The Iceland dust-up

by on July 5, 2012 at 8:13 am in Current Affairs, Economics, Uncategorized | Permalink

There has been enough coverage that I won’t summarize the entire debate, suffice to say that Krugman offered a picture like this:

Some writers from the CFR (I am not completely sure how to attribute authorship), offered this picture, along with some analysis and links (and further pictures).  The key point is that the second picture considers a longer time horizon, and all of a sudden the relative performance of the different countries has changed:

They argue that in this light, looking over the longer time horizon, the Icelandic story appears mediocre rather than impressive relative to some of the other small countries.  A few points:

1. Arvind Subramanian argues we should look at per capita growth and also PPP vs. market exchange rates, read here.

2. Arvind’s point aside, that the pictures give different impressions is more important than either picture taken alone.

3. The second picture brings value-added to the debate, and it suggests stories which the first picture taken alone does not.  I’ll come back to that.

4. The debates have mixed together a few different questions, such as “how well have the countries done?,” “how well have the countries’ policies done?,” and “should we be looking at both pictures?”  Since the answer to the first two is obviously agnostic — too soon to tell — I will focus on the last of these questions and the answer is yes, we definitely should be looking at both pictures.

Krugman’s response, once you get past the inappropriate insults, doesn’t serve up much.  He has arguments against the view “Don’t look at the first picture” but no good arguments against “Look at both pictures very carefully and integrate.”

Ryan Avent offers a more polite response here.  He focuses on convergence (maybe we should have expected the Baltics to have outperformed Iceland, so a tie means Krugman wins), but this estimate suggests the fixed exchange rate did not really cost the Baltics much in the way of convergence points.  In any case I fear the goalposts are being shifted and what we know about convergence and its speed is iffy anyway; for instance anyone worried about bad monetary policy, and opposing 1980s style RBC theories, should be a convergence pessimist for the short run.

Most importantly, there is still no argument against looking at both pictures in a serious way.

So what additional thoughts come to mind looking at the second picture, which you might not get so much from the first picture alone?  Here are two:

a. Some countries simply may be more volatile than others, and this may or may not have to do with their policy responses.  I’ll note Cowen’s Second Law, namely that there is a literature on this and the people who work in that literature consider this to be a plausible proposition (which does not mean it is here the operative explanation, however).

b. The size of an initial run-up, possibly bubbly at that, is correlated with the size of the later collapse and also the difficulty of recovering from that later large collapse.

There is a literature on that too, start here, it is not an absurd proposition by any means.

By the way, did I mention a 2009 IMF Staff Report which concluded that Latvian “output exceeded potential by 9 percent in 2007.”?  That supports the relevance of b) and possibly a) as well, and it discriminates against Krugman’s story of the peak being a point reattainable through policy management.

I take Krugman to be suggesting something like c): all the relevant information for understanding performance is contained in post-bust policy, so we needn’t look at earlier years and in fact doing so may mislead us.

Yet this attempts to pre-settle the dispute by putting all of the explanatory burden on post-crash policy.  In general commentators often overattribute results to policies and in any case we should not build in this bias a priori.

Are you a fan of Dani Rodrik’s “every country is different” hypothesis?  If so, you probably should think that both graphs are important.  Country characteristics don’t morph away overnight, so earlier data points should matter for understanding current policy results.

If we look carefully at both pictures, I say we still don’t know what is going on, but we do have a richer sense of the possibilities.

From now on these two pictures should be shown and considered together.

1 Ed July 5, 2012 at 8:19 am

The largest of these countries has a population of 6.3 million. Is there a reason why economists are obsessed with generalizing conclusions from the economic performance of countries whose population is smaller than that of a good sized city?

2 Orange14 July 5, 2012 at 8:38 am

+1, and I’ve been making this point as well. When you compare these countries either alone or in the aggregate to Spain or Italy one is left with the feeling that this is all much ado about nothing. The only thing that can be said in a positive way about Iceland is that they let the foreign depositors in their banks go down with the sinking ship whereas the government of Ireland bailed theirs out (which of course was a big mistake). Regardless of which chart you look at, none of the countries’ GDPs are back to the pre-meltdown levels.

3 R Richard Schweitzer July 5, 2012 at 9:49 am

Ed has it right.

