Carmen Reinhart and Kenneth Rogoff have posted an open letter to Paul Krugman to try to correct some of the misrepresentations of their scholarship that continue to be repeated by people who should know better.
Reinhart and Rogoff’s letter begins by making some points about the critique of their work by Thomas Herndon, Michael Ash, and Robert Pollin that should be very familiar to readers of Econbrowser. Among these is the observation that, even if one accepted all of the particular methodological criticisms raised by Herndon and co-authors (which I personally do not), the results are in fact virtually identical to those reported in a follow-up study by Reinhart, Reinhart, and Rogoff published in 2012, which study interestingly was never mentioned in Herndon, et. al.’s critique.
Apart from these technical issues through which we have previously waded in detail ([1], [2], [3], [4]), Reinhart and Rogoff’s open letter also addresses some of the even more irresponsible personal attacks that have been leveled against them.
An example of these is the claim most recently repeated by Paul Krugman in the
New York Review of Books that
Reinhart and Rogoff had not made their data widely available– and researchers working with seemingly comparable data hadn’t been able to reproduce their results.
But you can’t get away with this kind of careless mudslinging in the internet age. Reinhart and Rogoff used the Wayback Machine to get a copy of Carmen Reinhart’s web page exactly as it appeared to the world in October 2010. Try it yourself– you can click on the links to download whatever spreadsheet you like, or you can use the Wayback Machine to find versions that were publicly reported as of various historical dates (both at Reinhart’s own page, as well as the database’s subsequent migration to the public website for their book).
Another often-repeated and equally groundless charge has been that Reinhart and Rogoff failed to distance themselves from politicians and pundits who maintained that Reinhart and Rogoff’s research lent support to austerity measures as the preferred short-term fiscal policy. You can find countless vociferous repetitions of this false allegation in the comment sections to items [1]-[4] linked above. But again it is straightforward to use internet archives (as Reinhart and Rogoff now have) to uncover the actual public record of what they did and did not say at various historical dates. For example, in an interview with the BBC in July 2011 Rogoff stated:
The current strategy that calls for years of austerity and recession in the periphery countries is just not tenable.
Or see this reported summary of Carmen Reinhart’s position in The Guardian in April 2010:
The world’s best known female economist has warned cutting the deficit the Tory way would send the UK back into recession.
You will find links to literally dozens more examples of statements like these in the Reinhart-Rogoff open letter.
Not that facts matter to those who took up the cudgels. The smear campaign had only one purpose– to distract people from thinking clearly about the consequences of the current high debt loads of many of the world’s countries. On this fundamental question you can also find much to help set the record straight in Reinhart and Rogoff’s open letter.
But be forewarned– there are many folks out there who still think that if that if they just keep shouting lies and nonsense loudly enough they can prevent you from hearing Reinhart and Rogoff’s true message.
Very powerful stuff. Thank you, Professor, for posting this.
And the true message is? The debt is over 90%! We are all going to die!!!!
This post is very uneven and subjective.
This is not a “smear campaign”, it is a spirited debate over issues that guide policy that affect millions of lives. R-R have had ample time to admit they were wrong and put an end to it. Instead, they have chosen to defend and-or cloud the issue at every opportunity.
And, just listen to yourself: “Not that facts matter to those who took up the cudgels. The smear campaign had only one purpose– to distract people from thinking clearly about the consequences of the current high debt loads of many of the world’s countries.”
Utter nonsense.
It really damages your credibility to defend them in this way. And–no–I do not carry a brief for Krugman.
The letter is just stunning. It’s not that it’s just a window to see their warped perception of reality but it’s also not even written well.
How are these people at Harvard again?
Forget the fact that nothing PK did was ‘spectacularly uncivil’. Please, give these people a lesson in adverb abuse. I feel terrible for these adverbs. It’s just not fair to them and their rightful place in the English language. Poor little adverbs.
And well it has to be said…
Just because one’s feelings are hurt does not mean that the one who prompted these feelings was uncivil. Pointing out mistakes and deep failings, in method, conduct and theory, is not uncivil. It is ‘spectacularly’ impossible to take these people seriously after this ‘spectacularly’ misleading accusation of ‘uncivil’ behavior. This focus on (imagined) behavior does little more than lay bare their true aim with the letter: to build a new protective sheath for their bruised egos.
I’m sorry, but this just doesn’t cut it. I don’t understand their claim that “The Herndon, Ash and Pollin paper, using a different methodology, reinforces our core result that high levels of debt are associated with lower growth.” The HAP paper primarily concentrated on the post-war results, not the 200 year result in their 2012 paper. Does anyone really care about or take seriously debt/GDP numbers from George Washington’s first term? So America in 1791 was an “advanced” economy??? Looking out over 200 years is especially strange given that one of the HAP findings was that the relationship between high debt and low growth seemed to be degrading over time.
Reinhart claims that debt events over 90% are rare. Well, I don’t call 10% of the time “rare.” But notice that with this sleight of hand she has implicitly used exactly the same weighting methodology as HAP.
Reinhart also discusses a quote from Sen. Coburn’s book. Except that the quote she cites is not the one that captured everyone’s attention. And it doesn’t address the charge that in open testimony she did not make any effort to correct a couple of congress critters who attempted to resummarize her testimony.
She also misrepresents Krugman’s position regarding fiscal expansion in Europe’s periphery. His position is that fiscal stimulus is not a viable option there, which is why he has called for Germany to accept higher inflation. Reinhart’s claim that this won’t work because the ECB may very well try to contain any emerging German inflation may well be true…but there is no reason the ECB must behave stupidly.
Krugman did not say that R&R didn’t make their data available; he said that they did not make it widely available. One should not have to go to an obscure and unobvious place such as a personal website within a tab listing a professor’s course offerings. What’s wrong with a simple “Data” tab? That’s what most academics do. And in the footnotes of their paper they should have cited a website for their data instead of misleading readers into thinking it came directly from the cited data sources. And of course, the data were not available at the time their paper was making its biggest splash. But the biggest complaint coming from scholars wasn’t access to the data, but the difference between how they actually weighted the data and the way the weighting was described in their paper.
The quote from the Fareed Zakaria interview is shockingly dishonest and out of context. In fact, I linked to that very same interview showing the full context. Yes, Rogoff did say he would support infrastructure spending…but why didn’t she include Rogoff’s subsequent comment where he completely reverses himself? As they like to say…roll the tape!
Finally, Reinhart continues to perpetuate the strawman argument that critics claim higher debt/GDP ratios are something we don’t have to worry about. No one is saying that. The slam against R&R is that they presented and encouraged and reified (there’s a good Marxist word) the 90% “threshold” as a sharp break point. As R&R concede in a footnote in a paper behind a paywall, the 90% threshold was just an artifact of the way IMF organized data. So why not just come out and say the 90% number isn’t a firm threshold? And then when they do admit that, don’t try and walk it back. And for God’s sake, please have the courage to try and correct Rep. Paul Ryan with as much vigor as they seem to feel towards correcting Paul Krugman.
JDH,
I do agree that the smear campaign is designed to obscure R&R’s message. But I think Krugman has deeper reasons for engaging in it as well.
Krugman’s false and personal attacks were designed to impugn the credibility and reputations of R&R in order to damage their ability to influence policy makers. Krugman of course knows that these charges are false yet he has refused to retract them and apologize. What accounts for this? Is Krugman so ideologically obsessed that he can justify these tactics in his mind?
A few summers ago I was on the beach trying to relax. Unfortunately, I was also reading “The Conscience of a Liberal” and as each minute went by was getting less relaxed. I explained Krugman’s arguments to my wife-who knows nothing about economics-and she took a look at the book. Her impression was that Krugman is preaching to the choir; he’s making arguments and writing in a style that will only influence people who already agree with him. She predicted he’d have little real influence on policy makers.
I think Krugman has come to realize that himself. Despite his efforts on his blog, in his most current book, and in the NYT magazine, he has not been able to convince anyone who matters that they should adopt what Jeffrey Sachs has called “crude Keynesianism.” Instead, he has been dismayed to see the very real influence of R&R.
R&R have succeeded for three reasons: 1) They are doing important relevant research on today’s problems today, rather than relying on their reputations for what they did 20 years ago; 2) they are making arguments to both sides rather than treating their own side as right and the other as a pack of idiots; and 3) they explain their technical research clearly so that the non-technical reader/policy maker can follow it.
Krugman does none of this. There is a lot of sound and fury on his blog and is acolytes love the personal attacks and innuendo; but no one that matters really cares. It should not be any surprise that an angry, jealous, and bitter Krugman lashes out against R&R on his blog and in the pages of the NYT magazine.
What R&R failed to do is denounce the people who were using their data incorrectly. So for example, even though Paul Ryan was citing their results incorrectly in his budget plan, R&R found Ryan useful in their concern about debt and so remained silent. In other words, for R&R the end justified the deceitful means.
R&R were at the front in the lead up to the bitter debt limit debate in 2011. Here are things they said at the time:
“even advanced economies hit a ceiling where the pressure of rising borrowing costs forces policy makers to increase tax rates and cut government spending”
“we find that very high debt levels of 90% of GDP are a long-term secular drag on economic growth”
“If it is not risky to hit the 90 percent threshold, we would expect a higher incidence.”
Stuff like this added fuel to the fire for proclamations of an immediate debt crisis which required immediate austerity.
Now, you might say they R&R were just useful idiots exploited by dishonest politicians, but I think it is really the other way around. R&R were dishonestly taking advantage of useful idiot politicians in order to promote their genuine concerns about debt. The end justified the means.
Excellent post Jim Hamilton.
dilbert: And the true message is: (1) The US will labor under the burden of reducing debt until perhaps 2030 or beyond; (2) real growth may be degraded on the order of a percentage point from its pre-crisis level of 2½%; (3) rolling fiscal cliffs will be the order of the day each time the debt ceiling needs to be raised as the necessity to reduce deficits clashes with society’s desire to maintain entitlement spending; (4) fear generated by ongoing Social Security cutbacks and income tax increases will in coming years impel households to save more out of current income to provide for retirement years [read David Stockman’s book]; (5) interest rates may not normalize for years as central banks strive to get inflation above 3% so as default on Treasury debt via erosion of the purchasing power of fiat money; (6) the artificial pricing results of the regime of negative real interest rates will show up more and more as having malinvested society’s physical capital; (7) uncertainty will remain high and financial shocks from abroad will be more frequent; (8) recessions will be more like those of 1974, 1982, and 2008 since traditional policy has lost its potency; (9) preservation of purchasing power will become a dominant social theme in the not too distant future.
Dogbert: “And the true message is? The debt is over 90%!”
The federal public debt to GDP is the latecomer to the big debt party. The overarching constraint is the TOTAL local, state, federal gov’t AND corporate and household debt to real GDP per capita and wages.
Total corporate debt to GDP (differential order of exponential magnitude [DOEM] occurred from the early ’80s to ’08):
http://research.stlouisfed.org/fredgraph.png?g=iSV
Non-financial corporate debt to GDP (approaching the DEOM since the early 1950s, increasingly debt used for stock buybacks of late):
http://research.stlouisfed.org/fredgraph.png?g=iSW
Total local, state, and federal gov’t debt to GDP, excluding local and state employee retirement funds (the total debt has reached the DOEM since the ’70s-’80s):
http://research.stlouisfed.org/fredgraph.png?g=iSX
Total household, corporate, and gov’t debt to GDP (reached the DOEM in ’08 from 1950s-80s):
http://research.stlouisfed.org/fredgraph.png?g=iSZ
Total debt to GDP and real GDP per capita:
http://research.stlouisfed.org/fredgraph.png?g=iT1
Total debt to wages and salaries:
http://research.stlouisfed.org/fredgraph.png?g=iT2
Household debt to wages and salaries:
http://research.stlouisfed.org/fredgraph.png?g=iT3
The US reached the jubilee threshold of TOTAL debt to real GDP per capita and wages in ’08 (cumulative exponential order of magnitude of the differential rate of growth of debt to wages and GDP) from the early ’80s (1950s in terms of non-financial corporate debt). In order for real GDP per capita to have continued to grow, debt would have had to accelerate at a differnential super-exponential rate to GDP, which would eventually have been constrained by debt service costs.
Similarly for local, state, and federal gov’t debt. Debt service as a share of receipts for TOTAL public debt is on course to become a hard constraint as soon as ’17-’20, and sooner should a recession occur with the potential for deficits to reach 100% of falling receipts.
M2 + institutional MMFs + large time deposits less the difference of bank cash assets and vault cash:
http://research.stlouisfed.org/fredgraph.png?g=iT8
Velocity of same to GDP less total gov’t spending:
http://research.stlouisfed.org/fredgraph.png?g=iT9
Annual change of same:
http://research.stlouisfed.org/fredgraph.png?g=iTa
In a debt-money system, without growth of debt-money, and the ability to service the growing debt-money, there is no growth of real GDP per capita. Growing supply of debt-money no longer creates its own demand growth per capita. Demand must hereafter come from increasing wages to GDP (and to returns to capital) and falling debt to wages and GDP. However, US, EU, and Japanese wages cannot compete with the low wages of developing countries’ billions of workers. Neither can US, EU, and Japanese workers compete with low-cost developing workers’ wages AND the accelerating effects of automation that will constrain the wages even low-wage workers around the world.
http://research.stlouisfed.org/fredgraph.png?g=iT4
Running large federal deficits/GDP and increasing total gov’t debt to GDP by 50% in four years at best prevented nominal GDP from contracting further and the U rate from rising well beyond 10%, but it did not result in growth of debt-money supply after hoarded bank cash assets nor of real GDP per capita to date.
But the deficit spending will maintain only until debt service to receipts on total gov’t debt constrains the ability to fund spending less debt service.
Therefore, the issue is MUCH LARGER than US federal debt to GDP. The developed world is laboring under an unprecedented private and public debt burden and asset bubbles, the latter of which are held primarily by the top 1-10% of households as claims against wages, profits, and gov’t receipts in perpetuity.
Then there is the cost of liquid fossil fuels to GDP at $95-$120 oil that is further constraining growth of business investment, production, and after-tax spending for the bottom 90%.
Add still further the incremental costs of Obamacare, which effectively subsidizes the bloated, obscenely costly insurance and medical services cartels.
Thus, energy and medical services costs, as well as unprecedented debt-money claims against wages, profits, and gov’t receipts, concentrated to the top 1-10%, are preventing growth of real GDP per capita. Adding more public debt to GDP and trillions of bank reserves to banks’ balance sheets is only exacerbating asset bubbles and energy and debt-money service costs to the private sector.
An economy is not strong and resilient, and thus a society is not long to be a civil one, when 100% of wages, profits, and gov’t receipts are pledged in perpetuity to the top 0.1-1% to 10% of households while the bottom 90% must compete with billions of low-wage workers the world over for subsistence AFTER payroll taxes, debt service, and the rising cost of medical insurance services and energy.
Unless debt to GDP and wages can be reduced 40-50%, and/or wages rise and/or prices of goods and services fall proportionally to wages, real GDP per capita will not grow and the costs of private and public debt service, energy, and medical services will continue to constrain further real GDP per capita.
R&R could salvage further their reputations by writing a sequel book to include PRIVATE debt to GDP AND place the proper focus on the inherent flaws and vulnerabilities in the debt-money-based fractional reserve banking system and its contribution to credit bubbles and busts, including the debilitating wealth and income concentration that historically accompanies secular debt cycles as we have experienced since the early ’80s.
The people who defend Krugman need to unsubscribe from their Rules for Radicals mindset. Krugman has long been a joke to the majority of economists precisely bc of episodes like this. He lies. He argues for actions that will clearly have negative consequences, according to the majority of academic Econ literature, and then tells the world that only he and a few other genius (i.e. far left leaning) economists have the facts straight.
Daniel Kuehn is much more on point than this post.
Here is the “true” public record:
“Our empirical research on the history of financial crises and the relationship between growth and public liabilities supports the view that current debt trajectories are a risk to long-term growth and stability, with many advanced economies already reaching or exceeding the important marker of 90 percent of GDP.”
