Abstract
Prominent theories of decision under risk that challenge expected utility theory model risk attitudes at least partly with transformation of probabilities. This paper shows how attributing local risk aversion to attitudes towards probabilities can produce extreme probability distortions that imply paradoxical risk aversion.
Similar content being viewed by others
Notes
Formally, urn \(R_{g}\) is prospect \(\{\$ 900,0;p,1-p\}, p=(g+5)/100\) and urn \({S}_{g}\) is prospect \(\{\$ 900,\$ 200,0;p-1/2n,1/n,1-p-1/2n\}\), \(n = \)10. All prospects R and S considered in this paper are of these types.
The supposition that the safer prospect is weakly preferred to the riskier prospect for all \(g\) from {0, 5,..., 90} is made here for simplicity of exposition. Section 3 provides general results for cases when the weak preference for the safer urn is observed only for some subset of {0,5,..., 90} and strict preference for the riskier urn is observed outside the subset. Note that the latter preferences cannot be rationalized by expected utility.
For a power utility specification, \(x^{r}\) the supposition \(q>1\) is equivalent to \(r>0.461,\) which is plausible (Harrison and Rutström 2008). For smaller values of \(r\), one can construct an example with other prizes. For example, for \(r= 0.3\), the high, middle, and the low prizes that satisfy condition \(q>1\) are $900, $300, and $100.
To be able to construct numerical illustrations for implications of inequality (1), I need to choose a value for \(q\). Prizes used in the example, 900 and 200, are within the range of payoffs in Abdellaoui et al. (2008) study. In my numerical illustrations I use 2 as a lower bound of \(q\) since the value of 900 being at least three times the value of 200 is satisfied for the power estimates in their study. Their reported estimated (mean) power exponent on the gain domain is 0.86.
The supposition \(q>1\) requires that the high prize of €40 is valued at least twice as much as €10, which is arguably plausible; for a power utility specification, \(x^{r}\) this supposition is equivalent with \(r>0.5.\)
For \(q>1, \kappa (q,k^{*}-k,k-k_*)\) approaches infinity as \(k^{*}-k\) (see lemma 5.1 in the appendix). For given probabilities \(p\) and \(p^{*}\), verify that \(k^{*}-k\ge 2n(p^{*}-p)-1;\) so the larger the value of \(n\), the larger the value of \(\kappa (.).\)
Let 0.5M denote $500,000 and use subadditivity of \(v(.)\) and \(f(0.75)-f(0.5)\le 0.00004(1-f(0.5))\) to verify that \(v(0.5M)(f(0.75)-f(0.5))\le 25000v(20)\left( {f(0.75)-f(0.5)} \right) <v(20)(1-f(0.5)).\) Rearrange terms in the last inequality to get \(v(0.5M)f(0.75)<v(0.5M)f(0.5)+v(20)(1-f(0.5));\) hence, $0.5 million or $20 with equal probability is preferred over $0.5 million or $0 with probabilities 3/4 and 1/4.
References
Abdellaoui, M., Bleichrodt, H., & L’Haridon, O. (2008). A tractable method to measure utility and loss aversion under prospect theory. Journal of Risk and Uncertainty, 36, 245–266.
Allais, Maurice. (1953). Le Comportement de l’Homme Rationnel devant le Risque: Critique des Postulats et Axiomes de l’Ecole Américaine. Econometrica, 21, 503–546.
Cox, J. C., Sadiraj, V., Vogt, B., & Dasgupta, U. (2012). Is there a plausible theory for decision under risk? A dual calibration critique, Working Paper, Georgia State University, Atlanta, GA, USA.
Edwards, Ward. (1954). The theory of decision making. Psychological Bulletin, 51, 380–417.
Handa, Jangdish. (1977). Risk, probabilities, and a new theory of cardinal utility. Journal of Political Economy, 85, 97–122.
Harrison, G. W., & Rutström, E. E. (2008). Risk aversion in the laboratory. In J. C. Cox & G. W. Harrison (Eds.), Risk aversion in experiments. Research in experimental economics (Vol. 12). Greenwich, CT: JAI Press.
Kahneman, Daniel, & Tversky, Amos. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263–292.
Preston, Malcolm G., & Baratta, Philip. (1948). An experimental study of the auction value of an uncertain outcome. American Journal of Psychology, 61, 183–193.
Quiggin, John. (1982). A theory of anticipated utility. Journal of Economic Behavior and Organization, 3(4), 323–343.
Quiggin, John. (1993). Generalized expected utility theory, the rank-dependent model. Boston, MA: Kluwer.
Rabin, Matthew. (2000). Risk aversion and expected utility theory: A calibration theorem. Econometrica, 68, 1281–1292.
Starmer, C. (2000). Developments in non-expected utility theory: The hunt for a descriptive theory of choice under risk. Journal of Economic Literature, 38, 332–382.
Tversky, Amos, & Kahneman, Daniel. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5, 297–323.
Wakker, Peter P. (2010). Prospect theory: For risk and ambiguity. Cambridge: Cambridge University Press.
Yaari, M. E. (1987). The dual theory of choice under risk. Econometrica, 55, 95–115.
