Tuesday, December 15, 2009

Is Nate Silver the World's 22nd Best Economist?

Nate Silver likes to refer to Greg Mankiw as "the world's 23rd best economist." He also seems to relish every opportunity to take him on.

About six months ago, Mankiw made a somewhat disparaging comment about Sonia Sotomayor on his blog:
I once wrote a short paper... based on the premise that there are two types of people: Some save and intertemporally optimize their consumption plans, while others live paycheck to paycheck, spending their entire income as soon as it's received. Sometimes readers of the paper think of the two groups as rich and poor, but that interpretation is wrong. Some people with low incomes manage to scrimp and save (I always think of my grandmother), and some people with high incomes spend most everything they earn.

Apparently, the new Supreme Court nominee Sonia Sotomayor is an example of the latter...
My grandmother would have been shocked and appalled to see someone who makes so much save so little.
Here Mankiw is assigning Sotomayor to the group of individuals who do not "intertemporally optimize" based simply on evidence that her accumulated savings are small relative to her income. Using impeccable economic reasoning, Nate Silver countered as follows:
What are a person's incentives to save, rather than spend, money? The four basic ones are usually these:
  1. To protect against downside in one's income, particularly the risk of being fired.
  2. To save for retirement.
  3. To save for one's family and children.
  4. To save for an expensive purchase, such as a home or a nice car. 
Nos. 1-3 don't really apply to Sotomayor. With the possible exception of being a tenured professor at Harvard, few positions offer more job and income security than that of a justice on the Federal Circuit Court; Sotomayor would have to be impeached by the House and found guilty by the Senate to lose her job, something which has happened only a handful of times in American History. Sotomayor's federal pension is undoubtedly very generous, rendering #2 somewhat moot, particularly as she could also stand to make a significant "post-retirement" income in private practice or on the lecture circuit. And she does not have a children or a husband to support. It would be quite irrational if she had half a million dollars collecting dust and 0.01% interest in her Chase checking account.
Perhaps Mankiw's grandmother would find her more virtuous if she were saving up for a Lexus or a summer home in the Hamptons, but that doesn't seem to be her cup of tea. Her one real indulgence is the apartment she keeps in the West Village. Although virtually anywhere that would be a reasonable commute from her courtroom in Lower Manhattan would be relatively expensive, she could save a bit by living in the Financial District or perhaps in Brooklyn. But Mankiw, who lives in a zip code where the median price of a house is 1.65 million dollars, should not exactly be throwing stones from his undoubtedly very charming, New England Colonial home.
In other words, given her circumstances and preferences, it appears that Sonia Sotomayor is intertemporally optimizing perfectly well.

That exchange was back in May, and there the matter rested until Mankiw published a piece in the New York Times this week arguing for a change of course in fiscal policy:
Congress passed a sizable fiscal stimulus. Yet things turned out worse than the White House expected. The unemployment rate is now 10 percent — a full percentage point above what the administration economists said would occur without any stimulus.

To be sure, there are some positive signs, like reduced credit spreads, gross domestic product growth and diminishing job losses. But the recovery is not yet as robust as the president and his economic team had originally hoped.

So what to do now? The administration seems most intent on staying the course, although in a speech Tuesday, the president showed interest in upping the dosage. The better path, however, might be to rethink the remedy.

When devising its fiscal package, the Obama administration relied on conventional economic models based in part on ideas of John Maynard Keynes. Keynesian theory says that government spending is more potent than tax policy for jump-starting a stalled economy.

But various recent studies suggest that conventional wisdom is backward...
These studies point toward tax policy as the best fiscal tool to combat recession, particularly tax changes that influence incentives to invest, like an investment tax credit. Sending out lump-sum rebates, as was done in spring 2008, makes less sense, as it provides little impetus for spending or production.
This was too much for Nate to resist. In his response he notes first that tax cuts were indeed a large component of the original stimulus package.
First, let's take a look at what the stimulus package was actually designed to consist of... Some $288 billion -- 37 percent -- consisted unambiguously of tax cuts that were labeled as such. Meanwhile, $274 billion -- 35 percent -- consisted of traditional, Keynesian-type investments in infrastructure and related areas. These types of spending are easy to characterize...