Reinhard and Rogoff on Bloomberg (see link)
http://www.bloomberg.com/news/2011-07-14/too-much-debt-means-economy-can-t-grow-commentary-by-reinhart-and-rogoff.html
As everyone knows, the elephant in the room is the 90% claim.
What, Professor, do you say about that claim? Is this their “true message”
Is this the truth?
Quasi-genteel smears are still smears. I understand your frustration, even if some might say it smacks of unexamined bias, but one cannot reasonably deplore open criticism while making snide personal insinuations. As someone else would say, This isn’t bean ball. R&R enthusiastically entered the political forum and continue to participate in it. Criticism in that forum is not limited to throat clearing and “tut-tut, old chap.”
Shouting “deficit” in a global recession is not excused by whispering more measured messages. There is a false equivalence in pushing on opposite sides of the leaning tower of Piza.
Having been through many academic discussions that were a lot more brutal and personal than that between Krugman and Reinhart/Rogoff, I think that R&R are whining.
I also think that Professor Hamilton is making a mistake by defending them in the manner that he has.
I should point out that it was Thomas Herndon who said he could not access the data. So accusing Krugman of mudslinging on this point is ridiculous.
For that matter, as I understand it, there have been updates to the data over time. Therefore, making “the data” available doesn’t help unless it includes a clear indication of what revision was used for a particular calculation. Most of us are hesitant to judge work unless we are certain we are using the right revision, so it matters whether academics answer their e-mail.
The rancor in this argument comes down to “what does This Time is Different say”? No one disagrees with the general proposition that debt reduces growth. But two questions are important: By how much is growth reduced? and What is the consequence of not spending at a time of crisis in a specific instance? It could, after all, be even lower future growth, or (as in the case of the US in WW II) outcomes even worse.
What we have in the Krugman vs. R&R debate is a collision of two spheres, the public and the academic. In the academic sphere, R&R are guilty of a little sloppiness, a little non-responsiveness, and some questionable assumptions. These are fairly common sins, although few academics will admit to it. In the public sphere, there have been large and bad consequences to the idea of a debt cliff.
Had Krugman responded in the academic sphere, by publishing a rebuttal in a journal, it would have vanished like foam on the sea. But he’s perhaps the most influential public intellectual of our day. He responded to the public consequences of This Time is Different in the public sphere. But R&R are public figures. They should expect this.
Again, I think R&R are whining. We all do at one time or another, but it’s a mistake to do so publicly.
Here’s Dean Baker from July 4, 2010:
“Actually, much of the fodder for the public debate has been over a separate paper that claims that economies perform more poorly once their debt to GDP ratio exceeds 90 percent. Mr Rogoff and Ms. Reinhart have declined to adhere to standard ethics within the economics profession and have refused to share the data on which they base their conclusion with other researchers.”
http://www.cepr.net/index.php/blogs/beat-the-press/not-following-professional-ethics-matters-also
By July of 2010 the damage had already been done.
From the Herndon et. al. paper:
“On their website, Reinhart and Rogoff
provide public access to country historical data for public debt and GDP growth in spreadsheets with complete source documentation. However, the spreadsheets do not include guidance on the exact data series, years, and methods used in RR. We were unable to replicate the RR results from the publicly available country spreadsheet data although our initial results from the publicly available data closely resemble the results we ultimately present as correct. Reinhart and Rogo
ff kindly provided us with the working spreadsheet from the RR analysis. With the working spreadsheet, we were able to approximate closely the published RR results.”
One of the more revealing aspects of this affair is just how bad at economists are at data analysis.
I make my living doing it. I have to be honest: had R&R made the same mistake in any business they would have been fired. This has been defending as “a coding error” – but that is nonsense. It is the hallmark of sloppy work.
Here is what happens in the real world jdh – when you make as basic a mistake as they did you get dismissed – because people conclude that they cannot trust your analysis. Your credibility is shot. Destroyed. You don’t get a second chance.
If a lawyer made the same mistake he would lose clients. Any third year associate at an accounting firm would be fired.
What I don’t think R&R’s defenders get is just how badly disconnected they are from the real world – where mistakes like this destroy reputations.
I could go further and note the spinning they have done about the advice they have given – it is clear at best that they are shading the truth about the presentation they gave to the Senate in 2010, but the bottom line is you seem shocked.
But the truth is I just don’t think you get that mistakes like the one R&R would have made simply would not be tolerated. The fact that you defend it ultimately says more about the line between academia and business then anything else.
The Grad Student Who Busted Reinhart & Rogoff Explains How Open They Really Were With Their Data
Joe Weisenthal 5/26/2013
http://www.businessinsider.com/thomas-herndon-on-reinhart-and-rogoffs-data-availability-2013-5
. . .
Via email, we asked Herndon his precise take on what happened, and what he makes of their claim that their data was always made available.
Here’s his full email response to our query:
The question has a few different parts and I don’t want to assign wrong doing to any parties. Your memory is generally accurate in that before I got the spreadsheet, it was difficult for me to try to piece together the data from the publicly available sources, and that I spent a decent amount of time trying to. Their data for the GITD paper consisted of two variables: debt/GDP and real GDP growth.
As they say in their letter, they did make their debt/GDP data available on their website, and they linked to the Angus Maddison dataset for growth, which they got much of their growth data from.
However, they also got growth data from numerous other sources, and they did not make the spreadsheet where they “put it all together” available (and actually still haven’t), so that their actual calculations could be checked, and that you could see which sources of growth data they used. During the fall semester, I spent a bit of time trying to figure out what was the actual growth data they used and how they took the average, because some of my summary statistics weren’t matching, but in the end after trying a number of different sources I just ended up mainly using the Maddison data they cited on their website. So I guess much of the judgment depends on how strict standards you want to have for data release. I tend to think that as a general principle, where possible original scripts, calculations, data, and spreadsheets, etc. should all be made available. It saves a lot of time for replication, and can be a good way to make sure spreadsheet errors or typos are found early. However, there is a whole host of complicating factors and less than ideal conditions that might make this difficult, so there is room for reasonable people to disagree about what “complete” means regarding the release of data.
Another complication to the story is that I think Krugman might be referring to comments by Dean Baker (2010) here as well as Bivens and Irons (2010), here. From Baker, Bivens and Irons their comments, it seems they tried to get the data (I think Dean Baker tried, and then Bivens and Irons who were co-authors) from Reinhart and Rogoff, but were unable to get it, and I could be wrong, but I think that this might be what Krugman is referring too. That being said, the authors did write these in early 2010, and RR did make half the dataset (the debt/GDP data) available a little later in the year. So this is another complication that I’m not too familiar with, and you would probably need to ask Krugman, Baker, Bivens and Irons about their experiences.
Also, you’ve probably seen it but Brad Delong has a good post today here critiquing some of the arguments RR make in their jep paper that is worth a read.
Dr Hamilton,
putting aside the rest for a moment, Rogoff’s latest on europe was prefaced by an obvious straw man argument about “muddle headed keynesians”. And also without any real clarity on the point presented a debt forgiveness approach as in opposition to a core stimulus approach rather than seeing it as at least somewhat complementary. I don’t get why the strawman argument was allowable for an academic– you would never teach your students to do that, nor would professor Rogoff I hope.
juris debtor, I don’t see the text of Herndon’s e-mail at Business Insider. But even without it, this narrows the issue of what data was really available. One can easily see how much of a pain in the rear it would have been for someone trying to replicate the results to do so based on growth data drawn from several sources without any clear specifications as to what comes from where. My conclusion is that no, Reinhart and Rogoff really did not make the data widely available.
Dear Dr Hamilton
Have R-R apologized for the higher unemployment resulting from their testimony in Congress?
Have they shown remorse about the “spectacularly uncivil” hardship they have helped launch on thousands of their countrymen?
Until they do that, both them and you should really, really, not talk about unprofessional conduct. Economists indeed.
Regardless of whether R-R have legitimate complaints about some of the details of the criticism of their work and their presentation of it, it seems to me that their defense doesn’t really address the central point of the critique.
Using their own data, it is clear there is no particular significance to the 90% level, but they made the 90% number a benchmark and referred to it repeatedly.
They also failed to establish the causal direction of the negative relationship between debt/GDP and growth, but made claims that only make sense if you believe you know that direction.
I don’t think that they are bad economists, or bad people, or ideologues bent on wreaking austerian havoc on the peoples of the world, but I think they made two main mistakes. One was in writing a paper that didn’t establish what it seemed to be claiming, and one was promoting that paper to support positions which their data only vaguely supported (the slope of the debt/GDP vs. growth line is quite shallow). I think they may now be making a third mistake by not simply admitting that they made the first two mistakes.
Clearly Prof. Hamilton thinks that they have been unfairly criticized, and feels it is appropriate to defend them, but I can’t agree that all the criticism has been unfair. It would be interesting to know if he thinks any of the criticism is justified, beyond the minor problem of the presumably unintentional errors in the spreadsheet.
Krugman has no business smearing R&R given his terrible forecasting and policy record and it’s only the ignorance of his many followers that lets him get away with it. After all, Krugman has been wrong on just about every big macro question since the early 1980s.
If you go back to the time that he served on Reagan’s Council of Economic Advisors, he was completely wrong about the central question of that time–whether inflation had been defeated. Krugman (along with Lawrence Summers) sent a memo to Martin Feldstein in which they predicted that inflation will accelerate significantly, available here:
http://www.pkarchive.org/economy/KrugmanSummersCEAInflation1983Recovery090982.pdf
Inflation, as everyone knows, collapsed. I’m sure that Feldstein wisely put that memo in the circular file where it belonged.
Then, in January of 1991 Krugman predicted that the economy was sliding into a recession but that it would be short-lived. He expected that real growth would decline 2% in 1991 but would increase 5% in 1992 and 3% in 1993. But by the time of Krugman’s prediction, the economy had already been in a recession since July of 1990 that would soon end in March of 1991. Rather than declining 2% in 1991, real growth increased 2.6% in 1991, 3.4% in 1992 and 3.5% in 1993. If you had listened to him in early 1991, you would have very disappointed about how his prediction turned out that year.
Then, we go to the next recession, March 2001 to November 2001. In August 2002, he wrote a NYT article called “Dubya’s Double Dip?” in which he suggested that a double dip is likely. However, in 2003, real growth was 4%. Completely wrong again. Moreover, he issued disastrous policy advice in that article. Here’s the quote from the article available here:
http://www.nytimes.com/2002/08/02/opinion/dubya-s-double-dip.html
“The basic point is that the recession of 2001 wasn’t a typical postwar slump, brought on when an inflation-fighting Fed raises interest rates and easily ended by a snapback in housing and consumer spending when the Fed brings rates back down again. This was a prewar-style recession, a morning after brought on by irrational exuberance. To fight this recession the Fed needs more than a snapback; it needs soaring household spending to offset moribund business investment. And to do that, as Paul McCulley of Pimco put it, Alan Greenspan needs to create a housing bubble to replace the Nasdaq bubble.”
That’s right Krugman zombies out there. Krugman actually recommended that the Fed blow up a housing bubble, the same bubble that produced the financial crisis, the deep recessions around the world, and all that unemployment. That policy advice alone should disqualify Krugman from ever issuing any kind of macro opinion.
And now we have Paul Krugman smearing R&R. He’s doing it because he’s jealous of their influence but also because he knows justifiable concern about ever larger debt stands in the way of his desire to do more stimulus programs.
Given Krugman’s disastrous policy and forecasting record, why should anyone believe him on anything? In fact, people who matter don’t. But it’s a sad testimonial to the foolishness of Krugman’s many followers that R&R have to defend themselves against his scurrilous attacks. Krugman is fading into the obscurity he so richly deserves but these sorts of attacks revive attention on him, if only temporarily.
Bloomberg, July 2011, Rogoff and Reinhart: “Our empirical research on the history of financial crises and the relationship between growth and public liabilities supports the view that current debt trajectories are a risk to long-term growth and stability, with many advanced economies already reaching or exceeding the important marker of 90 percent of GDP. Nevertheless, many prominent public intellectuals continue to argue that debt phobia is wildly overblown.” Note: causation debt to low growth; Note: 90% theshold; Note: Krugman is right.
On the matters that Jim H. attempts to defend RR I think there are legitimate differences of opinion, although I think Jim a goes too far in his efforts at defense. I would say that on the matters of making data available and on correcting politicians distorting their message, reality lies between them and Krugman. They were not smeared, but some of the critical claims may be overdone.
However, on the two points that Krugman concentrates on, he is complete;y right, and Jim makes himself look bad by not recognizing this. One is the correlation/causation issue; the other is the threshold issue. RR simply do not give on these and Jim does not mention them, but evidence suggests if anything more cases are low growth to high debt/GDP than the opposite (but close call). And on the 90% threshold, this is one of the biggest frauds in a public economics debate in a long time, and they simply have not admitted this. This makes ideologues like Stryker here, who provided zero evidence or facts, not believable and looking complete silly.
Not being an economist, I feel a little bit like a lawyer or judge trying to see which opposing expert witness is most correct, and which expert witness is stretching the truth to make his case. I cannot judge for myself the technical aspects of the arguement. .. . I am simply not qualified to do so, as happens with most judges and lawyers when dealing with something of a technical nature.
Jdh, the thing you seem to be forgetting is that your take on this matter is not the only thing that I/people will read to make up my/their mind on who is most right. And in reading your post in comparison with the post issued by Delong as an example, i find yours to be signicantly more emotionally charged and accusatory, and his to be more analytical, putting the evidence out there for people to make up their own minds. Any guesses as to which one of these takes comes across as closer to the truth?
About how well r&r shared their data, I would suggest that you contact dean baker very soon and see his opinion about how well they shared their data. I would also assert that the better judge of how well data is shared would be by the person seeking the data, not the guy providing it. Krugman’s charge that you equate to often repeated mudslinging I believe is simply based on baker’s experience. Either baker is a liar or the charge has merit (in which case your condescending talk about checking the Internet way back machine ourselves makes you look even worse)!
Agarwal is spot on. And I couldn’t care less about Krugman or RR – I feel sorry for the people who are now suffering the consequences of unnecessarily high unemployment.
Charles II:
The text of the Herndon email has been removed from the original article. Not sure why.
Further proof that, to some, winning the argument is more important than being right. Truth isn’t the issue here anymore, it’s your interest in maintaining an investment in a failed ideology. Rogain and Braveheart committed nothing short of freshmen level errors, period. And their public statements confirm everything Krugman has claimed.
A number of Krugman’s apologists, taking their cue from the master himself, have tried to confuse the smear with the 90% threshold and the causation question. Of course Krugman wants to divert attention from the HAP smear that he encouraged and participated in by pretending that it is a an academic disagreement. That tactic seems to be working with at least some people. So let’s clearly differentiate the smear from the academic dispute.
The smear, started by HAP and continued by Krugman, is the insinuation that R&R are dishonest and incompetent researchers. The rap is that they had coding errors in their analysis which they combined with selective exclusion of data, both of which helped to make the case they wanted to make. Then R&R also added a non-standard estimation technique on top of all that, presumably to gin the results the way they wanted. However, they were caught by an intrepid graduate student who had to pry the data out of them to get to the truth.
However, R&R have acknowledged the insignificant coding error but decisively refuted the second point by showing that the data wasn’t available at the time for them to exclude. And JDH decisively refuted the third point in his discussion of fixed versus random effects. It’s a smear pure and simple. The fact that the graduate student went on the Colbert report to make fun of R&R should tell us all we need to know about their motives.
On the separate question of the 90% threshold and causation, Krugman is being his usual dishonest self. Why has Krugman never mentioned much less addressed the fact there already a number of academic studies that find a threshold around 90%? A number of commenters lament that they are not economists and can’t tell who is right. Well, don’t you think it is important for Krugman to tell you those studies supporting R&R exist and why they are wrong?
