tag:blogger.com,1999:blog-17232051.post5266353917206758100..comments2024-03-18T22:32:52.802-04:00Comments on Noahpinion: The most damning critique of DSGENoah Smithhttp://www.blogger.com/profile/09093917601641588575noreply@blogger.comBlogger135125tag:blogger.com,1999:blog-17232051.post-5964603581192323572017-03-15T10:23:30.701-04:002017-03-15T10:23:30.701-04:00Completely agree. Also, many of the best predictio...Completely agree. Also, many of the best prediction models that use VARs regularize coefficients use cross-validation to tune parameters, such as how to weight past observations. These techniques would make DSGEs more useful, but as you mentioned they have not been implemented in any software package yet.Andershttps://www.blogger.com/profile/14250388395636299155noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-3777710420061556962014-06-14T08:42:08.719-04:002014-06-14T08:42:08.719-04:00Someone just clicked a link here which sent him or...Someone just clicked a link here which sent him or her to my blog (thanks). So I re read the post. It is brilliant -- absolutely wonderfulRoberthttps://www.blogger.com/profile/14455788499385673507noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-84314174707461295792014-01-13T23:56:23.546-05:002014-01-13T23:56:23.546-05:00I have posted on this topic now, a bit late I gues...I have posted on this topic now, a bit late I guess, at Econospeak, econospeak.blogspot.com/21014/01/are-people-being-mean-to-those-doing.html , although I must note, Noah, that I am somehow finding myself unable to link to your posts, although this is now the second time I have tried to do so and failed. Not sure what is up.rosserjb@jmu.eduhttps://www.blogger.com/profile/09300046915843554101noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-21716903959607229702014-01-13T20:14:31.697-05:002014-01-13T20:14:31.697-05:00Jim,
CGE models have been around for a long time ...Jim,<br /><br />CGE models have been around for a long time and have been used by quite a few private consulting companies. However, they are not DSGE models. They are micro models used for looking at sectoral shifts or implications of policies.<br /><br />Barkley Rosserrosserjb@jmu.eduhttps://www.blogger.com/profile/09300046915843554101noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-91869072235825071182014-01-13T17:00:34.975-05:002014-01-13T17:00:34.975-05:00Daniels:
First, I just want to thank you for havi...Daniels:<br /><br />First, I just want to thank you for having such a reasoned discussion. I appreciate learning from you about some aspects of DSGE's I didn't know.<br /><br />On the BOE: I got this from looking at their website, so maybe I am wrong? But it does seem to imply that they still use their quarterly model, which looked a lot to me like a small-scale traditional model. I know for a fact that, at the Fed Board, the staff's main tool remains FRB/US (I have a chapter on Forecasting in the Federal Government coming out in a book soon(?), so I've spoken to people there about it). Not that DSGE's aren't used, but when it comes to organizing all the pieces of the toolkit, the Klein-type model is still at the core.<br /><br />I don't think you are correct about the IO characteristics of the economic forecasting industry. There have certainly been new entries over the years---Dismal Scientist/Economic.com came out of nowhere to pick up a significant market share in the 1990s, for example. And the big firms are always updating their models anyway. If there was a market advantage to be gained from using a new methodology, I think it would have been adopted. I know this because I've been there (and pushed successfully for the adoption of error correction in estimation, among other things). In fact, I heard that one of the large firms hired a DSGE modeler at one point, but they parted ways after a couple of years when it became evident that the firm's investment wasn't worthwhile. <br /><br />It doesn't take "dozens" of economists. Three Ph.D.'s and maybe one full-time manager could probably get it started. There's plenty of turnover among business economists, so lots of opportunity to find people who are ready for something new and different that they can sell to their users as more up-to-date than what others are using.<br /><br />Agree about the calibration issue--and especially the point forecasts. But that leads to the difficulty of getting users to see beyond the point forecast. Having been there a lot, I can assure you that it's non-trivial.<br /><br />You see, I would like to get beyond the idea that DSGE's aren't accepted because of market failures, or ignorance, or whatever, and look more closely at how and why decision-makers use information--and how economics supplies it to them. I think that's a really worthwhile area of thought that ought to play a role in driving the research agenda. And I think the survival of Klein-style models is a really interesting data point in that context.Daniel Bachmannoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-87880791263882277762014-01-13T16:37:39.253-05:002014-01-13T16:37:39.253-05:00A consultancy in New Zealand called BERL runs a CG...A consultancy in New Zealand called BERL runs a CGE model. I am not aware of anyone who takes their work seriously. I am aware of many people who think they are muppets.Jimnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-51199954844427498142014-01-13T15:39:38.383-05:002014-01-13T15:39:38.383-05:00I