Adjust either or both those displays to GDP per capita; or better yet, extract government expenditures from the GDP and then do per capita

4 TheAngryPhilosopher July 5, 2012 at 10:47 pm

Eh. Maybe mass migration has a role, but if so I haven’t heard anything about it. And barring a sort of labor exodus, switching from GDP to GDP per capita would have precisely nil effect since we are comparing to a benchmark over a very short time period. The real point Ed seems to be making is that the stories of these countries might not be very generalizable to much larger economies like Spain.

5 Merijn Knibbe July 6, 2012 at 3:40 am

See this . Emigration is important in the Baltics (surely when you compare the emigration data with the rather low fertility data). An interesting point: Emigration from Ireland, Portugal or Spain is not so much to other EU countries – but to countries with the same language (in the case of Portugal for instance Mozambique and Brazil).

By the way – potential long run growth for Ireland and Iceland is of coure lower than for the Baltics. Relative wages in the Baltics are much lower than relative GDP, thanks to neo-liberal policies which enable quite some rent extraction by foreign companies. One might consult a ‘classical’ economist (except for his chartal view on money) like Michael Hudson on this.

6 Rich Berger July 5, 2012 at 10:14 am

Sweden’s population is just over 9 million, and it is often used as an example of a successful welfare state.

7 Adam July 5, 2012 at 11:03 am

Because, as Krugman says, you take your natural experiments where you can get them.

8 Richard July 5, 2012 at 8:19 am

Hold on: Iceland has been at the ‘technology frontier’ for a long time while the other countries are basically east-european-catching-up. No wonder that the growth since 2000 is more like Germany and the growth of the other is more like ‘other east european countries’.

Fun part is: Ireland (=looks more like iceland than like latvia, estland and lithuania) has more-or-less the same growth since 2000 as iceland. 😉

9 AlanW July 5, 2012 at 9:23 am

+1 – Although that seems to be Cowen’s point: The second graph may show that these countries’ economies are too different for the first graph to mean what Krugman argues it does. Hard to draw policy prescriptions from “Rich countries are less volatile even in a crash,” other than to become as rich as possible.

10 Richard July 5, 2012 at 8:21 am

And why did you just include lithuania? Paul did not do that?

Why is that Tyler?

11 Danton July 5, 2012 at 9:38 am

Because they should be included with the rest of the Baltics?

12 Nick July 5, 2012 at 8:32 am

If you are talking about post-crisis responses then presumably you should be looking at the peak?

Given the bubble was larger in Iceland than Estonia (I think this is undisputed?) we can expect that Iceland’s GDP vs. potential was overestimated by a similar (or larger) margin.

The “lets go back to 2000” option has no more merit than the “lets go back to 1950” option. Or any other variation thereof.

In which case maybe the peak is the best (only?) point of reference for informing policy?

13 Pat July 5, 2012 at 12:01 pm

Nick, this seems to me a perfectly obvious point, and I cannot for the life of me figure out why Cowen insists on misunderstanding it. Isn’t the whole question about the best way to solve the financial crisis and recession/depression? What happened in 2004 pretty clearly wasn’t based on Iceland and Estonia’s reactions to events in 2007…

14 The Anti-Gnostic July 5, 2012 at 8:35 am

If you’re living beyond your means, then yes, when economic reality catches up with your leveraging schemes, your current GDP is going to reflect the artificial nature of your past boom The piper will be paid.

15 TheAJ July 5, 2012 at 9:45 am

Except it was Iceland, not Estonia or Latvia, that was by far more engaged in leveraging schemes.

16 The Anti-Gnostic July 5, 2012 at 10:11 am

Whatever their bailiwick–inflation, debt, MBS’s, Ponzi schemes, Amway, public works–the fact is they weren’t actually wealthy; they just spent a lot.

Parse the tea leaves all you want. If an individual, household or business can’t get away with it forever, neither can a State. The sum and substance of macroeconomics seems to be people’s desperate attempts to prove otherwise.

17 TheAJ July 5, 2012 at 10:45 am

Oh, my bad, I thought you were actually trying to make some comment relating to, you know, the topic. Maybe discuss why Iceland’s fake wealth didn’t really dissappear while a huge chunk of Latvia and Estonia’s did.

I didn’t realize you just wanted to throw some random financial terms in with Amway and Public works and participate in some good ol fashioned ideological ranting. My bad.

