Enrica Carbone

Which Error Story is Best

Two recent papers, Harless and Camerer (1994) and Hey and Orme (1994) are both addressed to the same question: which is the ‘best’ theory of decision making under risk? As an essential part of their separate approaches to an answer to this question, both sets of authors had to make an assumption about the underlying stochastic nature of their data. In this context this implied an assumption about the ‘errors’ made by the subjects in the experiments generating the data under analysis. The two different sets of authors adopted different assumptions: the purpose of this current paper is to compare and contrast these two different error stories—in an attempt to discover which of the two is ‘best’.

Stochastic choice with deterministic preferences: An experimental investigation

Abstract Most theories of decision-making under risk imply deterministic choice, yet actual choice appears to be stochastic, the resulting discrepancy being attributed to ‘error’. This paper explores an alternative explanation — in which preferences are deterministic but choice is stochastic. Whilst being more satisfying from a theoretical perspective, our results show that the empirical performance is far from satisfactory.

A Test of the Principle of Optimality

This paper reports on an experimental test of the Principle of Optimality in dynamic decision problems. This Principle, which states that the decision-maker should always choose the optimal decision at each stage of the decision problem, conditional on behaving optimally thereafter, underlies many theories of optimal dynamic decision making, but is normally difficult to test empirically without knowledge of the decision-makers preference function. In the experiment reported here we use a new experimental procedure to get round this difficulty, which also enables us to shed some light on the decision process that the decision-maker is using if he or she is not using the Principle of Optimality - which appears to be the case in our experiments.

A Comparison of the Estimates of Expected Utility and Non-Expected-Utility Preference Functionals

This paper extends the literature on the estimation of expected utility and non-expected-utility preference functionals (and the consequent exploration of the superiority of non-expected-utility over expected utility preference functionals) to a comparison of two different ways (pairwise choice and complete ranking) of experimentally obtaining data on such preferences. What is revealed is that the magnitude of the subject error is clearly conditional on the elicitation method used and, rather alarmingly, that the preference functional apparently employed by the subject may also be conditional on the elicitation method.

Discriminating between Preference Functionals: A Preliminary Monte Carlo Study

This article is connected with recent attempts to estimate EU and Generalised EU preference functionals using (complete ranking) experimental data and maximum likelihood estimation techniques. In particular we explore, using Monte Carlo techniques, the power of such procedures in correctly determining the true preference functional. We conclude that several of the more popular generalisations to EU are very difficult to disentangle, and that the techniques are rather poor at correctly identifying EU when it is the correct functional.

Estimation of Expected Utility and Non-Expected Utility Preference Functionals Using Complete Ranking Data

In recent years, there have been many empirical investigations into various alternative models of decision making under risk. Most of them have been experimental tests of the predictions (or the axioms) of the various theories — experiments testing the empirical validity of the various new theories against each other and against Expected Utility Theory (EU). In contrast, a number of recent experiments have followed the alternative route of estimating preference functionals, to discover whether the more general preference functionals explain observed behaviour significantly better than the less general functionals. The experiment discussed in this paper belongs to the second group of experiments and follows two previous experiments (Hey and Di Cagno (1990), Hey and Orme (1992)), which estimated the preference functionals implied by several competing models. The first of these, Hey and Di Cagno (1990), reported on an experiment in which 68 subjects were asked 60 pairwise preference questions involving random prospects from four Marschak-Machina Triangles (see Machina 1987).

Understanding intertemporal choices

Previous experimental results show clearly that many subjects do not optimize when solving a life-cycle consumption problem. What do they do? This paper attempts to resolve this question, looking at the discounting, hyperbolic and rolling models as possible explanations. Data from two experiments (one an experiment with a typical subject pool and the second an experiment with subjects from the CentER panel) is used, and the advantage of having experimental data is exploited, which means that one can actually estimate the hyperbolic model. It is shown that the (exponential) discounting model appears to give the best explanation – suggesting that subjects do look ahead (as they should) but increasingly less as time passes (as they should not in the context of these experiments).

Discriminating between Preference Functionals: A Monte Carlo Study

This paper reports on the results of a Monte Carlo investigation into the power of commonly employed procedures for identifying the ‘correct’ preference functional of individuals, and hence for discriminating between the large number of preference functionals now advocated in the theoretical literature. The paper also asks which of two commonly employed experimental procedures might be the most efficient in this respect. The results show that several of the ‘newer’ preference functionals are difficult to distinguish empirically-at least on the basis of conventional experimental tests-and that the Complete ranking experimental design might be better than the Pairwise Choice design. The conclusion of the paper is that more thought should therefore be given to the question of the experimental design.

Are groups better planners than individuals? An experimental analysis

We present the results of an experiment comparing group and individual planning in the domain of lifecycle consumption/saving decisions. Individual decision making is compared to two group treatments, which differ based on the presence of a rematching rule. We find that individuals and groups differ in how they solve the intertemporal consumption problem, but not in how they improve their consumption planning within a sequence. Individuals’ performance improves across sequences, groups without rematching perform approximately the same, while groups with rematching do significantly worse. Our main finding is that while groups perform better than individuals in the first sequence, this difference seems to disappear in the second lifecycle. Results show that in the second sequence groups in the rematching treatment deviate substantially more from optimum than groups that are left stable across sequences.

The Effect of a Short Planning Horizon on Intertemporal Consumption Choices

Previous experimental results (Ballinger et al. (2003) and Carbone and Hey (2004)) have found that many agents fail to correctly take into account the length of the planning horizon also finding some support (See Carbone (2006)) for descriptive models, such as the Rolling Model. This paper presents an experimental analysis on the effect of a short planning horizon on intertemporal consumption choices. The purpose of the study is to test whether very short horizons are more easily perceived by agents, allowing them to plan optimally. This experiment tests a somewhat implicit assumption of the Rolling Model, or of similar descriptive approaches, namely that people might be able to use the optimal strategy if they are faced with shorter planning horizons. Moreover, this hypothesis is tested in the cases of decision making under certainty, risk and uncertainty, in order to analyze how these environments may affect the perception of the length of the planning horizon. Results suggest that planning periods have a significant effect on deviations from unconditional optimum in all sequences and all treatments. This finding has been interpreted as evidence of participants not using the optimal strategy. When conditional deviations are considered, results are confirmed only in the case of decision making under uncertainty. This second finding has been interpreted as suggesting that uncertainty on income seems to prevent participants from improving their decision making.