don't get what you're saying about about...I don't get what you're saying about about the BoE? The core of their old forecasting framework was a DSGE model. The difference relative to a VECM is that instead of just having convergence to some long run static EQ, their error correction terms are responding to the deviation between a dynamic equilibrium model (a DSGE model, or maybe rather a DGE since it seems the BEQM core is non stochastic). Their new core model is a DSGE model similar to an open economy version of Smets and Wouters' model+some credit constrained households. <br />The post says this has not proven to be the case so far very frequently in the private sector. But I don't know that Klein type models were useful in the private sector in the 1950's either, which didn't mean they're not useful today. If I understand correctly the leaders in using large macro models were a macro consulting firm established by Klein himself and the Fed through Modigliani's MPS model. At the time it must have been much easier probably to gain clients just by showing up with a model when there was nothing comparable. Nowadays, the market is still oligopolistic but more crowded, and it would be harder for someone like Larry Christiano to open up a macro consultancy business at the same level as say Oxford Economics or IHS: it wouldn't be enough for him to quit his university job, he'd need to find dozens of other like minded economists, coordinate them, hire secretaries and business development managers...in short there are barriers to entry into the field, and like in other (semi) scientific fields it's not always obvious to convince non expert clients that a new method can do better than something that's been around for decades and seemed to provide a satisfactory though highly imperfect service. But who knows, maybe Bernanke and Gertler will get bored with academia (not sure what Bernanke's plans are now) and they'll team up to provide macroeconomic advice based on their New Keynesian DSGE+Financial accelerator model? I personally wouldn't mind listening to Bernanke's macro advice.<br />By the way, Nate Silver and my comment weren't about how come the point forecast from business economists didn't predict Lehman brothers or other recessions. The point is that typical private sector forecasts seem to be badly calibrated: their confidence/uncertainty bands around the forecast (e.g in the SPF density forecasts) are too narrow and underestimate the risks facing the economy (and it's true this was also an issue for DSGE models estimated on great moderation samples when the financial crisis hit) - and why is everyone so fixated on the point forecast when you know it's an (almost) probability zero event? With all the uncertainty around, it's the confidence bands and the scenarios which should be more important.danielshttps://www.blogger.com/profile/01799942447501959179noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-38985893967404440162014-01-13T14:55:59.360-05:002014-01-13T14:55:59.360-05:00This is consistent with everything other people ha...This is consistent with everything other people have told me!Noah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-84508023577760058312014-01-13T10:35:56.948-05:002014-01-13T10:35:56.948-05:00I can speak from the model based macro hedge tradi...I can speak from the model based macro hedge trading perspective.<br />1) Most of the biggest macro funds are fundamental, NOT technical. And I mean the ones that are running models.<br />2) This is a case of having no faith in the DSGE or Fed models. Speaking from personal experience (over 20 years), I can say that a lot of the traditional theory is suspect. Recent examples being - no balance sheet or debt view of the economy has led to inflation and growth being lower than forecast. Interest rates do not work in the real economy as advertised in traditional models. Why should I use them if they do not work? It is clear to us in the real world that the world does not work like the Fed describes.<br />3) Dont even get me started on currency trading. The fact that academic models have such a bad track record should tell the macro guys something.<br />4) Ultimately, the continual denial by the Fed that asset prices do not swing between overpriced and underpriced AND that they have nothing to do with it just plain does not make any sense.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-752206805386982782014-01-13T08:33:27.341-05:002014-01-13T08:33:27.341-05:00This is interesting. I quite agree about the lack ...This is interesting. I quite agree about the lack of confidence in business economists forecast--this is well demonstrated in the literature, as well as by Nate Silver. At this point, nobody in the economics world really expects much ability in forecasting turning points (educating our customers is another matter). This is, of course, where we really need a good tool. I don't believe DSGE's did that well in the last downtown. But how could any econometric model predict that the Lehman would fail when it did?