Indeed, it's probably fair to think of the stimulus as consisting of about half tax cuts. About 37 percent of the stimulus consisted of tax cuts which were labeled as such. But also, some fraction of the 10 percent labeled as transfer payments functioned like tax cuts, and some fraction of the 18 percent allocated to state and local governments had the effect of offsetting tax increases (or perhaps financing a tax cut proper in a few cases).
Does this mean that tax cuts are ineffective in stimulating production? Not really:
In other words, the stimulus -- in terms of its potential impact on the economy thus far -- looks at least as much like Mankiw's alternative as the "conventional economic models" that he trashes.
All of which might indict Mankiw's thesis -- except that he also may be wrong about his other conclusion: that the stimulus is not working. Mankiw cites studies describing the impact of various types of stimulus on GDP -- not unemployment -- and as we pointed out yesterday by GDP standards the stimulus has done quite well, with 3Q growth coming in at 2.9 percent versus a consensus forecast of 0.7 percent and 4Q GDP poised to print at about 4.0 percent versus a forecast of 1.9 percent...
So, to summarize: Mankiw is wrong that the stimulus consists mostly of Keynesian-type investments. So far, it has been closer to the tax cut end of the spectrum. But he's also wrong that the stimulus is not working. By the benchmark that he implicitly endorses -- GDP -- it's done very well. Mankiw is so wrong, in other words, that he may actually be right: the stimulus looks a lot like one he might have designed, and it's helping the economy.
As Nate concedes, the substance of Mankiw's argument (that tax cuts are more effective than spending increases in stimulating production) may well be correct. The empirical studies cited by Mankiw do seem to support his position, and do indeed run counter to the conventional Keynesian view. This is an important debate to have, both for immediate policy purposes and for our theoretical understanding of the manner in which fiscal policy operates. But there seems to be no reason to make it a partisan exercise.

What can one learn from all this? First, that couching an economic argument in overtly moralistic or political terms can considerably diminish its impact. And second, Nate Silver may well be the 22nd best economist in the world.


Update (12/16): There's yet another noteworthy exchange between our two protagonists, dating all the way back to January 2009. It started with an article by Mankiw arguing (just as he did in his more recent piece) that the government spending multiplier was much smaller than conventional wisdom would have us believe, while the tax multiplier was significantly larger. Silver responded by claiming that the empirical findings on which Mankiw's argument was based could not be generalized to the current economic environment because they referred to the effects of exogenous changes in  taxes and expenditure. Mankiw replied that without focusing on exogenous changes one could not possibly hope to identify causal effects of any kind. In this case both were making valid points: Mankiw about identifiability and Silver about external validity, as Andrew Gelman helpfully pointed out.


Update (12/18):  Mankiw makes a much more effective (and considerably less partisan) case for an investment tax credit on his blog.


  1. I advise that you be very careful of deceptive tricks with Greg Mankiw (like, for example, he did here, comparing a number in one study not to a number in the same study, but in a different study), and look at the details.

    Please consider this from Princeton's Ewe Reinhart on Mankiw on this subject (at: http://economix.blogs.nytimes.com/2009/01/16/can-economists-be-trusted/ )

    Writing in The New York Times, for example, the Harvard professor N. Gregory Mankiw, former chief of President Bush’s Council of Economic Advisers, makes a case for stimulating the economy through tax cuts rather than added government spending.

    First, he suggests that government usually spends money on things people do not want or need – like bridges to nowhere, or digging ditches and then filling them in again. To buttress his case further, he then cites an empirical study by Valerie A. Ramey, according to which the $1 of added government spending will ultimately increase gross domestic product by only $1.40, while according to another recent study by Christina and David Romer, $1 of tax cuts over time increases G.D.P. by $3.

    Noneconomists may ask, of course, exactly how a $1 cut in taxes would translate itself into a $3 increase in G.D.P. at a time when traumatized households, whose wealth has been eroded, might use any new tax savings merely to pay down debt or rebuild their wealth through added savings, rather than spend it, and when businesses unable to sell their output even from existing capacity might hesitate to invest such tax savings in more capacity.

    But never mind this fine point.

    More interesting is that Christina Romer is to be the head of President-elect Barack Obama’s Council of Economic Advisers. In that capacity, last Saturday she released an analysis of fiscal stimulus alternatives, with a co-author, Jared Bernstein. Curiously — or perhaps not — for that analysis, the two authors assume a much larger four-year multiplier effect for added government spending (1.55) than for tax cuts (0.98), although they do confess to a high degree of uncertainty on the actual sizes of these multipliers.

    So there you have the flexibility, shall we say, that economists enjoy when they apply their professional skills to affairs of state in what may seem, to outsiders, like purely scientific analyses.

  2. Also, Menzie Chinn makes a crucial point, "It matters what assumptions are made" (at: http://www.econbrowser.com/archives/2009/03/the_great_multi.html ). This is especially true with freshwater research where they can make some amazing whoppers and then interpret the conclusions very literally, or completely literally even.

  3. If tax cuts are so effective, why were the Bush tax cuts followed a mere five years later by the largest economic collapse in 70 years? Why were the enormous tax cuts of the early to mid-20's followed by the Great Depression, again starting a mere five years later? And let's not forget the only significant tax increase of the last 30 years was subsequently followed by the economic boom of the late 90's and the only balanced federal budget in my lifetime.

    I realize economies are complicated and there's far, far more going on than simply tax rates. But when I look at 100 years of data, I see that major increases in X don't seem to have any significant negative affect on Y, and in fact have been followed by greatly improving Y in many cases. Major decreases in X have preceded the two largest drops in Y in the entire data set, and overall seem no more likely to increase Y than decrease it.

    How does one conclude from such information that decreasing X is the best way to increase Y???

  4. Richard, I don't believe for a moment that anyone here is being deliberately deceptive. Mankiw clearly believes that (suitably targeted) tax cuts are a more effective fiscal policy instrument than increases in government expenditure, and that the empirical evidence on the matter supports his position. But his attempt to distance himself from administration policies and to cast these policies in a poor light made his own argument less effective, and left him vulnerable to Nate's critique. That's all I was trying to say.

