Lots of people are unhappy with what Lucas et al. invented to replace the "old macro". But few would argue that it needed replacing. Identifying correlations in aggregate data really doesn't tell you a lot about what you can accomplish with policy.
Because of this, I've always been highly skeptical of John Cochrane's claim that if we simply launched a massive deregulatory effort, it would make us many times richer than we are today. Cochrane typically shows a graph of the World Bank's "ease of doing business" rankings vs. GDP, and claims that this graph essentially represents a menu of policy options - that if we boost our World Bank ranking slightly past the (totally hypothetical) "frontier", we can make our country five times as rich as it currently is. This always seemed like the exact same fallacy that Lucas et al. pointed out with respect to the Phillips Curve.
You can't just do a simple curve-fitting exercise and use it to make vast, sweeping changes to national policy.
Brad DeLong, however, has done me one better. In a short yet magisterial blog post, DeLong shows that even if Cochrane is right that countries can move freely around the World Bank ranking graph, the policy conclusions are incredibly sensitive to the choice of functional form.
Here is Cochrane's graph, unpacked from its log form so you can see how speculative it really is:
DeLong notes that this looks more than a little bit crazy, and decides to do his own curve-fitting exercise (which for some reason he buries at the bottom of his post). Instead of a linear model for log GDP, he fits a quadratic polynomial, a cubic polynomial, and a quartic polynomial. Here's what he gets:
Cochrane's conclusion disappears entirely! As soon as you add even a little curvature to the function, the data tell us that the U.S. is actually at or very near the optimal policy frontier. DeLong also posts his R code in case you want to play with it yourself. This is a dramatic pulpification of a type rarely seen these days. (And Greg Mankiw gets caught in the blast wave.)
DeLong shows that even if Cochrane is right that we can use his curve like macroeconomists thought we could use the Phillips Curve back in 1970, he's almost certainly using the wrong curve. You'd think Cochrane would care about this possibility enough to at least play around with slightly different functional forms before declaring in the Wall Street Journal that we can boost our per capita income to $400,000 per person by launching an all-out attack on the regulatory state. I mean, how much effort does it take? Not much.
And this is an important issue. An all-out attack on the regulatory state would inevitably destroy many regulations that have a net social benefit. The cost would be high. Economists shouldn't bend over backwards to try to show that the benefits would be even higher. That's just not good policy advice.
(Also, on a semi-related note, Cochrane's WSJ op-ed (paywalled) uses China's nominal growth as a measure of the rise in China's standard of living. That's just not right - he should have used real growth. If that's just an oversight, it should be corrected.)
Cochrane responds to DeLong. His basic responses are 1) drawing plots with log GDP is perfectly fine, and 2) communist regimes like North Korea prove that the relationship between regulation and growth is causal.
Point 1 is right. Log GDP on the y-axis might mislead 3 or 4 people out there, but those are people who have probably been so very misled by so very many things that this isn't going to make a difference.
Point 2 is not really right. Sure, if you go around shooting businesspeople with submachine guns, you can tank GDP by making it really hard to do business. No one doubts that. But that's a far, far, far cry from being able to boost GDP to $400k per person by slashing regulation and taxes. Cochrane's problem isn't just causality, it's out-of-sample extrapolation. DeLong shows that if you fit a cubic or quartic polynomial to the World Bank data, you find that too much "ease of doing business" is actually bad for your economy, and doing what Cochrane suggests would reduce our GDP substantially. Is that really true? Who knows. Really, what this exercise shows is that curve-fitting-and-extrapolation exercises like the one Cochrane does in the WSJ are silly sauce.
Anyway, if you're interested to read more stuff I wrote about regulation and growth, see this post.
