• PAGE ONE
• INDEX
• HOME

Economics Nobel: Discrete choices

P. R. Brahmananda

THE BANK of Sweden Prize in Economic Sciences, in memory of Alfred Nobel, has been awarded to Prof James Heckman for developing the theory and methods to analyse selective samples, and to Prof Daniel McFadden for developing the theory and methods to anal yse discrete choice. The broad area in which the prizes have been awarded is microeconometrics. Messers Tinbergen, Klein and Haavelmo were awarded prizes for macroeconometrics. Ragnar Frisch was one of the founders of econometrics. But his work was prima rily in microeconometrics. Messers Stigler and Becker were the two Marshallians, who applied the partial equilibrium approach to studies in industry and human behaviour. Both Prof Heckman and Prof McFadden were concerned with the methodology studies of m icro units, be it households, firms or units of labour, transport and choices.

What is the selection criteria of candidates for the Nobel Prize in economics? Why has economics been included in the category of scientific subjects for which the Nobel Prize is declared? Note that the prize in economics is for economic sciences. A scie nce is characterised by analysing material phenomena by means of controlled experiments or observations. The analysis involves methodologies, techniques, concepts and models largely unique to that particular science. At any given point in time, there is near unanimity among the scientists about the methodologies and tested hypotheses concerning the science's subject matter. The sum of the contents, generally accepted in the scientific community are known as the paradigm of that science. When the paradig m changes, there is said to be a revolution in that science. The term paradigm is often applied to branches of a particular science as well. The Nobel Committee felt that economics also had the characteristics of subjects such as physics, chemistry and p hysiology/medicine. But controlled experiments are not possible in economics. We depend on statistical information and on history concerning particular economic phenomenon. Kuhn in his seminal work on scientific revolutions notes that in any science prog ress consists of closing the gaps in that area, as also exposing errors and substituting the latter by corrected versions of the reality of the science. Some times paradigm shifts also occur when the manner of perception of reality is suddenly transforme d by new discoveries and inventions.

The Nobel Committee is always in search of new methodologies, techniques and ideas in any subject. It looks for personalities originally involved with the discoveries and inventions. The particular contributions have to be identified with specific scient ists.

Professors James Heckman and Daniel McFadden fulfilled the tests adopted by the Committee. Prof Heckman's earliest papers concerning selection bias date from 1974, though the writings of R. Gronau and H. G. Lewis also appeared about the same time on the same theme. But it is Prof Heckman who has continuously worked on exposing that bias in different areas and finding statistical/econometric solutions to overcome the bias.

Similarly, Prof McFadden's earliest papers on discrete choices and the econometric techniques to deal with them seem to date from 1974, though R. D. Luce seems to have theoretically explored that area in 1959 in his Individual choice behaviour: A theoret ical analysis. Incidentally this book drew a review from the great Debreu in 1960. Both Profs Heckman and McFadden have been identified as filling a very important gap in microeconometrics. In filling the gap, they brought about an almost paradigm shift in that branch.

A number of new applied problems of policy could then be more accurately tackled by their contributions. The work of the two Nobel awardees are highly technical, though in the US their work seems to have gathered a lot of follow-up. In the following we o ffer some lay comments on the contribution of Prof Heckman.

Prof Heckman's contribution has a wide sweep. In statistical investigations, data constraints prevent the sampling of the universe of potential observations necessary to analyse the behaviour of micro units. The choices of excluded portions of the univer se introduces a selection bias. The exclusions occur because of administrative instructions for the collection of information and/or because of the huge costs involved in the sampling of a whole universe. In India in earlier years, agricultural crop prod uction was estimated by reference to eye estimates of village officials. Such estimates, apart from others, involved selectivity bias. It was only later that random crop-cutting experiments could be introduced as a basis for aggregate estimates.

Another example is that of the National Sample Survey of consumer expenditure which concerns itself with studying households in different villages and tehsils and States with the sample size being broadly proportional to population sizes. Those who live outside households in streets, parks and forests, for example, are excluded. If we assume that these persons are mostly poor, estimates of poverty based on the NSS would be underestimating the headcount numbers of the poor, particularly when the sampled results are blown up for the whole population.

There is a certain bias in favour of expenditures made mostly out of regular incomes or sources thereof. Those who live on the basis of irregular or transitory receipts are out of the picture. In addition, the aggregate food expenditure derived from the NSS data would be larger than had the entire universe been sampled and the results thereof blown up. One should expect the NSS to derive poverty estimates by deriving the sample from receivers of income or other receipts whether they are in households or not. But this probably is not feasible. Many of our sector-wise estimates of contribution to NDP involve heavy selection biases. Prof Heckman's work should be of interest to the Statistical Commission headed by Dr C. Rangarajan. In fact, I would suggest the Commission invite Prof Heckman to scrutinise the basis of many of our aggregate estimates and suggest methods to overcome or correct the selectivity bias.

The selection bias is common in most statistical enquiries concerning partial equilibrium studies. The basic work of Prof Heckman seems to be in the extension of statistical methodology from an economist's angle to encompass missing observations. To stud y the relation between wages and higher education, we take the employed persons as the universe for enquiry. But we do not know how wages would rule under conditions when the others got themselves educated and offered to take up the jobs. In that case, t he wage hierarchy with respect to education might have been less steep than otherwise. This means that in reality, taking only those who are employed and have had higher education would imply that we are overestimating the education's contribution. We ma y conclude that the returns to education are higher than what they are and that there is underinvestment from a policy angle in higher education.

Prof Heckman's studies seek to correct the selectivity bias in such investigations. The contribution is general and has been applied to women in the labour force, migration and remittances, demographic economics and family behaviour. He has applied the u tility maximisation hypothesis to explain why many do not migrate or why many do not take up higher education or why women prefer to stay in households rather than offer themselves up for work. But there could be other factors acting as barriers and the utility maximisation hypothesis may not always provide a satisfactory explanation. My observation is especially relevant for developing countries such as India, where Prof Heckman's approach may have to be modified. But, before that can occur, we must ex plore the scope for applying the Heckman methodology to many of our investigations and enquiries where the selection bias is very common. We know that such biases (self-selection bias) exist in the case of individual household choices. But they are also common in almost all our methods for collecting information and deriving deductions therefrom. For example, the company statistics data analysed by the RBI. To what extent is there a selection bias if we blow up the data for the whole economy. Wherever r andom sampling is not possible or is not applied, there is preliminary scope for applying the Heckman methodology.

I wonder whether the Nobel Committees would subject their awards to be examined methodologically by Prof Heckman since heavy selection biases are involved in the Nobel awards!