Friday, 28 August 2009

Three kinds of economic story

Anthony J. Evans (who kindly linked to my camel story earlier in the year - or as he notes, perhaps not all that kindly) has written a paper about economic counterfactuals. He confirms what's now becoming a mantra among economists (if not their public): economists can't make predictions.

However, he says, rather than giving up completely on exploring the future, they should do so with two tools of the economist's imagination: counterfactual analysis and scenario building.

Counterfactuals are stories about the past - a retrospective prediction if you will - considering what might have happened had a different decision been taken at a key moment. For example, what if the financial sector had not been rescued by governments in 2008? Or what if governments themselves had defaulted on their debts?

Scenarios are possible alternative futures - constructed not so that you can predict which one will happen, but so that you can prepare to understand events as they unfold. Each scenario lays clues as it starts to emerge: the V-shaped recession has certain indicators, hyperinflation has others, and so forth. By knowing what to look for, you'll have some forewarning of what is happening.

These exercises are undoubtedly valuable to test and explore one's model of the economy. But I fear that the interpretations can become self-confirming.

Evans outlines four scenarios, potential future paths for the economy. Against each, he indicates several signs that might indicate this scenario is becoming real. However one principle of science is that you can only disprove a theory, never prove one. It would be more scientifically convincing to tell us how to spot that the scenario is not happening.

I appreciate these scenarios are not meant to have the force of scientific theory - they are more of a guide to understand what's happening in the world - but if we hold them to higher standards, we'll have more confidence in them if they pass.

The counterfactuals are, by their nature, impossible to disprove - but we can apply some logical analysis to them to see whether they stack up.

By this standard, a couple of the examples in this paper are suspect. Evans proposes the following remedies for wage rigidities and the adaptability of business:
  • eliminate corporation tax
  • raise retirement age (to reduce pension obligations)
  • repeal minimum wage laws
  • ease restrictions on migrant workers
I can't for the life of me see how corporation tax causes wage rigidity, nor how it has any major impact on corporate flexibility. Raising the retirement age has an equally vague connection with the intended outcome. While these ideas might have their own merits, the interpretative nature of counterfactuals can lead us to mix in our own policy preferences instead of rigorously maintaining the logic of the argument.

My own interpretation is that wage rigidity is a psychological phenomenon as much (or more) than an institutional one, though both factors do make a contribution. But perhaps I'm just revealing my own bias there.

In Evans' third counterfactual, the UK acts to slow credit growth, increasing interest rates and reserve requirements, and seizes control of the financial industry. He points out that this version of events does have a few echoes in the UK government's actual actions, as well a much closer match to the policies of Croatia, Ireland and Poland (I do not agree with him that it describes Germany, despite Angela Merkel's rhetoric). However it bears a much closer resemblance to another set of events in real history: the Federal Reserve's actions at the start of the Great Depression.

Most economic historians agree that the Fed's tightening of credit after the Wall Street Crash was the main trigger for the depression of the 1930s. Remember that they did not know they were at the start of a depression - all they could see was that a credit-fuelled boom had precipitated a crisis. Perhaps it seemed natural to cut back on credit, even if it was a little too late. But the crisis was only indirectly caused by the growth of credit - in fact its proximate cause, just like in 2008, was the too-sudden withdrawal of credit when lenders and savers panicked. For the central bank to compound the problem by restricting credit even further, was a disaster - and would be again today.

As my camel story - perhaps clumsily - showed, a third tool of the economic imagination is analogies. An apposite analogy here is a patient who hyperventilates, leading to an asthma attack. Even though the cause of the attack may have been too much oxygen, it doesn't follow that the correct response is to reduce oxygen intake even more. A long-term plan to manage the problem might indeed include tools to stabilise the patient's oxygen consumption, but it cannot be right as an emergency response.

Like counterfactuals and scenarios, analogies are always imperfect approximations to life - and indeed I'm wandering from the epistemological discussion about how to use these tools, into an indulgence in my own example.

So even though I disagree with most of the models and interpretations that Evans chooses in this paper, its key point on "the economic method" provokes some useful thoughts. These three techniques are all different forms of story-telling, and stories have a capability to illuminate and explore issues which is powerful though - critically - has the potential to be highly misleading.

Stories are very convincing to the human mind, which makes them incredibly useful as political (or sales) rhetoric; but should warn scientists to treat them with great care.

Despite this caveat, the paper is a useful contribution to finding the right role for economists and setting the right expectations for the authority of what they say.

2 comments:

PunditusMaximus said...

If economics cannot make predictions, it is not a science. It is a religion. I'm not sure our more . . . right-thinking friends want to go there.

aje said...

@PunditusMaximus

Saying that 2+2=4 is not a prediction. Would you argue that mathematics is a religion?