases like Britain today, or America in 2011, and think it fairly obviously that a bit more AD would be helpful.</i><br /><br />These same "famous" macro-economists had no clue the housing market was about to collapse, and those same economists call for more inflation the same way children call for more food and water. It's never enough. Why? Because inflation is what is preventing the real economy from correcting, so it's macro-economists chasing their own tails. They call for more inflation, that prevents corrections of capital misallocation, which hampers economic recovery, which is responded to with "more inflation."Major_Freedomnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-42075102481366334342012-06-11T12:50:10.342-04:002012-06-11T12:50:10.342-04:00Now this may be going a bit far afield, but what a...<i>Now this may be going a bit far afield, but what about the Lucas Critique? If NGDP hasn't been targeted, how do we know that the gut-instinct relationship between NGDP and AD will hold if we try to use NGDP as the lever to control the economy?</i><br /><br />I've asked this many times, but I am seemingly not worthy enough of an answer (but I am worthy enough for contempt and scorn apparently). <br /><br />Hopefully he'll answer it now that you have asked it.<br /><br />In fact, I would suggest a response to Goodhart's Law (on which the Lucas Critique is derived) applying to NGDP targeting. I find it a clearer, more succinct "version" of the same intuition.Major_Freedomnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-66451645023239642452012-06-09T11:11:56.779-04:002012-06-09T11:11:56.779-04:00Scott! Thanks for dropping by!
What I've done...Scott! Thanks for dropping by!<br /><br /><i>What I've done is pointed out to the profession "Look at all those times where your gut instinct tells you more AD is needed. Do you notice that they...correlate very well with NGDP relative to trend?" Of course that may be a bad argument, because the gut instincts of macroeconomists may be wrong. But I think that's why my proposal has been somewhat persuasive.</i><br /><br />Agree.<br /><br />Now this may be going a bit far afield, but what about the Lucas Critique? If NGDP hasn't been targeted, how do we know that the gut-instinct relationship between NGDP and AD will hold if we try to use NGDP as the lever to control the economy?<br /><br /><i>Thanks for the Box link. I've been saying "all models are false, but some are useful" in my blog, thinking that I made it up.</i><br /><br />Argh! My pet peeve. It's just a really bad quote. It basically deprives the word "right" of all useful meaning. I will write a post about this soon... ;)Noah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-32860950356667508592012-06-09T10:25:13.478-04:002012-06-09T10:25:13.478-04:00Noah, Interesting post, which I'll have to th...Noah, Interesting post, which I'll have to think about. Here's an initial reaction:<br /><br />1. There are two issues here. Why did some famous macroeconomists apparently find my National Affairs article to be persuasive? And should it have been persuasive? (I.e., was it persuasive for the right reasons?)<br /><br />2. I think at this late date most macroeconomists have basically made up their minds about the "do nominal shocks have real effects" debate, which has been going on for 100s of years. We've gone over the data exhaustively. Most have concluded nominal shocks do have real effect. I'm speaking to that group. <br /><br />3. I also think that most famous macroeconomists have a "I know it when I see it" attitude toward demand shocks. Thus most look at cases like Britain today, or America in 2011, and think it fairly obviously that a bit more AD would be helpful. Indeed this consensus is so powerful that conservative opponents of fiscal stimulus in 2009 usually didn't argue that it will "work," and hence would be bad, but rather that it will fail to boost AD due to crowding out. That's very revealing. Tacit admission that more AD would be nice.<br />4. What I've done is pointed out to the profession "Look at all those times where your gut instinct tells you more AD is needed. Do you notice that they don't correlate very well with inflation (which has been above 2% in Britain during the recession, and was above 2% in the US during 2011) and they do correlate very well with NGDP relative to trend?"<br /><br />5. Of course that may be a bad argument, because the gut instincts of macroeconomists may be wrong. But I think that's why my proposal has been somewhat persuasive.<br /><br />PS. Thanks for the Box link. I've been saying "all models are false, but some are useful" in my blog, thinking that I made it up.Scott Sumnerhttps://www.blogger.com/profile/15864819372390187247noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-78309459501427630572012-06-01T11:58:48.114-04:002012-06-01T11:58:48.114-04:00I think in most cases, pure qualitative prediction...I think in most cases, pure qualitative predictions are useless. They must come with some quantitative sense.<br /><br />Suposse I say: if you print money, you will have a boom, and then a bust.