18 derek July 5, 2012 at 10:15 am

Are you certain about that? Iceland had a speculative banking failure that passed liabilities onto the government and taxpayers. From memory, Latvia and other Baltic states had large fiscal deficits and private debt loads often denominated in foreign currencies.

This stuff is silly. Different countries with different sources of economic activity would fail in different ways and require different strategies to come out of it. I know that an economy is the sum of transactions made between individuals and/or businesses each driven by the specific needs and wants of those individuals and businesses. The ability to superimpose a graph on those things is at best interesting and at worst misleading and food for hubris.

19 indrek July 5, 2012 at 5:03 pm

Estonia and Latvia had huge booms right before the crash, this is exactly what the graph starting earlier should show. Private debt went up from next to none to over 100% of GDP during short time, nominal wage growth reached 20% per annum for the boom years. There was a huge influx of money that suddenly stopped, huge real estate bubble that went bust right before the international financial crisis (which could not have come at a worse time for Baltics). No large fiscal deficits though, Estonia had a constant surplus of couple of % of GDP in good years, Latvia a very small deficit – this is why government debt was very low. But current account deficits reached 15%-25% of GDP during the boom years.

I would thus be very careful in evaluating that Iceland was in worse position than the Baltics before the crisis or that the peak GDP has much value as a comparison point.

20 derek July 6, 2012 at 10:13 am

Thanks. Sounds like a description of Spain.

21 CPV July 5, 2012 at 8:44 am

It’s kind of like looking at graphs of high beta and low beta stocks. Here the beta is to the bubble, or overall leverage, I suppose. Looks to me like the return is pretty strongly correlated with the volatility. How about a calculation of “risk adjusted GDP growth” over the period?

22 Cliff July 5, 2012 at 11:36 am

Keep in mind beta is not a real property, it is just an ex-post description of recent volatility. I don’t know if that makes your comparison more or less valid.

23 jdm July 5, 2012 at 8:51 am

“Most importantly, there is still no argument against looking at both pictures in a serious way.”

Isn’t this a shocking statement? The GDP of several small countries are plotted either from 2001 or 2007 to now. Looking at one of another of the plots, one speculates about causal relationships – this line goes above that line and it must be because (or not because) of such and such. Or not.

Doesn’t this procedure seem crazy, like reading tea leaves or bird entrails? Isn’t it embarrassing that the field is in such a state that debates over such meaningless bits of data occur regularly? I’m certain that economists know better – they know that these plots are not based on the results of experiments natural or otherwise, and so one can learn nothing from them about causal relations, and yet despite knowing this, they pretend that by looking at a handful short univariate time series, they can say something meaningful about what caused them to have their particular patterns. I find this very worrisome.

24 Adam July 5, 2012 at 11:07 am

The basic weakness of macro is the inability to run actual experience. So when you get sets of countries that take radically different policy responses, it’s as close as you’re going to get.

25 Adam July 5, 2012 at 11:16 am

Darnit. Experiments.

26 a July 5, 2012 at 9:33 am

Whilst the differences in real GDP growth rates are ambiguous, the difference in unemployment rates is pretty clear.

The unemployment rates in Iceland peaked at 7%-8%, whereas the unemployment rate peaked at 21%-22% in Latvia, and the unemployment rate in Ireland is currently just under 15% and still rising.



27 Mark Thorson July 5, 2012 at 9:39 am

I’m reminded of the great little book How To Lie With Statistics.

28 Noah Smith July 5, 2012 at 9:43 am

I was going to write this post, but Tyler has written it for me. And probably better than I would have done.

29 Orange14 July 5, 2012 at 10:53 am

Get back to work on your thesis!!! 🙂

30 Noah Smith July 5, 2012 at 9:51 pm

Just waiting for my committee to get back in town so I can defend…

31 Uffy July 5, 2012 at 7:04 pm

You’ve got to be kidding.

How is data from 2003 relevant to the “natural experiment” of differing post crisis responses? What of the huge disparity in peak unemployment rate between the two camps?

32 Spencer July 5, 2012 at 9:50 am

For another perspective, start the second chart at about mid-cycle –around the end of 2004 — and a rough eyeball comparisons implies that they would now all be at about the same place. This says the difference between Iceland and the Eastern Europeans really was in the early 2000s when Irish growth was really very weak. If this is so, the post crises results are not significant.

