Bloggingheads and lampposts

by Henry Farrell on December 18, 2006

I’m on “bloggingheads again”: with Dan Drezner. Dan and I had a long discussion about Krugman and whether or not academics should get engaged in broader political debates, dipping into Krugman’s recent “piece”: on inequality as we went along. One of the things I mentioned was the bit in Krugman’s _Peddling Prosperity_ where he talks about the way in which people can cherry-pick economic statistics in order to prove what they want to prove. Krugman is talking about aggregate growth statistics, but nonetheless the point travels.

by choosing your years carefully and talking a good game, you can seem to prove whatever conclusion you like … We learn that a clever propagandist, right or left, can always find a way to present the data on economic growth that seems to support her case. And we therefore also learn to take any statistical analysis from a strongly political source with handfuls of salt. Someone once said about partisan analysis that they use economic data the way a drunkard uses a lamppost: for support rather than illumination (Peddling Prosperity pp.110-111).

Cue “Alan Reynolds”: in comments at Mark Thoma’s place, defending a rather dubious-sounding WSJ “editorial”: attacking claims that income inequality has been growing since _1980._

there is no clear evidence of a sustained and significant increase in inequality since 1988 by any other measure. I very carefully did not say there was no such evidence about 1981-87.


Reynolds goes on to defend his choice of periodization, but it would appear that he has a bit of a “track record”: (to extend Krugman’s metaphor) for employing lampposts not only to provide support, but to meet those other needs and imperatives that drunks are subject to while weaving their way home after a convivial evening.

(Note by the way that Krugman’s criticisms come in the midst of a longer discussion of how chancers at think tanks rather than proper economists have come to dominate debate; while academic peer review doesn’t serve as a perfect protection against this sort of cherry picking, it does make it considerably more difficult to get away with).



leederick 12.18.06 at 3:13 pm

This reminds me of a paper by Danny Dorling about the Labour Party website on “What Labour’s done in your contituency”.

The idea is you type in when you live, and they give you various indicators telling you how how things have got better. And they do tell you that things have got better (everywhere on every indicator) because if things didn’t get better they used a different time period or changed the geographical area to one where things had improved.


P O'Neill 12.18.06 at 4:28 pm

I hope you weren’t billed as the Mickey Kaus stand-in again.

Seriously, your point reminds of a recent example from someone who should know better — Ed Prescott, Nobel Prize and all, writing in last week’s WSJ (11 Dec, no free link)

Myth No. 2: GDP growth was extraordinary in the 1990s. Even though I referred to the expansion of the ’90s as a boom, inasmuch as it was a period of above-trend growth, and I noted the strong gains due to unmeasured investment, we have to put things into historical context. So let’s return to the data. GDP growth relative to trend in the early 1960s was 12%, and in the famous 1980s boom (from the end of 1982 to mid-1989) it was a very impressive 9.7%.

And how about the boom from the previous decade? From 1996 to 1999, GDP grew 3.8%, about in line with the 3.9% growth of the early 1970s and less than the 5.5% growth of the mid-1970s expansion. Even when we account for unmeasured investment and add four percentage points, the 1990s growth spurt — fueled by rapid growth in tech industries — still falls short of the 1980s boom and does not approach the 1960s, both of which were fueled by tax cuts.

So he’s dating the Reagan boom from the trough in 1982, but making Bill Clinton wait until the recovery was already mature before starting the count.

I don’t know whether Prescott tries this stuff on the academic side as well, or if he’s just something he thought would impress the true believers at the Journal.


tebbitt 12.18.06 at 4:40 pm

It’s easy to get over $200,000 in income with two wage earners in a household.

— Ed Prescott


radek 12.18.06 at 5:01 pm

Is the interview in the Republic where Prescott says this available anywhere online for free?


P O'Neill 12.18.06 at 5:23 pm

Here’s a closely related version of Prescott’s WSJ piece that’s free access. It’s not quite the same piece but the same dodgy dating is there.


Ken Houghton 12.18.06 at 5:32 pm

Doesn’t appear to be anywhere but the Republic’s archives.