<br /><br />So the question is whether DSGEs provide another useful tool. As the original post points out, so far this has not proven to be the case. (Again keep in mind that while Central banks have DSGEs, they also use Klein-type models. The Bank of England, as far as I can tell from what they show on their web site, haven't replaced the Quarterly Model with a DSGE.) <br /><br />We had a session on models a couple of years ago at AEA (sponsored by NABE, so it was relegated to an inconvenient location). One of the Fed people at that session made an interesting point: That central banks (and especially the Fed Board) have a lot more resources than the private sector, so they simply buy more of all types of forecasting, without necessarily making strict cost-benefit analysis of what they are buying. A nicer way to put it might be that central banks are more willing to buy new, risky technology than the private sector.<br /><br />In any event, I welcome the possibility that DSGE models could provide a useful tool. I am currently a purchaser of a traditional model, but would welcome a sales call from somebody who could convince me that their DSGE model would be a better tool. <br /><br />Any takers? If not, I think the original post stands.Daniel Bachmannoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-6471049417884638202014-01-13T08:27:13.938-05:002014-01-13T08:27:13.938-05:00Noah: "But notice that Bayesian beliefs depen...Noah: "But notice that Bayesian beliefs depend on your priors"<br /><br />It may be so, but your priors may be wrong and you should punish yourself more the more confident you were in these mistaken priors. I had this discussion with Scott Sumner and god, it is frustrating.<br /><br />Ok, lets have this example: you have 3 theories that may explain some phenomena. You need to be more than 50% confident in order for you to publicly subscribe to a theory. Imagine that your priors are as follows: 51% for theory A, 48.99999999999 for theory B and 0.00000000001 % for odd theory C that you deem almost impossible.<br /><br />Now new evidence comes up that makes you update your priors as in the following 2 scenarios:<br /><br />1) New evidence supporting theory B appears so that the posterior probabilities are as follows: A - 49% ; B - 51.99999999999% and C - 0.00000000001 %<br /><br />Since it is now theory B that passed the threshold you make public announcement that effective now you are "completely" changing your opinion and now support theory B. However from Bayesian point of view it is not complete change of opinion. You reduced probability of theory A and B by approximately 4%. We may say that it is not a large change. You made a mistake, but not a large one.<br /><br />2) New evidence supporting theory C appears. Now it looks like this: A - 51% ; B - 46.99999999999% and C - 2.00000000001 %<br /><br />Now your public "opinion" did not change much. No bold statements, you still support theory A, right? But this is HUGE, really HUGE from bayesian point of view. You just elevated something from being labeled as "really, really incredibly not likely" to just "not likely". You made a huge change in your bayesian scoring scale. You should really berate yourself for having such terrible priors.<br /><br />So Bayesian is something else from what you think. The most useful concept here is that of information entropy, a concept from information theory. This concept can help you measure how "surprising" a change in probability was, how much new information you gained with a new evidence. The point being that the lower the probability of a surprise the larger the information value it has. Promoting low probability even by orders of magnitude means a lot of information was gained. And by this account the scenario 2 has a much larger impact then scenario 1. <br />Anonymoushttps://www.blogger.com/profile/14853090724221729923noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-80234174470452511792014-01-12T22:47:45.368-05:002014-01-12T22:47:45.368-05:00That's why mechanism design application (like ...<i>That's why mechanism design application (like auctions) work well.</i><br /><br />Yep.<br /><br />So of course I was being extremely tongue-in-cheek with my examples...naturally you can't do what I described, which is why people don't. ;-)Noah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-20315581029455123182014-01-12T18:54:22.423-05:002014-01-12T18:54:22.423-05:00Noah,
Thank you, although you forgot a couple of ...Noah,<br /><br />Thank you, although you forgot a couple of z's. <br /> Anyway, the cases you quote correspond to situations in which the economic structure is very transparent, the optimization does not interact with the (estimation) noise and the underlying DGP is stationary and very stable and you clearly control the rules of the game. That's why mechanism design application (like auctions) work well. In actual markets, where the fundamentals are not very transparent, your ability to filter noise away are very limited, econ models will typically fail.<br /><br />Let me give you a perfect example from commodities. Around 10-15 years ago, GE was marketing a great product: GE maps. It was great since it perfectly captured the fundamentals in the local electricity market in Pennsylvania: PJM. How perfectly? GE maps was used by the system operator to actually dispatch the system in real time. So, they knew the underlying supply and demand structure pretty much with perfect precision. Yet, they couldn't predict or even explain past prices. Neither levels nor volatilities. Consider how strange that is. Electricity is the simplest commodity. No storage, hence (almost) no inter-temporal optimization in the underlying production decision. Fundamentals as transparent as could be. Yet, no success. Krzyshttps://www.blogger.com/profile/15794655390770135247noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-30138054616743930332014-01-12T16:23:35.249-05:002014-01-12T16:23:35.249-05:00Actually I think Moody's has started looking a...Actually I think Moody's has started looking at DSGE models as part of their credit risk scenarios modeling recently. 