Skip to main content

“Behavioural economics is a dead-end for development policies”

Interview with Jean-Michel Servet

14 April 2019

[English] [français]

Published initially on April 11, 2018.

Jean-Michel Servet just published ’’L’Economie comportemantale en question’’ (Editions Charles-Leopold Mayer 2018), a critical assessment of behavioural economics as it is applied in international development studies and other fields.

Inst. Veblen : Behavioural economics is often said to offer an alternative to standard economic theory, a return to a more empirical approach. And yet, you seem very critical about it?

J-M Servet : Without any doubt, this research community has been very successful in promoting its approach. To understand why, we must go back to the years after the 2007/08 global financial crisis. Mainstream economists came under massive critique for being unable to anticipate the crisis and to address its causes properly. That’s precisely when this new behaviourism, until then a little-known academic school of thought, emerged as a promise of some sort of « good » economics. For the new behaviourists, the starting point of academic enquiry is always a concrete, practical problem that needs to be solved. Once they identified the problem, they either study their test subjects in front of their computers or conduct field experiments.

At first sight, these experiments seem very similar to randomized clinical tests used to test the effectiveness of new medication. The whole procedure might thus seem very straightforward. However, when we go a step further than simply reading the general summary of a test or an experiment, when we take time to dig into technical appendices and especially into econometric explanations of collected data, the whole matter becomes much more complicated.

Behavioural economists claim it’s their statistical tools that make their results more convincing than qualitative surveys carried out by socio-economists or indeed any other social science, where the goal is to understand mechanisms and systemic causes and not simply to measure. In behavioural economics, on the other hand, this focus on precise measurements is easily taken for scientific rigour – by the general public, by the media, by decision-makers, but also by the expert community itself. However, this image of behavioural economics as “pure science” fades away as soon as we take a look at how these studies are actually produced: we then discover all the possible bias induced by the choice of test-persons for instance, or by the way in which some test-persons are replaced within the sample, by the choice of geographical location to carry out the surveys, by implicit presuppositions of questions asked by the interviewers, or simply by experts who know nothing about the cultural setting of the community they are supposed to study.

In sum, no matter how strict the statistical processing of collected data might be, the processes leading up to such results is a true “black box”. Behavioural economists state facts, they don’t explain them. And when they do explain, their assumptions reveal how close they actually are to most mainstream economics dogmas they pretend to reject. You can see it through the questions they ask during the tests, or through the comments they make about their results. The homo œonomicus might have got a nice facelift here, but it’s still the same homo œonomicus. And behind the facade of “pragmatism” lies the same neo-liberal ideology.

Inst. Veblen : You devote a whole chapter to a specific case study, the randomisation experiments carried out among impoverished populations in India and policy proposals based upon these experiments. What specific issues did you want to raise in this way?

J-M Servet : Indeed, I did not want to remain at a level of general critique, I wanted to get concrete. So, I dig into the analysis of an experiment on which I could collect enough information. I chose an Indian one because I’ve studied this country myself. I also contacted several researchers from the French IRD-Institute (Institut de Recherche pour le Development) and the CIRAD (especially with Isabelle Guérin), who know this field very well, just as some of my former PhD students and friends. [1]

What’s most appalling is the unwillingness of behaviourial researchers to dialogue with other researchers. For instance, in the chapter about India I focus on one single behavioural study, a study which examined the effects of handing out money to poor populations in two different ways: as physical cash or as deposits. I was hoping that, since one of the authors of this study actually works at the same research institute where I used to work, we could easily meet and talk over this particular case. I sent her the text with my technical critique, but I never got any answer. Years earlier, I experienced the same kind of unwillingness from Esther Duflo’s team located in India when it was carrying out its first randomizations.

Bear in mind that this is not a personal issue. In my book I mention another example, a double study in Morocco commissioned by the French Development Agency (AFD). It was carried out by two parallel groups: randomization researchers and socio-economists using other methods. The results of the randomization experiments were sent out by the Afd to the socio-economists. But when these socio-economists invited the randomizers for a meeting to compare their respective findings, the randomizers refused. They even forbid the other group to be present at the presentation of their study! Same scenario with the scientific articles published after the study: the socio-economists cited the randomizers’ work, but not the other way around. So, the issue is not personal, the problem lies in the somewhat sectarian way in which new behaviourists choose to work.

In sum, I see three limitations to this approach. First, the randomization technique itself, all the uncertainties linked to the choice of test-persons. Secondly, there is a kind of hegemonic thinking among economists about their relations to other sciences. And finally, the ideology itself, the way in which these researchers think about the individual agent they study: behavioural economists tend to ignore the web of interdependences between the individual and the society as such, just as they ignore all the systemic effects which may be positive for some and negative for others among their test-persons.

Considering these methodological limitations, you might wonder if these surveys are really worth the money, given they are so much more expensive than other research techniques. Can social sciences really afford to spend half a million euros for producing such limited results? Especially if it is at the expense of other methods, which are, in my opinion, actually more useful for the society?

I don’t know how long the attraction with randomization and clinical test will last. It is interesting to notice, however, that their objects of study evolve over time: they are less determined by a global scientific project than by some public or private institutions.

J-M Servet : If behavioural economics is this expensive scientific dead-end as you claim it to be, what other ways would you recommend for redefining development policies? Especially for addressing challenges such as climate change and depletion of natural resources?

J-M Servet : I dedicate a part of the last chapter to discuss this question. If we want to address the main issues of our time, we need a systemic thinking capable to integrate interdependences into the scientific method. Which means that our knowledge must necessarily come from various disciplines and various theories. This thinking must be substantive in the meaning of Karl Polanyi, i.e. the emphasis must be on what enables us to understand the constraints impacting human behaviour.
This is obviously not a new issue; we could go back to Aristotle or Plato and find some answers! But if we limit the scope to the invention of social sciences since the 18th century, this rethinking of development must be based on true collaboration between various disciplines, where each ones’ assumptions and limitations are acknowledged. “Interdisciplinarity” does not mean creating a magma of accumulated opinions and results, but rather confrontation and cooperation where each discipline’s contribution and limitation will appear.

Translated from French by Marie Laot and Wojtek Kalinowski

[1Thanks to Solène Morvant-Roux from the University of Geneva for instance I was able compare this particular study with other randomizations carried out in Mexico. I also consulted an anthropologist colleague from the London School of Economics who carried out his own research on local practices in the same geographical zone in India where the behavioural studies were conducted.

Subscribe to the newsletter