There have been worse-fitted curves used to support conservative positions. Like this one from the WSJ (and, I guess, Kevin Hassett): http://scienceblogs.com/goodmath/wp-content/blogs.dir/476/files/2012/04/i-52f10b92b127efe3b084854fd413e78e-ED-AG112_1corpt_20070712182433.gifReplyDelete
Actually ran a simple cross-validation check and all four models have more or less the same prediction error. So I suppose it can't really be said that Cochrane was using the WRONG functional form.ReplyDelete
Code is here: https://www.dropbox.com/s/of8gxyjozwkkjze/cv_check.R?dl=0
It'd be nice to have somebody double-check
Right, the point is not that he's wrong, just that his conclusion extrapolates out so far that it's super sensitive to the model.Delete
Are you talking about temperature or gdp?Delete
And guess what: Doing Business indicators are suspiciously inconsistent with enterprise surveys.ReplyDelete
You need to read more carefully. That is not at all what Mary is saying in her paper. They measure quite different things with some topical overlap.Delete
I would fear Cochrane might be right and with much worse air pollution we would be spending a whole lot more on medical care, more gdp and the worse for it.ReplyDelete
"As soon as you add even a little curvature to the function" - sorry guys, not true - the quadratic form is little different than linear (there seems to be a small error in Brad's code that makes the linear fit not run/plot, so orange = quadratic)ReplyDelete
fyi : the link of Greg Mankiw points to this same page.ReplyDelete
Thanks, will fix later.Delete
A loess specification, which is by design more unsure of extrapolating beyond the existing Business climate, shows both the curvature down at the high end and the extreme uncertainty of prediction at 100 or 110:ReplyDelete
lo <- loess(ln_GDP_per_capita ~ Business_climate, surface='direct')
p_have <- predict(lo, se=T)
p_out <- predict(lo, se=T, newdata=data.frame(Business_climate=c(100,110)))
On top of that, deregulation would not necessarily improve "ease of doing business". The subcategories in the World Bank's metric include "protecting minority investors" and "enforcing contracts" - two categories dragging down the USA scores. To bring them up might require more regulation not less.ReplyDelete
I used to like DeLong. I've never vote Republican and was for Obama since he won the Iowa primary in 2008 and again in 2012. But both DeLong and Krugman jumped the shark during the 2016 primary. Now I hate them both. Future progressive candidates shouldn't put out any policy papers that are not incredibly vague like Hillary's.ReplyDelete
But Krugman supported Hillary in 2008. Did you hate him then too, and if not, then why are you hating him now?Delete
Campaigns also can't have anybody cite Academic economists' analysis of the plans they do have. Otherwise it could be called a "fairytale".Delete
Peter is mad because BDL and PK criticized Bernie.Delete
Really ... that's it. Now he's all over the econ blogosphere ad hom-ing them.
"But few would argue that it needed replacing."ReplyDelete
You mean "few would deny."
"Few would argue" means, "few would make the argument that ..."Delete
It can also mean "few would dispute that . . ." Which is probably the sense in which it was intended here. Arguably not the best usage, as it is ambiguous, but not incorrect.Delete
thank you for the first graph, which makes clear what lucas sargent prescott did for the rest of us; I really do appreciate itReplyDelete
What deLong doesn't do, his commenters do.ReplyDelete
How to leave pulp economics behindReplyDelete
Comment on Noah Smith on ‘Brad DeLong pulpifies a Cochrane graph’
“When Bob Lucas, Tom Sargent, and Ed Prescott remade macroeconomics in the 70s and 80s, what they were rebelling against was reduced-form macro. So you think you have a ‘law’ about how GDP affects consumption? You had better be able to justify that with an optimization problem, said Lucas et al.”
From the accurate realization that there is no such thing as behavioral macro laws Lucas/Sargent/Prescott drew the wrong conclusion that there must be something like behavioral micro laws and marched even deeper into the woods.
There is no such thing as a behavioral law, neither in macro nor in micro. But there are structural laws of the economic system. So the right methodological move is from subjective-behavioral macro to objective-structural macro.
Strictly speaking, economics is since Jevons/Walras/Menger on the wrong microfoundations track. In what has been called Keynesian Revolution, Keynes attempted the paradigm shift from microfoundations to macrofoundations. This attempt failed because Keynes got the formal foundations of macro wrong.
