Risk Attitudes and the ‘Vicious Circle of Poverty’

Abstract

The article reconsiders the view that the poor are trapped into poverty as a result of their risk aversion, precluding the level of investment needed to lift them out of poverty. The starting point is the experimental methods of Binswanger who found no significant association between risk aversion and low income. Using experimental data from Uganda, Ethiopia and India the article confirms these early findings, and provides estimates of the correlation between the variables involved in the ‘vicious circle of poverty’. The paper concludes that there is generally little relationship between risk aversion and an income measure of poverty, but a strong relationship between the latter and asset levels and returns.

Cet article reconsidère l'idée que les pauvres sont enferrés dans la pauvreté du fait de leur aversion au risque, qui empêche la constitution de l'investissement nécessaire pour les sortir de cette situation. Le point de départ de l'analyse repose sur les méthodes expérimentales de Binswanger, qui a mis en évidence l'absence de corrélation significative entre l'aversion au risque et les bas revenus. A partir de données expérimentales sur l'Ouganda, l'Ethiopie et l'Inde, l'article confirme ces résultats initiaux, et fournit des estimations sur le degré de corrélation entre les variables impliquées dans le ⟨cercle vicieux de la pauvreté⟩. L'article conclut qu'il y a généralement peu de rapport entre l'aversion au risque et une mesure de la pauvreté [agrave] ; partir du revenu, mais en revanche une relation étroite entre cette dernière et le niveau et la rentabilité des actifs détenus.

Notes

Paul Mosley is Professor of Economics at the University of Sheffield, UK, and Arjan Verschoor is Lecturer in Development Studies at the University of East Anglia, UK. The authors gratefully acknowledge the support of the Department for International Development under their research programme on ‘Risk, Incentives and Policies for Pro-poor Growth’ (R7614/7615/7617).

 1. At this stage, and even in Nurkse [1952], the vicious circle was macro in nature, running from low incomes to low capacity to save to low investment to low incomes once again.

 2. Notably Bauer [1966] and Lal [1983], whose argument was based on the proposition that some poor people and countries manage to escape from the vicious circle.

 3. Krugman's term [Krugman, 1993].

 4. As support for this view Lipton cites Kalecki's Theory of Economic Dynamics (London, 1954), which also presents the proposition in axiomatic form.

 5. This statement is presented in terms of MVPE (marginal value product equalisation), which represents conventional ‘optimising’ behaviour under certainty. Under uncertainty, the corresponding optimising postulate becomes MEVPE (marginal expected value product equalisation). Later in his article, however, Lipton makes clear that his strictures apply to MEVPE as well.

 6. ‘As households move closer to extreme poverty and destitution they become very risk averse: any drop in income could push them below their survival point’ [World Bank, 2000: 145].

 7. Binswanger inferred that differences in behaviour – in particular, reluctance to invest in modern inputs – were due rather to ‘limitations on credit or on access to modern inputs’ [1980: 395]. This has implications for the design of microfinance.

 8. Interestingly, Vepur, one of the Indian research locations, is only 40 miles from ICRISAT headquarters at Patancheru from which Binswanger launched his surveys in 1980.

 9. In India participants were asked to step into a booth when their name was called, where they would be asked to point to the bag of their choice. Our enumerators suggested this device, because it would ensure anonymity of choice, and because frequent periodic elections in this part of the world mean that participants are very used to polling booths.

10. In the Ethiopian and Indian experiments, question numbers were written on pieces of paper and put in a bag (or box in India) only after participants had stated all their preferences, and one of the participants was invited to draw a number; enumerators found that this way of doing it enhanced the experiment's transparency and credibility in the eyes of the participants.

11. In India winnings were paid in vouchers, redeemable at the local bank (with no transaction costs to the participants).

12. In expected utility theory, risk aversion is related to the concavity of the agent's utility function, and can be expressed as a combination of some or all of: its first- and second-order derivatives, initial wealth, and the stochastic variable that determines increments to wealth [e.g. Laffont, 1989]. If we assume that the utility function is twice differentiable, the Arrow-Pratt approximation allows us to disentangle the respective effects on welfare of risk and preferences, as follows. Let U(W) be a suitably behaved concave utility function in wealth, and let Z be the prize of the lottery, α the probability of winning the prize, and λ the reservation price. A measure of risk aversion ρ equal to minus the second-order derivative divided by the first-order derivative of the utility function can then be deduced by developing a Taylor expansion of U(W−λ) and U(W + Z - \lambda ) around U(W) and solving for ρ [for details see Gollier, 2001].