1 possible advantage with DSGE models, is that with current software such as dynare it's quite easy to do for example a 3rd order Taylor approximation of the dynamics, so you can better capture the important nonlinearities that may matter if you're trying to simulate a more extreme scenario. I'm not aware of anything equivalent in the Klein/Cowles comission style models. <br />I'm not implying business economists are ignorant, though there is a general lack of confidence in business economists' forecasts (exemplified for example in the chapter on economic forecasting in Nate Silver's recent book the "Signa and the Noise")- to be fair forecasting the economy either as a point forecast or with scenarios/well calibrated confidence bands is going to be hard whatever you do. <br />In academic applications, you never see add factors, because that's not part of what you want to do in a journal article where the goal is to see how far a model or theory can go before failing without add factors (I doubt Tobin or Solow or Modigliani used add factors in any way when writing an academic article using some IS/LM model). But in policy environments, DSGE models use variants of add factors quite a lot either through the shock processes or via explicit model combination (treat the DSGE model as a core model, then add extra dynamics/adjustment via a non-core block), and presumably you'd do the same in private sector macro consulting. In some cases, the add factors/fixes from the perspective of one DSGE model can be justified as compensating for missing factors/mechanisms that are present in other auxiliary DSGE models that are richer in some dimensions but may be harder to work with. The best exposition I know of is the recent BoE explanation of their forecasting framework, <br />http://www.bankofengland.co.uk/research/Documents/workingpapers/2013/wp471.pdf<br />.<br />The core/non core approach is also given a good exposition by the BoE as part of their previous model,<br />http://www.bankofengland.co.uk/publications/Documents/other/beqm/beqmfull.pdf<br />.<br />Other central banks may be using similar procedures, but are less transparent.danielshttps://www.blogger.com/profile/01799942447501959179noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-45214999754816244592014-01-12T15:32:46.878-05:002014-01-12T15:32:46.878-05:00Stephen,
1. I do quant commodity trading.
2. I l...Stephen,<br /><br />1. I do quant commodity trading.<br />2. I like Econ for its qualitative insights. It's always a useful check. Plus I always look around for stuff I might be missing.Krzyshttps://www.blogger.com/profile/15794655390770135247noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-63795699624417208532014-01-12T14:47:35.985-05:002014-01-12T14:47:35.985-05:00Also:
2. Why are you reading Noah's blog, any...Also:<br /><br /><i>2. Why are you reading Noah's blog, anyway?</i><br /><br />Well gee, thanks! ;-)<br /><br /><i>I assure you that my tastes are soooo cool that no one likes what I do.</i><br /><br />NOW I MUST KNOWNoah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-40467881035402879292014-01-12T14:45:25.381-05:002014-01-12T14:45:25.381-05:00If you claim to be condemning all of modern macro,...<i>If you claim to be condemning all of modern macro, then that's essentially damning all of modern economics to hell, as it's now impossible to say where micro stops and macro begins.</i><br /><br />No. False. Not even slightly true. First of all, even if the techniques used are similar, the questions they try to answer are different and the data and empirical methods they use to validate their theories are completely different.<br /><br /><i>So now I think you're saying that "storytelling" might be useful.</i><br /><br />People do enjoy stories! I'd rather models make usable quantitatively accurate predictions about stuff, but stories are better than nothing. Heck, that's all cosmologists and a lot of evolutionary biologists get from their models.<br /><br /><i> And there's a tendency for more contemporary models like what came out of Schmets-Wouters and Christiano-Eichenbaum-Evans to grow into monsters that really look much like the Lawrence Klein models. And those monsters are really quite useless, I think, though the people who construct those things and work with them will argue otherwise, of course.</i><br /><br />I suspect you're probably right. You should write this blog post! "Monster models" is a great title. I will link approvingly. :-)Noah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-67263006336467619562014-01-12T14:29:17.662-05:002014-01-12T14:29:17.662-05:00The point about people in the private sector being...The point about people in the private sector being old is precisely the red herring that academics have clung to since the RE revolution. I was trained in the 1980s in RE/real business cycles and had to relearn quite a bit to become an effective business analyst in the 90s. This is still the case. It's not a question of age, it really is a question of DSGE's not being useful in the business environment. <br /><br />The point of the original post is that the private sector doesn't use DSGE's (or, for that matter much of the structure of post-1960s macroeconomics). And this is also true for government by the way...when push comes to shove the discussion will be understandable by anybody who took intermediate undergraduate macroeconomics in the 1970s.