So the right methodological move at the critical After-Keynes junction would have been to put macrofoundations right and not to regress to microfoundations which had been dead in the cradle already 140 years ago.
Both, the Walrasian and Keynesian axioms are faulty and have to be replaced by the correct macrofoundations. This is achieved as follows.
(A0) The objectively given and most elementary configuration of the (world-) economy consists of the household and the business sector which in turn consists initially of one giant fully integrated firm.
(A1) Yw=WL wage income Yw is equal to wage rate W times working hours. L,
(A2) O=RL output O is equal to productivity R times working hours L,
(A3) C=PX consumption expenditure C is equal to price P times quantity bought/sold X.
This objective structural axiom set leads logically to a testable stable macro employment law, a.k.a. Phillips curve (2012; 2015), which is entirely FREE of brain-dead green cheese assumptions like constrained optimization, rational expectations, equilibrium, etc.
Not by any stretch of the imagination did Lucas/Sargent/Prescott remade macroeconomics — they pulpified it further.
You say: “But few would argue that it [old macro] needed replacing.” Yes, but some are quite sure that new pulp economics needs replacing even more urgently.
Kakarot-Handtke, E. (2012). Keynes’s Employment Function and the Gratuitous Phillips Curve Desaster. SSRN Working Paper Series, 2130421: 1–19. URL
Kakarot-Handtke, E. (2015). Major Defects of the Market Economy. SSRN Working Paper Series, 2624350: 1–40. URL http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2624350.
There he goes again, that scourge of the After-Keynesians, good old Egmont, spouting accounting identities as being "axioms." What he did not tell you in this post is that he claims these accounting identities imply that inventories never change and that the most important issue of all, that he thinks nobody is aware of, is that firms invest out of retained earnings. Wow!Delete
Econ101, a crash course for Barkley RosserDelete
(i) Economics is a failed science.
(ii) The fault of a theory lies in most cases in the premises = axiomatic foundations and not in the logical derivation of the implications.
(iii) Both, the Walrasian axioms (microfoundations) and the Keynesian axioms (macrofoundations) are provably false and have to be replaced.
(iv) When the underlying theory is false empirical testing in most cases degenerates to an inconclusive curve fitting exercise.*
(v) Therefore, nothing less than a paradigm shift will do, i.e. a replacement of the axiomatic foundations of the accustomed approaches.
(vi) The correct macrofoundations as enumerated above consist of nominal AND real variables. Therefore, they are NO accounting identities, which consist of nominal variables alone (BR’s 1st idiocy).
(vii) The structural axiom set contains output O and quantity sold X. The difference is the change of inventory. Obviously, there is NO claim in the premises that ‘inventories never change’ (BR’s 2nd idiocy).
(viii) From the elementary structural axioms as enumerated in the post above follows that the profit theory is false since Adam Smith.** It does NOT follow that nobody is aware ‘that firms invest out of retained earnings’ (BR’s 3nd idiocy).
The main conclusion is: as long as economists do not get the structural macroeconomic relationships between the foundational concepts profit/income right they will not get above the proto-scientific level or what Feynman famously called cargo cult science.***
For the specific issue under discussion, Longtooth’s summary is spot on: “Parenthetically I believe the highly subjective ‘ease of doing business’ data points used to show deregulation can improve GDP is about as unscientific as it gets. Plotting a highly subjective composite of highly subjective terms against an objective variable is laughable.”
It is more than laughable: for economics as proto-science and economists as incompetent scientists there is no alternative to the expulsion from the sciences.