Acknowledgments
Financial support was provided by the National Science Foundation (Grant Number SES-0849590). The author is grateful to James C. Cox, Glenn W. Harrison, Peter P. Wakker, and two anonymous referees for helpful comments.
Author information
Authors and Affiliations
Corresponding author
Appendix: Proof of the Theorem
Appendix: Proof of the Theorem
We first prove one proposition and one Lemma. Then we use these two results to prove the main theorem.
Let set \(\Psi (k_*,k^{*})=\{k_*/2n,(k_*+1)/2n,..., k^{*}/2n\}\) be defined for any given two integers \(k_*,k^{*}\in N\) such that \(1\le k_*<k^{*}\le 2n-1.\) For any given non-negative prizes \(h>m>\ell \) let \(q\) denote \((v(h)-v(m))/(v(m)-v(l))\).
Proposition 5.1
Let \(n\in N\) and non-negative prizes \(h>m>\ell \) be given. Suppose that \(q>1\) and \(S_i \;{\succeq }\;R_i ,\) for all \(p_i \in \Psi (k_*,k^{*})\). Then \(\forall \hbox {k}\in \Psi (k_*,k^{*})\)
where \(\kappa =\kappa (q,k^{*}-k,k-k_*)\).
Proof
If \(S_i \;{\succeq }\;R_i \) for some \(i\in \{k_*,\ldots , k^{*}\}\) then according to statement (2), the agent has revealed
for all\(\;\;i=k_*,\ldots , k^{*}\). Adding and subtracting \(v(m)f(i/2n)\) on the left-hand-side of the above inequality, rearranging terms and using notation \(q=(v(h)-v(m))/(v(m)-v(\ell ))\), one has
If \(S_j \;{\succeq }\;R_j \) for all \(j=i,\ldots , i+t\) where \(i,i+t\in \{k_*,\ldots , k^{*}\},\) then application of inequality (5.3) \(t\) times gives
To show statement (5.1) and complete the proof it suffices to show that \(\forall k\in \{k_*,..., k^{*}\},\)
and
because the last two inequalities imply that
rearrange terms and use notation \(\kappa =1+\sum _{j=0}^{k^{*}-k} {q^{j+1}} /\sum _{j=0}^{k-k_*} {q^{-j}} \) to obtain statement (5.1)
Inequality (5.5) follows directly from inequality (5.4) and some rearrangement of terms
Similarly, for inequality (5.6) verify that
\(\square \)
Lemma 5.1
If \(q>1\) then \(\mathop {\hbox {lim}}\limits _{t\rightarrow \infty } \kappa \left( {q,t,s} \right) =\infty \)
Proof
It follows from \(\kappa (q,t,s)=1+\sum _{j=0}^t {q^{j+1}} /\sum _{i=0}^s {q^{-i}} =\frac{q^{2+t+s}-1}{q^{1+s}-1}>q^{1+t}\)
\(\square \)
Proof of the Theorem
Part a. Suppose that \(S_i \;{\succeq }\;R_i ,\) for all \(p_i =i/2n\in (p_*,p^{*})\) for some \(n\in N.\) Let \(p\in (p_*,p^{*})\) be given. Without any loss of generality let \(p_*\) be written as \(p_*=(k_*-1)/2n\) for some \(k_*\in N.\) Let \(k\) denote the smallest integer larger than 2np, that is \(k=\left\lceil {2np} \right\rceil \) and \(k^{*}\) denote the largest integer smaller than 2np \(^{*}-1\), that is \(k^{*}=\left\lfloor {2np^{*}-1} \right\rfloor .\) By the supposition in the theorem and the construction of \(k,k_*,k^{*}\) one has \(S_i \;{\succeq }\;R_i ,\) for all \(p_i \in \Psi (k_*,k^{*})\). By Proposition 5.1 and construction one has
where \(\kappa =\kappa (q,k^{*}-k,k-k_*),\) which completes the proof.
Part b. For any given \(\varepsilon \in (0,p^{*}-p_*),G\in N\) and \(q>1\) take \(n^{*}\ge \left( {1+\ln G/\ln q^{2}} \right) /\varepsilon \). Let \(n\ge n^{*}\) be given and suppose that \(S_i \;{\succeq }\;R_i\), for all \(p_i =i/2n\in [p_*,p^{*}].\) Construct \(k=\left\lceil {2np} \right\rceil ,p=p^{*}-\varepsilon \) and \(k^{*}=\left\lfloor {2np^{*}-1} \right\rfloor ,\)and apply Lemma 5.1 and \(n\ge n^{*}\) to get
Next apply subadditivity of \(v(.)\), the last inequality and inequality (*) in part (a) of the theorem to get
Hence, \(v(zG)f(p^{*}-\varepsilon )\le v(zG)f(p_*)+v(z)\left[ {f(p^{*})-f(p_*)} \right] ,\) which completes the proof for it reveals that
Rights and permissions
About this article
Cite this article
Sadiraj, V. Probabilistic risk attitudes and local risk aversion: a paradox. Theory Decis 77, 443–454 (2014). https://doi.org/10.1007/s11238-013-9410-3
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11238-013-9410-3