Moreover, if you fairly read Rogoff I don’t think you can say that Rogoff ever made such a strong claim about the 90% as do the papers who support him. As I understand him, he is saying that 90% is an important indicator of being in a high debt regime. The reason is that at that point real growth is about a percentage point lower. The important point that Rogoff has made, and that Krugman is silent on, is that the slow growth episodes last decades in which 1 percent lower growth accumulates into a substantial output loss. Rogoff has also made the point that causation runs both ways: slow growth produces higher debt and higher debt produces slower growth. The reason for that, in Rogoff’s view, is that the slow growth periods last too long for causation to go in one direction only.
Krugman never acknowledges Rogoff’s real arguments. He doesn’t mention the empirical papers supporting the 90% threshold or Rogoff’s points about the lengths of the slow growth episodes that R&R discovered in the empirical data.
Instead, Krugman smears R&R so that his readers will think them dishonest and incompetent. Then, he pretends that they have made assertions about the 90% thresholds and causation for which they have no evidence, assertions that Krugman can easily show to be false by the simple charts he posts in his blog. Krugman doesn’t tell you about the evidence favoring R&R that he surely knows about as a professional economist. Nor does he refute Rogoff’s real arguments. Instead he relies on the smear to soften up his ignorant and gullible readers so that the academic points are more plausible.
The comments I keep reading remind me that we live in a sad world in which charlatanism works.
Rick Stryker Krugman has certainly made plenty of bad calls. And to his credit he encouraged “Bobby” to archive his writings and posts up to the time that the NY Times went behind a paywall. That’s how you were able to find the Krugman and Summers memo to Feldstein. You forgot to mention that Krugman admits that he was very wrong about interest rates during the mid-2000s. Krugman has never denied that he was wrong about plenty of things. In fact, he kind of wears it as a badge of pride.
As to the double-dip memo calling for a housing bubble, I’m afraid that (yet once again) you have badly misunderstood the joke. This was just one in a series of comments Krugman was making about bubbles and recoveries. At the time he was also talking about the railway bubble in the 19th century and the electrification bubble in the early 20th century. Note, those were examples of the kinds of pre-war bubbles that led to typical pre-war recessions. Contrary to your clumsy misinterpretation, Krugman was not advocating a housing bubble anymore than he truly believes in an imminent attack from space aliens; he was being ironic. His point was that the latest recessions were not created by deliberate Fed decisions to deflate the economy, so we shouldn’t expect the Fed to have the tools to undo the new type of recession. His point was that if you were looking only to the Fed for solutions to a double-dip, then Greenspan’s only option would be to create a bubble, which was the pre-war solution. He wasn’t advocating a bubble creation, he was advocating a fiscal solution. His point (which you clearly missed) was that your options were restricted to a bubble solution if you were only willing to look at the tools available to Alan Greenspan. I have no idea how you could have possibly missed the dripping irony. You must be “Literal Man.” Also note Krugman’s reference to consumers being tapped out. And oh by the way, we did get a lot of fiscal in 2003 as the deficit more than doubled from 1.5% of GDP in 2002 to 3.4% in 2003. Recall that in 2002 and 2003 Krugman was arguing that the one good thing coming from the two wars was the added fiscal stimulus. He took a lot of heat from liberals on this point too. So once again you have completely blown the story.
One area in which Krugman has been both correct and 10 years ahead of the rest of the macroeconomics profession is the rediscovery of liquidity trap economics. Almost every economist concedes that Krugman was right about Japan despite all the abuse he took 15 years ago. And the Japan model sure looks like something that we should be paying attention to today.
Now how about answering the two main charges against R&R? R&R oversold (or passively allowed to be oversold) the 90% threshold finding and they oversold (or passively allowed to be oversold) the implied direction of causality. Those are the substantive charges against R&R. To date they still have not directly and truthfully answered any of these charges. What we get instead is misdirection, evasion and grossly out-of-context quotes. They should take a lesson from Krugman and not only admit past mistakes, but have some grad student archive them to a website along with all the correct calls.
So you are saying that it is OK for unemployment to be above 25% in Spain, and because they are now approaching 90% debt load ( same as Germany, and Spain was at half that load when the crisis hit with Spain at the tail of the whip of interest rates).
How can you fit the crisis in Spain into this model? And if you can’t fit Spain, how is the model relevant to anything?
Reinhart Rogoff did seminal work. First, they conceptualized what may or may not prove to be a major gap in macroeconomics. They then laboriously assembled a database from diverse original sources. Care was taken to assure the data was reliable. In some cases they chose to not include certain data in their overall studies as it did not pass the test of reliability. With yet further effort, they then found new information on some of these borderline cases that swung them to the side of sufficient reliability. Their dataset grew organically in this manner. Literally thousands of decisions had to be made every step of the way. It defies logic to think that in a Platonic sense each and every one of these decisions was correct. The only thing we can ever ask is that we do our human best. In the case of scientific work, we know that over time additional research will asymptotically home in on the ultimate “truth.” Scientific progress over all the centuries tells us this.
Reinhart Rogoff then took the amassed data and demonstrated a hitherto unknown (overlooked) correlation between debt and growth. In the earliest stages of their research they had to find a way to make sense of what was not a nice neat sample, but rather a blurred fuzzy one across diverse countries and time periods. They conceptualized and chose a method of dividing the data by level of debt ratio into 4 buckets, with breakpoints at 30%, 60%, and 90%. There was hardly anything sacred about 90%. But almost surely by having an holistic comprehension of the data, the data spoke to them to do it this way. About the main result – that debt is inversely correlated with growth in a robust and statistically significant way – there is no question.
The furor, in my judgment, has arisen because the Reinhart Rogoff result promises to knock a hole in the current economic paradigm. The credit bubble of the 1920s does not play a very large role in the conventional explanation of the Great Depression; and the recent credit bubble that wrought so much damage certainly did not. Debt is the tail of a coin with credit as its head. Not the one without the other. A salient and vastly underappreciated fact is that the flow of credit affects the real economy in close time, whereas the stock of debt accumulates over a long expanse of time. And then according to the findings of Reinhart Rogoff, casts a long shadow into the future. It is easy to see that today’s slump in economic growth causes a deficit which causes an increment to the debt in the very next time period. It is far harder to see the consequences of today’s debt, say in RR’s highest debt bucket, on growth one, two, and n periods out. Who of us knows all the channels whereby debt affects growth? Prior to Reinhart Rogoff, hardly anyone even asked the question!
Carmen Reinhart is explicit that now the frontier question for research is the issue of causality. We can already see causality from debt to growth playing out in private sector deleveraging. We see deleveraging casting a multiyear shadow, and must imagine it has more years to go. The channels from household debt to growth are, in general, probably not the same as the channels from sovereign debt to growth. But they can provide clues, and the channel effects will be spread over time too.
It is fruitful to think in terms of acute and chronic. A chronic condition can fly a long time under the radar screen debilitating the organism all the while. The household debt ratio reached the acute stage in 2007, and then cascaded down on growth quickly. In effect some of the hot potato then got passed off from household to sovereign debt. Now both are in a chronic state, and together the joint effect is (along with other factors) visibly keeping growth slow. If debt can be handed off like this, which it certainly was or otherwise the recession would have been lengthier and more severe, the effect of debt on growth had to be handed off too. Given an existing level of overall debt, there is likely to be an economic law analogous to the first law of thermodynamics: that “effect” can neither be created nor destroyed, it can only be transferred. Similarly, securitization didn’t eliminate credit risk, it just spread it around. What makes the effect of debt so difficult to see is it gets spread over a long period of time. Putting causality aside for a moment, this is exactly what Reinhart Rogoff found for the high debt-slow growth episodes. They lasted on average 23 years.
The interesting question is why PK is so fixated on R&R. I find the most plausible and thoughtful explanation to be Warsh’s
http://www.economicprincipals.com
Warsh was a Boston economics and business journalist for a long time, so he has the sort of Inside Baseball perspective that you need to make sense of this stuuf
In summary, the offence few epsilon deviation to trend depending on the data collection , and mind you no one has been wondering how these Epsilon would look like should the GDP deflators be harmonized during the contemplated period through investigation. The outcome would resolutely be sustaining R&R thesis slow growth with higher debts.
Good practice and healthy Puritanism may as well drive attention on policies makers, theories, mathematical models which contents and lay out do not specifically fit with the problems when the implementations become universal with a universal collateral damages « An equilibrium model of global imbalances and low interest rates » The implications of more relevance who is responsible for the misuse, the applications of such models? In Europe the presidium of the Europe supreme is still managing supply and prices of housing at the expenses of the households.
A good movie « One Flew Over the Cuckoo’s Nest «
Professor Hamilton, when you have clowns like Stryker and Ricardo (who write remarkably similar prose) on your side, I would take pause. First, this is not just between Paul Krugman and Rogoff and Reinhart. Martin Wolf, Dean Baker, Mike Konzal, Robert Waldman, Tim Duy, and Mark Thoma have all taken R&R to task for their inadequate sharing of data, for implying casuality where at most is correlation, and, most importantly, at least during the first six months of 2010, stating that 90% debt to GDP ratio was some catastrophic cliff that a nation crosses at their peril. As for hurt feelings and damaged reputations, you and they have done the most to yourselves by not being willing to acknowledge errors and the horrible human catastrophe that is unfolding in Europe and the U.S. through chronic high unemployment and falling real median wages for the 90% of the people in OECD countries even as these economies continue to grow wealthier and more prodcutive through technoligical progress and productivity. I refer you to Robert Waldman who rebuts your post in detail. http://angrybearblog.com/2013/05/what-reinhart-and-rogoff-should-do-now.html
Rick Stryker You’re desperate. First, Krugman has never focused on the coding error. He’s been pretty forgiving about that. In fact, he’s repeatedly said that this was the least significant sin even though it was the one that got the most attention. So you’re just wildly off base on this one. Second, Krugman did not say that R&R were incompetent and dishonest. As he said just yesterday, they have done good work in the past and he is sure they will do good work in the future. Go read his post. And while I have accused them of being intellectually dishonest, Krugman has not. His argument is that they were co-opted and seduced. Third, your comment “he is saying that 90% is an important indicator of being in a high debt regime” is simply wrong. R&R were saying that the 90% figure was a “threshold” with special significance. Big difference and if you don’t understand the difference you have no business even being in this discussion. Fourth, regarding your comment: “is that the slow growth episodes last decades in which 1 percent lower growth accumulates into a substantial output loss.” Go read Delong’s comment on this. Fifth, “Rogoff has also made the point that causation runs both ways: slow growth produces higher debt and higher debt produces slower growth.” R&R do make this comment, but only in a footnote in a paper behind a firewall. That is not the message they were telling Congress. Sixth, “ Krugman never acknowledges Rogoff’s real arguments. He doesn’t mention the empirical papers supporting the 90% threshold. Wrong again. Krugman and others have argued that while it may be true that higher debt is associated with lower growth, there is nothing magical about the 90% figure that would justify calling it a “threshold” or “important marker.” In fact, R&R are strangely silent about the insignificance of the 90% figure as a stark threshold…except (once again) in a footnote in a paper behind a paywall, where they admit that the 90% number was just something they selected for convenience and consistency with another analysis done by someone else. Seventh, “ if you fairly read Rogoff I don’t think you can say that Rogoff ever made such a strong claim about the 90% as do the papers who support him.” You’re rationalizing and stretching credulity. Reinhart had plenty of opportunities in open congressional testimony to correct the record on this and she demurred at every opportunity. No one would have paid any attention to this paper if it weren’t for the hardness of the 90% threshold. You yourself seem hopelessly confused and befuddled by the 90% threshold. At some points in your post you emphasize the importance of the 90% threshold, and then a few sentences later you try and downplay it all as little more than an interesting marker that may or may not mean anything in particular. Sorry, but you can’t have it both ways.
“And JDH decisively refuted the third point in his discussion of fixed versus random effects.”
Hmmmm…while a random effects model is generally preferred over a fixed effects model, a random effects model is only appropriate under some very narrow assumptions. In this case it seems highly likely that the independent variable will be correlated with the random effects intercept component of the composite error term, so this is likely to be an econometric dead end. And a lot of the countries only had one or two data points in the +90% range. In any event, one of the problems that HAP and others complained about was that the weighting technique that R&R described in their paper did not match the weighting technique that they actually used. That was one of the reasons researchers were unable to replicate R&R’s results.
The basic charge still stands. R&R should admit unequivocally that the 90% threshold is no threshold at all and that Rep. Paul Ryan and others in Congress misinterpreted their actual research. They should further admit their culpability in allowing the 90% threshold to take on a significance beyond what is justified by the data.
Stryker,
Please provide earlier studies supposedly finding a 90% threshold. Krugman overshoots and his 2001 remarks on housing look silly now, although he did see that the housing bubble was a bubble before Jim H. did. The 90% threshold is the big issue here, overshadowing all the rest, which are just a sideshow.
Bottom line remains: There is not and never was a 90% threshold. Period.
Natural scientists get fired and debarred from future grant funding for the kind of academic misconduct in which R&R engaged. And you certainly won’t find the miscreants being defended by their colleagues, quite the contrary. The contrast between this and the intellectually dishonest defenses of R&R- as dishonest as their own work- that we’ve been seeing from Prof. Hamilton and many other economists, tells you everything you need to know about the real status of economics as a scholarly discipline. Namely, that it’s a sorry parody of one.
I provide you the following large quote, which you can also find in the “Wayback Machine” called the internet, and bring you this large quote from R&R in July 2011, and you can see how it fed the “debt” hysteria that gripped political elites, a particularly useful hysteria if you want to decimate the welfare state. If that is also your desire, you should plainly state it and also state that “deficits don’t really matter” if what we are financing are wars and tax cuts for rich people. If on the other hand you and R&R state that perhaps you have overestimated the threat of public debt and underestimated the economic and social consequences of prolonged unemployment and underinvestment in public goods (education, roads, sewage, public health, railroads, and telecommunications), then perhaps criticism will have less of an edge then it does when you refuse to admit any error.
Carmen Reinhart and Ken Rogoff (July 2011):
http://www.bloomberg.com/news/2011-07-14/too-much-debt-means-economy-can-t-grow-commentary-by-reinhart-and-rogoff.html
Too Much Debt Means the Economy Can’t Grow: As public debt in advanced countries reaches levels not seen since the end of World War II, there is considerable debate about the urgency of taming deficits with the aim of stabilizing and ultimately reducing debt as a percentage of gross domestic product.
Our empirical research on the history of financial crises and the relationship between growth and public liabilities supports the view that current debt trajectories are a risk to long-term growth and stability, with many advanced economies already reaching or exceeding the important marker of 90 percent of GDP. Nevertheless, many prominent public intellectuals continue to argue that debt phobia is wildly overblown. Countries such as the U.S., Japan and the U.K. aren’t like Greece, nor does the market treat them as such.
Indeed, there is a growing perception that today’s low interest rates for the debt of advanced economies offer a compelling reason to begin another round of massive fiscal stimulus. If Asian nations are spinning off huge excess savings partly as a byproduct of measures that effectively force low- income savers to put their money in bank accounts with low government-imposed interest-rate ceilings — why not take advantage of the cheap money?
Although we agree that governments must exercise caution in gradually reducing crisis-response spending, we think it would be folly to take comfort in today’s low borrowing costs, much less to interpret them as an “all clear” signal for a further explosion of debt.
Several studies of financial crises show that interest rates seldom indicate problems long in advance. In fact, we should probably be particularly concerned today because a growing share of advanced country debt is held by official creditors whose current willingness to forego short-term returns doesn’t guarantee there will be a captive audience for debt in perpetuity.
Those who would point to low servicing costs should remember that market interest rates can change like the weather. Debt levels, by contrast, can’t be brought down quickly. Even though politicians everywhere like to argue that their country will expand its way out of debt, our historical research suggests that growth alone is rarely enough to achieve that with the debt levels we are experiencing today.
While we expect to see more than one member of the Organization for Economic Cooperation and Development default or restructure their debt before the European crisis is resolved, that isn’t the greatest threat to most advanced economies. The biggest risk is that debt will accumulate until the overhang weighs on growth.