  5. Another key thing to consider in any analysis of the value of a stimulus measure is what the actual increased(short-run, cyclical)GDP goods will be.

    If a tax cut will increase (short-run, cyclical) GDP by $1 trillion in highly positional/context/prestige consumption goods, like yachts, $5,000 watches, etc., then this will do little to increase total societal utility, as these things have enormous positional/context/prestige externalities. They're very zero sum utility game, and provide relatively little additional non-positional, or intrinsic, utility. Moreover, they are of little productive investment value.

    On the other hand, if a fiscal stimulus creates the same $1 trillion in (short-run, cyclical) GDP, but it is in things like alternative energy, basic medical research to cure cancer, education, smart infrastructure, and the like, these essentially non-positional/context/prestige goods would create far more total societal utility, and be far more productive investments, and thus would increase GDP far more over the long run.

  6. Mankiw has a history of being intellectually dishonest when he puts his pundit hat on. Fortunately I had already read all four of the papers he referred to that purportedly show tax cuts are supperior as fiscal stimulus to spending. They are popular fodder for those who suffer from tax cuts can cure any economic problem dementia. Two of the papers Mankiw (purposefully?) missinterprets and the other two suffer from theoretical and empirical defects as I will now note:

    1) The Romers' paper did not suggest that tax cuts were uniquely effective. They used special techniques to classify tax changes based on intent that have yet to be applied to spending changes. Only noncountercyclical tax changes that decreased a fiscal balance were found to be statistically significant in their effect on GDP growth. Not only is this not a supply side result the paper suggested that tax cuts to fight recessions had no statistically significant effect on GDP. David Romer described the results as "Hyper-Keynesian."

    2) Andrew Mountford and Harald Uhlig's paper was based on a theoretical Neo-Classical model that assumed Ricardian Equivalence. The assumption of Ricardian Equivalence is virtually guaranteed to achieve the results they reported. Ricardian Equivalence is also not supported by the vast majority of the empirical economic research literature.

    3) Alberto Alesina and Silvia Ardagna's paper was a cherry picked study of 91 fiscal stimuli. Only two of those incidents included a financial crisis/liquidity trap incident (Sweden early 1990s and Japan late 1990s) under which monetary policy has become less effective and spending based fiscal stimuli are empirically known to be highly effective.

    4) Olivier Blanchard and Roberto Perotti's study did not find that tax cuts were more effective than spending increases in increasing output. As they said "the results consistently show positive government spending shocks as having a positive effect on output."

    In my experience it's best not to take Mankiw's punditry word for anything.

  7. We also have an interesting laboratory experiment in progress on the relative effectiveness of tax cuta and spending as fiscal stimuli. The CBO has examined both the 2008 fiscal stimulus and the current one (ARRA).

    They concluded that the $150 billion 2008 stimulus lifted consumer spending by 2.3% at an annual rate in Q2 2008 and thus GDP by 1.6% at an annual rate. The overall effect was to lift GDP by about 0.4% in each the second and third quarters of 2008, or a total of $30 billion. Thus we can estimate that the fiscal multiplier was only about 0.2.


    The CBO just released its estimates on the effect of the 2009 stimulus (ARRA) on third quarter GDP. The midpoint of its estimate range is that the stimulus added 2.2% to quarterly GDP or about $80 billion:


    ARRA is being spent (direct spending + tax cuts) at a fairly steady clip of $100 billion a quarter from Q2 2009 through Q3 2010. Thereafter it will tail off rapidly. Thus we can estimate that the fiscal multiplier for the third quarter alone was about 0.8.

    Its impact however will not peak out until Q3 2010 according to the CEA when it will be roughly 250% of what it was in Q3 2009. The most recent CBO analysis thus implies it should add 5.5% to GDP or about $200 billion on a quarterly basis by then, producing a fiscal multiplier of about 2.0 for that quarter by itself. I estimate at this pace that the fiscal multiplier averaged throughout the three years or so of ARRA should in fact turn out to be about 2.

    So, why is the fiscal multiplier for ARRA turning out to be so much higher than for the 2008 stimulus? Well, for one thing ARRA is being implemented during a financial crisis/liquidity trap which empirical studies have shown results in a higher fiscal multiplier. But more importantly the 2008 fiscal stimulus consisted entirely of a tax cut.

    The IMF has a good summary of the economic research literature on fiscal stimuli (about 200 papers are featured):


    Here is what it says about the empirical research in its conclusion:

    "Estimates of fiscal multipliers are overwhelmingly positive but small. Short term multipliers average around half for taxes and one for spending, with only modest variation across countries and models (albeit some outliers)."

    In short, on average, spending provides double the fiscal stimulus of tax cuts.

  8. Mark, I appreciate your comments on the content of these papers. But as I said earlier, I am not willing to question the integrity of Mankiw, Silver, or anyone else for that matter. And (until I've had a chance to look closely at the issue myself) I will remain open minded about the effects of tax cuts versus government spending.