* See also post ‘From mathiness to empiriness: forget it!’ http://axecorg.blogspot.de/2016/04/from-mathiness-to-empiriness-forget-it.html
** See paper ‘How the Intelligent Non-Economist Can Refute Every Economist Hands Down’ http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2705395
*** See post ‘A science without scientists’ http://axecorg.blogspot.de/2016/05/a-science-without-scientists.html
It would be fun to see the reaction of someone like Acemoglu after hearing that some guy is being criticized for suggesting (in an op-ed) that institutions cause growth and no the other way around. Imagine the laughs of the crowd in the NBER growth meetings after some guy comes screaming that all empirical growth literature is wrong because they are fitting linear relationships instead of cubic ones.ReplyDelete
What this debate really shows is that some guys get a PhD in Economics without ever running a growth regression or understanding the difference between econometrics and statistics. Unfortunately, this same guys have probably been so very misled by so very many things in life that they have no problem on talking with authority about things they have no clue about.
All the microfoundations trash to suggest that easy-to do business is at its optimal level based on a cubic trend. Magisterial ass-licking.ReplyDelete
Wait, what? Noah wasn't arguing THAT easy-to-do is at its optimal level and justifying it with a cubic trend. He said that a cubic trend WOULD offer that justification to the same extent as Cochrane's linear trend justifies destroying regulations. But his bigger point -- indeed, the whole point -- is that the graph doesn't really justify either, because it's so ridiculously sensitive to which one you pick.Delete
off topic but you mentioned the Phillips curve in the context of discussion of Cochrane. The idea that nonsense like Cochrane's was challenged by Lucas is fiction.ReplyDelete
Lucas added two things. One is that when considering the effect of policy on expectations (which all the people he criticized did) one must assume that agents know the joint probability distribution of everything that has and might ever happen. And second, that completely implausible assumptions about tastes and technology are OK because reasons.
You have explained very well why Lucas's alleged methodological contribution is worse than worthless.
Also this post is brilliant as is Brad's post.ReplyDelete
I read you, Cochrane, and Delong on a semi regular basis. I find each of you to be smart, insightful, and thought provoking, even though your disagreements imply you cannot all be right on every issue.
One thing I observe is the tendency (and I have to say this happens more with left wing economists, maybe taking a page out of Krugman's playbook) to respond to the stupidest version of each other's arguments. Here it seems obvious to me from reading Cochrane's original WSJ piece that he is not literally saying we should cut all regulations to 0 and get to 400k per capita gdp. It is true that he does this extrapolation in his chart and it is a bit silly, and he should have been clearer in the article about the dangers of extrapolating far outside the range of observed data. But the purpose of the chart is simply to open readers' minds to the possibility that we can achieve higher growth by deregulating. Furthermore, Delong's curve fit is far dumber because it essentially posits a sigmoid shape based on one or two data points (Singapore) instead of just using the simpler log linear fit that the bulk of the data implies. Moreover Delong unlike Cochrane can't offer an economic justification for his curve shape; it would be interesting for instance if he took a closer look at what is happening in well known pro growth environments like Singapore and Hong Kong posited an idea as to why, past a certain point, deregulation can have negative effects. Absent something along those lines I think the right prior ought to be much closer to Cochrane's view that deregulation always improves growth, even if we are unwilling to fully endorse the exponential implications of his fit.
But more generally, Cochrane's point that we should at least be thinking about the cumulative impacts of regulation on growth seems to me to be worth considering and not just dismissed out of hand by pointing out the silliness of his chart, as you and Delong are both doing. To limit the validity of Cochrane's argument to simply saying North Korea would be better off if they stopped shooting their businessmen is reductive to the point of stupidity. Your own colleague at Bloomberg, Megan Mccardle, wrote a piece last week discussing the possibility that individual regulations might be positive in isolation, but the combination of regulations past a certain point leads to paralysis. Luigi Zingales has done interesting work suggesting that a more heavily regulated state actually favors large corporations and enables greater rent seeking behavior, as those corporations a) seek to sculpt regulations to their advantage, and b) are better equipped for economy of scale reasons to comply with regulations. There is also a good paper by Brink Lindsey (http://www.cato.org/publications/white-paper/low-hanging-fruit-guarded-dragons-reforming-regressive-regulation-boost-us) that discusses the ways in which things like occupational licensing (also a form of regulation) mostly serve no useful purpose and exist instead to allow rent seeking. Cochrane's main point is that we should be taking a closer look at the impacts of the regulatory state instead of allowing it to expand unheeded; I wish you and Delong would actually respond to the substance of this instead of making snide remarks about 'intellectual quality'.
