Friday, March 4, 2011

Scientists vs Engineers?

The House Republicans' plan to cut federal spending by $61 billion for the remainder of the fiscal year (i.e., the period between now and the end of October) will be a drag on the economy and reduce employment, according to both Goldman Sachs and Moody's Mark Zandi, who says:
The House Republicans’ proposal would reduce 2011 real GDP growth by 0.5% and 2012 growth by 0.2 percentage points This would mean some 400,000 fewer jobs created by the end of 2011 and 700,000 fewer jobs by the end of 2012.
This shouldn't come to a surprise to macroeconomics students, who know that a decrease in government purchases reduces aggregate demand and - outside of the special "classical" case of vertical aggregate supply - output.

John Taylor disagrees, however.  Ezra Klein explains:
Mark Zandi says the GOP's proposed spending cuts will cost about 700,000 jobs. John Taylor says they will "increase economic growth and employment." Both are respected economists who immerse themselves in data, research and theory. So how can they disagree so sharply?

The dispute comes down to how much weight you give to "expectations" about future deficits. Taylor's argument is that Zandi's model -- which you can read more about here -- doesn't account for the upside of deficit reduction -- namely, that when the government spends less, the private sector will spend more. Taylor thinks individuals and businesses are hoarding their money because they're afraid of the high taxes, sharp spending cuts and assorted other nastiness that deficit reduction will eventually require. "The high unemployment we are experiencing now is due to low private investment rather than low government spending," he writes. "By reducing some uncertainty and the threats of exploding debt, the House spending proposal will encourage private investment."
A similar disagreement is playing out over monetary policy.  In a recent NY Times column, Christina Romer wrote:
The debate is between what I would describe as empiricists and theorists.

Empiricists, as the name suggests, put most weight on the evidence. Empirical analysis shows that the main determinants of inflation are past inflation and unemployment. Inflation rises when unemployment is below normal and falls when it is above normal.

Though there is much debate about what level of unemployment is now normal, virtually no one doubts that at 9 percent, unemployment is well above it. With core inflation running at less than 1 percent, empiricists are therefore relatively unconcerned about inflation in the current environment.

Theorists, on the other hand, emphasize economic models that assume people are highly rational in forming expectations of future inflation. In these models, Fed actions that call its commitment to low inflation into question can cause inflation expectations to spike, leading to actual increases in prices and wages. 
She sides with the "empiricists" and argues that the influence of the "theorists" has held the Fed back from taking bolder, more effective action.  Stephen Williamson begs to differ:
Romer says some things about economic history in her piece, but of course she is very selective, and seems to want to ignore the period in US economic history and in macroeconomic thought that runs from about 1968 to 1985. Let's review that. (i) Samuelson/Solow and others think that the Phillips curve is a structural relationship - a stable relationship between unemployment and inflation that represents a policy choice for the Fed. (ii) Friedman (in words) says that this is not so. There is no long-run tradeoff between unemployment and inflation. It is possible to have high inflation and high unemployment. (iii) Macroeconomic events play out in a way consistent with what Friedman stated. We have high inflation and high unemployment. (iv) Lucas writes down a theory that makes rigorous what Friedman said. There are parts of the theory that we don't like so much now, but Lucas's work sets off a methodological revolution that changes how we do macroeconomics. 
The divides between Goldman/Zandi and Taylor over fiscal policy and between Romer and Williamson over monetary policy both reminded me of Greg Mankiw's distinction between "scientific" and "engineering" macroeconomics. The models used by the "engineers" - the people in Washington and on Wall Street who need to make practical, quantitative assessments of the impact of policy alternatives on the economy - are more elaborate versions of the "textbook" Keynesian IS-LM aggregate supply and demand framework that most of us (still) teach our macroeconomics students.  As Williamson points out, the models used by academics - Mankiw's "scientists" - in our research are fundamentally different.

The engineering models are built on relationships among aggregate macroeconomic variables like the Phillips curve, which relates inflation and unemployment, and the consumption function, which connects consumption and disposable income.  As Williamson alludes to, Robert Lucas and others won a methodological war in the profession (or at least the academic branch of it) in the 1970s and 1980s.  The result of their victory is that the macroeconomic models published in leading journals today are expected to be grounded in the optimizing, forward-looking behavior of rational individuals.

Such individuals might believe, for example, a reduction in government spending today implies that their future taxes will be lower (because the government will be servicing a smaller debt burden).  The resulting increase in their lifetime disposable income means that they will immediately increase their consumption.  So any negative impact of a cut in government purchases is offset by an increase in consumption.  Rational, forward-looking optimizers might also recognize that any monetary expansion will erode their real wages and demand an offsetting increase in nominal wages.  This means that employment will remain unchanged (because the real cost to the firms of a worker is the same) even as inflation rises.

At its most extreme, the assumption of dynamic optimization under rational expectations was once believed to imply the Lucas-Sargent "Policy Ineffectiveness" proposition, which Bennett McCallum explained in a 1980 Challenge article:
Macroeconomic policies - sustained patterns of action or reaction - will have no influence because they are perceived and taken into account by private decision-making agents. Thus, the adoption of a policy to maintain "full employment" will not, according to the present argument, result in values of the unemployment rate that are smaller (or less variable) on average than those that would be experienced in the absence of such a policy.
Of course, in a world of rationally optimizing people, where prices adjust to clear markets, it is hard to explain how we could get to such large deviations from the natural rate of unemployment in the first place...

More generally, while macroeconomic science has continued on the methodological path established by Lucas, many of its practitioners have worked to re-incorporate real effects of monetary policy.  This is a large part of the "New Keynesian" project, which is arguably now the reigning paradigm and best hope for reuniting "science" and "engineering" (and arguably is as much "monetarist" as it is "Keynesian").

Fiscal policy has received less attention - the implausibility of managing aggregate demand through the slow, cumbersome and messy budget process means that, in general, the focus has been on the Fed.

That has started to change as the global slump has pushed conventional monetary policy to its limits (and beyond into unknown worlds of unconventional policy), and governments around the world have made fitful attempts at fiscal policy.  For example, recent papers by Christiano, Eichenbaum and Rebelo, Gauti Eggertson and Michael Woodford have shown that it is possible for fiscal policy to have significant multiplier effects when monetary policy is at the zero lower bound (as it is today) in New Keynesian models.

So, while, at a superficial level, it appears that the split between "scientists" and "engineers" persists, some of the "scientific" work being done today is finding that the remedies proposed by the "engineers" are not wholly inconsistent with forward-looking rational behavior after all.

1 comment:

Anonymous said...

Empiricism requires rejecting the entirety of Phillips curve based on the single failure of the 1970's. Romer had to get cute with this but it's important to remind her of "modus tollens", something they furiously hide from econ students.