Schedule - Parallel Session 2 - Choice in Context

Engineering F1.10 (note the change of room from F105-106 to F1.10) - 14:00 - 15:30

Anchors Or Targets? An Examination of Credit Card Statements

Daniel M. Bartels; Abigail B. Sussman


Anchors and reference points differ in important ways. Notably, reference points serve as targets, with motivational properties, whereas anchors act as neutral starting points for subsequent judgments. One area where these differences may be particularly critical is in the domain of personal finance and specifically, choices of credit card payment amounts in response to values appearing on credit card statements. Research designs are often unable to disambiguate whether these minimums are anchors or reference points, and researchers have been agnostic regarding this distinction. Instead, this literature often refers to minimum values as anchors despite the fact that they appear to have many properties of reference points (e.g., Stewart, 2009). However, this distinction may be important for understanding which values can effectively be used to increase chosen payments. From a practical perspective, we investigate how information provided on credit card statements influences subsequent payments. Importantly, we aim more generally to better understand how people rely on cues in their environments. We put forth methods for distinguishing anchors from references points and highlight the importance of this distinction. Additionally, we examine how people use external cues as goals to motivate themselves. We begin by examining how participants select payments as a function of suggested values proposed to them on credit card statements (Study 1). In Studies 2 and 3, we investigate whether suggested values take on properties consistent with reference points, namely loss aversion and diminishing sensitivity. We test for loss aversion by examining whether people feel different amounts of satisfaction if an identical payment is above or below the suggested amount. We also examine whether people have higher motivation to pay when they are close to (vs. far away from) their goal. Study 4 tests whether disappointment tracks the same patterns as motivation does. To test for the ecological validity of these findings, we turn to a large data-set of Chase cardholder payments. We examine distributions of payments around values that have been explicitly selected by cardholders as goals, and examine how inclusion of these values alters payments (Study 5). After our detailed look at values on credit card statements, we broaden our findings by revisiting existing datasets on anchors and reference points across domains (Study 6). We conclude with a discussion of how distributions of anchors and reference points vary and describe how our more nuanced understanding of values on credit card statements can be used to encourage debt reduction.

Daniel M. Bartels

Assistant Professor, University of Chicago

A Revealed Preference Test of Quasilinear Preferences

Mikhail Freer; Marco Castillo


Quasi-linearity is an essential assumption underlying important results in different areas of economics (e.g. the revenue equivalence theorem, implementation in dominant strategies and the rotten kid theorem). Brown and Calsamiglia (2007) propose a test of the joint hypothesis of quasi-linear and concave preferences with linear budget sets. We develop a test of quasi-linearity of preferences that relaxes both assumptions: concavity of the utility function and linearity of budgets. That is, 1) observed demand can be generated by maximization of utility function that represents quasi-linear preferences if and only if revealed preference relation passes the test and 2) the test imposes minimal requirements on budget sets, hence it can be used to test quasi-linearity of preferences in strategic environments like auctions. The test we propose has combinatorial nature and can be executed in polynomial time. The paper presents several applications of the test. In the context of Arrow-Debreu securities as in Choi et al (2007) and Choi et al (2014), the condition provides a test of subjective expected utility with risk neutral preferences and constant probabilities. We also test for quasi-linearity of preferences in context of consumption goods. For this purpose we use experimental data from Mattei (2000) to compare the predictive success (Beatty and Crawford, 2011) of the assumption of quasi-linearity in (at least) one good versus the generalized axiom of revealed preference.