To read Cochrane's piece, knowing the extent of his training and the many other pieces he has written on the subject, and knowing the space constraints that op-ed writers operate under, and to then conclude that the entirety of the evidence for Cochrane's hypothesis that further deregulation may promote growth is a single regression, and the observation the North Korea kills some of its own people, is to be as uncharitable as humanly possible. Not to mention that, as an economist, Noah ought to be aware of the very large, academically rigorous literatures that stand behind the broad claims Cochrane makes, whether one ultimately agrees with said literatures or not.
Note that Krugman routinely posts graphs made in Excel with small sample sizes, debatable inclusion/exclusion decisions, non-obvious choices of transformations and variable definitions, and results from simple linear regressions, in support of his pre-existing positions (e.g., austerity is bad for growth). These posts are routinely ignored or cited uncritically by Noah and others. Why does Cochrane's similar exercise provoke such cheap shots and gleeful "excoriations", while Krugman's posts are A-OK? Is it maybe because Noah and others want to signal to their audience that they consider Krugman a good guy, and Cochrane a bad guy? Note the difference in tone, substance, and the level of engagement in Cochrane's response, and ask yourself who in this debate is being the unscholarly one.
". . . ask yourself who in this debate is being the unscholarly one."Delete
Okay, I did that. I'm going to go with the one who drew that line in the graph, published it on the op-ed page of the Wall Street Journal, and wrote "the evidence from the graph is strong."
Hello, fellow anonymous. You've posted a very thoughtful note, but IMHO, a wrong one.Delete
1) Main thrust of Prof. Cochrane's pleading is not that reducing regulation can improve growth somewhat, but that it can improve it drastically. The scale is important here because the argument goes more or less like this "sure, regulation is doing some fine things, but it is so detrimental to growth and growth is so good, that we better leave with the problems of unregulated capitalism". Small gains from small deregulation is not what animates people like Cochrane.
2) It seems that US is doing pretty well in not constraining business too much in the realm of what other countries are actually doing, not wild projections of what might have been in the ideal world. Keeping in mind that US peer group is not North Korea or even Singapore, but OCED countries and even better English speaking countries like Britain, Australia, or Canada. In this group, US is pretty lightly regulated country, which means that, practically speaking, it will be very hard to do much better by deregulating even more.
3) I would be surprised if either Krugman, DeLong, or Noah Smith are for keeping regulations for the purpose of increasing rent-seeking. They simply are skeptical that there is a lot to gain in terms of gdp from reducing them. In fact, it is possible that bad social consequences rather than economic ones that militate against much of the occupational licensing.
I read you, Cochrane, and Delong on a semi regular basis. I find each of you to be smart, insightful, and thought provoking, even though your disagreements imply you cannot all be right on every issue.Delete
But the purpose of the chart is simply to open readers' minds to the possibility that we can achieve higher growth by deregulating.
Large-scale deregulation has benefits and costs. By "suggesting" extremely overstated benefits while utterly ignoring costs, Cochrane encourages people to get the cost-benefit calculation wrong. To me, that's not good policy advice, and it should be actively fought and rebutted in the public sphere.
Furthermore, Delong's curve fit is far dumber because it essentially posits a sigmoid shape based on one or two data points (Singapore) instead of just using the simpler log linear fit that the bulk of the data implies.
If the sigmoid shape is "dumb", why don't the cubic and quartic fits look like the linear fit? DeLong did not force those curves to follow a sigmoid. Cubic and quartic polynomials can be monotonic. But with this data, the cubic and quartic polynomials chose to look like a sigmoid, to best fit the data you see on the graph. It wasn't DeLong.
I think the right prior ought to be much closer to Cochrane's view that deregulation always improves growth, even if we are unwilling to fully endorse the exponential implications of his fit.
To see my own priors, check out this "tweetstorm" (series of threaded tweets).
