I have been asking this question of bureaucrats who make economic policy, and economists who provide the fig leaf to them, for the last 25 years. I am yet to get a satisfactory answer.
Overall, however, theory has made way for empirics as the idea of "proof" in economics has shifted from a logical argument to often dodgy statistics because while statistics by themselves are seen to be conclusive, logic by itself is not.
More From This Section
But is this method foolproof? If not, why not?
The answer lies in a simple fact: all economic data have a context. That's to say, the operating environment influences the nature of the data and can render it unfit for use in a different context.
The same experiment in physics and chemistry can be replicated by others in a different context, producing nearly identical results. In economics, you cannot perform the same experiment twice. Or, as Confucius is supposed to have said, "You cannot step in the same river twice."
Look at the context…
More crucially, the context-free nature of data in the physical sciences has an impact on the veracity of the theory. But economics, despite all its pretensions, is not a physical science. Human beings are involved.
This means that the regulatory, external, technological, environmental, social or the political environments, singly or together, do influence the nature of activities. So, it is necessary to be very careful about how policy uses data. If the context is not identical we can - and do - end up getting the wrong policies.
But then, all past data have some regulatory, natural, external, technological, social or the political context. Indeed, in India the problem is not confined to the contextuality of the data, which is a universal problem. Here it is exacerbated by bad data as well.
Does this mean policy must not use past data at all? Obviously, that is not the solution either.
So, here's an utterly heretical suggestion: don't use time series data without first examining its context. If you do, there is a very high probability that you will get bad, if not necessarily wrong, answers.
The best example of this the substitution of automatic traffic signals with hand-waving policemen. The former uses theory. The latter uses data about traffic flows and volumes visually collected by the policeman's brain. It is this shifting context that informs the policeman's hand-waving.
… and the theory
What about economic theory, then? How useful would it be for policy purposes? The most succinct answer came from an economist who is an out-and-out data man.
"In the beginning," he said, "since there was very little data about anything there had to be assumptions. But now that there is data, there is no need for assumptions. You can look at the facts."
This is the view that has informed policy-making in India, at least, for the last two decades. That's fine, except that the quality of the data has steadily worsened even as its quantity has increased. And, of course, the context has changed.
So, in the absence of accurate or very current data, the gap (at least in my view) has to be filled by some amount of theory. Sadly, theorists are a dwindling breed and almost never consulted by anyone.
The result is that we often have policy that is neither informed by good data nor by good theory.
It is forgotten these days that that some of the deepest insights in economics have come from theory. There are several theoretical insights that do not have an iota of empirical content. Yet, the things the theorists said have been verified over and over again in a variety of contexts.
Thus, Harold Hotelling's location model has always worked.
Von Neumann's and John Nash's concepts were entirely theoretical, influencing everything from political science to evolutionary biology.
In the last decade or so, Paul Klemperer and Co have designed the mechanisms behind every successful European auction, from spectrum to medical permits, purely on the basis of theoretical insights.
The Turkish revolution in some aspects of theory has transformed the kidney transplant market there, not to mention the school admissions process in the US.
I could go on giving examples, but the key issue is that without theory, it is hard to design a decent experiment because how do you know what you are testing for? How would you know if your conclusions are right or wrong? Theory provides the yardstick by which the data are judged.
In sum, research in India needs to focus on theory as well.