Should economists be “imagineers” of our future?
By Mark Thoma
The opinions expressed are his own.
Roger Martin is unhappy with the state of economics. One charge is that:
[an economist] predicts a future that is based on the past. And when it is anything but, he returns to the same tools to do it again, believing that in doing so he is being meritoriously scientific. … Extrapolating the future to be a straight-line projection of the past is neither accurate, nor is it helpful in creating better understanding and newer ideas.
As I will discuss further below, I agree that macroeconomists need to fix their models. But I don’t think that predicting the future based upon “a straight-line projection of the past” is the problem. Let me explain why, first in a relatively narrow sense, and then more broadly.
This year’s Nobel prize award to Thomas Sargent and the previous award to Robert Lucas were partly in recognition of their development of the tools and techniques that economists need to go beyond simply trying to extrapolate the future from the past, a procedure that can lead forecasters astray.
Prior to Robert Lucas, economists analyzing policy interventions by monetary or fiscal authorities did exactly as charged above, they extrapolated based upon the past and an assumed unchanging future. But the (often false) assumption that the future would be like the past is at the heart of what is known as the Lucas critique.
To understand the critique, suppose that the government is considering imposing a new tax on a particular industry. Based upon the government’s estimate of profits in the industry, it expects to collect a large amount of taxes and solve its revenue problems.
But when the tax is actually imposed, profits do not turn out to be as large as expected and tax revenues come in far short of projections. What happened? The firms took steps to reduce their tax exposure, e.g. they used the usual accounting tricks to inflate costs and lower reported revenues to reduce taxable profit. To the extent they were successful, tax collections were lower than expected.
The lesson from this example is that people change their behavior in response to changes in the conditions they face. And this is one of the things that separate what researchers in the hard sciences do from the work of economists. If I tell my TV set that I am going to smash the screen with a baseball bat, it will just sit there. It won’t take evasive action. But a human in the same situation will do their best to get out of the way and avoid harm. When harm is expected, whether it’s physical harm, higher taxes, more work for less pay — whatever — people try to avoid it.
This means that backward looking extrapolative estimates — the straight line projections objected to above — can give wrong answers about changes in economic policy. If, for example, the Federal Reserve uses estimates based upon past data to alter its policy rule and does not account for the fact that people will change their behavior in response to the change in the rule, then it will get policy very wrong.
Along with Robert Lucas, Thomas Sargent was in the forefront of developing tools and techniques to incorporate the response of people — the change in their expectations — to changes in policy. Sargent borrowed heavily from the engineering literature, for example the engineering literature on optimal control is useful to macroeconomists trying to develop monetary and fiscal policy rules to optimally control the economy. However, if that was all he had done — borrowed from another discipline — he would not have received a Nobel prize. As noted above, what makes economics different from engineering is that people can respond to changes in their environment. They will use all the information available to them to anticipate the future as best they can – they will form rational expectations in this sense — and then take action to avoid anything that will make them worse off.
Incorporating rational expectations into macroeconomic models increases the level of complexity by an order of magnitude over what was already a difficult problem in the engineering literature, and much of what Lucas, Sargent, and others did was to find a way to forge forward despite the technical difficulties. That was an important contribution, but for our purposes it is the conceptual contribution — the loud, clear message that simple extrapolation from the past can lead to problems — that was important.
But the objection about extrapolating from the past raised by Roger Martin is broader than this. He is referring to using the same models again and again even after they fail, i.e. returning to “the same tools” again and again rather than learning from failure and abandoning models that do not work.
I am in agreement with the argument that our models need to be improved. But the problem is not mindless extrapolation from the past. As the discussion of the recent Nobel prize awards above shows, we are well aware of the limitations of that approach. Nor is the problem the failure to abandon models and move on to new ones when they cannot adequately explain the data. In my career, the attempt to find models that can explain past data and predict future data with more accuracy has caused Old Keynesian models and New Classical Models to be replaced by Real Business Cycle and New Keynesian models.
And it’s time for that to happen again. One of the big problems with the existing class of macroeconomic models is the failure to adequately incorporate the possibility and consequences of a meltdown in financial intermediation. There are technical reasons for this, and also the fact that many macroeconomists did not think a Great Depression style financial meltdown was possible and hence it wasn’t important to invest time asking questions related to such an event. That was a mistake, when financial intermediation began to malfunction we did not have the models we needed, and this was far from the only mistake we have made.
Presently, there is no shortage of work trying to fix the problems with our models. I don’t know if we will succeed — the next model will work until it doesn’t — but we are certainly trying. And there is also no shortage within economics of “imagining things other than as they are,” a phrase the author uses repeatedly. From recommendations on how to fix markets, address pollution problems, stabilize the economy, put people back to work, to models of comparative economic systems that imagine societies with different institutional structures and societal relationships, economists are constantly imagining how to improve social conditions. In fact, for the most part we are charged with trying to do too much social engineering, not too little. In any case, we don’t think of the economy as an unchangeable “hunks of granite,” we understand that social relationships are at the heart of what we study and that those relationships are not etched in stone.
The end of the second essay calls for “attempts to create a future that does not now exist, rather than mindlessly crunching the numbers that do exist.” There are plenty of number crunchers in the profession, and as an applied econometrician I’ll certainly defend their value in grounding theorists in real world data. But there are also plenty of “imagineers” — people who play with toy models and toy ideas to envision worlds that do not now exist, but could — and perhaps one of them will discover the “blueprint for a better way” that Roger Martin hopes will emerge from the broader conception of science he writes about in his essay.