33 Tom Davies July 5, 2012 at 10:02 am

As all the data in the first picture is also in the second picture, how could anyone look at the second but not the first?

34 JohnReardon July 5, 2012 at 10:38 am

Excellent point. Expand it to, why bother with the the one that is fully contained in the other, if not to “lie with statistics” as one commentator above put it.

35 Arthur July 5, 2012 at 10:24 am

I want Karl Smith to get to do a post on this so he can get some sense into the discution.

Look, it does not matter how they were growing before, or even after. It does not matter if fixed currency is good or bad in the long run. GDP growth is not the aim of fixing exchange rate, or not fixing it.

Devaluating is a way to let prices adjust downward. So the labor market clears.

Has anyone looked at their unemployment rates?:

GEO/TIME 2008 2009 2010 2011
Estonia 5,5 13,8 16,9 12,5
Ireland 6,3 11,9 13,7 14,4
Latvia 8,0 18,2 19,8 16,2
Lithuania 5,8 13,7 17,8 15,4
Iceland 3,0 7,2 7,6 7,1

36 JohnReardon July 5, 2012 at 10:39 am

As usual people are being too nice to Krugman (there is no “too mean” to Krugman).

He lies with statistics, then in a blustering nasty way defends those lies. He’s a joke that you all tell too much.

37 Lou July 5, 2012 at 10:52 am

I have to agree with this viewpoint. Krugman and hacks like him have trashed the reputation of economics as an academic discipline.

38 Adam July 5, 2012 at 11:15 am

You know, the more I look at that second chart, the more I conclude that it’s highly misleading. It still shows Iceland being much, much closer to it’s pre-crisis output than the Baltics, which is Krugman’s point. And wasn’t 2000 Iceland vastly richer than the 2000 Baltics, a point which is obscured in this presentation?

What I see in the second chart is three Baltic nations experiencing rapid catch-up growth from depressed East Block economies, perhaps then mixed with some frothiness. Given the difficulty of distinguishing one from the other, I have a hard time seeing the second chart as particularly useful.

39 Lou July 5, 2012 at 10:56 am

I have to confess I haven’t followed this debate, and it’s too much for me to catch up on in a short period of time, but shouldn’t the argument be that a flexible currency would mean the trough is not as bad, because it is a “shock absorber” for the real economy? It has nothing to do with the long run growth of GDP. In that case, I would look at the chart indexed to the peak and conclude that the theory is correct. Generalizing to Greece, it would mean that the ensuing depression would not be as bad if they exited from the EMU.

40 Adam July 5, 2012 at 11:02 am

I think you’ve overstated Krugman. I would take his position to be that he’s trying to isolate the effects of post-crisis policy, thus little is added to that particular exercise by looking at the earlier periods.

Your response, that what came before the crisis may be as or more influential than the policy response is a valid point, but not really a refutation of Krugman’s method. Maybe a bigger bubble is harder to recover from, or maybe Iceland’s policy choice has worked out better.

41 Yancey Ward July 5, 2012 at 11:08 am

Any attempt to defend looking at only the first graph but not the second must fail since the second graph contains all the same information within the first.

42 MikeF July 5, 2012 at 1:15 pm

Any attempt to defend only looking at the second graph but not a graph of GDP since [insert earliest date for which data is available] for those countries must fail for the same reason.

43 TheAJ July 5, 2012 at 11:12 am

I look at the second graph and the second thought that comes to mind is, wow, after all that, all the post-communist opportunities, those two countries were only able to reduce their income gaps relative to Iceland by perhaps 500 basis points. They had somewhat better growth than one of the top 10 richest countries in the world, which you would expect to have slow growth.

44 TallDave July 5, 2012 at 11:14 am

Iceland is too small to draw any conclusions anyway; see #4 and #5 from yesterday.

45 TheAJ July 5, 2012 at 11:22 am

I applaud Tyler for finally having the courage to look at the Baltic states and realize that “maybe every country is different” instead of heralding them as miracles like he used to. He should be thanking Krugman for helping him expand his world view and not fixate on calling out miracles just by focusing on short time frames. For example, nobody would ever call Ireland a miracle based on one quarter’s data, nobody other than a hack or a dishonest liar. Its fruitless sometimes to draw conclusions from small countries, so I’m sure he’ll stop acting like a little liar and stop posting about Latvia and Estonia.