Barry 12.18.06 at 5:39 pm

Trough-to-peak vs last few years to peak. It was a very, very common thing in the WSJ editorial page all through the 1990’s. It made Reagan look really good, especially as the 1982 recession was extremely severe. There’s no excuse for it anywhere in a business paper, and Prescott is merely showing that the Nobel in economics is awarded for *a* thing that was done, and not for character.


Jane Galt 12.18.06 at 6:21 pm

If you’d bothered to actually look at the figures, you’d find that Mr Prescott was going out of his way to be charitable by cherry-picking the fastest years of the Clinton expansion. Depending on whether you pick 1992 or 1993 as your base year, Mr Clinton’s full term–all of which took place during the recovery, which started in 1990–averaged either 3.45% growth, or a hair under 3.7%.


Barry 12.18.06 at 6:23 pm

Um, trough-to-peak, Jane?


jj 12.18.06 at 7:36 pm

There is a French (France-based, anyway) democratic theorist by the name of Slobodan Milacic who writes about public opinion as the presumed expression of the collective will and the problems with this. Among a great many are the fact that something like 90% of all public opinion polls are never published. Those that we do see are cherry-picked and presented for very specific political purposes.


saurabh 12.18.06 at 8:09 pm

Um, isn’t the appropriate name for that site “Bloggerheads”? What gives?


P O'Neill 12.18.06 at 11:46 pm

So apparently it’s news to The Economist that trend growth rates should be calculated from similar points in the economic cycle. Heh Indeed.


Brandon Berg 12.19.06 at 4:25 am

I very carefully did not say there was no such evidence about 1981-87.

I know you linked to the post and all, but I still think it’s bad form to quote that sentence, which sounds rather weaselly in isolation, without quoting the sentence immediately following it.


stostosto 12.19.06 at 7:28 am


If you’d bothered to actually look at the figures, you’d find that Mr Prescott was going out of his way to be charitable by cherry-picking the fastest years of the Clinton expansion. Depending on whether you pick 1992 or 1993 as your base year, Mr Clinton’s full term—all of which took place during the recovery, which started in 1990—averaged either 3.45% growth, or a hair under 3.7%.

OK, I bothered to look them up. The numbers for US annual real GDP growth 1980-2005 from The Economic Report of the President 2006:

1980 -0,2
1981 2,5
1982 -1,9
1983 4,5
1984 7,2
1985 4,1
1986 3,5
1987 3,4
1988 4,1
1989 3,5
1990 1,9
1991 -0,2
1992 3,3
1993 2,7
1994 4
1995 2,5
1996 3,7
1997 4,5
1998 4,2
1999 4,5
2000 3,7
2001 0,8
2002 1,6
2003 2,7
2004 4,2
2005 3,5

In terms of which president presided over which record when, in average annual growth:

Reagan I (1981-1984): 3,075
Reagan II (1985-1988): 3,775
Bush Sr. (1989-1992): 2,125
Clinton I (1993-1996): 3,225
Clinton II (1997-2000): 4,225
Bush Jr. I (2001-2004): 2,325
Bush Jr. II (2005- ): 3,5

For reference:

Reagan I+II: 3,425
Clinton I+II: 3,725

Of course, GDP growth is only a very partial story. To see whether the growth is healthy and macroeconomically sustainable, you need to factor in unemployment, inflation and, notably, the budget and the current account deficits. In that light, the Clinton terms outperform the others by a wide margin. The Republican record of the last three decades is one of borrow-and-spend. This goes emphatically for the current president whose economic policies are borderline disastrous.


stostosto 12.19.06 at 7:31 am

Oh, bugger. The preview played tricks with me. Here is the first series from above again:

1980: -0,2
1981: 2,5
1982: -1,9
1983: 4,5
1984: 7,2
1985: 4,1
1986: 3,5
1987: 3,4
1988: 4,1
1989: 3,5
1990: 1,9
1991: -0,2
1992: 3,3
1993: 2,7
1994: 4
1995: 2,5
1996: 3,7
1997: 4,5
1998: 4,2
1999: 4,5
2000: 3,7
2001: 0,8
2002: 1,6
2003: 2,7
2004: 4,2
2005: 3,5


Barry 12.19.06 at 8:29 am

Thank you very much for that work, stostosto. I’d just like to add that:

1) The trough-to-peak measures for Reagan give him extra credit, because the 1982 recession was the most severe since the Great Depression (imagine what the first two terms of FDR would look like!).