<br /><br />That is useless unless I have some Idea of how large the boom will be, how large the bust will be, when they will happen and how much money I have to print for that to happen.<br /><br />If the FED prints $10 I'm pretty sure that the prediction mentioned above will not hold. But if the FED prints $1000000000 trilion then I'm pretty sure shit will happen without any boom.<br /><br />So, the "qualitative" prediction you refer to may actually be quantitave in the sense that it referes at least to order of magnitudes, but they are just not very precise. Otherwise, I think qualitative predictions are not useful at all.Carloshttp://analisereal.wordpress.comnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-53530138843695278322012-06-01T11:44:04.316-04:002012-06-01T11:44:04.316-04:00Very interesting.Very interesting.Carloshttp://analisereal.wordpress.comnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-41032827128355297502012-06-01T11:39:50.814-04:002012-06-01T11:39:50.814-04:00My point is that if you have a model that is consi...My point is that if you have a model that is consistently making predictions that are very wrong, then an honest researcher is perfectly within his rights to question the model. Imagine 2 models describing a relationship between A and B: Model 1 says A=B*B, Model 2 says A=5*B*B*B. If your measurements say that A=2*B, then model 2 should be discarded, model 1 might be used until something better comes along, but neither one is really right. What would be wrong is to try to make the measurements look like model 1 is right, and even worse would be claiming that model 2 was in any way correct.<br /><br />As far as the EMH is concerned, I can't buy strong-form EMH even conceptually, for the simple reason of information asymmetry. On a micro scale, of course, I have every reason to lie in order to sell a $5 item at $30. But my buyer, now seeing what I've sold him, also has every reason to convince another sucker to buy the same $5 for at least $30, leaving the market price at $30 even though the underlying product is still worth $5. There's no reason to think that with a large enough market and nobody looking to carefully, the price would remain $30 and a mythical $25 worth of value comes out of thin air.Dave Knoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-75038072525850434892012-06-01T11:08:24.111-04:002012-06-01T11:08:24.111-04:00Noah,
The first thing one must decide is to what ...Noah,<br /><br />The first thing one must decide is to what use the model will be put on. <br /><br />(i) if it is for prediction in policy matters, a very compelling way to judge the model is to just set up a loss function, and evaluate its performance with a reasonable separation of the calibration and validation data. <br /><br />The step above is likely to doom almost any model for most practical purposes. One could dispute the loss function. But, so what, that's the point. The model is not real and if it is going to be used for prediction you have to decide how big the discrepancy must be untill you decide that it is useful or not.<br /><br />And this is not a validation of the "truth". It is relative, and open to the fallacy of affirming the consequent. <br /><br />But it is very reasonable in most cases.<br /><br />Now, I don't know what you really meant sugesting the statiscal test for the model. But, unless you also define how big the discrepancy must be to reject it, you will just conflate statistical significance with economic significance. And that is certainly a trivial (but unfortunately common) mistake.<br /><br />(ii) if it is a "Hal Varian" model, to improve understanding of a mechanism, then it is intuition, comom sense and persuansion that will do the trick. The rules here are just: donĀ“t commit logical fallacies.<br /><br />And so on.<br /><br />I don't know if you're familiar with the literature of modelling in earth sciences, but there's a paper in Science vol 263, 1994 "Verification, Validation, and Confirmation of Numerical Models in the Earth Sciences" that is very useful for economists.<br /><br />Thanks<br /><br />CarlosAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-42077006661285081552012-05-31T07:40:03.931-04:002012-05-31T07:40:03.931-04:00This Cosma Shalizi research proposal would seem to...This Cosma Shalizi research proposal would seem to show one way forward,<br /><br />" 5000+ words, and many equations, about a proposal to improve, not macroeconomic models, but how such models are tested against data. Given the actual level of debate about macroeconomic policy, isn't it Utopian to worry about whether the most advanced models are not being checked against data in the best conceivable way?"