46 Tyler Cowen July 5, 2012 at 11:44 am

Sorry guy, never heralded them as miracles or anything close. Learn how to read! Really.

47 TheAJ July 5, 2012 at 12:23 pm

Why is there such a strong aversion to currency devaluation within the conservative/libertarian spheres? I’m trying but I can’t find anything in the ideology that makes them so stringent on this. Why is that not okay while devaluing wages or pensions is?

48 tt July 5, 2012 at 1:32 pm

because curreny devaluation hurts the Kochs more than the poor

49 Dave T July 5, 2012 at 1:41 pm

The Koch’s are still going to be able to pay rent and buy food after the inflation hits.

50 IVV July 5, 2012 at 3:50 pm

In more down-to-earth terms, savings lose value when there is a currency devaluation, but not when there is a wage/pension devaluation. The overriding concern for a goldbug is that amassed savings will continue to provide the same (or better) expected purchasing value over time. Inflation erodes this value, and so it is abhorred.

There are secondary effects as well, such as a desire to see the profligate suffer–if the spendthrift does not suffer in bad times, or only suffers as much as the saver, it makes the desire to hold onto savings seem foolish. Furthermore, the schadenfreude of saying, “I told you so,” is strong.

51 JWatts July 5, 2012 at 5:21 pm

“Why is there such a strong aversion to currency devaluation within the conservative/libertarian spheres? … Why is that not okay while devaluing wages or pensions is?”

I’m not sure why you think that currency devaluation is considered okay ‘within the conservative/libertarian spheres’ while devaluing wages or pensions.

52 tt July 5, 2012 at 6:48 pm

aversion = “not okay”

53 Benny Lava July 5, 2012 at 7:32 pm

Oh no, not miracles. Just lucky! Hahaha, you cad you.

54 Lord July 5, 2012 at 11:38 am

It is not Krugman that talks about the Icelandic miracle. He is only responding to others that assert a Baltic miracle. Showing their shortcomings may be unpleasant but it is more honest, but when shown their shortcomings, they become defensive and assert Krugman is making the same claims they are. They may be doing what they must but that doesn’t mean they are doing well.

55 JWatts July 5, 2012 at 5:44 pm

“It is not Krugman that talks about the Icelandic miracle”

That’s a silly statement. Here’s the title of Krugman’s article: “The Icelandic Post-crisis Miracle” June 2010

56 Really Curious July 5, 2012 at 12:06 pm

Krugman’s insults notwithstanding, his point still stands: austerity in the Baltics has not performed any magic that simple currency devaluation has not done. Not only that but what it has done is drive enormous percentage of Lithuanians, for example, to emigrate. Clearly, those two facts combined point to it as being a very poor solution relative to currency devaluation.

57 Careless July 5, 2012 at 12:23 pm

A huge percentage of Lithuanians have emigrated and their GDP has increased that much from the bottom? That is very impressive, if true.

58 Lord July 5, 2012 at 12:51 pm

Perhaps what should be said is look, austerity is not a total disaster.

59 Really Curious July 5, 2012 at 3:08 pm

The ones that have emigrated found jobs elsewhere, say Germany, and do send remittances back home as well as do not claim benefits and therefore are not a fiscal drag on the Government. Plus it’s GDP per capita remember ?

60 Careless July 6, 2012 at 3:31 am

No, it isn’t gdp per capita, which is why I was so amazed by your earlier claim.

61 dirk July 5, 2012 at 12:46 pm

What is Cowen’s First Law?

62 MD July 5, 2012 at 1:20 pm

Restaurants with fancy decor and beautiful women drinking cocktails may be relying on decor and women, rather than food, to attract customers.

63 Tony Cohen July 5, 2012 at 1:01 pm

I am a complete economic neophyte with no real knowledge. Can someone explain to me how higher growth rates in developing countries relative to slower growth rates of developed countries before the crash is a rebuttal to the concept that the post crash policies of the Baltic states are worse than Iceland and Ireland’s?

64 byomtov July 5, 2012 at 10:38 pm


65 RZ0 July 5, 2012 at 1:20 pm

When goalposts move, shouldn’t the ground shake?

66 Jason July 5, 2012 at 1:49 pm

The longer term perspective makes the same mistake as is made in this graph (where I commented before):

You should take the log of the data or use a shorter time scale. Logs are how you show growth rates, not normalizing to a given time for long term data. For very short term data it roughly works out (like Krugman, 3 years … do a Taylor expansion), but over any period where growth no longer appears linear (I emphasize *like the graph in this post*), normalizing creates a changing scale over time.