2) Reagan did what he did through increasing the national debt by amounts thought preposterous. Molly Ivins said that she figured, way back when, that the one good thing about electing Reagan was that a Republican would take care of that nasty $70 billion deficit. And he did. Being able to incur debts of 1-3% of the GDP each year certainly helps one look better. Meanwhile, George HW Bush and Clinton had to take the hit of reducing the defict – note, with the vociferous opposition of the econopundit wing of the GOP. If Megan McArdle had been writing back during 1992-3, she, rather than Phil Grahm, might have come up with the saying ‘this budget is a one way ticket to a recession’.


aaron 12.19.06 at 9:42 am

13. That’s common form here. Always click the link. Often, they do not even support the conclusion the author is drawing.


aaron 12.19.06 at 9:45 am

Sometimes they are even counter to it.


Henry 12.19.06 at 10:33 am

aaron, what a _surprise_ to see you again. When last heard of, you were complaining (with rather idiosyncratic punctuation) that Crooked Timber had lost all value as an academic blog and exiting stage right. I guess that you were just impelled back by stupid, like the moth to the flame.

brandon – I did say in the post that Reynolds follows with a defence of the periodization. But the point I was making is that _even on his own terms_ the stats don’t do what he assiduously implies that they do – i.e. provide evidence that inequality hasn’t increased since 1980. A half-sentence qualification in the original piece – something along the lines of “Although inequality may have increased between 1980 and 1987 because of the recession” – would have made the article substantially less hackish. But he didn’t do that, and I don’t think that it’s at all unreasonable to question his motives in not so doing. As the BdL post that I link to demonstrates, he has a long track record of making bogus and mutually contradictory claims about the statistical evidence in order to suggest that inequality isn’t increasing. When someone’s basic claims are fixed, but the purported empirical justifications for those claims are sometimes nonsensical, and swing wildly back and forth, there’s strong _ex ante_ reason to suspect that this person isn’t committed to evidence-based reasoning but instead, as the Krugman book suggests, to using the evidence (and where necessary abusing it) to support a predetermined position in a rather hackish manner. Which is all to say that I might have been prepared to take his defence of the periodization as a bit of justified special pleading if it had been referenced in some way in the original article, or even if it had come from someone who was committed to figuring out the truth, albeit with clear ideological commitments. But when it comes from someone who demonstrably (and I don’t see how one can easily defend some of his antics in the past) doesn’t have such an interest, it’s so much camouflage and persiflage.

To put it another way – I had a conversation with a prominentish right of center blogger recently where he suggested that we had a broadly similar approach to understanding the world, but disagreed strongly about the evidence. The implication of this, as I understood it, being that if he demonstrated to me that one of my beliefs was empirically unfounded, I would change my belief accordingly (and the converse would be true too). I hope that’s right – clearly it’s always hard to change beliefs in practice, but it _is_ possible. But I don’t think it would be right for Reynolds – if past practice is anything to go by he would stick to his position, and try to find another empirical justification, however bogus, for maintaining it.


anon 12.19.06 at 3:42 pm

Have think tanks recently emerged as dominant players in policy debates? Is academic peer review a suitable check against cherry picking – or an ideological thumb on the scale? There certainly seems to be lots more tankers and tanks but peer review only screens out those things the peers get peeved about.

I don’t quite trust economists with the topic of inequality, income or economic more generally. Partly because economists aren’t trained to see it. Theoretically, income inequality is assumed natural – as in the natural distribution of skill or whatnot. And theory becomes a moving target. Inequality is no big deal insofar as it represents just reward, productive incentive, and the workings of competition. Wasn’t there an AER article some years ago defending inequality? And economists do theory smoke and mirrors and let policy talk pass as professional econ talk – as in inequality isn’t the problem so long as no one is worse off. So inequality isn’t the issue so much as it is poverty. And of course poverty is about what people bring to the market (see above).