<br /><br />This is a fact, which has been driving people crazy for a long time. It doesn't necessarily invalidate DSGE's or real business cycles as a tool for analysis, but it does raise some serious questions. It also suggests that the experience of analysts in the government/business world might be useful for developing theoretical approaches. But instead I typically see some variant of "business economists are ignorant."<br /><br />Sorry for the mini-rant. I quite agree that a toolkit is useful, and, of course, nobody sane gives a decision-maker a purely model-based forecast. Which is the type of thing that should give us all pause. Except (I have to get this in) the Klein-type model system recognizes this formally, with addfactors, while the DSGE approach...well I don't see it in the literature. <br /><br />Same thing with risk. If DSGE's are better for scenario analysis, why don't Macro Advisors and Moody's Economy.com have competition from this direction in supplying tools for stress testing. Because I'm not aware of anybody supplying a DSGE based tool for stress testing. It does seem like a potentially lucratuve market niche.Daniel Bachmanhttps://www.blogger.com/profile/08559214766265313671noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-70091861425859404402014-01-12T14:22:47.462-05:002014-01-12T14:22:47.462-05:00B) Weirdly, I don't like either Mozart or Radi...B) Weirdly, I don't like either Mozart or Radiohead. I know, I'm strange.^H^H^H^H^H^H^H wrong<br /><br />Didn't say I liked them these characters either. I assure you that my tastes are soooo cool that no one likes what I do.Steve Williamsonhttps://www.blogger.com/profile/16629774961390533020noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-81128272730791863762014-01-12T14:14:57.463-05:002014-01-12T14:14:57.463-05:00Krzys,
I'm not sure that we're even discu...Krzys,<br /><br />I'm not sure that we're even discussing the same thing. Two questions for you:<br /><br />1. What exactly is the line of business you're in.<br />2. Why are you reading Noah's blog, anyway?Steve Williamsonhttps://www.blogger.com/profile/16629774961390533020noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-19556852440550867742014-01-12T13:36:45.601-05:002014-01-12T13:36:45.601-05:00Noah, I think a lot of this is a case of big menta...Noah, I think a lot of this is a case of big mental adjustment costs. Private sector macro to me still seems dominated by people who got most of their training in the 70's and 80's, at a time when DSGE models were rudimentary. The state of the art DSGE modeling with everything from financial sectors, to search frictions to all sorts of imperfect information/learning are probably too recent for many people to be aware of outside academia and central banks. And I think switching from the simultaneous equations model with all variables observables, to a state space framework where you aknowledge the importance of latent,unobserved state variables and try to measure them is a big improvement in perspective about the macroeconomy. But things like Kalman filtering are still not part of the standard training of most Masters in economics students I think, so the technical baggage required for using DSGE models may be missing sometimes (though it's probably there in some quant/finance applications- e.g for stochastic volatility modeling.).<br />Macro forecasting is not exactly a business with a great reputation for accuracy either in conditional or unconditional forecasting, so while you can't clearly prove their forecasts and scenario analysis would improve with using DSGE models, it's hard to argue that they couldn't improve if they diversified their toolkit.<br />A lot of DSGE models are about giving insight, sharpening intuition by taking a more theoretical perspective, which then you can complement via your judgement/add factors +statistical models like VAR's to get a final forecast. It's not meant to be the single forecast producer, and neither should a large old style model (forecast combination over several models almost always dominates individual model forecasts). And much of the insight can also be obtained by going over all sorts of DSGE models/papers, without necessarily using one as a key forecasting tool.<br />Otherwise, I suspect things like Bayesian VAR's or dynamic factor models or sometimes just univariat ARIMA's will always dominate in pure forecasting at 1-2 years horizon. The more relevant question is whether DSGE models can be useful in risk/scenario analysis. You seem to suggest no, because the microfoundations are wrong, but that's a bit of a straw man. After all, it's not like the typical consumption or investment function in a big old style models are true either. They're also surely missing many variables (that must be the case since they're limited to thinking only of observable variables or using very rough proxies for unobservable things like potential output), using the wrong linear functional form, too hasty in assuming error terms are IID and don't contain richer dynamics, and in general often seeming to ignore key supply/demand considerations that should matter beyond the shortest forecasting horizon. And the notion that they take into