67 Jason July 5, 2012 at 2:17 pm

I’d like to emphasize: I really think this is a serious problem and an impediment to constructive dialog.

Growth rates of NGDP are on the order of a few percent to ten percent per year. That means after 10 years, your parameter r*t is going to be order 1 and the use of a linear expansion (normalizing to compare graphs) is seriously flawed. If you keep this to a couple years, your error is order a few percent (~ r^2*t^2), but over ten years, your error is order 1 (requiring you to include more parameters in your Taylor expansion).

Someone with a more prominent voice than me should address this.

68 Jason July 5, 2012 at 4:32 pm

I can make any of the countries look like they are faring best in the recession by choosing the appropriate quarter to normalize to …

Iceland 2007 Q3
Estonia 1997 Q3
Latvia 2011 Q1
Ireland 2011 Q2
Lithuania 1997 Q2

Note that normalizing the data to 2000 (specifically Q2) makes Iceland look the worst. (I am suddenly not just worried about innumeracy, but very suspicious of deliberate manipulation — what are the chances that the CFR would pick the date that makes Iceland look the worst according to available CFR data? The data goes back to 1997 for all five of the countries depicted, so there are 60 quarters of data to choose as the normalization point and 15 years … why choose 2000?)

Normalizing to peak RGDP doesn’t give one any way to massage the data. The peak is a well defined point.

69 Jason July 5, 2012 at 4:34 pm

*available Eurostat data

70 Lou July 5, 2012 at 4:37 pm

Thanks for that excellent exposition.

71 Jason July 5, 2012 at 5:02 pm


72 Jason July 5, 2012 at 4:58 pm

The most natural choice, % difference from pre-crisis trend, shows Iceland faring best during the recession — by far.

In fact, since choices based on features of the data (trends and peaks) show Iceland faring better, while arbitrary normalization points do not (and can show anyone doing better), Tyler’s assertion:

“3. The second picture brings value-added to the debate, and it suggests stories which the first picture taken alone does not. I’ll come back to that.”

is demonstrably false. The second picture is value-subtracting.

73 indrek July 5, 2012 at 5:29 pm

Why would the % difference from pre-crisis trend be the most natural choice? Half of the answer why the recession might have been so steep in the Baltics is in a huge boom just before the crisis, which would make it one of the worse choices. 10%+ growth rate with 20% current account deficit is not really something to consider normal. Not considering it while evaluating how large was the bust, might be wrong, which is just the point of the longer term graph.

74 Jason July 5, 2012 at 6:21 pm

It is a long term trend over almost 20 years so no particular boom or recession will dominate the data; since 1993 all of the countries listed in the graphs above have very linear trends in log space. I suggest you look at the data … much like US GDP see here:

We are not talking about some non-obvious linear trend.

Your argument if it applied to a fit to all pre-crisis data would basically imply that the entire post-cold war economies of Estonia, Latvia, etc have been a boom that we shouldn’t count.

75 indrek July 5, 2012 at 7:40 pm

I happen to know the data rather intimately (economist from Estonia) and I can assure you – there was one huge overheating going on from ~2005 till mid 2007 with a full scale property bubble and everything. As I mentioned – you do not expect CA deficits above 20% of GDP to be sustainable, CPI inflation as well as wage growth went up to double digits during this period etc. While Estonia had a very respectable growth record previously too (indeed it was the best in the region for long time), it went out of hand during this period. While a difference between 7% and 10% is not very visible on graph, it can be the difference of being able to adjust internally and going crazy.

This might also have been one of the reasons why the austerity measures were feasible – no one considered this boom era levels to be a normal state of the world, there were a lot of warnings in the public during the boom by central bank etc. And this is why the comparison with the peak tends to puzzle us. We would not have called it a bubble if we would have considered it normal.

76 Jason July 5, 2012 at 8:03 pm

Given your information indrek, I did the fit again using data from 1997 to 2004. Previously I had done it from 1997 to 2007. I get the same result. (1997 was chosen because Iceland data only goes back that far.)

The key point is that the variance from a linear trend in log space is not very much (even the overheated economy growth rate is only a few percent higher than the average growth).

