For a microeconomic study on the evolving household level consumption patterns, in a given geographical location in India, a colleague and I put our heads together on the issue of executing the perfect analysis. While it is widely acknowledged that no empirical or experimental research is free from limitations (for they’re always subject to constraints on feasibility and design), it is frequently advisable to not leave any stone unturned. In the scientific circles, after all, good research is tagged as ‘valid contribution’, and bad research is (politely) tagged as ‘literature’.
Pressures clearly remain on people contributing to the sciences—try telling somebody (who visibly wants to demonstrate his scientific prowess) that his scientific work is, in fact, a great piece of literature—and you’ll know what I mean here.
Coming back to our story, a significant branch of microeconomics is the study of consumer behaviour, which is often modelled as a choice (of consumption bundle) problem given exogenous parameters such as income, tastes, preferences and market prices. Now, although, tastes and preferences can be, and often are, captured by observable household characteristics, a major issue remained. So what was our problem? There was no data on prices—tell me about producing the perfect and complete research! Specifically, we were studying the consumption patterns of food, say, rice. A first attempt to get an estimate of average prices was to simply divide the total expenditure on rice, by the total quantity consumed. But this approach was prone to further problems: think about this—a household with five family members for a given income may report expenditure on rice that is significantly less than that with just three members with the same income, simply because the latter prefers to consume rice of higher quality (say, basmati) than the former. After all, all households are not the same. We were sure that researchers had come across similar problems, so we felt the need to dig deeper here.
The exact issue, along with elegant statistical and econometric methods to deal the same, was addressed in the book titled “The Analysis of Household Surveys”. This is how I was first introduced to the works of Angus Deaton—the author of the book. If you are a student of economics and still wondering what the solution was, I’d recommend that you first try and work out a solution to the above problem for yourself. I strongly recommend this, so that when you reason your way through each step and begin to get around the problems that subsequently emerge, you will begin to appreciate more the volume of thought put into this (simple?) analysis … and I’m talking about material that cover only about ten pages (from over 400 pages) of the book. While a detailed discussion of the issue at hand, that I have chosen to discuss as an example, is beyond the scope of this article (since it involves what the New York University economist William Easterly calls “tortuous details”), it suffices to say that the method discussed captured some levels of household heterogeneity via geographical variations—a beautiful approach.
That is what sciences are all about. To any dedicated researcher, science is more about the approach to arrive at answers, than the nature of those answers themselves. For example, the famous physicist Sir Eddington’s experiment that confirmed the deflection of light, predicted by Einstein’s theoretical understanding of our universe, had several merits—the most crucial of them being that it was conducted during a solar eclipse; ideal conditions given the purpose of study that clearly bettered our understanding of how light behaves as a response to the sun’s gravitational field. Of course, both sciences and the scientific approach have further evolved since then, although we still admit that, as of even today, we do not have a complete understanding of everything about this universe—and it is almost certain that we never will, for there will always remain scope for improved understanding, given the limitations of the human mind. To therefore say the least, the key merit of the scientific approach is in the ability to discard explanations that do not fit well with what is observed. Keeping this in mind, it is now fitting to discuss some of the contributions of Deaton.
Of the most important things that we learn from Deaton’s work concerns the previously held belief, that any economic agent (say, a household) would curtail its present consumption if his income was expected to fall in an immediate future (and by the same logic, increase consumption, were its future income to rise). The well-established wisdom on the matter today is what is famously known as the Deaton paradox—that consumption actually varies very little in response to household incomes (and more in response to, say, costs of borrowing), effectively revealing a preference for varying individual consumption in response to lags in income (contrary to the idea of raising consumption immediately when incomes rise). Deaton’s contribution to this field of study can be well understood in the words of the Royal Swedish Academy of Sciences which emphasises that an understanding of an individual’s consumption choices is a prerequisite to any economic policy aimed at poverty alleviation—the academy sums it up by saying that “more than anyone else, Angus Deaton has enhanced this understanding.”
There are theorists and empiricists in any field of science. Deaton’s work will continue to inspire both the categories of ‘valid contributors’ for many generations to come, for his contribution spans not just the important understanding of microeconomics (and therefore micro-founded development economics and welfare studies)—the theorist’s concerns—but also the meticulous and careful methodological contributions that concern the empiricists.
By Subrato Banerjee
The author is a research fellow at the Indian Statistical Institute, a member of the Econometric Society, the Royal Economic Society of England, the European Economic Society, and the American Statistical Association