account bounded rationality isn't very serious to me beyond the fact that they include lags of variables (but so do policy oriented DSGE models, and it's hard to know you've really separately identified habit formation, from adjustment costs, from adaptive expectations when you add lags of consumption for example). And I don't see anything reflecting prospect theory or other state of the art psychological theory in old style macro models either. So any model you use is false, but we can disagree about whether it contains enough reality to be useful (I'd say a modified version of the permanent income consumption model that allows for some demographic factors, a proportion of households who just consume all their income and some discount factor shocks isn't so bad. You'd probably disagree).danielshttps://www.blogger.com/profile/01799942447501959179noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-17430838085227985512014-01-12T12:50:04.699-05:002014-01-12T12:50:04.699-05:00John H. Cochrane sez: " Pure "structural...John H. Cochrane sez: " Pure "structural" weather models don't do as well in making actual forecasts as statistical approaches which combine different structural models, simple correlations in the data, and forecaster judgement. " - but taking your analogy one step further: the really important stuff for planet Earth is in fact 'structural weather' models of CLIMATE as opposed to 'whether it is going to rain this weekend' WEATHER forecasts.Ray Lopezhttps://www.blogger.com/profile/11134761834999705305noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-17925876671155462462014-01-12T11:23:28.009-05:002014-01-12T11:23:28.009-05:00"And aren't simple models, like little OL..."And aren't simple models, like little OLG models, more useful for storytelling than these big infinitely-forward-looking monstrosities?"<br /><br />Now you're actually getting down to business. If you claim to be condemning all of modern macro, then that's essentially damning all of modern economics to hell, as it's now impossible to say where micro stops and macro begins. So now I think you're saying that "storytelling" might be useful. That seems a bit on the pejorative side, but to put a nice spin on it, we could broaden our notion of storytelling by saying that we have some little models that give us some sparks of insight into how the world works.<br /><br />Possibly the area of agreement we can find here is the following. I find various models useful, sometimes for quite different reasons. Maybe they tell me something about how monetary policy works; maybe they provide some insight into why some people in the world are so poor and some are so rich; maybe this just provides me with another tool - I'm going to learn how this thing works, and I'm going to file that away for later use when I'm struggling with some problem.<br /><br />So, I have a toolkit - it's some array of neoclassical growth model/search and matching/mechanism design/information economics/OG model/Lagos-Wright/etc. etc. And then I use that for everything I do, whether I'm writing academic papers, blogging, writing textbooks, giving advice to people who make policy.<br /><br />But there is a class of monster models, which started with 1960s Lawrence Klein type 400-equation macroeconometric models. Those were the models Lucas was critiquing. And there's a tendency for more contemporary models like what came out of Schmets-Wouters and Christiano-Eichenbaum-Evans to grow into monsters that really look much like the Lawrence Klein models. And those monsters are really quite useless, I think, though the people who construct those things and work with them will argue otherwise, of course.Steve Williamsonhttps://www.blogger.com/profile/16629774961390533020noreply@blogger.comtag:blogger.com,1999:blog-17232051.post-44562658354278825892014-01-12T10:11:57.978-05:002014-01-12T10:11:57.978-05:00From what I've seen, investment decisions are ...From what I've seen, investment decisions are based on a plausible narrative of what will happen and why. If you can't tell that story within 10 minutes and 3-pages, then it won't fly at the top levels. And that narrative had better fit the political view, personal experience, and rule of thumb of the top guy. <br /><br />Complex models tend to obscure the narrative, so no one is going to stake their fortune on it. Nor should they, since too many programming mistakes can be made and too many variables aren't measured timely or precisely. Once the narrative has been chosen, then maybe models will be used to quantify the magnitude and timing of the expected outcome. The complex models are often used for showmanship rather than strategic decisions.<br /><br />Investors' expectations of inflation for the past 4-years were simply because MBA's didn't know the difference between the monetary base and the money supply, and that the latter was created by commercial bank lending more than central banks. It was a confusion of terms, they didn't really rely even on simple spreadsheet models.Kent Willardhttp://www.kentwillard.comnoreply@blogger.comtag:blogger.com,1999:blog-17232051.post-89990089143796803982014-01-12T00:17:04.497-05:002014-01-12T00:17:04.497-05:00Not sure I buy that, Ksrzyksrykz. McFadden's m...Not sure I buy that, Ksrzyksrykz. McFadden's models famously predicted BART ridership, and there's a lot of stuff based on those, right? Noah Smithhttps://www.blogger.com/profile/09093917601641588575noreply@blogger.com