At what point does indebtedness become a problem? In our study “Growth in a Time of Debt,” we found relatively little association between public liabilities and growth for debt levels of less than 90 percent of GDP. But burdens above 90 percent are associated with 1 percent lower median growth. Our results are based on a data set of public debt covering 44 countries for up to 200 years. The annual data set incorporates more than 3,700 observations spanning a wide range of political and historical circumstances, legal structures and monetary regimes.
We aren’t suggesting there is a bright red line at 90 percent; our results don’t imply that 89 percent is a safe debt level, or that 91 percent is necessarily catastrophic. Anyone familiar with doing empirical research understands that vulnerability to crises and anemic growth seldom depends on a single factor such as public debt.
2slugbaits,
At that time in 2002, Krugman was advocating both monetary and fiscal policy. He supported the Fed’s reduction in interest rates but worried that it would not be enough, even if the Fed reduced rates to zero. Pauly-one-note was talking about a liquidity trap even back then and wanted fiscal stimulus too. But Krugman rejected the military action in Iraq as being inneffective stimulus.
All this policy advice was based on Krugman’s mistaken view that the economy was in a period of long slow growth, a la Japan. As growth kept coming in strongly in 2003, Krugman kept denying it. For example, he claimed the strong increase in stock prices in 2003 was not a reflection of strong economic growth but rather a bubble.
By 2004, it was clear that Krugman had been completely wrong. Rather than admit it, however, he instead shifted his criticism to questions of distribution of the economic growth and continued weakness in the labor market. But subsequent developments in job growth would soon show that Krugman was wrong about that too.
I think Krugman’s article plainly shows that he thought the Fed should blow up a housing bubble. Krugman has defended himself on this charge using a different argument from the one you’ve given. He’s claimed that he was just stating McCulley’s belief but wasn’t necessarily endorsing it.
Who knows what the truth is with Krugman? He writes with a studied imprecision that allows him to wriggle out of whatever claims he’s made–so you can never be sure just what he really believes.
However this much is clear. Let’s grant for the sake of argument that he really didn’t mean that. Didn’t Krugman have an obligation to warn McCully and Greenspan of the effects of the housing bubble that the Fed was creating? Why was he silent? Many commenters have attacked R&R about allegedly not being aggressive enough to warn policy makers about misuse of their findings. What about Krugman then? Shouldn’t Krugman have warned about the coming economic debacle that would result from the housing bust if he knew about it in 2002?
On your last point, like Krugman you are attempting to shift the discussion to a different subject. I don’t think anyone would be upset about this if the charge were that R&R oversold the 90% threshold or direction of causation. That is not what anyone is upset about. People are upset about HAP and Krugman maligning the reputations of R&R by insinuating that they are incompetent and dishonest researchers. You need to call for Krugman to apologize for his libelous remarks before we talk about changing the question to whether R&R oversold their results.
JBH Reinhart Rogoff did seminal work.
Correct. Their most original contribution found a significant difference between garden variety recessions and financial recessions. Their finding that financial recessions tended to last a very long time was something that should have answered critics of a more aggressive fiscal policy who “worried” that the fiscal stimulus was likely to kick in just as the economy would be getting back on its feet. There was no chance of that happening.
Reinhart Rogoff then took the amassed data and demonstrated a hitherto unknown (overlooked) correlation between debt and growth.
This is demonstrably wrong. Go read footnote 5 of their 7 Jan 2010 draft version of “Growth in a Time of Debt” where they refer to other studies that looked at debt and growth. Note also that they say their four “buckets” of debt was simply selected for convenience. They further recognize the possible meaninglessness of their four “buckets” when they say: “Sensitivity analysis involving a different set of debt cutoffs merits exploration as do country-specific debt thresholds along the broad lines discussed in Reinhart, Rogoff, and Savastano (2003)” So when talking to academics in the footnotes of draft working papers behind a paywall they are properly cautious about the significance of the 90% threshold. But in the later versions of their work these cautionary footnotes are removed. They clearly recognized that others had done similar studies on debt/growth correlations. So you’re wrong about this being “hitherto unknown” territory. And they clearly recognized the dubious bucketing that has caused so much outrage. Yet they chose to hide that recognition when talking to politicians and writing op-ed pieces.
The rest of your post is just warmed over Austrian Austerian mumbo-jumbo.
It’s just not correct to say that the Herndon et al. results are identical to the R&R JEP paper, in which the authors really do imply a causal relationship between debt going over 90% and slow growth. Sure, R&R don’t back this assertion up with the proper data, but that’s what this is all about…
The endgame for Reinhart and Rogoff is that they are advocating for austerity now on the grounds that research supports the notion that high debt loads leads to lower growth. But research does not support the proposition, and austerity is now not the right policy for Europe or the US, and it wasn’t the right policy when R&R testified in front of the Senate on the basis of fundamentally flawed research.
Barkley,
Amazingly enough, Krugman himself posted a reference to this paper
“Public Debt, Economic Growth, and Non-linear Effects: Myth or Reality” by Balazs Egert.
to buttress his claim that there is no evidence for a 90% threshold. But that paper provides such evidence and cites 7 additional recent papers by prominent researchers who do find evidence for such a threshold in support of R&R. Either Krugman didn’t read the paper before he posted it or he’s confident that he’s readers will never check him. Here is a quote page 5 of the paper.
“Many recent empirical papers sought to pin down and explain the possibly nonlinear negative
relationship between public debt and growth. Most of these papers broadly confirm that the turning point beyond which economic growth slows down sharply is around 90% of GDP. Cecchetti et al. (2011) find a threshold of 86% of GDP for a panel of 18 OECD countries and for the period from 1980 to 2010. Padoan et al. (2012) report similar effects for a similar group of countries but a longer period (1960 to2010). Covering a mix of advanced and emerging market economies, Kumar and Woo (2010) finds a turning point at 90% of GDP. Checherita and Rother (2010) and Baum et al. (2012) report similar results for a set of euro area countries. But Caner et al. (2010) and Elmeskov and Sutherland (2012) show that the tipping point is probably lower: 77% for a set of 77 countries, and 66% for a dozen of OECD countries,respectively. Finally, in a recent contribution, Panizza and Presbitero (2012) argue that a negative correlation between debt and growth does not imply causality, as lower growth can result in a higher public debt to GDP ratio.”
The foundation of the entire global push for austerity and debt reduction in the last several years has been based on a screwup in an Excel spreadsheet and poorly constructed data by R&R.
This is the economics profession, what’s new about that ?
My hat tip to this brave, young guy named Herndon at U Mass Amherst for getting to the bottom of this insanity named politics/economics.
Sherparick,
The link you posted engages in the usual diversionary tactics. Waldman disputes every sentence practically in the R&R letter so that by the end you forget what we are actually talking about. It makes me think that Waldman is actually 2slugbaits. Let’s focus back on what we are actually talking about. Here is how Krugman repeated the HAP smear in his NYT Review of Books article:
“So the revelations in April 2013 of the errors of Reinhart and Rogoff came as a shock. Despite their paper’s influence, Reinhart and Rogoff had not made their data widely available—and researchers working with seemingly comparable data hadn’t been able to reproduce their results. Finally, they made their spreadsheet available to Thomas Herndon, a graduate student at the University of Massachusetts, Amherst—and he found it very odd indeed. There was one actual coding error, although that made only a small contribution to their conclusions. More important, their data set failed to include the experience of several Allied nations—Canada, New Zealand, and Australia—that emerged from World War II with high debt but nonetheless posted solid growth. And they had used an odd weighting scheme in which each “episode” of high debt counted the same, whether it occurred during one year of bad growth or seventeen years of good growth.”
Krugman’s clear implication in this passage is that R&R are incompetent at best and dishonest at worst. That’s why R&R are justifiably angry.
These are Krugman’s so-called errors:
1) The failure to share data
2) The coding error
3) The failure to include Canada, New Zealand, and Australia
4) the odd weighting scheme
The only error here is one that R&R acknowledged and that JDH showed to be small in previous posts: the coding error. (Apparently Krugman is reading JDH’s posts.)
The rest of these points have been shown to be false over and over. R&R shared their data on a website. But so what if they didn’t? They have no obligation to do so.
R&R didn’t exclude CAD, NZD, and AUD data because it wasn’t available at the time. And, jdh’s discussion of fixed vs. random effects showed that if anyone was using an odd weighting scheme it was HAP.
You and Krugman’s other acolytes continue to try to defend the indefensible by throwing irrelevant charge after charge out. But you won’t stick to the points we are actually talking about because you can’t. You’d have to admit that Krugman and HAP are wrong. You want to talk about anything other than points 1 – 4.
I’ve tried to do a virtual intervention for you Krugman zombies by going through the history of how Krugman has been wrong about just about every major macro issue since the early 1980s. See my earlier comments. Krugman is wrong about this too. But what’s particularly distressing is that besides being wrong Krugman is now practicing the economics of personal destruction and none of you will call him on it.
Sherparick,
What’s your point in posting that quote? Rogoff has always said that 90% is an important indicator of a high debt regime and that bad consequences can follow. I’ve explained what I think he means by that in previous comments.
So what?
Rick Stryker rejected the military action in Iraq as being inneffective stimulus.
This is flat out wrong and something of a play on words. No one opposed the Iraq war more than Krugman, but he clearly said that the defense spending would be stimulative. Now it is true that he didn’t regard the war spending as a particularly optimal way to engage in fiscal stimulus, but sub-optimal is not the same as ineffective.
As to his housing bubble comment, I used to be a regular on that NYT blog back when it was realtime and before the paywall. No one…not even Krugman’s harshest critics interpreted the housing bubble comment the way you are interpreting it. Everyone understood that he was making an argument for the limited potency of monetary policy without fiscal stimulus. Even NYT blog poster “cantab” recognized as much; and “cantab” didn’t cut Krugman any slack. In one post around that time Krugman even joked that we might have to find another bubble to get out of the recession; but no one understood that comment to be an expression of support for bubbles. There’s such a thing as being completely tone deaf, and I think you may need to have your hearing checked.
I don’t think anyone would be upset about this if the charge were that R&R oversold the 90% threshold or direction of causation. That is not what anyone is upset about.
You are completely missing what’s going on here. That is all that this discussion is about. If R&R would simply admit that they oversold and allowed others to oversell their research, then all would be right with the world. The reason they’re in hot water is that they refused to acknowledge any culpability or responsibility. They chose to go outside of pure academic writings and enter the world of public intellectual when they decided to write op-ed pieces and testify before Congress. At that point they lost the right to hide behind footnotes; but yet that’s exactly what they’re trying to do even now. They played both ends of the game. They wrote careful and cautious working drafts with careful and cautious footnotes behind the NBER paywall. Then they wrote careless and reckless op-ed pieces that exaggerated the results of their research. They got caught and now they are trying to pretend that it’s all Rep. Paul Ryan’s fault that those nasty politicians misused their research. To redeem themselves all they need to do is make a forthright and unequivocal repudiation of their earlier op-ed pieces and admit that they were derelict in allowing their work to be oversold.
Krugman maligning the reputations of R&R by insinuating that they are incompetent and dishonest researchers.
Again, that is not what Krugman and others are saying. The charge is that they allowed themselves to be seduced and acquiesced in letting others oversell their research. That is not a charge of dishonesty or incompetence. Now I do happen to believe they were also being dishonest, but that is not Krugman’s opinion and he has been pretty clear about this. And if you think Krugman is smearing Ken Rogoff’s competence, then how do you explain Krugman’s very recent statement that Rogoff is the #1 international macroeconomist in the world today? Is that your idea of a smear?
Oh, BTW. While I did study ancient Latin and I’m only a few semester hours away from an undergrad degree in the classics (I considered it once as a major), I don’t speak modern Italian. So it’s unlikely that I’m Robert Waldmann.
@ Stryker: dude, Not everyone who disagrees with you is a Krugman zombie, a Krugman acolyte, etc. And RR did mess up by (passively) backing up austerity in an improper moment.
Reinhart and Rogoff had not made their data widely available– and researchers working with seemingly comparable data hadn’t been able to reproduce their results.
Seems like there is another area of heated academic debate in which people have noted the lack of data sharing. *cough* Michael Mann *cough* I sorta doubt whether Krugman has the same opinion about data sharing in that debate.
2slugbaits,
I’m desperate all right–desperate for closure on this. My goal in that comment was to separate the policy points from the smear points. It seems that everyone agrees, Krugman included, that the coding error was minimal, thanks to JDH’s efforts. You haven’t mentioned the exclusion of data point. Of course not–who could defend Krugman on that one? The only thing left is Krugman’s claim that R&R’s weighting method was odd. Once we settle that one, you’ll be ready to publicly call for Krugman to apologize. Then we can deal with Krugman’s other false claims, one by one.
Let Y(i,t) be the GDP growth rate of the ith country at time t and consider the following model:
Y(i,t) = a(i) + e(i,t)
where e(i,t) is a random error term with mean zero and constant variance u.
HAP make the assumption that a(i) = a, an unknown constant, and estimate
Y(i,t) = a + e(i,t)
The optimal estimate is the average of all the observations for each country over time, just as HAP did it. In other words, if you do an OLS regression, you sum all the Y(i,t) and then divide by the total number of observations to obtain HAP’s estimate a.
In panel data, the way HAP are doing this is non-standard, since they are assuming away in their estimate any differences in GDP growth rates.
In a fixed effects model, you wouldn’t do that:
Y(i,t) = a(i) + e(i,t)
where in this case the a(i) are fixed constants to be estimated. If you estimated this by OLS, each a(i) would be the average growth rate for the ith country. You could then average them together. This is what R&R did. HAP then is a special case of R&R. In the context of the fixed effects model, R&R have used the more sensible estimation.
JDH made quite a fair-minded point however. Although what R&R did makes more sense in panel data, it’s still the case that there is something to HAP’s estimate.
Going back to the fixed effects estimator, the R&R estimate assumes that all the variation of the data is contained in the differences across countries. The HAP estimator assumes that all the variation in the data is across time. JDH’s suggestion was to combine both using a random effects estimator.
So, you let
Y(i,t) = a(i) + e(i,t)
where now the a(i) are iid random variables with mean a and variance v. We then need to estimate a, which we could do using feasible generalized least squares. JDH’s suggestion was to run random effects in one regression with dummy variables for each category of the data, i.e. for the 30%, 60%, and 90% as follows:
y(i,t) = a0(i)*d0(i,t) + a30(i)*d30(i,t) + a60(i)*d60(i,t) + a90(i)*d90(i,t) + e(i,t)
This discussion makes clear that there is nothing odd about the way the R&R weighted the data, contrary to Krugman’s assertion. R&R did the sensible thing while HAP’s estimate was odd.
Will you now call for Krugman to apologize?
2slugbaits,
You are wrong on the facts yet again. Krugman discussed whether the war would stimulate the U.S. economy in this 2002 article:
https://www.nytimes.com/2002/09/13/opinion/stocks-and-bombs.html
Here is a relevant excerpt:
“The idea that war would actually be good for the economy seems like just one more step in this progression. But one must admit that there are times when war has had positive economic effects. In particular, there’s no question that World War II pulled the United States out of the Great Depression. And today’s U.S. economy, while not in a depression, certainly could use some help; the latest evidence suggests a recovery so slow and uneven that it feels like a continuing recession. So is war the answer?
No: World War II is a very poor model for the economic effects of a new war in the Persian Gulf. On balance, such a war is much more likely to depress than to stimulate our struggling economy.
There is nothing magical about military spending — it provides no more economic stimulus than the same amount spent on, say, cleaning up toxic waste sites.
The reason World War II accomplished what the New Deal could not was simply that war removed the usual inhibitions. Until Pearl Harbor Franklin Roosevelt didn’t have the determination or the legislative clout to enact really large programs to stimulate the economy. But war made it not just possible but necessary for the government to spend on a previously inconceivable scale, restoring full employment for the first time since 1929.