I wish you and Delong would actually respond to the substance of this instead of making snide remarks about 'intellectual quality'.
Well, check out this other thing I just wrote.
I think the conclusion to draw here is that you shouldn't try to do empirical work in a WSJ article. The dangers of over-interpreting reduced-form regression coefficients were known in the 1950s, long before Lucas thought about this in the context of the Phillips curve. The scatter plot that Cochrane is showing us is suggestive, and nothing more. Fitting any kind of curve to the points in that scatter plot is a pointless exercise, and it's even more pointless to argue, by fitting different polynomials to the data, that someone is more stupid than you are. A higher degree of pointlessness is writing a blog post about how the second guy seemingly "pulpified" the first one.ReplyDelete
Really? Look, I don't have a PhD in econ, or a professional reputation anything like yours, but this doesn't look complicated to me: a guy made a far-fetched, politically-motivated claim by stretching a bunch of data points beyond all reason, and somebody else called him on it. You can try to re-frame it as an abstract dispute over mathematical methods, but that's what is really going on here, and why the smackdown tone of the responses is not inappropriate.Delete
It seems a mistake to treat this as a simple academic debate. Articles like John's have a political impact, as do the responses (or lack thereof). It may not be of interest academically, but addressing it isn't completely useless.Delete
And honestly, even ignoring politics, it should probably be called out anyway, in the interest of professional integrity. You can't control what every PhD says, but it does reflect on the profession.
The titles are a bit sensational, but to be fair, it's not like they're pointing out a reasonable mistake. A bit of name and shaming might not be the worst thing ever.
Your perspective makes perfect sense if John had just made an honest mistake or couldn't get the point across in an article (which is admittedly not the best format), but I think anyone would be hard put to defend it as such. It's just flat out intentionally misleading.
Has no one plotted the fitted curve (the linear model) with confidence bands? Presumably this would highlight the greater uncertainty at 100 or 110..ReplyDelete
Why wouldn't the causality be reversed? In other words, why wouldn't a higher GDP indicate a greater ability for a country to afford the institutional infrastructure to make it possible to do business easily? (Also, the Lucas critique is garbage and relative to systems theory, completely wrong.)ReplyDelete
Would it be possible for you to respond to the substance of Cochrane's claims instead of focusing on this chart? I went back and read the column you link to where you say you are "highly skeptical" of Cochrane, but even in that article you seem to acknowledge the possibility that deregulation can lead to a large one/off increase in per capita gdp. Your arguments focusing on whether or not we can get us to 400k per capita gdp, or whether deregulation gets us a one/off level increase vs a growth rate increase, seem to me to be mostly besides the point. I think most Americans would be very happy if Cochrane's proposals led to a one/off level shift to something like 85k per capita gdp followed by an ongoing 2% growth rate. Why not address that point directly rather than focus on the most outlandish versions of his arguments?
Of course everyone would love 2% more growth. But if it's not possible (at least not with what Cochrane is suggesting, if it all), that isn't beside the point- that *is* the point.
The discussion isn't over whether people would want more growth (the answer is yes, of course). There also isn't any debate that regulation can cost some amount of lost GDP. Any reasonable person will nod their head and agree,so there's nothing left to discuss.
The debate is whether it's possible to deliver as much as Cochrane claims, and that's exactly what Noah addressed.
No one is disagreeing that regulation can cost us GDP. But there's magnitudes of difference between a one off bump, and something like compounding 2% extra growth. Perhaps it looks minor, but it really isn't- it's the difference receiving a 2% one off return on your investment, vs 2% growth interest annually. You need only look at your bank account to see the difference.
It's not beside the point, it *is* the point. When you're talking about different orders of magnitude, that matters. It's not a minor academic quibble over what's reasonable.