13. The World Bank study of Voices of the Poor found that income was not found by most poor people globally to be a significant concept by which to measure their feelings of deprivation [Narayan et al., 2000].

14. There is as yet no consensus as to what constitutes an ideal index of vulnerability [Gamanou and Morduch, 2002:I]. We include here a perceived measure for three of the concepts of vulnerability which they review (past variability, risk of future change in poverty status, and ability to cope) and add a fourth measure attached specifically to risks associated with entrepreneurship. We do not include assets in this measure but rather estimate their influence separately (cf. Table 4). However, in Table 4 below we do consider the effects of interactions between assets and subjective vulnerability measures.

15. These standard axioms are: ordering (individuals are willing to state preferences across all pairs of alternatives, and these preferences are transitive, in other words, having once preferred A to B, and B to C, they do not then prefer C to A), continuity (there are no kinks in indifference curves) and independence (for any three prospects A, B and C, if A is preferred to B then an x per cent probability of A combined with a (1−x) per cent probability of C will always be preferred to an x per cent probability of B combined with a (1−x) per cent probability of C, whatever the probability x may be; in other words, once third options are introduced, they do not, however they are presented, alter the structure of an individual's preference between any two basic options). It is this independence axiom on which most debate has focused, and in relation to which violations of EU theory have been most often noticed, starting with the famous ‘Allais paradox’ [Allais, 1953]. This conjecture suggested that, given a choice between two lotteries of equivalent expected value, individuals would not always choose the ‘riskier’ one (with the higher probability of getting nothing) or always choose the ‘safer’ one, but might flip-flop between the two depending on the size of the payoffs in the lottery – which according to the independence axiom should be irrelevant. This conjecture, which has often been confirmed empirically, is one example of the more general ‘common consequence effect’, which is that under an expected utility approach choices between ‘risky’ and ‘safe’ gambles should not vary according to the value of the ‘common consequence’ if things turn out badly. This ‘common consequence effect’ is tested in our experiments below.

16. Starmer in his recent review claims that these models ‘now number well into double figures’ [Starmer, 2000: 1].

17. This statement is presented in terms of MVPE (marginal value product equalisation), which represents conventional ‘optimising’ behaviour under certainty. Under uncertainty, the corresponding optimising postulate becomes MEVPE (marginal expected value product maximisation). Later in his article, however, Lipton makes clear that his strictures apply to MEVPE as well.

18. The remaining 52 members of the sample (26 pairs) played an ‘insurance game’, a modification of the trust game, on which we report below.

19. Note that if the first player chooses to buy insurance s/he sacrifices not only a ‘tax’ on gains, in the shape of the insurance premium, but also the possibility of maximum gains, since it is no longer possible to invest the maximum stake of 4,000 UGS in the second player – 1,000 UGS must be sacrificed to pay the insurance premium.

20. These considerations are not far-fetched. Anyone supposing that illiterate and semi-literate small farmers in a poor-country context do not engage in such mind games would be mistaken: from the debriefing interviews we know that our subjects reasoned precisely along these lines.

21. Indeed, effective demand for insurance appears to be kinked (Figure 2 above) – there is very little demand for it at low levels of income.

22. Physical and social capital may be converted into liquid assets in times of hardship; human capital not only raises permanent income but also increases access to non-agricultural income sources (quantified for Uganda in Appleton [2001a]).

23. For Ugandan evidence see Mosley et al. [2003, chapter 7]. In Andhra Pradesh risk-averse attitudes in a multiple regression similar to those in Table 1 had a regression coefficient of 0.009, significant at the 5% level, on adoption of hybrid seed.

Additional information

Notes on contributors

Arjan Verschoor

Paul Mosley is Professor of Economics at the University of Sheffield, UK, and Arjan Verschoor is Lecturer in Development Studies at the University of East Anglia, UK. The authors gratefully acknowledge the support of the Department for International Development under their research programme on ‘Risk, Incentives and Policies for Pro-poor Growth’ (R7614/7615/7617).