By contrast, this time around Congress is eager to spend on domestic projects; if the administration wants to pump money into the economy, all it needs to do is drop its objections to things like drought aid for farmers and new communication gear for firefighters. In other words, if the economy needs a burst of federal spending, neither economics nor politics requires that this burst take the form of a war.
And in any case it’s not clear how much stimulus war would provide. One assumes that the necessary munitions are already in stock, so there will be no surge in factory orders. There will be spending on peacekeeping — won’t there? — but it will be spread over many years.
Meanwhile there is the potential economic downside, which may be summed up in one word: oil.”
Krugman is consistently wrong. Just about the only correct thing he said recently is the Rogoff is the number one international macroeconomist in the world. I’ll give him credit for getting that right. Too bad he thought it was OK to smear the number one macroeconomist in the world.
as a matter of form, in scholarly, academic, polite debate, it is not sufficient to give a url as a source, if within that url the relevant point is not obvious.
I think careful scholarly practice requires that one say such and such a url in the wayback machine, and on that page, see……
I downloaded “the data” as a csv file (from C Shalizi, via B Delont). I simply plotted GD vs debt, threw out a few outliers, and…it is really hard to see what is what
one can torture the data with complex statistics, but one wonders why a real scientist would not prefer to understand the data points…
Much recent discussion of the R&R 2010 paper has accurately emphasized that “correlation is not causation.” Still, many who have made that point quite confidently assert a direct line of causality running from R&R’s work to both the terrible plight experienced by the long-term unemployed in the US and much of the slump-horror in Europe. Have these commenters forgotten that “correlation is not causation”?
In a similar vein: those who feel Jim Hamilton has written “utter nonsense,” etc. wrt things R&R, it seems, are comfortable with ad hominem attacks.
What’s getting ignored in the austerity debate is that the US economy seems to be handling quite a lot of it with very little consequences.
I have never downplayed the contractionary effect of austerity, or overplayed the boost to growth that can come from greater confidence. My argument has always been that it is often better for the longer run to take some pain in the short run.
What we’re seeing in the US is a surprisingly stable low growth trend despite significant austerity in the form of the 2% increase in payroll tax and the sequester. Those who were warning of calamity have gone quiet on the topic or are directing attention to the UK and Europe.
It’s still too early to be sure growth won’t slow. In any case, the compensation has come from a lowering of the personal savings rate to less than 3%, which is unsustainable and would probably be destructive if it were sustained too long. We need to know more about the demographics of that shift and how and why it happened.
But it’s interesting to see how the anti-austerity crowd is simply ignoring the news. No appetite for self reflection there.
Stryker,
Will look at your sources. I do note that if one looks at RR’s own data, they present a figure, and it is a straight line beyond 30%. It is a downward sloping straight line, but it is straight, no kinks, no inflection points, no discontinuities, no thresholds.
Jim,
You seem to be changing the game. There is certainly a stronger argument for a causal link if one looks at total debt/GDP ratio, but that is not what RR did. They were simply looking at national government debt to GDP and coming up with this artificially constructed threshold that does not even appear to be there in their data. Maybe those Stryker mentions find such a threshold, but it is not in RR’s data. Your discussion of household debt going to sovereign debt is serious, but it has nothing to do with this debate, nothing.
Jim,
You just keep digging yourself into a bigger hole.
Are you dishonest, or just stupid? You wrote the book on time-series econometrics.
How much are the Koch brothers (and everyone else who can’t stand a low interest rate) paying you to post this nonsense?
Stryker,
Looked at Cecchetti study. Focuses on household debt. I note that there has been deleveraging of household debt during the Great Recession. Many would say that opens room for expansion of public debt to offset the contractionary effect of this. BTW, all the studies you cite are since the problematic RR paper in AER P&P. Their earlier book and their more recent JEP papers are much better.
It is rather telling to read the comments attacking Rogoff and Reinhart, and Professor Hamilton for defending them. In the Keynesian view, the notion that government spending cuts can be beneficial is so harmful that it must be fought with all necessary means. Proponents must be shown to be bad actors, hacks, liars.
If small cutbacks are beneficial, larger cuts may be proposed, and the next thing you know, the entire Keynesian edifice may be in danger. If people began to ask whether specific governmental expenditures are worth diverting funds from private use (through borrowing or taxation), then you have a problem. I think this fear is what drives the vehemence of the Keynesian crowd’s attacks on R&R and anyone who would defend them.
Can an economy have ‘excessive’ debt (howsoever defined), a fiscal deficit and a goods surplus?
Obviously yes, for that is the u s today.
What are we arguing about?
Barkley,
The cecchetti study focuses on non-financial debt broadly, including household and government debt. It’s true these studies came after R&R 2010 paper, but that just shows how R&R’s results prompted more investigation which tended to confirm their original discovery. As I have said in my other comments, Rogoff never made the claim that the 90% was a threshold in the sense of a non-linear relation in the data. It’s rather an indicator of a high debt regime. Subsequent research has provided evidence that the 90% threshold has a stronger interpretation than what R&R originally claimed.
“If small cutbacks are beneficial, larger cuts may be proposed, and the next thing you know, the entire Keynesian edifice may be in danger. If people began to ask whether specific governmental expenditures are worth diverting funds from private use (through borrowing or taxation), then you have a problem. I think this fear is what drives the vehemence of the Keynesian crowd’s attacks on R&R and anyone who would defend them.”
Nailed it.
Karl, (AKA Nick),
I must admit I’m disappointed in your ad hominem attack on JDH. You have obviously learned nothing from Krugman. If it’s one thing Krugman is good at it’s the ad hominem. Go back and reread his posts and articles. Note how he’s mastered the insinuation and innuendo, all while maintaining deniability if called on it.
When you do the attack so obviously, it’s inneffective. People feel sympathy for the victim and you hurt your cause. Krugman just provided another classic lesson on how to do this in his recent post about how “its about the policy.”
After suckerpunching R&R and backing his manure truck over them a few times, he drops the truck load over their dazed bodies. Then he gets out of the truck and says, “It’s not about hurt feelings or a few broken bones. This is about the policy.”
If you do that with great sincerity, people forget what you just did with the truck. People like me will say that policy is the last refuge of a scoundrel, but no one will listen.
You don’t have any arguments so you must rely on the ad hominem attack. That’s fine. You are in good company judging from the other commenters. But you’ve got to learn to do it better.
@Rick “there is no right to strike” Stryker
Yes, I’m aware that you can see names by hovering over them. Which is why I always use the same email address 🙂
You mentioned in your first comment in this thread that Krugman is merely preaching to the choir. If what he says really doesn’t matter, why are you wasting all of your precious energy defending clowns like R&R?
Get back to managing your wealth, and stop worrying Paul K. God forbid you lose a few bucks when we finally get the top marginal rates back up to 50%.
Rick Stryker You are wrong on the facts yet again. Krugman discussed whether the war would stimulate the U.S. economy in this 2002 article
UGH! Can you not read? Krugman was NOT arguing that defense spending wouldn’t help stimulate aggregate demand. In fact, as you can clearly read, Krugman says: “There is nothing magical about military spending — it provides no more economic stimulus than the same amount spent on, say, cleaning up toxic waste sites.” And in terms of aggregate demand it doesn’t provide any less. Krugman is trying to turn military Keynesianism on its head and convince those who want to argue for defense spending as stimulus that what’s good for the goose is good for the gander. This is obvious. Once again you have botched it. Krugman’s point was that while defense spending increases aggregate demand, wars (as opposed to mere spending) also have adverse supply side effects. You know…it kills and maims people, which hurts long run growth. Read what he wrote. He didn’t say defense SPENDING hurt growth; he said WARS hurt growth. I hope you understand the difference. But it’s not the spending that causes those losses, it’s the actual business of fighting wars. Also notice Krugman’s comment about how it isn’t clear that defense spending would provide much of a punch because of the large stockpile of munitions at that time. But that was in 2002 and long before the war spending got really cranked up. Well after the Iraq was was underway he reminded his liberal allies that as much as he hated the war, the fact is that defense spending does increase demand. And by then there was no large munitions stockpile.
Rick Stryker I think you have misunderstood the panel effects/random effects discussion. R&R did not do a fixed effects panel analysis; they did something that was in the spirit of a fixed effects panel analysis. They first collected debt/GDP data into one of four buckets. Then within each bucket they computed a country average. Then they computed an average of the country averages. This is sorta kinda like a fixed effects model with a slope of zero. What HAP did was to group debt/GDP observations into one of four buckets just as R&R did. But rather than take an average of each country’s average, they just treated the data as cross-sectional and took a simple average of all observations that fell in each bucket. What JDH was proposing was a more formal random effects model that would combine the long run aspects of a cross-sectional model and the short run dynamics of a fixed effect model. JDH hardly needs me to confirm this, but such a random effects model would almost certainly yield an answer that was somewhere between the R&R and HAP estimates, and that’s because a random effects model combines the long run and short run aspects of cross-sectional and fixed effects models respectively. While I haven’t done any formal Hausman test, my hunch is that a random effects model, while superior if feasible, is only feasible under some fairly strict assumptions that probably don’t apply here. It seems very likely that the explanatory variable will be highly correlated with the composite error. A random effects model teases out the random effect intercept from the usual error term where the observed error is a composite. As I said before, the odds are that in this case a random effects model would probably be an econometric dead end because the data will not cooperate. And then there’s the problem that many of the countries only had one or two observations, so good luck with that. A simpler approach would be to take a geometrically weighted mean of each country’s mean (similar to R&R) but weight each country by each country’s variability so countries with a lot of variability get less weight. A more sophisticated weighting approach would be to create two new categories; one for countries with chronically high debt/GDP ratios and another for countries with only occasional lapses. Then separate by those regimes that control their own currency. And then include a time variable to capture HAP’s finding that the relationship between debt and growth seems to be degrading over time. Oops…just ran out of degrees of freedom. Here’s a better idea: junk the four buckets, which don’t really add any value because all greater than 90% get truncated.
Now calm down and take a deep breath.
“Reinhart and Rogoff used the Wayback Machine to get a copy of Carmen Reinhart’s web page exactly as it appeared to the world in October 2010. Try it yourself– you can click on the links to download whatever spreadsheet you like”
Where on the wayback machine is the speadsheet with the infamous error? According to Thomas Herndon, its not there.
I have written about this at my place.
I find it very unfortunate Mr Hamilton has found himself lead down a garden path and nowhere to hide.
This is obviously a blindspot for him and he appears not to know it.
I am curious why anyone would write about this concerning Krugman and NOT talk about how Krugman clearly seperates the book from the article.
I think most sensible peope would conclude the date was not made avaible widely as has been shown here.
Personally I think Robert Waldeman is devastating in his analysis.
Mr Hamilton you are far too good to stubbornly remain here.
walk away and get back to the high heights you were.
2Slugbaits,
He doesn’t understand and doesn’t want to understand. He simply hates Krugman.
P.S.
I regularly on Fridays ,Sydney Time, put out a review of the week of blog pieces ( links really).
I have always had You and Menzies on it and will continue to do so!
Karl, (Nick)
I do think Krugman is ineffective and am not worried that he’ll influence policy. I’ve never commented on Krugman’s blog because I don’t think it worth it. I stopped reading his books after “The Conscience of a Liberal” because they’ve become tedious and off-putting. He’s an advocacy journalist now.
I’m defending R&R because I’m outraged about how they’ve been treated. It’s not about ideology. As I’ve said before, I don’t necessarily agree with them on some issues. But I think it’s important that economists of the stature of R&R influence the economic debate. That can only happen if economists who disagree with each other do so fairly and civily. Otherwise, good economists on all sides will stay out of real world policy questions and we’ll all be poorer for it.
R&R have had their professional and personal reputations sullied by false accusations. They’ve been mocked on the Colbert report by a fellow economist. They’ve been maligned in the pages of the NYT magazine by a fellow economist. They don’t deserve that and the economics profession shouldn’t tolerate it.
I admire and respect JDH for standing up to this bullying. He obviously knew he would take a lot of personal abuse in doing so. Many people understand what’s really going on but don’t want to speak out because they are afraid that they’ll be abused too. I am just trying to help out in my own small way.
2slugbaits,
I posted the text of Krugman’s article where he plainly says he doesn’t think the war will be stimulative. And then you go off into all kinds of obfuscation. Marco, you asked why I keep referring to Krugman zombies? Here’s your answer.
But I want to focus on the second comment. In this one, you repeat what I said back to me as your own comment and then say I don’t understand what I just said that you repeated back. Then you go off on a huge tangent. I don’t think you understand this at all. As always with you, it’s necessary to narrow the point to pin you down. Let’s forget the merits of the random effects estimator.
To repeat, I said:
Let Y(i,t) be the GDP growth rate of the ith country at time t and consider the following model:
Y(i,t) = a(i) + e(i,t)
where e(i,t) is a random error term with mean zero and constant variance u.
HAP make the assumption that a(i) = a, an unknown constant, and estimate
Y(i,t) = a + e(i,t)
The optimal estimate of a is the average of all the observations for each country over time, just as HAP did it.
However, if you allow the a(i) to differ by country and estimate a(i) on this model
Y(i,t) = a(i) + e(i,t)
and then you average the estimated a(i) you get the R&R estimate.
Here is a very narrow question:
Do you agree or disagree with this? Whether you agree or disagree, please provide the detailed mathematical justification for your answer. This is a trivial exercise. If you can’t do it, there is no point in discussing further.
and the correct answer is…?
Libs bash R&R because they don’t want their handouts to get cut.
Paul,
I’m curious to see how 2slugbaits will proceed here. If he does it, he’ll have demonstrated that there was nothing peculiar about what R&R did. In fact, it was HAP who made the peculiar assumption that the averages are all the same in panel data. That will take care of the last leg of HAP’s original smear that Krugman repeated in the NYT. 2slugbaits can then publicly call for Krugman to apologize.
If he doesn’t do it, he will have demonstrated that he doesn’t know what he’s talking about.
My bet is he’ll pick a third way: obfuscation. That’s the historical pattern. We’ll see.
By the way, I thought your link to the David Warsh piece on Krugman’s motives was spot on.
Rick-
You are a very patient man and your responses are thoughtful and to the point. As far as Mr. baits, his usual approach consists of one part jargon, one part condescension and one part disdain. You don’t have a discussion with him, but rather you must stay put while he lectures you.
Krugman seems to be showing signs of some kind of growing personality problem, which is unfortunately overshadowing by a clear margin his analytical abilities. That’s not a personal attack, but rather a fair observation, I think. His tone is becoming increasingly strident and mean, really beyond reasonable limits.
The NYT should let him go and so should ABC, for he is reflecting badly upon them, as he is upon our economics profession. We need serious, sober commentators when we have such serious problems to solve. If someone started a petition to ABC/NYT to return him to full-time academic work, I would sign it.
Perhaps some of you would be interested in this:
Quartz on Kimball and Wong
Rick Stryker First, Krugman distinguished between defense spending, which he said was stimulative, and the war itself, which hurts economic growth. I’m sorry if you don’t know the difference.
As to your R&R versus HAP stuff. You’re simply wrong unless you want to call your “a(I,t)” a constant. I guess you could do that, but it would be kind of stupid. How about just calling it “C”? For example, “In fact, it was HAP who made the peculiar assumption that the averages are all the same in panel data.” This is wrong because HAP didn’t attempt a panel analysis. What they did was a cross-sectional analysis in which the observations were pooled into four separate buckets. If you want to put what they did in the form of a linear regression, then try this:
Y(i,b) = C + e(i,b)
where “i” indicates the observation and “b” indicates the bucket. In other words, they just did a simple average and if you want to express that as a regression against a constant, then be my guest.
As to what R&R did, they didn’t do a true fixed effects panel data analysis either. They simply took an average of each country’s observations and then took a global average of the individual country averages. Or, if you prefer:
Mean(g,b)
where (g,b) = Mean(i,b)
where (g,b) = global mean for each bucket
and (i,b) = country “i” mean for each bucket.