This graph shows exactly how linear the data are:

If you have better years for me to choose, let me know and I will run the regression.

FYI, here is the percent difference result (Iceland in blue):

more here:

77 Jason July 5, 2012 at 8:05 pm

Left off the fact that in this graph

Iceland is blue, Estonia is red and Latvia is orange.

78 Indrek July 5, 2012 at 10:27 pm

And now do the same exercise for Ireland considering growth in the late 90-s for example. You see what I mean. Extrapolating the same growth for Estonia would have made her richer than Germany in what – 15 years (considering PPP)? While I would certainly be happy with this, I am finding it really hard to take it as standard scenario.

This is exactly why we need to go a bit deeper into the matter. And why Tyler included the reference to IMF Staff Report warning that Latvia was 9% over its potential output (I think the estimates may have gone even upwards since). The bubble was very very real – do a quick search of “Baltic bubble”, you will find a lot of contemporaneous warnings. From 2001 to 2007 private sector credit grew over 50% points (as in % of GDP) in Estonia, close to 70% points in Latvia. The levels achieved might not have been too exceptional, but there is no way you can expect this kind of capital inflow to continue and project it to the future. At least we did not.

This is what the longer term view should give – the previous growth was at break-neck speed. This was no simple catch-up growth with couple of percentage growth differential which would have been expected (Arvind Subramanian (link in point 1 in Tyler’s posting) did some calculations considering catch-up growth effect as well – definitely be a better basis than pure extrapolation). Even this kind of contraction did not take the Baltics back more than couple of years in time, but if you look a bit deeper into the macro imbalances before the crisis (and also the openness of the economies), you would not be surprised at all that the crisis hit Baltics the hardest. There are not so many good policy measures to keep the number of construction workers up on construction boom levels indefinitely.

79 Jason July 7, 2012 at 6:59 pm

indrek, I understand how different periods in the economic history of any country have different average growth rates that may or may not be sustainable. None of that information is represented well by the second graph in Tyler’s post either!

However, I absolutely disagree with the idea that the “longer term view” expressed above is meaningful. It is arbitrarily normalized to 2000. If I chose a different year, I would get a different result. If I choose 2003, all the countries appear to do the same. If I choose 2000, Iceland appears to be doing terribly. If I chose 2006, Iceland is on top. How can that graph convey information if the information it conveys depends on my arbitrary choice?

1) Talking about shorter term periods near the peak and using the peak as a normalization is mathematically correct, but may obscure the phenomena you describe.

2) Talking about longer term periods and subtracting out the trend is mathematically correct, but may obscure the phenomena you describe.

3) Talking about longer term periods and normalizing to an arbitrary year is **mathematically incorrect** and not only obscures the phenomena you describe but allows the person who made the graph to obscure all kinds of other features. And that is my point.

Any graph will obscure some phenomena. (Even plotting raw data on a linear axis!) The key to representing data in an unbiased way is to a) understand what you are trying to represent and b) describe carefully what you have done.

Subtracting the trend based on all previous data is the most natural representation because that gets at what a recession is: a short term deviation from trend growth. Countries have recessions regardless of being poor or rich, so the level doesn’t matter, only the relative (fractional) deviation.

Normalizing to the peak and looking at short run data is somewhat less natural, but still valid. Normalizing to the peak doesn’t allow the creator of the graph to manipulate the normalization and since short run RGDP data is approximately linear, normalization does not obscure the slope of the curves (it does obscure the intercept, but again in a recession we are worried about relative deviations).

Normalizing to an arbitrary year (2000) doesn’t have any meaning. You have obscured the level and the trend. Since long run RGDP data is exponential, you have obscured the relative sizes of deviations from the trend, **especially at points far from the normalization point**! That means you’ve obscured what you mean by recession!

The only piece of information that is not obscured (actually emphasized) by the second graph in Tyler’s post is the fact that countries grow faster from a low base. That is all it shows. And it’s not even shown well because it doesn’t say that! It would be better shown by a graph of RGDP per capita on a log scale! It shows nothing meaningful about the recession, the relative sizes of booms and busts or even basic information about the level RGDP level.

