Essays in Behavioural Economics: Representation and Weighting of Numerical Information

Abstract

This thesis consists of four essays. The first two essays (Chapters 2 and 3) present experimental tests of how numerical information is represented in the brain, and the implications of this representation for economic decisions. The latter two chapters use field evidence to infer how decision makers weight numerical information in economic choices that include a social dimension. Chapter 2 experimentally investigates systematic bias in intuitive summation. The traditional (implicit) assumption in economics is that rational agents can add a sequence of numbers without error. Or, if errors occur, that they are random and hence cancel out in aggregate. Instead of random error, we find consistent underestimation, even though accuracy is incentivised, and the task is framed to be familiar and economically meaningful. Underestimation occurs both when people are asked to generate a best guess for the sum, and when asked to compare their impression of the sum to a given number. Underestimation bias can help explain anomalies in consumer choice. Chapter 3 tests the explanatory power of underestimation bias, and a limited- attention phenomenon called concentration bias, on willingness to pay for energy- efficient investments. These cognitive mechanisms can underlie behaviour generally attributed to discounting. In a pre-registered experiment on a large, nationally repre- sentative sample of car buyers, we elicit willingness-to-pay (WTP) for an improvement in fuel economy. The results support the pre-registered hypotheses and suggest that the proportion of the energy-efficiency gap attributed to time preferences may be exaggerated. Chapter 4 explores the nature of status-signalling in the market for new cars. Tra- ditional models of signaling assume that all agents are fully attentive. Empirical evidence of limited attention suggests status signals need to be salient and have an obvious meaning to be inferred correctly. In this paper, variation in salience comes iii from random differences in the presentation format of age identifiers on licence plates in Ireland and Great Britain. The age identifier is more salient in Ireland. Results indicate that the difference in salience has a causal effect on market demand both in terms of when purchases occur, and the type of cars purchased. Premium makes such as BMW - conventionally status signals - are less popular when the age identifier is more salient, implying substitution between status attributes of car make and age. These findings can inform labelling policy to nudge consumption in a pattern that generates positive externalities. Chapter 5 investigates how groups of experts weight a quantitative attribute when deciding how to allocate a scarce resource, namely literary prizes. I analyse the popu- lation of shortlisted novels for three literary prizes covering a time span of 1963-2021. I show that longer novels are more likely to win. The result is robust to controlling for author gender and Goodreads rating, and to whether one uses absolute page length or relative length on the shortlist. The size of the effect suggests other valid cues are underweighted in the process of selecting a winner. Judgment and decision making research suggests several causes of the bias. One is the representativeness heuristic: Other explanations include an effort heuristic and the effects of accountability in decisions. Heightened sensitivity to magnitude under joint-evaluation is another po- tential cause. These results may explain previous findings that Booker Prize winners are not higher quality than shortlisted novels. The findings have broader implications for the inferred quality of expert judgment.