So one more time….R&R did NOT do a fixed effects panel data model. What they did was weight the country observations in a way that tried to capture some of the intuition of a fixed effects model, but they did not actually run a fixed effects model.
One thing that is not clear is how R&R handled their median estimates. For each bucket did they take the global median of country means, or did they take the global median of country medians?
Rick Stryker In fact, it was HAP who made the peculiar assumption that the averages are all the same in panel data.
Aside from the fact that HAP never attempted a panel data model, your calling it “peculiar” is in itself kind of peculiar. One reason they took a cross-sectional approach for each bucket is because that’s how R&R described their approach in their paper. As many critics have pointed out, one of the reasons people couldn’t replicate the R&R results is that the weighting technique they actually used did not match their description. There are good arguments for doing things the way HAP did them. There are also good reasons for weighting things the way R&R did them. And if R&R’s thesis is correct, then you would expect a certain amount of robustness regardless of which weighting technique you use. I don’t know if you do a lot of quantitative work, but if you do I sure hope you run your analyses lots of different ways.
And as I said before, I don’t think either HAP or R&R did the weighting correctly. I would have used a weighted geometric mean with the country weights being the reciprocal of the std-dev.
Thanks very much Rich.
I do follow your comments on this blog as well and appreciate them.
It is amazing how many personal attacks there have been here on Krugman that aren’t personal attacks.
This must be an indictment on the US education system.
The other thing I would add if you wish to allege a smear is made then show it do not generalise!
2slugbaits,
As I suspected you don’t have the slightest idea about what’s going on. I posed a simple and clear problem. Given the fixed effects model
Y(i,t) = a(i) + e(i,t)
let a(i) = a. Then derive the optimal estimate of a and show that it is equal to the HAP estimate, providing the relevant mathematical detail. This is very easy.
Then, take the general fixed effects model
Y(i,t) = a(i) + e(i,t)
and derive the optimal estimate of each a(i). Show that the average of the estimated a(i) is the R&R estimator, providing all relevant mathematical details. This is also very easy.
2slugbaits, if you can’t do that, then you can’t understand this point. You’ve had a day and half. You could have looked up the answer by now or asked someone. Then you could have pretended that you understood it. It appears that your knowledge of statistics is basically zero.
This is why I keep calling people Krugman zombies. Here is a guy that’s vehemently defending Krugman on a technical issue and attacking R&R but has no idea what the technical issue is.
Why in the world would you attempt to dispute something like this on the internet when you have no idea about what’s going on? I don’t get it. This is the kind of behavior that Krugman inspires. It’s surreal.
Poor Ken Rogoff. Krugman has unleashed an ignorant, angry mob on him.
Mr Stryker,
you cannot actually show an example of a smear and you have been caught not knowing what you are on about.
Me thinks you protest too much.
move on
http://qz.com/88781/after-crunching-reinhart-and-rogoffs-data-weve-concluded-that-high-debt-does-not-cause-low-growth/
Matt Wilbert, your link didn’t work, so thought I would put one on that did. Of course, as our blog host, Mr James Hamilton put it, this is just one gigantic smear campaign by ignorant professors from the universities across America. To quote exactly:
“Not that facts matter to those who took up the cudgels. The smear campaign had only one purpose– to distract people from thinking clearly about the consequences of the current high debt loads of many of the world’s countries. On this fundamental question you can also find much to help set the record straight in Reinhart and Rogoff’s open letter.”
“But be forewarned– there are many folks out there who still think that if that if they just keep shouting lies and nonsense loudly enough they can prevent you from hearing Reinhart and Rogoff’s true message.”
Or as Rick Stryker just put it: “Poor Ken Rogoff. Krugman has unleashed an ignorant, angry mob on him.”
I just can’t wait to see the response to this latest article! This Krugman orchestrated smear campaign knows no bounds, all in the attempt to get us to not think clearly about the consequences of debt (considering how successful Bush and the Republicans were in increasing our debt to begin with, even without Krugman’s help, we can only imagine the sh*t storm that would erupt if R & R were not on the case right now). And this latest duo, from the University of Michigan nonetheless, are almost machiavellian in how polite they come across in their writing. Without shouting, they put their professional reputations on the line and bullsh*t their way for paragraph after paragraph of mumbo jumbo, only to slip the lie in at the end: “Done carefully, debt is not damning. Debt is just debt.”
Nottrampis,
Well, that’s ironic. You’ve just made a very general comment. If you have been reading JDH’s posts and the comments, you’d know that the claims on the smear have been very specific. You can hardly come along at this point and say nothing specific has been said.
Case in point: I just wrote down a very specific statistical model to illustrate to one of Krugman’s more vociferous defenders how wrong Krugman is to claim that R&R’s estimation technique was “odd.” If you want to contribute to defending Krugman, you could comment on that.
I also posted links to Krugman’s consistent failed predictions and wrong-headed analysis since the early 1980s: his miscall on inflation; his miscall on the 1990-1991 recession and his miscall on the 2002 recovery. I posted links to his actual words illustrating the point that Krugman has been wrong over and over again. If you want to help to defend Krugman, comment on that.
But don’t say people are generalizing.
In addition to the article quoted in my post above (http://qz.com/88781/after-crunching-reinhart-and-rogoffs-data-weve-concluded-that-high-debt-does-not-cause-low-growth/), I thought I would add a few co-conspirators to Krugman’s smear campaign of the most honorable and careful researchers, Rogoff and Reinhart. Since our blog host has claimed that there are many who are spouting lies to lull us all in to a sense of complacency about our debt:
Dean Baker:
http://www.cepr.net/index.php/blogs/cepr-blog/reinhart-and-rogoff-trip-over-data-krugman
http://www.cepr.net/index.php/blogs/beat-the-press/reinhart-and-rogoff-are-not-being-straight
“In addition to misleading the public about the role their work has played in policy debates, they also are not quite straight about two strictly factual points. The first sentence begins by referring to the publication of their article in May of 2010. This might lead readers to believe that this is when their claims about high debt slowing growth first began to affect public debate on stimulus and deficits.
This is not right as I know since my first e-mail requesting their data was written in January of 2010. In fact, their work first made a splash in international debates when they put out a version of this article as a National Bureau of Economic Research working paper in January, 2010. Their findings were already widely known by the time the paper was published in May, 2010.”
Most Uncivil.
James Kwak and Simon Johnson:
http://baselinescenario.com/2013/04/19/fatal-sensitivity/
“The question to ask is this: If a method produces results that can drastically change by the addition of a few more data points, are those results worth anything? The answer is no.”
How Vulgar.
Brad DeLong
http://delong.typepad.com/sdj/2013/05/accurate-and-inaccurate-ways-of-portraying-the-debt-and-growth-association.html
http://delong.typepad.com/sdj/2013/04/reinhart-rogoff-weblogging-no-their-argument-for-austerity-now-didnt-make-sense.html
And even before the famous excel error that anyone would have made was found, we have this . . . Rogoff and Reinhart are the most dangerous economists ever!?!?! With hacks like Koo and Robert Shiller chiming in?
http://www.businessinsider.com/reinhart-and-rogoff-dangerous-debt-ceiling-2011-8
As I said above, Mr. Hamilton, when there is this much stuff out there, and you come out with a quote like – “Not that facts matter to those who took up the cudgels. The smear campaign had only one purpose– to distract people from thinking clearly about the consequences of the current high debt loads of many of the world’s countries” – you are making a pretty strong statement about the ethics of a great many of your fellow colleagues. This simply doesn’t pass the smell test . . . do you have anything else to defend your opinion?
Rick Stryker: “This is the kind of behavior that Krugman inspires. It’s surreal.”
As I passer-by who is reading casually, may I suggest that this is not a wise supplement to a post that starts with:
“As I suspected you don’t have the slightest idea about what’s going on … “.
While that you say may be true, as I see it 2slugbaits gave an explanation as to why he was proceeding as he did. Valid, or not, you did not deal with that, but simply declared – as he did with you – that you were missing the point. However he explained why and proceeded from an identifiable point of departure. You did not.
Given that we now have two of you “missing the point” how is a non-expert to assess the arguments regarding this little mathematical challenge? You give the appearance – for that is all I can assess – of failing / not wanting to deal with what he has to say.
This non-expert thus finds your exposition less convincing than that of 2slugbaits.
In making these comments I do not consider myself to be part of “an ignorant, angry mob”. Just think about it – what did you intended by making such a comment ? As far as I can see there is a fairly ignorant and angry bunch of folks on both sides of the argument to be found across the Internet. So what?
You can only be judged by what you write and the arguments you make.
I am not an economist so I will not comment on the substance of the RR hypothesis or the criticisms of that hypothesis.
However, I do think it is quite unfair to criticize them publicly, as Mr. Krugman has done, for failing to make public their data when in fact they have done exactly that.
Nergy,
I think you missed the point here. I made a specific assertion that 2slugbaits denied while peppering me with a number of irrelevant points. My specific assertion was that R&R is an average of fixed effect estimator and HAP is a special case of a fixed effects estimator in which all means are equal.
I wanted to stop him from inundating me with irrelevant comment after irrelevant comment. So I narrowed the question and challenged him to prove I’m wrong by deriving the average of the fixed effects estimator and the estimator of the special case and to show that these estimators are not equivalent to R&R and HAP. That requires mathematics not words but if 2slugbaits really knows what he’s talking about, he can easily do it. The problem for him, though, is that if does do it he’ll see he’s wrong.
He responded with more words, not equations–just more obfuscation. The technique works for him and other Krugman defenders (and Krugman himself) because they know that people such as yourself who aren’t familiar with the issues can’t tell who is right.
Just another aspect of the smear campaign.
River,
Naturally, you are trying to change the subject from the smear points to some people’s technical disagreement with R&R. That’s been a common tactic of Krugman’s defenders. No one wants to defend the smear points of course because they are indefensible.
The smear points made by HAP and Krugman are:
1) R&R selectively excluded data
2) R&R used an odd estimation technique that led to an error
3) R&R did not share their data
Those assertions have been demonstrated to be false. So let’s stay on point and demand that Krugman apologize for the smear before moving on to new questions.
Nergy,
I understand the problem that many commenters aren’t familiar with technical arguments and so don’t know how to judge. Let me try to explain the weighting issue intuitively.
Let’s forget economics and look at a simple situation. Suppose we have to decide what what the legal drinking limit is going to be, i.e., the blood alcohol level before it’s unsafe to drive. We take 6 men and every day give them enough to drink to raise their blood alcohol level to some point, let’s say it’s 1 on some scale. Each day we measure each man’s reaction time. A reaction time greater than 10 is unsafe to drive. We want to know if a reading of 1 means that you are driving drunk.
We are lucky enough to do the experiment on the first man for 100 days but can only get 10 days of data each for the other 5
Here are the measured 100 reaction times of the 1st man.
8.6 8.2 10.0 7.5 9.8 7.5 10.6 8.2 7.7 8.7
9.2 9.2 8.2 10.1 10.1 7.4 9.0 10.3 8.2 9.6
8.6 9.2 9.1 9.3 9.0 7.0 9.5 9.0 7.9 9.9
9.9 8.4 9.4 9.0 7.5 8.3 9.5 8.3 6.8 10.5
8.2 9.3 7.4 8.6 10.4 9.0 10.4 8.5 10.0 8.6
8.1 8.3 9.9 8.8 9.2 9.2 10.0 8.9 6.5 8.7
8.2 9.3 9.7 6.8 8.6 7.5 8.9 12.1 8.9 9.6
9.0 10.3 10.1 7.4 9.7 7.5 9.2 7.3 8.0 9.1
8.6 8.8 7.7 8.0 8.6 10.2 8.5 9.2 9.9 8.3
9.5 11.8 8.5 9.2 7.8 6.8 8.9 10.1 8.8 9.0
You can see on many days he’s too drunk to drive but not on all or even the majority of days.
Here are the reaction times of the other 5 men.
2 3 4 5 6
1 10.4 11.5 10.1 11.3 12.5
2 9.0 11.3 8.3 12.9 13.2
3 11.4 11.6 10.0 13.2 14.0
4 11.2 12.8 9.3 11.5 12.1
5 9.5 11.1 8.7 9.9 12.8
6 11.4 10.5 8.9 13.1 12.5
7 11.4 11.7 11.4 12.6 13.1
8 11.2 12.5 10.9 11.5 13.2
9 10.8 13.2 9.2 11.5 12.5
10 12.2 12.5 10.0 11.4 13.9
Now how do we summarize our findings? If we assume that each man’s capacity to hold his liquor is the same as every other and that the only variations are in what they ate that day, etc., you would just take all 150 data points and average them. If you did that, you’d get a reaction time of 9.7. Thus, you’d conclude a blood alcohol level of 1 is OK.
However, what if you looked at the individual averages of reaction times? Here’s what you’d get for each man.
1 2 3 4 5 6
8.9 10.8 11.9 9.7 11.9 13.0
Here it becomes obvious from the individual averages that the first guy is different from everyone else–he’s much better at holding his liquor. In fact, everyone is different as we might expect but the majority are drunk on average. So, averaging the first man’s 100 data points in with the 50 of the other 5 men will exaggerate the first guy’s influence and make it look like they can all hold their liquor.
It would be better just to average the averages, in which case you’d get 11 and you’d conclude that a blood alcohol level of 1 is unsafe to drive. That summarizes what’s actually going on better.
HAP did the first estimation and assumed that all the men were the same. R&R did the second method and assumed that the men were in fact different. You can see that the second method is more justifiable if you have any reason to believe that the men are different. Since R&R are talking about growth rates of countries, we certainly have reason to believe they are different.
Moreover, the assumption that the averages are different is the standard starting assumption when analyzing data that is both cross sectional (different men) and time dependent (different days).
Strangely enough, HAP and Krugman accused R&R of doing something non-standard and an “error” by using the second method. I hope it’s clear intuitively how this is wrong from this example. In fact, what HAP and Krugman are proposing is non-standard.
What I asked 2slugbaits to do would have established that the way we did the average over the 150 data points would have fallen out of the simpler model I gave him if he had done the math. That’s HAP. And the way we did the average of average estimates would have fallen out of the fixed effects model, if he’d done the math. That’s R&R.
Hope this helps.
I love how mad R&R get liberals. Libs are terrified of any evidence that says the debt gravy train must stop.
This will be my last comment on this now too-long thread. First I want to partly defend Jim H. in his effort to defend R&R, although he may not like how I do so. I think a good deal of it has to do with a long past of Krugman misrepresenting things, claiming too much for himself, and engaging in silly and inappropriate attacks on others. What has been annoying to many is that he has gotten away with much of this thanks to a friendly press that decided he was the “emperor” of the new international trade and new economic geography theories, and promoted him as such, even as there were others uncited by him and forgotten in all the praise. He also made a complete ass of himself in his attacks on Laura Tyson 20 years ago when Clinton picked her for a position Krugman clearly wanted, and then there was his inane attack on Brian Arthur regarding increasing returns, which even Ken Arrow took him to the woodshed for. I think that former journalist David Warsh was somewhat siding with R&R in this matter partly because he himself was one of those initially puffing Krugman up back when he wrote for the Boston Globe.
As for what has gone on here, I have already said that I think that Jim overdid his critique of PK, even though PK oversimplified what is apparently a very complicated and mixed situation regarding both the matter of how willing R&R were to share their data and the nature of their interactions with politicians and their specific actual policy recommendations (mostly pretty reasonable). I excuse his intensity on this matter to his awareness of much unpunished past wrongdoing by PK.
However, on the matter of substance regarding fixed and random effects that Jim makes and that Stryker has continued to attempt to pound, there is simply nothing there. In fact in the troubled AER P%P paper R&R used no such methods. The 90% threshold was simply arbitrarily picked out of thin air, or more precisely, picked by a team of IMF researchers working for R&R there wherein the other cutoffs were 30% and 60%. As it is, if one has a roughly straight-line relationship, one will find different averages for the different boxes, and the issue Jim raised about averaging over whole sub-samples versus averaging averages within the sub-samples almost certainly amounts to a big nothing. And, Stryker, the later studies you mentioned all came off from this original one basically using the same flawed methods, a giant pile of now-discredited garbage.