Mikhail Freer

PhD Student, George Mason University

The Influence of Experience on Bidders' Cognitive Biases in Public Procurement

Timo Tammi; Jani Saastamoinen; Helen Reijonen


Although cognitive biases as deviations of rationality are well-known and empirically identifiable phenomena in many decision-making environments, only a bit, if anything, is known about cognitive biases among firms in the context of public procurement and about how they are associated with firm behaviour in tendering for procurement contracts. According to the so-called discovered preferences hypothesis (Plott 1996; Cubitt et al. 2001; Bardsley et al. 2010) cognitive biases will disappear in situations where incentives are salient, decisions are made repeatedly and decision-makers get feedback. Competitive bidding in public procurement is, clearly, an environment where firms are able to learn due to repetition, feedback and salient incentives. Consequently, we may hypothesize that experienced bidders exhibit less cognitive biases than non-experienced bidders do in public procurement. In this paper, we study loss aversion, overconfidence bias (in the form of ‘better than average effect’) and risk attitudes among firms who have participated in competitive bidding in public procurement for different time periods and have therefore different levels of experience of bidding in this particular context. The analysis is based on survey data of nearly four hundred firms registered in the most-widely used electronic tendering management system in Finland, which have made one or several bids, or have planned to do so. We analyze how loss aversion, confirmation bias and risk attitudes differ and are distributed among firms having various levels of experience of bidding in public procurement. The survey data is being currently (December 2015) collected by with an electronic questionnaire. Loss aversion is measured by using a multiple price list format (MPL) along the lines of Fehr and Goette (2007), Gaechter et al. (2010) and Koudstaal et al. (2015). Concerning overconfidence, we employ instruments used by Szyszka (2013) (see also Stotz and von Nitsch 2005) in asking firms to self-evaluate their competencies in bidding in public procurement. In the measurement of risk attitudes, we use two instruments. The first is the survey-based measure used by Dohmen et al. (2011), Koudstaal et al. (2015) and Vieider et al (2015). Respondents were asked to express their self-perceived willingness to take risk in general and in more specific domains of driving, finance, leisure and sport, occupation, health and faith in other people. Another measuring tool is a choice based risk elicitation task applied in Menkhoff et al. (2006). The analysis provides preliminary information on evaluating the discovered preferences hypothesis in a natural environment outside a laboratory setting. It also gives guidelines regarding pooling the survey data with the bidding data available via the electronic tendering management system. As a practical application, the findings of this study can help to attenuate cognitive biases in tendering.

Timo Tammi

Senior lecturer, University of Eastern Finland Business School

Why Accommodate Decision-Making Heuristics in Discrete Choice Experiments?

Seda Erdem


The discrete choice experiment (DCE) is a technique for eliciting individuals’ preferences. Notwithstanding the appeal of DCEs, there are some issues raised in the literature that might be important. For example, in DCE studies, the typical assumption that individuals consider and trade-off between all attributes within the choice sets is often questioned. Indeed, a number of studies show that many respondents exhibit signs of adopting a range of simplifying mental processing rules, which are referred to as decision-making heuristics. Making decisions based on certain criteria, such as the cost thresholds, is one of such heuristics. ‘Elimination-by-aspects’ (EBA) and ‘selection-by-aspects’ (SBA) like behaviours are examples of such cases where respondents make choices by eliminating or selecting some alternatives based on some decision criteria. Despite the increased attention on decision-making heuristics within the stated preference literature, EBA and SBA behaviour have largely been overlooked. This paper furthers this line of enquiry and explores EBA and SBA behaviour in the context of public preferences for health service innovations. To do this we use empirical data obtained from a DCE survey administered in the UK exploring public preferences relating to health service innovation investment decisions. The modelling approach we use if flexible and capable of addressing EBA and SBA choice behaviour, whilst addressing preference heterogeneity. We use the approach to investigate the extent to which respondents eliminated alternatives based on who the health service innovations were aimed at or limited their choice to those alternatives that targeted a certain population. The empirical application of our modelling approach shows that it has important implications for model fit and welfare analysis. Our findings reveal that accommodating EBA and SBA concurrently proved to give a richer insight into respondents’ behaviour and suggested that as few as 40 percent of respondents did not present these heuristics. In line with previous studies, we found that assuming homogeneous preferences in respondents was inappropriate. Going beyond this, we also show that each segment of respondents differed not just in their preferences, but also in the decision-making heuristics they adopted. To conclude, failing to account for EBA-like and/or SBA-like choice behaviours was not optimal and has implications for welfare analysis. Where our models accounted for such behaviours they outperformed those based solely on the no-heuristics assumption. Although there are a large number of other decision-making heuristics and processing strategies that can also be considered and there is the potential to uncover additional random taste variation, the models illustrated here provide additional insight into the manner in which respondents arrive at their final decision. This represents an exciting research challenge.

Seda Erdem

Assistant Professor, University of Stirling