<br /><br />http://cscs.umich.edu/~crshalizi/weblog/700.htmlA Hhttps://www.blogger.com/profile/06916657901677009228noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-81125859926852889672012-05-30T18:45:38.284-04:002012-05-30T18:45:38.284-04:00Yes. There's a distinction between a theory be...Yes. There's a distinction between a <i>theory</i> being quantitative and the tests of the theory being quantitative. Tests will always have to be quantitative, even if the theories are qualitative "tub/tube" type of things.Noah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-21330486035042947472012-05-30T18:09:23.167-04:002012-05-30T18:09:23.167-04:00oops, the correct statement is "managed funds...oops, the correct statement is "managed funds should not consistently over-perform index funds".CAnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-36622147190261850112012-05-30T18:06:47.703-04:002012-05-30T18:06:47.703-04:00Dave, as I mentioned, and Kuhn has demonstrated, n...Dave, as I mentioned, and Kuhn has demonstrated, no science has a way of disproving a theory. They have ways of testing particular aspects of a theory, but failure of the data to comply generates further research into why this is so. It does not lead to discarding the entire theory. The fact that such unresolved issues are often kept secret by those who like the policy implications of a theory is truly troublesome. But so is the fact that others use these discrepancies to justify an "anything goes" attitude and discredit policies that are based on predictions of the theory that do comply with the data. For example, one prediction of the EMH is that managed funds should not consistently under-perform index funds. Several studies have shown that this is the case. Fund managers do not like this, after all their income depends on their clients belief that there advice does actually have value, so they often use the failure of the EMH in other aspects to also attack this particular prediction. But are they correct to do so?CAnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-75252363648201199422012-05-30T18:06:05.846-04:002012-05-30T18:06:05.846-04:00Noah, I agree very much with your post. I do have ...Noah, I agree very much with your post. I do have one addition however about your exact biology analogy in the update. In the late 1980's I believe, the econometricians Soren Johansen and Katarina Juselius spent some time at John Hopkins to test competing models of the liver. They ultimately rejected the competing "bath tub" model for the "tube" model, which went on to be a Nobel prize-winning idea (aside: in the former you represent it as a tub which dilutes contents, in the latter as a tube which "eats" some of the contents passing through it to varying degrees depending on its functioning). In fact, biology does rely on a means of rigorously testing competing models (bio-stats), and the abstract models are also mathematical in nature. <br /><br />I do however agree that useful scientific predictions in economics are in general qualitative (Friedman's essay has been the basis for the use of models with sharp predictions but his example was simply when demand goes up price goes up). In our mathematical models the parameters are generally symbols which we only constrain qualitatively (or on occasion assuming long-run homogeneity or something). We don't say in our math that beta equals 2.55, we estimate these parameters assuming our model is correct (and generally reject them if they return insignificant results or those with the wrong sign). We then pretend these parameters are time-invariant, often without testing for parameter constancy, and typically find that they, in reality, break down over time (results are sample dependent). Pretending any one model provides exact prediction (up to a random error of known distribution) forever into the future probably accounts for part of the Wall street blow-up (and relates directly to Hayek's Nobel speech about the pretense of exact knowledge, which he referred to as charlatanism).Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-16428672208064228922012-05-30T16:24:40.986-04:002012-05-30T16:24:40.986-04:00If economics doesn't develop an idea of dispro...If economics doesn't develop an idea of disproving a theory, then it can't qualify as even a semi-scientific discipline. What honest economics should be doing is developing models and predictions based on those models, along the lines of "If policy A, then outcome B", and then evaluating those models based on the degree to which A in fact leads to B.<br /><br />The trouble is, that much of the profession practices not honest economics but dishonest economics. In this scenario, instead of the goal being a dispassionate search for an accurate model of economics, the goal is to convince people that policy A is a good idea (whether or not it actually is). This leads to patently bogus arguments (Paul Krugman's "zombie lies") being brought up repeatedly no matter how often they're discredited. And as long as there are lavish rewards for policy A, some economists will be paid very good money to practice dishonest economics.Dave Knoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-35506106490462173332012-05-30T16:04:46.123-04:002012-05-30T16:04:46.123-04:00Wait.. After all this people still believe in EMH...Wait.. After all this people still believe in EMH? That's sort of similar to Intelligent Design. It might be a nicely packaged theory that would be useful to believe for