I don't understand why we are being so literal minded about this. From a practical standpoint, there are not 'magnitudes of difference' between a one/off level bump and extra compounding. Yes over a sufficiently long period an extra 1-2% of compounding will trump a large one/off level increase. But why focus at all on the most extreme version of Cochrane's claims? A one/off level increase due to deregulation that takes place over the next ten years and boosts per capita gdp by 20-50% would be an enormous win for everyone if it were achievable. Whether you want to call that a level change or 10 years of higher growth that then peters out seems to me to be irrelevant. You say the debate is about whether it's possible to deliver as much as Cochrane claims. I say it is about whether it's possible to deliver even a fraction of what Cochrane claims, as even that amount would be enormously valuable.Delete
Yes Cochrane's own writings on the subject lend themselves to hyperbole, and I suppose there is some marginal value in pointing out the absurdity of his extrapolation. But surely we can be charitable here and focus on the meat of what he is saying. I would much rather see Delong and Noah address where they think per capita gdp would go if we substantially deregulate, instead of just saying the 400k number is wrong.
"Would it be possible for you to respond to the substance of Cochrane's claims instead of focusing on this chart?"Delete
But those are one and the same. A massive deregulatory push would undoubtedly have costs as well as benefits. By obviously overstating the benefits, Cochrane makes a reasonable cost-benefit calculation impossible.
Also, see this: http://noahpinionblog.blogspot.com/2016/05/regulation-and-growth.htmlDelete
Shouldn't the Y-axis be median household income instead of GDP per capita? Doesn't GDP per capita overweight the incomes of the wealthy and allow for greater inequality to skew the results?ReplyDelete
In a perfect world, yes. But you need to have information regarding income at the Household Level. Household surveys are quite difficult to carry out and many countries don't do them, or don't do them every year. GDP per capita is still a good proxy, and most countries already have a system of National Accounts, and provide information on this regard on a more timely matter. That is why GDP per capita is more frequently used.Delete
The economists debating this topic are all missing the most important issues:ReplyDelete
a) 99% Of WSJ article readers don't have a clue about logs relations to non-log. What they see in Cochrane's chart looks entirely reasonable to them because they haven't a clue about how it really extrapolates in the standard values they are familiar with. But nobody has yet published in the same or any other major newspaper (NYT? WP?) a different form of the chart that might give readers of the WSJ article some reason to suspect Cochrane's chart is leading them to believe in a fantasy rather than a semblance of reality.
b) Cochrane is proposing a fantasy in his WSJ article. It is in fact a pure propaganda piece (intent to mislead). He published intentionally in the WSJ rather than an economics journal or paper for economists to debate for the express purpose of story telling to the unwitting public for use to promote financial deregulation to their equally (even more?) clueless representatives in congress.
The issue is therefore rightfully a public relations propaganda campaign and it is only fitting that Delong or others publish in the same WSJ (or at least in the NYT or WP) the counter propaganda (which can be as equally misleading as Cochrane's was, btw).
Parenthetically I believe the highly subjective "ease of doing business" data points used to show deregulation can improve GDP is about as unscientific as it gets. Plotting a highly subjective composite of highly subjective terms against an objective variable is laughable. Yet, economists keep doing this dumb stuff and still want to be viewed as "scientific". Get a clue.
Not an economist, but from my point of view, it appears that Cochrane was attempting to make a social/political point (a lassaiz faire one)whereas the critics are attempting to make an academic/economics one. Cochrane produced a rocket launch graph because escape speed is important when attempting to influence anyone or go into space. And the critics, correctly in my view, said 'whoa buddy dial it back some.'ReplyDelete
Should someone write a full page response on how this article is misleading in forgetting to mention that there are other things wrong with North Korea in addition to shooting business-people? But sure win the race to bottom (that Cochrane started).Delete
Fitting logarithms of data without the appropriate weighing and some theoretical reason for logarithms to be the right quantity is just bad, because logarithms strongly deemphasize large numbers and strongly emphasize small numbers.ReplyDelete
Double a quantity? Just add the logarithm of two. It doesn't matter whether we're talking about 1, 100, 100000, etc. Log x -> infinity as x -> infinity is literally true, but has to be taken with a grain of salt in practice. For example, log_10 of Avogadro's number is about 23.8.