I don’t have a major opinion about the underlying argument. Icelandic recessions may be less severe for the same reason a boom, a bust or even growth is less severe in rich countries. All fluctuations and rates may be suppressed at high levels of RGDP per capita. As a physicist with a particular fascination with how people obscure information or lie with representations data, I do have a very strong opinion about the second graph in Tyler’s post. It is a misrepresentation. It’s not necessarily a deliberate misrepresentation. Someone probably decided to look at the data and graphed it over a longer period and normalized it at the beginning. It then seemed to show what they already believed and so they put it out in the world without questioning it.

80 indrek July 8, 2012 at 6:29 am

Jason, I am absolutely with you in that who ends up higher at the end of this graph is not meaningful (and I think we agree on most of the other points as well). I could not care less if any other normalization would have been chosen and I would have preferred that real levels had been used instead of change from some point (omitting logarithmic scales would be even bigger problem then of course). It is just about whether there is an additional value from knowing previous growth or not.

Looking only at the short run graphs you could not make the comments Tyler made: possible inherent volatility and high previous growth, which may have added to the deepness of latter fall (nr of years lost was actually rather similar). Indeed we have long considered our economic policy to bring growth that is more volatile but hopefully higher in the long term – fixed exchange rate for a very small and investment-hungry economy is supposed to do exactly this (exchanging some real volatility for higher investors confidence through smaller exchange rate risks). If there are long term considerations, then you should look at the long term.

Yes, there are a lot more that is missing from the picture and yes, none of the graphs is perfect as they are. The longer term view (wherever the reference point) strongly hints that there are other questions to ask, the shorter term one does not.

But I understand you concerns. My whole point in coming to argue was the one about whether there was a separate and clearly unsustainable boom building up before the crisis (which may not have been visible from deviation from the trend as the previous growth was high as well).

81 Jason July 9, 2012 at 4:38 pm

I certainly agree that a longer run of data can tell a more complete story and that there are several economic factors at play.

I believe at the beginning we had a disagreement that comes down to my use of the word “natural” in fitting the data to the previous trend and looking at the percent difference. I meant “natural” in a way that the axes and normalizations didn’t depend on arbitrary points or theories of the underlying data — your point that the trend itself was unnatural, so that a fit to that trend could not be natural.

I guess I should have used the word “naive” instead of “natural”. Or maybe “prima facie”.

The second graph in Tyler’s post is definitely not naive. It either accidentally or intentionally misleads in its point.

82 Benny Lava July 5, 2012 at 9:39 pm

Yeah, Tyler does this all the time. He should leave macro econ to the men and go back to writing about food or handbags or whatever.

83 Brian Donohue July 6, 2012 at 3:40 pm

Your comments are worthwhile and deserving of a response from the author of this blog.

84 RGregory July 5, 2012 at 2:11 pm

I am both appalled by Krugman’s shoddy presentation and the criticisms thereof. What this all points out is just how impossible it is under current standards to falsify any hypothesis in Policy Macro.

85 Max July 5, 2012 at 2:53 pm

When you grow faster you have bigger crashes, whats so hard about this?

86 Firat Uenlue July 5, 2012 at 8:25 pm

Seriously, what kind of debate is this supposed to be? Obviously governments can boost short-term GDP by spending above their means and make themselves look good so that Paul Krugman might dance with them on the equivalent of an economist’s prom night. None of these graphs has any value unless you are willing to dig deep into the data and see how various measures of leverage within the economies have changed. Can we also have a small look at ROI? Where is all that fancy schmanzy math now?

Incredible, by this measure Greece spending itself to bankrupcty would have been superior to a fictional Greece that lived within its means for years and years – until the final dust-up obviously.

87 Ronald Brak July 5, 2012 at 10:09 pm

The best example of how to deal with the GFC is still Australia.

88 Bender Bending Rodriguez July 6, 2012 at 12:17 am

Sell your country to China by the shovel-full?

89 Ronald Brak July 6, 2012 at 8:07 pm

People always bring up mining when I mention Australia. It’s an extremely cyclical industry. Check what happened to the price of coal, iron and other exports to China. The mining sector contracted more than other parts of the Australian economy after the GFC, so Australia’s large mining sector was a minus, not a plus when it came to avoiding recession. Thermal coal fell from a peak of $190 a tonne to $62 dollars a tonne.

90 santcugat July 6, 2012 at 6:48 am

I think a bigger point is that the critics who said that Iceland was dooming itself by not bailing out its banks were obviously wrong.

Comments on this entry are closed.

Previous post:

Next post: