Sunday, June 13, 2010

Stimulus is Perennial Bad Advice


Nobel prize winner Paul Krugman may say it's econ 101 to have the federal government spend more in recessions, and I agree, but it's also wrong. The highly rigorous yet ultimately naively simplistic macro models have not been useful in rectifying this incorrect intuition, which remains dominant. Gone are the days like in 1920, when during a nasty recession, the federal government cut expenditure by half from 1920 to 1922, and the economy rebounded nicely. Historically, whenever we are in a recession, economists generally call for more fiscal spending, all the better if it is deficit financed. Yet as Milton Friedman noted, "The fascinating thing to me is that the widespread faith in the potency of fiscal policy … rests on no evidence whatsoever." Evidence no, theory, yes (currently, unemployment is 9.7%)

You can create neat models where you have consumers maximizing their utility over time, and producers maximizing profits, and if you made some very heroic assumptions, you can anthropomorphize this situation as one single individual, supposedly represents millions of people acting this way. Add some assumptions about time series shocks to productivity and tastes, and you can then calibrate the interesting variables(output, investment), and say "I have a viable model of the macro-economy".

This always seemed a bit strained, but it highlights a lot of models I see, where one evaluates a model based on it being able to generate output with similar 'moments': means, variances, covariances. Correlation does not imply causation, but it's dirty little secret that probably 90+% of actual science in practice involves mere correlations with ex post theorizing. I remember most of my fellow grad-school classmates thought these models were rather naive, as if anthropomorphizing the economy as a single person explains growth rates and business cycles, but over time those who stayed with the program got sucked in like the way individuals are absorbed into the Borg collective, and the rest of us, most of us, decided to chuck macro entirely.

Many scientists spend their life on these models, where an intuitive story motivates an algorithm--ideally one with maximizing behavior by decision makers and a supply=demand condition--that is then jerry-rigged to fit the output. It's very easy given the many degrees of freedom, to generate models that seem to work really well because it can create time series that when you eyeball them, looks like a desired time series (this seems to encourage a lot of econophysics efforts). This is a really bad way to evaluate theories, because it's too hard to falsify, and too easy to convince yourself the model works.

Note how totally irrelevant the representative agent model was to the latest recession. No one thinks some mechanism, highlighted by the representative agent framework, was really important. To some extent, the low interest rates of 2002-5 were relevant, and this effect falls within the representative agent models, but I don't think this was really essential part of the crisis, and the representative agent models don't make this connection clearer or more measurable.

There have been several Nobel prizes in this modeling dimension (Lucas, Kydland, Prescott), and I think they were valiant efforts, but it's important to recognize failures and move on, because wasting time on a dead end only helps aging researchers who are incapable of developing a new toolset maintain their citation counts. I'm rather unimpressed by the common tactic of using citations to validate research: suggestive, but hardly definitive. Unfortunately, bad ideas are rarely rejected but rather orphaned, ignored by young researchers. After all, proving something is wrong is hard, especially if that something is not so much a theory (the representative agent model), but rather a framework. It's best to look at the gestalt and say, I'm just not going there, and leave it at that. So, the models of risk and expected return persist because they are intuitive on some level, and can generate results that seem to match broad patterns. They merely can't predict, nor do they underly actually useful fiscal policy.

The worst thing is the focus upon mathematics can allow a researcher to avoid these problems, because there's always the hope that even if one's model dies, the mathematical techniques used in creating it will live on. It allows one to revel in rigor, and dismiss any particular model's empirical failures. Many economists I think dream of being like string theorist Ed Witten: Fields Medal winner and creator of understanding an applied subject (not coincidentally, an untestable theory).

The insiders will say, you're just a player-hater and don't understand let alone appreciate the math. It's true critics will not understand the tools as well as the true believers, because if you think it's a waste of time, you aren't going to get really good at these tools. So the bottom line is, do they generate useful predictions? Do large banks hire top-level macroeconomists? No and no. Hard core macroeconomic theory has been pretty irrelevant to the latest recession, as it always has been.

For example, the stimulus bill was debated by famous economists Paul Krugman (trade theorist), who argued for government spending money on anything, and Robert Barro (macro theorist) saying the government should cut taxes. Robert Barro may be a full-time macroeconomist who imposes more of the representative agent logic on his theories, but his advice is considered as relevant as the musings of Paul Krugman, who has never done actual macroeconomics at the peer-reviewed level (in the words of Barro, 'His work is in trade stuff. He did excellent work, but it has nothing to do with what he's writing about.') But that's the result of having a framework that explains everything and nothing, not even economists feel constrained by it. They know, that even though it's not their full time gig, it's no constraint on their opinions, it's not like some macroeconomist will embarrass them by pointing out a flaw in their logic. Such theory can justify any standard Keynesian, Monetarist, or Laissez-fair prejudice, making it irrelevant.

10 comments:

Unknown said...

Very nice post. I was an engineer and initially pursued a PhD in a classic economics department. I have A's in all my microtheory, math, and game theory courses. However, the macro courses infuriated me for glossing over the huge assumptions made in model development that were mathematically beautiful but held no relationship to the real world. I have since switched (happily) to an applied economics department which focuses on real world problems, and only uses theory as needed to complement problem solving. While I enjoy the mathematical structure of economics, too many economists wish they were mathematicians instead of social scientists.

Thanks for the post.

Anonymous said...

Your points about mathematical formalism are well taken, but your post is based on the debate about fiscal policy, and it is simply not accurate to portray it as a debate between Friedmanite empiricists and Keynesian theorists. I mean, really, you think that all of economic theory is just a distraction from the fact that the Harding administration definitively settled fiscal policy once and for all? There is theory and evidence on both sides of this debate.

Anonymous said...

It's worse than that. The default position of a rigorous theoretical model is Ricardian equivalence, i.e., Keynes was wrong. Fiscal policy is ineffective and monetary policy affects only the price level. You can modify the model to get effects that look sort of Keynesian, but you can't rescue most of the corpus, because it really doesn't make sense.

Of course, that's just theory, but empirically, fiscal policy has never been shown to work. Every instance wherein a fiscal stimulus was followed by economic growth also had a monetary stimulus to confound it. The reverse is not true: there have been effective monetary stimuli (Roosevelt's devaluation is the shining example, as Scott Sumner has pointed out repeatedly) not matched by fiscal policy. The reasonable conclusion is that monetary policy sometimes works, fiscal policy most likely doesn't.

Despite his economics being mostly wrong, Keynes is still popular because lowbrow versions of his theory provide cover to spend other people's money.

The above was cut and pasted from a comment I left on the Marginal Revolution blog.

Anonymous said...

as economics is not a science where one can undertake controlled experiments to prove causality, the honest answer is that, "we don't know" whether the fiscal multiplier is even positive let alone above one, or whether ricardian equivalent holds (etc).

but that doesn't get you tenure. or a column in the NYT/WSJ.

Drewfus said...

In Australia, there is widespread belief that the federal governments 'stimulus' spending saved the country from recession. Technically, recession was avoided in Australia. That most of the world engaged in deficit spending with often poor results, seems to have escaped the attention of the media, and supporters of the government.

Nobody wants to be seen to be against the stimulus spending because it's been regarded as a winner. Rather like the lead-up to the GFC, no one wants to spoil the party, or in this case, the now high confidence levels in the governments demand management.

Btw, there's a huge debate going on 'down-under', regarding the governments proposed Resource Sector Profits Tax - 40% tax on all profits above the long-term bond rate, and 40% compensation on all losses. Would be interested to hear your thoughts on this, Eric.

Charles said...

Any increase in the structural deficit or the long term burden of government on the private economy will cause consumers to save more and entrepreneurs to invest less. It just is not the case that deficit = stimulus. When we look at the Obama stimulus package we see economic relief for the unemployed, which has positive and negative aspects to it but little stimulus, we see tax cuts in the form of rebates without prospective cuts in marginal tax rates, which do nothing but move debt from private balance sheets to the national balance sheet, we see payoffs to Democratic constituency groups, we see subsidies for wealth destruction and for shifting around transactions that would happen anyway, we see money for the states to postpone a needed scale back in expenditures. None of this stimulates the economy. A small amount of money will go to shovel-ready construction projects, some of which need to be done and some of which represent pointless boondoggles that have negative long term budget impacts.

Add to a non-stimulating stimulus a tsunami of new regulation, huge minimum wage increases, long term increases in taxation of wealth, investment and high incomes, cap and trade, card check and Obamacare and you have a prescription for a depression.

Anonymous said...

I'm confused...how does one rail against the mathematical intricacies of modern macro and use Krugman as an example?

What are you really arguing? You haven't actually shown that stimulus is bad advice and in fact, you've engaged in the same linear correlation/causation fallacy that you are claiming other to have engaged in.

You take one data point (unemployment) - say "unemployment hasn't gone down, therefore stimulus doesn't work." And then go on a diatribe about mathematics in economics (something Paul Krugman has, in fact, spoken out about already in his column - http://www.nytimes.com/2009/09/06/magazine/06Economic-t.html?_r=1&em=&pagewanted=all). Perhaps you should read more than just a cursory snippet of Krugman.

Now lets address your ridiculous (an overly simplistic argument made by right wing pundits all of the time). Perhaps we should look at the negative effects of state belt tightening - http://capitalgainsandgames.com/blog/bruce-bartlett/1791/fundamental-fiscal-error

Perhaps the way business is hiring has become much more lean - squeezing productivity out of less workers.

We might also say that there are structural problems with our financial sector which make it difficult to get financing. Coupled with high debts and low expected inflation, spending is subdued.

Perhaps the Fed should have a higher inflation target? Gauti Eggertson provides compelling evidence that the New Deal's price floors were actually expansionary - http://www.imf.org/external/np/seminars/eng/2009/macro/pdf/gbe.pdf - because they successfully increased the velocity of money when monetary policy was no longer effective (at the zero lower bound). Perhaps the fed should commit to buying more long-term, unusual assets to bring down the real interest rate (which by all accounts should be negative right now).

I just can't see how one can suggest that giving people work - and money - will not increase employment at all...this is just absurd. It is to suggest that spending money - and creating demand - does nothing and that employment is some exogenous random process where managers decide to hire and fire willy nilly.

You're right, stimulus may have not brought down unemployment, but it hasn't really added much to our structural deficit, and is financed at an extremely low rate of interest. The public and private consensus among forecasters (moody's, MacroAdvisors, many macro economists) seems to be that stimulus does in fact work. While there may be offsetting factors it does not kill the need for stimulus. It, in fact, severely increases the need for stimulus.

Eric Falkenstein said...

My point is that macro has not affected perenial debates, intuitions. A science should sharpen our focus. Instead, it's the same stuff argued in the 1930's (that I think stimulus is bad advice is really incidental to that main point).

Frank said...

Uh...that's quite an obtuse point to make...

I mean, I could take your point and suggest that Keynes ultimately had it right.

Instead you chose to go the route of suggesting that it's all wrong because we don't have a definitive answer. I think there is a clear answer - some of us just don't want to believe it because they have an inherent bias against government intervention of any sort.

The fact of the matter is, suggesting somehow that spending an assload of money won't do anything is denying the entire foundation of economies... you might argue at the margin (which most economists do), but to outright dismiss that the government putting 800 billion into the economy will do absolutely nothing (or worse, hurt the economy) is scary silly.

Anonymous said...

Topical research from HBS:
http://hbswk.hbs.edu/item/6420.html