What is most embarrassing to me at this point is not Krugman joking that he is a “meanie,” or well-intentioned and informed people like Jim or Dave Warsh overdoing efforts to stand up for R&R, but the pathetic ongoing efforts by R&R themselves to continue to try to defend the discredited 90% threshold. Their argument amounts to, “not many countries have been above this level, so it must be bad.” I leave this with reminding everybody that in 1815 the UK had a debt/GDP ratio of about 300%, with most of that debt owed abroad (to Dutch banks), with the following century known as the “Pax Brittanica” due to British economic/political/military dominance during that period. Ahem!
Hi Rick,
could you pls. elaborate a bit more about your alco-example. After a few beer, still could not understand that … Thx.
PLEASE NOTE THAT A POST, OSTENSIBLY MADE BY MYSELF AND TIMED AT 14:42, WAS NOT SUBMITTED BY ME.
THE ASSOCIATED E-MAIL ADDRESS IS CORRECT.
I FIND THIS STRANGE AND DISCOMFORTING. CAN SOMEBODY PLEASE CLARIFY EXACTLY WHAT IS GOING ON HERE?
Barkley,
There is nothing there on fixed vs. random effects? Are you kidding? I explained the issue non-technically and now I’ll do it technically. (Nergy, better have a few more beers!) Here is how 2slugbaits should have answered my question:
Let Y(i,t) be the GDP growth rate of the ith country at time t and consider the following model:
Y(i,t) = a(i) + e(i,t)
where e(i,t) is a random error term with mean zero and constant variance u. To be specific, assume that there are M countries where i = 1 to M and that the ith country has Ti data points. The total number of data points T = T1 + T2 + … TM
HAP make the assumption that a(i) = a, an unknown constant, and estimate
Y(i,t) = a + e(i,t)
The optimal estimate of a is obtained by minimizing the sum of squared residuals between a and the Y(i,t), i.e., choose the constant a to minimize the function
sum(over all T points) (Y(i,t) – a)^2
To find the minimum, differentiate the sum with respect to a and set to 0. We get
sum(over all T points) 2(Y(i,t) – a) = 0
or
sum(over all T points) Y(i,t) = sum(over all T points) X a
or
sum (over all T points)Y(i,t) = T X a
so
a = sum(over all T points)Y(i,t)/T
This is the HAP estimator.
The fixed effects model is just slightly harder. We must minimize the sum of squared residuals with respect to all the a(i), i.e., choose the a(i) to minimize
sum(over all T points) (Y(i,t) – a(i))^2
= min [sum(over all T1 points) (Y(1,t) – a(1))^2
+ sum(over all T2 points) (Y(2,t) – a(2))^2 + ..
+ sum(over all TM points) (Y(M,t) – a(M))^2
To find estimates for each a(i), we need to take the partial derivatives of this expression with respect to a(1), a(2), … a(M) and set to zero.
We then have M equations with the ith equation being
sum(over all Ti points) Y(i,t) = sum(over all Ti points) X a
or
sum (over all Ti points)Y(i,t) = Ti X a
or a(i) = sum (over all Ti points)Y(i,t)/Ti
Thus, each of the a(i) is the individual country average. R&R calculated each individual country average and averaged. This, shows that R&R used a fixed effects estimator.
I would also have accepted other answers from 2slugbaits. If he didn’t understand what I just did, he could have opened an econometrics book and looked at the general linear model
y = Xb + e
where y is a (T X 1) vector of dependent variables, X is a (T X M) data matrix, and b is a M X 1 coefficient vector. e is a T X 1 vector of error terms.
The optimal estimate of b is
b = inv(X’X)X’y
where X’ is the transpose of the matrix and inv() is the matrix inverse operator. If 2slugbaits had specified the matrix X properly, he could have worked through the simple matrix algebra and shown the correct result.
Another elegant demonstration would have been to use GMM.
I wrote this out so that people can see how far 2slugbaits was from answering the question I posed. There is so much smoke being blown by Krugman’s defenders it’s hard to breathe and now you are doing it too.
Did you even read the papers I gave you? They are very different from R&R.
I am with Barkley on this.
Mr Stryker you do not appear to understand the word smear!
As promised I have publicised this as well as two of Menzies’s articles
here.
Please note your blog is highlighted on my side-bar so I hope you see it is an argument amongst friends
Rick Stryker You have completely wasted your time. First, what you described is not what normally passes for a fixed effects panel model. For starters, you either have to establish a separate dummy for each panel (which eats up degrees of freedom) or you have to subtract the global mean from each observation before regressing…which eliminates any time invariant variables and is one of the reasons why random effects models are preferred. You didn’t do either one, so yours is not a fixed effects model. So what you seem to think is a fixed effects model is not. Second, neither HAP nor R&R ran a fixed effects panel model. R&R just lumped things into four buckets, took a simple average of each country’s observations within each bucket, and then took an average of the country averages. That’s it. That’s all they did. HAP skipped the second step and just took an average of all observations within each bucket. That’s why I said that if you wanted to replicate what HAP and R&R and do it in an overly complicated way by treating it as a regression, then you should just regress the observations and against a constant. Which is exactly what you did:
HAP make the assumption that a(i) = a, an unknown constant, and estimate
Y(i,t) = a + e(i,t)
where “a” is a constant. But there is no need for a “t” subscript unless that is supposed to represent one of four buckets…in which case the natural choice would a “b”.
But why would anyone in his right mind do that? Why not just say “take the average”? Third, you have obviously misinterpreted what JDH was talking about when he said that R&R took a panel approach. He clearly did not mean that literally…and if he did then he should have his license revoked. JDH meant that the R&R approach captured the intuition of a fixed effects panel model in that it tried to pick-up each country’s unique features.
And really…we all know how to derive a regression using matrix algebra.
BTW, plenty of new number crunching on the R&R data came out today…and all of it crushed the core of the R&R argument. See especially Miles Kimball’s work. And he originally very sympathetic to R&R’s position but has reluctantly concluded that their work is deeply flawed and worthless.
Nottrampis,
Well let’s see. A smear is a concerted and conscience effort to malign the character, reputation, and credibility of a person or group. It is characterized by distortions, half-truths, or outright falsehoods and is designed to make the victim defend his own reputation rather than the policy he was promoting.
Hmmm. Well the shoe does seem to fit. But you might be right. I guess “slimed” could work also. Maybe “mud slinging” would be more accurate? How about “swift boated?” “Character assassination?” I guess “yellow journalism” is a possibility too since this happened on the pages of the NYT.
2slugbaits,
I’m certainly wasting my time trying to explain this to you. You are back to your semantics. What we call the models doesn’t matter. I stated precisely what the models were and asserted that estimation of one would lead to R&R and a special case of that same model would lead to HAP. I challenged you to derive the estimators and confirm or deny my claim. I could see that you didn’t seem to understand and wanted you to demonstrate some comprehension of these issues. I was very clear in what I asked you to do. You couldn’t do it. I gave you over 2 days. Now, I’ve shown you exactly how to do it and you still don’t understand. You obviously know nothing about the issues you comment about, not that that stops you or any of Krugman’s other defenders.
The point of all this was for you and others to see that when HAP and Krugman claimed that R&R did something “odd” and an “error” they were just flat out wrong. But you will not to see.
This is why I keep talking about Krugman zombies. The level of illogic and irrationality is breathtaking.
My last contribution here as well.
It is not a smear to say the data was not widely available.
It wasn’t. That has been conclusively shown here.
Hendon and co had to go through a zillion hoops to get the data.
If people cannot get round that then they simply have comprehension problems.
Saying some data was left out is not a smear either UNLESS it is alleged it was done so deliberately.
I have yet to see anyone allege that. Sloppy is the word used and it is the correct word.
Finally if you are alleging some-one is smearing then you have a duty to show it. You simply cannot say they smeared by doing …
You have to show the smear.Quote them
2slugsbaits is correct Kimball etal , Dube etc all show the causation the other way.
Rick Stryker Case in point: I just wrote down a very specific statistical model to illustrate to one of Krugman’s more vociferous defenders how wrong Krugman is to claim that R&R’s estimation technique was “odd.”
Well, here’s one reason that it’s “odd.” Look at the error term in your convoluted regression to estimate a simple mean. What does the error term mean when there is only one observation, which is a real problem with some of the countries? And why would you take an average when the population is one? That’s odd. The R&R approach is also wasteful of information because it effectively throws away the country specific variance. That’s a bad feature of any model. The HAP model at least doesn’t throw away information. Finally, presumably the reason R&R put countries into different groupings was that those groupings captured what R&R already considered the relevant differences between countries. If they truly believed that, then why bother trying to tease out quasi-fixed effects? And if you don’t believe your grouping is sufficiently homogeneous then why not make a new grouping? Are differences between the US, UK and Canada so important that it makes sense to throw away information in order to capture what are likely trivial fixed effects differences?
This whole sorry saga of R&R is reminiscent of a similar issue with Martin Feldstein back in 1974 in which he claimed to show that Social Security reduced private savings. Like R&R, he used these results politically to push his pet cause, in his case a campaign against Social Security. Once again, years later, two other economists, after a long struggle to get the data, found a coding error in the computer program which when corrected, caused the claimed results to disappear.
Rick Stryker I stated precisely what the models were and asserted that estimation of one would lead to R&R and a special case of that same model would lead to HAP.
So what’s your point? What I said was that the way you were approaching this is flat out stupid and convoluted. It is certainly possible to take data into something like EViews, run it through a pooled cross-sectional fixed effects model and get an answer that exactly matches the way R&R and HAP did things. But you will get exactly the same answer by taking an average of each cross-sectional unit and then taking an average of all cross-sectional units, which is how R&R actually did their analysis. Now if you want to call the former exercise a fixed effects approach, then be my guest, but it’s a mighty odd one. When people talk about fixed effects models they usually have in mind a model that has slope coefficients as well as just a constant and fixed effects deviations. No one estimates two-dimensional pooled cross-sectional data as a special case of a fixed effects model. JDH was not saying that R&R actually did anything as stupid as run the numbers through a fixed effects model. JDH’s point was that their approach tried to capture some of the intuitions of a fixed effects approach, but they did so in a more straightforward way; i.e., just simple averaging in Excel. Doing things your way makes about as much sense as wanting to go from New York to Chicago by heading east. It’s possible to do that, but not very bright. Same with your crazy example of finding a simple mean by regressing against a constant. Yes, you could also call a simple mean a special case of a linear regression, but no normal person would do that…except I will note that you in fact did just that. Go figure.
With your latest tangent I take it that you have given up trying to defend R&R’s analysis. Both of their key points have been fatally undercut. The 90% “threshold turns out to no threshold at all. And the causality issue has also collapsed. Not only do high debt/GDP ratios fail to predict lower future growth, weak exogeneity tests failed to show a causal relationship.
2slugbaits,
I know this is a waste of time to continue to discuss this with you, but for the benefit of whoever is not bored silly with this and wants to learn something, I’ll try again.
The question I want to answer is, “Is Krugman right that R&R used an odd estimation technique?”
In order to answer that, we need to understand what the underlying assumptions are in each estimation method. So we need to write down the conceptual models that are equivalent to the estimation techniques. I’m not saying that R&R and HAP literally ran these conceptual models using statistical software, but rather these models are equivalent to what they did. The advantage of writing the models down is that we can see the underlying assumptions clearly.
I asserted that the R&R method is equivalent to estimating the model
Y(i,t) = a(i) + e(i,t) (1)
and averaging the estimated a(i). I also asserted that the HAP method is equivalent to estimating the model
Y(i,t) = a + e(i,t) (2)
If we can agree on that, then we can immediately see that HAP is a special case of R&R in which all means are assumed to be equal. We can also see that if anyone is making an odd assumption in cross sectional data, it’s HAP not R&R. We need to resolve this question because Krugman has asserted yet again in his latest post the unsubstantiated claim that R&R used an odd estimator.
You responded to this argument with a series of points that were irrelevant. For example, the fact that R&R and HAP didn’t literally run these models is irrelevant to the argument.
To keep us on track, I narrowed the point to just the question of whether the models I wrote down are equivalent to the estimators as I asserted. I asked you to derive the estimators. That way, it’s clear whether I’m right or not. If you derive the estimators and show that they aren’t equivalent to R&R and HAP, then my argument fails. But if you derive them and get HAP and R&R, then you will have demonstrated to yourself that a key assertion in my argument is correct.
But despite my request, you did not derive the estimators. Instead, you responded again with the irrelevant point that R&R and HAP didn’t actually run these estimators. At this juncture, I realized that you really don’t understand the point at all and can’t derive these simple estimators. I was frankly annoyed that I was wasting my time with you. I was also quite irritated that you are attempting to defend Krugman when you don’t understand these issues at all.
I gave you a day before I said anything. I thought that you might try look up the solution in an econometrics book. After 2 days, I did the derivation for you.
Amazingly enough, despite the fact that I laid out the derivations for you, you are still fundamentally confused. That’s why you need a conceptual model–to avoid confusion. For example, in your penultimate comment you said
“The R&R approach is also wasteful of information because it effectively throws away the country specific variance. That’s a bad feature of any model. The HAP model at least doesn’t throw away information.”
If you look at the models I wrote down and understand the derivations, then you can see that this statement is wrong. Look at the random effects model I wrote down:
Y(i,t) = a(i) + e(i,t)
where now the a(i) are iid random variables with mean a and variance v. Now, let v, the country specific variance, go to zero, i.e., throw away the country specific variance. What do you get? Not R&R as you claimed, but HAP!!
I think the argument I have laid out is exactly what JDH was saying. It must be frustrating for him too to watch this. He can blame me for not being clear enough in explicating it but his original point on fixed vs. random effects is absolutely right.
Also, I noticed that Krugman has backed away from one of the elements of his and HAP’s smear in his latest post, and is now saying that R&R’s excluded data was not intentional and perhaps unavoidable. But he still is claiming that the R&R estimator was “odd.” I wonder if he will back away from that assertion too? He should back away from both completely but that’s not enough. He should apologize.
Yes, R&R gave links to the data that they used for their paper. The page that you link to has links to four spreadsheets that contain the Debt-to-GDP ratios and to a Angus Maddison spreadsheet that contains the GDP data they used to calculate GDP growth. However, they provided no clear description of exactly how they used that data to arrive that the numbers in their paper. The only way that Herndon was able to replicate their numbers was to request the spreadsheet containing their calculations. Why couldn’t they have supplied this spreadsheet to begin with? If they had, others would have likely found the highly questionable items described in the next paragraph.
At the URL linked to my name below, I’ve posted all 71 data points that R&R used plus 39 that were excluded (25 from Belgium due to the infamous Excel error and 14 others). The R&R weighting gave equal weighting to 7 countries. One of those countries, the U.S., had only 4 data points and one, New Zealand, had only one data point! As can be seen, the -7.6% growth in New Zealand in 1951 was preceded in 1950 by 14.7% growth and followed in 1952 by 4.3% growth. Did R&R do anything to correct for this obviously unrepresentative outlier or even mention it in their paper? No! This is why we need to demand that all calculations (i.e. the spreadsheets) for any economic papers that is to be taken seriously be released to the public. Why aren’t they already? I suspect that one reason for that is that they don’t want any more number crunchers looking at their work. Peer review is good for catching some things but public release is invaluable for catching many other things, especially basic mathematical errors. If we consumers of economic studies start to ignore those studies for which the calculations are not made public, I suspect that economists will be more than happy to “show their work”.
Rick Stryker To be honest, your comments just don’t make any sense. You say that what HAP and R&R actually did is irrelevant, and then go on endlessly about something that you admit neither one of them actually did.