various reasons, but it doesn't make the theory true - or useful. <br /><br />Replicatable profits in the financial sector are derived from the creation of inefficiencies, especially through information gaps. Morgan Stanley's most recent successful efforts in persuading everyone that Facebook was worth $38 a share is a classic example of EMH in action. "Everyone" knew that Facebook was worth at least that much. Now everyone knows that Facebook isn't worth that much. But the people that were informed that Facebook wasn't worth that much have made a lot of money shorting the stock in the interim. In other words, EMH is a finance marketing tool used to sell financial products to people that Goldman Sachs terms "muppets". <br /><br />Of course people can make a career in sales.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-53500093397362292572012-05-30T15:10:13.149-04:002012-05-30T15:10:13.149-04:00I like your line of thinking. Making things mathem...I like your line of thinking. Making things mathematical does not make them scientific, it only makes them internally coherent. But that internal coherence can be as useful in reality as Borges short stories: great entertainment for those with the predilection, but fiction.jsnhttps://www.blogger.com/profile/18326058179873660321noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-8371819243280029612012-05-30T13:57:59.817-04:002012-05-30T13:57:59.817-04:00Reason and Zathras - Oh my God, don't you peop...Reason and Zathras - Oh my God, don't you people even read? Do a CTRL-F search for "George Box" in my original post. FACEPALMNoah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-58269617016878058312012-05-30T13:33:17.998-04:002012-05-30T13:33:17.998-04:00Noah has previously warned people against this (I&...Noah has previously warned people against this (I'm a statistician, and it's a corny old saying).Barry DeCiccohttps://www.blogger.com/profile/04735814736387033844noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-1687511061345771922012-05-30T13:31:46.943-04:002012-05-30T13:31:46.943-04:00"This is an organ, this is a gene, this is an..."This is an organ, this is a gene, this is an individual."<br /><br />IIRC, these definitions get more untidy every year. Genes may be transcribed in different ways, they interact with each other quite strongly. Developmental biology is changing the picture of the simple gene to protein model. <br /><br />Organs again are a bit fuzzier, since they interact with other things (i.e., the effect of A/a depends on B/b). <br /><br />And in many 'species', there's massive gene flow among individuals, and individuals might have most of their survival ability depend on the others around them.Barry DeCiccohttps://www.blogger.com/profile/04735814736387033844noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-62505117997175352032012-05-30T13:01:04.982-04:002012-05-30T13:01:04.982-04:00Correction: I meant "better" - not "...Correction: I meant "better" - not "best".Thiago Maciel Oliveirahttps://www.blogger.com/profile/05065258947093775842noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-19266378564223346652012-05-30T12:54:45.430-04:002012-05-30T12:54:45.430-04:00Maybe Sumner's point is that is best to be &qu...Maybe Sumner's point is that is best to be "generally right than precisely wrong"?Thiago Maciel Oliveirahttps://www.blogger.com/profile/05065258947093775842noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-5161903204722282902012-05-30T12:27:00.383-04:002012-05-30T12:27:00.383-04:00No, it makes plenty of sense without it. The key p...No, it makes plenty of sense without it. The key phrase is "in the end". You compare the predictions of a theory with data for the purpose of testing a theory against an alternative. As Kuhn pointed out, the scientific community hardly ever gives up a theory that fairs poorly against the data UNLESS they have a better alternative.CAnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-92194524370659684332012-05-30T12:20:00.506-04:002012-05-30T12:20:00.506-04:00Ok Noah, here's the post as promised.
ht...Ok Noah, here's the post as promised. <br /><br /> http://diaryofarepublicanhater.blogspot.com/2012/05/whats-in-macro-model-sumner-vs.htmlMike Saxhttps://www.blogger.com/profile/01360689916550576484noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-75987379909840445692012-05-30T11:58:53.255-04:002012-05-30T11:58:53.255-04:00Also demographics - births, deaths, household form...Also demographics - births, deaths, household formation, immigration, education etc. The volumes of these things are macro-economic variables. If you don't close your eyes to the dramatic demogrpahic changes in the 1970s, the change from the 1960s to the 1970s is not so extraordinary.reasonhttps://www.blogger.com/profile/10958786975015285323noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-30913173787797596722012-05-30T11:53:41.942-04:002012-05-30T11:53:41.942-04:00"All models are wrong; some models are useful..."All models are wrong; some models are useful."Zathrashttps://www.blogger.com/profile/14916717939276205773noreply@blogger.com