Sebastian holsclaw 12.19.06 at 6:17 pm

“there is no clear evidence of a sustained and significant increase in inequality since 1988 by any other measure. I very carefully did not say there was no such evidence about 1981-87.


The standard inequality storyline is that it has been increasing (and alarmists suggest greatly increasing) as a function of some underlying phenomenon in the way that the US currently does things. I would say that no evidence from 1988 to present (18 years) suggests that the standard inequality storyline is incorrect. Do you disagree?


Henry 12.19.06 at 8:15 pm

Sebastian – even this interpretation only works on Reynolds’ own terms. The debate goes as follows. Most everyone in this debate – including respected right wing economists (see the Mark Thoma post referenced above) believes that income inequality is increasing, and use income tax data to discuss these issues. There is serious econometric work (way too serious for me to be able to evaluate professionally – but trustworthy people say it’s legit) by Picketty and Saez showing increases in inequality over time. Reynolds points to genuine problems in income tax data – but says that we should be using census data, which is way more problematic (e.g. it counts everyone with more than $1 million income as having $1 million income) instead. He also ignores the various technical fixes that Picketty and Saez have employed to try to correct for errors in tax data. There is good reason to believe that he is cherry picking for the results that he wants to find – he has changed his tune on the data pretty substantially (as well as at least one genuinely dunderheaded error). So Reynolds’ claims only work, as I say, on his own terms (if you rely on a set of data which seems to most professionals to be more flawed than what we have for current purposes) and don’t work, contrary to his implications, for a substantial chunk of the period under consideration even then. So the short answer is: yes, I disagree. So do Feldstein, Hubbard, Lazear, Bernanke and Paulson.


Michael Sullivan 12.20.06 at 1:51 pm

Jane Galt says:

If you’d bothered to actually look at the figures, you’d find that Mr Prescott was going out of his way to be charitable by cherry-picking the fastest years of the Clinton expansion. Depending on whether you pick 1992 or 1993 as your base year, Mr Clinton’s full term—all of which took place during the recovery, which started in 1990—averaged either 3.45% growth, or a hair under 3.7%.

The problem is that the figures he’s using aren’t annual, but total. they look smaller because he’s referring only to growth *above trend*, which is positive.

What that means is that a longer boom will have a bigger number. So the fact that he included the whole of the 80s boom and only the last 4 years of the 90s boom makes the 90s boom look smaller for no justifiable reason. If you include the first 4 years (92-95) as well, you probably get a boom of about the same magnitude as in the 80s, maybe a litttle bigger. But there were no significant tax cuts in 91 or 92 either.

You can’t use aggregate numbers when you’re looking at periods of radically different length and expect comparable results, but that’s *exactly* what Prescott did. Only one year in the 1982-89 period outpaced any of the last 3 years of the 90s.

That’s either dissembling or a very amateurish error on Prescott’s part. I vote for the former.


spencer 12.20.06 at 1:59 pm

Sebastian holsclaw — reynolds says

“there is no clear evidence of a sustained and significant increase in inequality since 1988 by any other measure. I very carefully did not say there was no such evidence about 1981-87.

But if you go to the census data you find that from in 1988 the top 5% of the populatin received 18.3% of income. In 2001 this share was 22.4% if you use the actual data and 20.3% if you throw out the 1993 jump he complains about.

Now it looks to me like Reynolds is using the technique of describing the actual data in a way that is very misleading under the assumption that it is safe to do so because noo ne will actually check the data.

Do you agree that an increase from 18.3% in 1988 to 22.4% in 2001 as compared to an increase from 16.5% in 1981 to 18.3 in 1987 — over 4% vs under 2%–is not evidence of an increase in inequality?

Of course this is also inconsistent with census data that shows the share going to the middle three quantiles falling from 49.8% in 1987 to 46.3 in 2001.


MQ 12.21.06 at 1:18 am

Yet more evidence of exactly how reliable “Jane Galt” is when she makes any fact claim that touches on politics at all.

Comments on this entry are closed.