It appears to me that you did not include the very last step that R&R did, which was to apply your HAP formula to the results for you R&R formula. If I’m understanding your notation correctly you are only doing one pass against the data in your R&R reconstruction. What you seem to have forgotten was that R&R applied a second pass in which they took a simple average of each country’s average.
Look, here’s an example of what each did. Here’s data for each country within one of the four buckets. Note that country B only has one observation
Country…..Obs#1…..Obs#2…..Obs#3….Obs#4
…A………1.2…….2.3……-0.7……3.3
…B………0.5…….na……..na…….na.
…C………3.3……-0.7…….2.3……1.2
The HAP method would give us an average growth rate of 1.41. Under the R&R method the average growth rate is 1.18.
The problem with the HAP weighting is that there might be fixed effects differences between the 3 countries (in this case they are also negatively correlated). In effect, the HAP method treats each observation as unique…the data might as well be observations against completely different countries. That’s the nature of pooled cross-sectional approaches. The problem with the R&R approach is that an outlier gets the same weight as a country with many observations. Whether or not an observation is an outlier is important information which is discarded in the R&R approach. In the R&R weighting a country with 19 observations get exactly the same weight as a country with only one observation that also happens to be a clear outlier even for that country. Instead of using 20 observations R&R would use 2.
I don’t believe you understand how to do a fixed effects model using pooled cross-sectional data. If R&R had regressed each year’s growth rate against each year’s debt/gdp ratio by country and pooled them under a larger group called “advanced countries,” then this would have been a true fixed effects approach. The model output would be a constant, a slope coefficient, and fixed effect deviations for each country. The fixed effects coefficients would sum to zero because the global mean is subtracted from the observations. All of the countries would share the same slope. The only difference would be in the position of the implied intercepts (i.e., the constant plus the fixed effect deviation). Now it is true that you could run a trivial two dimensional fixed effects model with a zero slope and only a constant, but it will give you exactly the same answer as just taking simple averages. But that’s pointless. It’s like going to Chicago from to New York by heading east.
Do you actually know what a random effects model is? I don’t think so. It sounds like you think the difference between a fixed effects model and a random effects model is that the random effects model includes a variance term and a fixed effects model does not. This is just wrong, wrong, wrong. Just to make things very simple, a fixed effects model is one in which differences are either aborbed in dummy variables or (equivalently) by first subtracting the global mean from each observation. Most software does the latter because it saves degrees of freedom. One of the downsides is that all of the time invariant factors are also absorbed in the global mean. But there is still an error term in a fixed effects model and it has the usual interpretation.
A random effects model treats the error term as a composite of the usual error term and a random intercept. A random effects model has a lot of desirable features, but can only be used if the explanatory variable is uncorrelated with the composite error. Typically you first run a Hausman test to check for this.
The question I want to answer is, “Is Krugman right that R&R used an odd estimation technique?”
Then try focusing on just that question. I have already listed some of the reasons that what R&R did was considered odd. One of the reasons is that what R&R actually did was at odds with the plain English interpretation in their paper. It was also odd in giving outliers equal weight with countries that had many observations.
Rick Stryker I might have just had one of those “AHA!” moments regarding what you’re trying to say. So hopefully we won’t be talking past one another. You’re right that what HAP did is equivalent to Y(i,t) = a + e(i,t). That’s a very convoluted way of expressing it, but you would get the same answer as just taking the simple mean, which is what HAP actually did. And it is true that R&R do something that is mathematically equivalent to Y(i,t) = a(i) + e(i,t). Again, a very convoluted way of expressing it, but mathematically it will get you there. So at this point what you have is HAP giving you a global mean and R&R giving you country specific means. Agreed? Where you went off the tracks and what took me awhile to understand was that this is where you stopped your analysis. The problem is that this is not where R&R stopped their analysis. R&R then took each one of the country specific means and computed the mean of those means; or if you prefer the convoluted formulation, they applied a second filter equal to the HAP Y(i,t) = a + e(i,t). And it’s that second pass filter that is what’s getting them in hot water. That’s the weighting approach that is considered odd. That’s where they effectively throw away all of the interesting information. Instead of taking an average of observations, they are taking an average of the first statistical moments. The second moments have an entirely different interpretation. And this is also why the median is their preferred estimate…because the mean value is essentially meaningless. Still, that’s the one they talked about. But even the median has problems because it loses the information captured in the within country variance.
2slugbaits at June 1, 2013 08:41 AM: No, what you wrote down is the HAP conception of the question. RR’s and my conception is Y(i,t) = a(i) + e(i,t) where the goal is to calculate the average value of a(i) across i. The HAP formulation assumes that a(i) = a for all i, and that’s one way you could choose to proceed. But it’s surely not the only way, and it’s very inappropriate to suggest that any alternative to imposing a(i) = a is misleading, unscientific, or in any way wrong. I maintain that the best way to proceed is to allow a(i) to be different for different countries i but also take into account that you have more observations on some countries than others. I gave what I regard as the correct formula for doing that in comments in one of the earlier RR threads. The correct jargon to describe the estimator I proposed is a “random-effects estimator”.
Rick Stryker No, what you wrote down is the HAP conception of the question. RR’s and my conception is Y(I,t) = a(i) + e(I,t) where the goal is to calculate the average value of a(i) across I. Note that there is only one subscript. They treated each observation as a separate and unstacked cross-sectional data point.
No, this is just flat out wrong. HAP effectively did this:
Y(i) = a + e(i)
and then stopped. Note that there is only one subscript. They treated each observation as a separate and unstacked cross-sectional data point. In this formulation they are regressing each observation against a constant, which is equivalent to what they really did, which is to compute a global mean weighting all observations equally.
What R&R did was a two-step procedure. In the first step they did something equivalent to this:
Y(I,t) = a(i) + e(i,t)
This is equivalent to computing a separate mean for each country. Note: two subscripts. It’s basically what you do when estimating deterministic seasonal dummies. You just regress each country’s set of observations against its own constant. That would give you the separate country values that you see in R&R’s table.
So far, so good. No problem. And if R&R had stopped there they would have been okay. But they didn’t stop there. The number that got all the attention by Team Paul Ryan was the second step. And here R&R did the mathematical equivalent of this:
Y(f) = a + e(f)
where the Y(f) where the “f” subscript denotes each country’s quasi-fixed effect mean. Note: only one subscript. Using your preferred and convoluted approach, they effectively regressed each country’s mean against a constant in the same way that HAP effectively regressed each observation against a constant. But with the HAP approach what you get is an average of all observations. With the R&R approach what you get is an average of each country’s average. That is an odd weighting technique by anyone’s standard. Try and put it in baseball terms. Imagine a National League pitcher that has 100 at bats over the season and has 10 hits. His average is 0.100. Then an American League pitcher has one at bat in an interleague game and gets a lucky hit, so his average is 1.000. Using the HAP methodology the average across both pitchers is (10 + 1) / 101 = 0.109. Under the R&R approach the average across pitchers would be (0.100 + 1.000) / 2 = 0.550. Is that a meaningful average? No. The weighting is bizarre. I don’t think you could convince Theo Epstein that this is how he should evaluate hitting talent. And this is exactly why JDH does not defend R&R’s mean calculation. Instead he focuses on the median, which is more meaningful; but he mainly wants to just concentrate on the individual country differences. That is afterall the main reason you do a fixed effects analysis. The main interest is in finding how each country’s intercept deviates from the global mean.
Given that you were demanding Krugman apologize, and given that you clearly did not understand what R&R were actually doing, I think it’s only right that you use this forum to publicly apologize to Krugman. But I won’t hold my breath.
2slugbaits at June 1, 2013 04:12 PM: I believe you were responding to me, not to Rick Stryker.
The random effects approach views a(i) as having some distribution across countries. The mean of that distribution is denoted a. You are correct that HAP insist that the model is Y(i,t) = a + e(i,t). You are also correct that RR’s estimate is not the optimal estimate of a under a random effects specification. But you are not correct if you suggest that there is something ill-specified or inappropriate about a model in which Y(i,t) = a(i) + e(i,t) and for which the object of interest is the average value of a(i) across countries i.
JDH As I said, R&R did do the formulation you presented. But they also went further than that and presented a single average of the country averages. That is the point of contention. And that is where they got into trouble because it was that single number and not the country specific value that got all of the attention. I agree that had they stopped at Y(i,t) = a(i) + e(i,t) they would have been in better shape. But they didn’t stop there. Instead they chose to come up with a single number and in coming up with that single number they gave each country’s fixed effect estimate the same weight. When Krugman and Noah Smith and DeLong and Konczal and Thoma and Waldmann and God only knows how many others all complain about the odd weighting, that’s what they are talking about. And surely you know this.
The correct jargon to describe the estimator I proposed is a “random-effects estimator”.
Yes. If you will check further up this thread that is exactly what I said. You also said that you thought the R&R fixed effects estimator had some advantages over the HAP formulation. I also said that a random effects approach is generally preferred, but is only applicable in special cases.
JDH But you are not correct if you suggest that there is something ill-specified or inappropriate about a model in which Y(i,t) = a(i) + e(i,t) and for which the object of interest is the average value of a(i) across countries i.
At this point I’m not sure which posts are in the queue and which ones are not. Anyway, to answer your question, see my baseball pitcher reference whenever it does show up. If you have a lot of observations for each country, then you could probably get away with taking an average of an average. But when a lot of the countries only have one or two observations and it is known that some of those observations were extreme outliers and one-off problems (e.g., New Zealand), then taking an average of an average is dubious. I believe that is why you prefer to present R&R’s median value rather than their mean value. The median is a little less prone to those kinds of problems.
2slugbaits,
In your “AHA” comment and subsequently, you seem to think I neglected to include the R&R step of averaging the individual country means. But, to quote myself,
“I asserted that the R&R method is equivalent to estimating the model
Y(i,t) = a(i) + e(i,t) (1)
and averaging the estimated a(i).”
In JDH’s comment to you, he suggested that you go back and look in the previous comments for his description of his proposed estimator. You really should do that before commenting further. JDH’s comment was in response to reader AS’s request for more details and is in JDH’s post following HAP’s response. In that comment, JDH showed that the HAP and R&R estimators were limits of either the within country or across time variance going to zero, respectively, in the random effects model. Thus, R&R is in no way “odd” or wrong.
Since JDH’s comment was brief, allow me to expand on it. I’ve danced around this with simpler models so far, not wanting to make it complicated, but it seems we must. Consider the random effects model
Y(i,t) = a(i) + e(i,t)
where now a(i) are random variables with mean a and variance v. e(i,t) has mean zero and variance q.
The first thing to realize is that OLS is not the optimal estimator of the parameter a. The reason is that the random variable a(i) acts across the ith country. As a result, the variance covariance matrix has terms v off the diagonal and we can’t use OLS. We must use GLS, generalized least squares, for estimation.
One way to do GLS is to transform the model so that the variance covariance matrix satisfies the requirements of OLS. You can do that by premultiplying the model by a matrix
(1 – theta)P + Q
where theta = 1 – q/[sqrt(Tv^2 + q^2)]
P + Q = I, the identity matrix, and the matrix Q transforms the model to the fixed effects estimator.
JDH focused on 2 cases. In the first case, let v = 0, i.e., the within country variance is zero. Then theta = 0, and the transformation is P + Q = I, the identity matrix. That means that v = 0 corresponds to OLS, which is the same as HAP. I made the same point more intuitively to you in a previous comment when I mentioned that HAP not R&R throws away the within country variance.
The other extreme is to let q, the across time variance, equal zero. In that case theta = 1 and the transformation is Q. That corresponds to the fixed effects case, which is R&R. Note that you will also get R&R if v or T is large.
I had not wanted to get into this kind of technical detail since it’s not really appropriate for comments on a blog in which most readers are non-technical. But it seems to me important that you realize the depth of JDH’s point if you want to continue to defend Krugman. You’ll need to show that this analysis is wrong if you want to keep this up. Anything else, baseball analogies, etc. is irrelevant.
I hope that you will at look at this seriously and realize that HAP were not thinking very carefully when they claimed that R&R’s estimation technique is wrong. And Krugman isn’t thinking carefully on this issue either.
Sorry, in my comment above I treated v and q as standard deviations rather than variances in the formula for theta. The formula should read
theta = 1 – sqrt(q)/[sqrt(Tv + q)]
Rick Stryker I agree that in a random effects estimator OLS is not appropriate…as I said, the error is a composite of the usual error term and the intercept. So yes, you use GLS. No one is disputing this. In fact, if you’ll look up this thread you will see that I was the first to point that out. I also said that you cannot always use a random effects approach if the data refuse to cooperate, which happens all too often.
I made the same point more intuitively to you in a previous comment when I mentioned that HAP not R&R throws away the within country variance.
Sorry, not buying it. Remember, in their study R&R simply took an average of each country’s debt/GDP. Then they took an average of each country’s average. If all you know is the average of the average, the only thing you can tease out is the variability of that final average. You have lost the within country variability, which is valuable information. And it’s especially valuable when it turns out that many of the within country averages were driven by extreme outliers. If you think I’m wrong, then show us how you can recreate the within country variability using only the within country means and using R&R’s actual methodology. Show us the numbers.
Finally, I would submit that even JDH has backed himself into supporting a position that he originally did not support. On several occasions JDH emphasized the R&R median as the “preferred” statistic over their mean numbers. Indeed. And with good reason. So I find it a little hard to reconcile his former commitment to the median and his later comment “you are not correct if you suggest that there is something ill-specified or inappropriate about a model in which Y(i,t) = a(i) + e(i,t) and for which the object of interest is the average value of a(i) across countries i.”
Consciously deflecting attention away from the mean and towards the median is evidence that in this particular case there is something ill-specified and inappropriate in R&R’s weighting technique. If each country had 30 observations, then there wouldn’t be a problem with taking an average of within country averages. But one of the reasons for attacking their weighting was that there were too many cases of single outliers being the only within country observation. Here’s Mike Konczal’s comment on this:
In case that didn’t make sense, let’s look at an example. The U.K. has 19 years (1946-1964) above 90 percent debt-to-GDP with an average 2.4 percent growth rate. New Zealand has one year in their sample above 90 percent debt-to-GDP with a growth rate of -7.6. These two numbers, 2.4 and -7.6 percent, are given equal weight in the final calculation, as they average the countries equally. Even though there are 19 times as many data points for the U.K.
Now maybe you don’t want to give equal weighting to years (technical aside: Herndon-Ash-Pollin bring up serial correlation as a possibility). Perhaps you want to take episodes. But this weighting significantly reduces the average; if you weight by the number of years you find a higher growth rate above 90 percent. Reinhart-Rogoff don’t discuss this methodology, either the fact that they are weighing this way or the justification for it, in their paper.
Konczal complaint wasn’t that R&R just did an average of an average, but that they took an average of an average when that approach was clearly inappropriate. It would be the same thing as my baseball pitcher example. If you don’t find that baseball example odd, then my advice is that you shouldn’t hang around the phone waiting for Theo Epstein to call.
Lord 2slugbaits,
When invited to look at JDH’s actual argument and respond to it, you didn’t do it. I also wrote out the argument in a different way, arriving at the same conclusions, and you didn’t respond to it. The reason is that you can’t. You couldn’t do the simple derivation I asked you to do and you have no clue what JDH is saying, which is far more sophisticated than that. You are just throwing out technical terms to give the impression that you understand what’s going on. But it’s clear you don’t.
As an example, you deny the point that you get HAP when the within country variance goes to zero in the random effects model. But JDH showed that fact in his discussion of the formula for the weights in his comment on random effects. I also showed it above by pointing out the pre-multiplication matrix in GLS is the identity matrix in that case. And it’s intuitively obvious. If the random variable a(i)’s variance goes to zero, it becomes a constant, which is OLS, i.e., HAP. The fact that you still can’t see this after all this discussion shows that don’t have the slightest idea what’s going on.
I think you have distinguished yourself among Krugman zombies by showing a particular immunity to facts, logic, and reason. But when you deny even mathematical results themselves to defend your master, then you’ve earned the right to become a Krugman zombie lord. Congratulations Lord 2slugbaits.