Economic Experiments on Impulsive Urges, Control, and Irrationality
One of the standard assumptions underlying microeconomic theory is that more opportunities from which to choose cannot be a bad thing. Indeed, it is logical from a welfare perspective and a common sense perspective that things like more employment opportunities, including the social mobility that goes along with it, allows a person to realize a higher level of life satisfaction in multiple realms. Logically this freedom to choose should extend to the decision situation, in that a person should not seek to restrict the type of environment she makes her decisions in. This thesis contains studies in which decision environments have potential negative effects on the decision-maker’s welfare. We believe that if the decision maker were to be made aware of the type of effect the environment had on her decision, she would seek to restrict her environment, in other words the way information is transmitted to her.
More generally, the thesis looks at different types of control. All studies have to do with control, whether it be internal control, such as dealing with one’s own temptations, or external control, such as an irrational aversion to ceding another person control over your own payoff (for example, allowsing someone else to be in the driver seat despite equal abilities, or an unwillingness to let a computer determine one’s own lotto ticket numbers). It looks at human decision making through the lens of dual processing models, specifically the interplay between an impulsive, fast, and biased decision-making system; and a slow, deliberative, rational one (2004). Much behavioral research is pointing to the importance of understanding these two motivations in everything from exercise patterns to reciprocity in social interactions (DellaVigna & Malmendier (2006), Knoch, et al., (2006)).
This thesis is divided into four chapters. The first three chapters concern behavioral biases and inconsistencies brought to light by psychologists and behavioral scientists. The fourth chapter concerns the processing of conflicting motivation in social decision making. The first two chapters concern intertemporal decision making. Chapters 1, 2, and 4 are joint work with Urs Fischbacher.
Chapter 1 (Projection Bias: The Price for Food Craving) is about biases in how people predict their future preferences. Most decisions people make involve consequences for the future, and as a result the decision maker must make an estimate about what her preferences will be at that time. Of course these predictions are just that, an estimate, and are subject to error. They are also subject to biases. Projection bias (Loewenstein, O'Donoghue, & Rabin, 2003) is the tendency for predictions about preferences in a different “state” from the one at the moment of deciding are biased toward the one a person is in at the moment of deciding. For instance, Read & van Leeuwen (1998) show that people who are hungry when making a decision are more likely to choose an unhealthy snack to be received at a future date than people who are not hungry when making the decision. Indeed, if a person expects to be hungry in the future it makes sense to pick the more calorie rich food item, and people do. However, current levels of hunger should have no effect, but there is a large impact. Of those who are hungry when making their decision, 78% (56%) choose the unhealthy snack when they predict they will be hungry (satiated) at the time of receipt, whereas for those who are satiated when decided, only 42% (26%) choose the unhealthy, calorie-rich snack.
Our study also deals with projection bias under hungry and sated states. We use an exogenous and randomly assigned treatment variable to manipulate hunger levels of our subjects, whereas past studies have not. In our experiment, subjects participate in a Vickrey auction for high quality chocolates to be received on a later day, in which it is optimal for a subject to bid her true willingness to pay for a good. We are therefore not only able to show the existence of projection bias, we also show its effect on willingness to pay for products. Projection bias in the marketplace affects consumption decisions, and therefore lifetime patterns of consumption, saving, and overall well-being. It is also a potential source of gain for firms, and for these reasons understanding how projection bias translates into willingness to pay is important. We show that hungry subjects are willing to pay 58% higher prices for a small box of chocolates than sated subjects.
Chapter 2 (Battling Impulses: Intertemporal Choice in the Short Term) contains three experiments about self-control. Many studies look at impatience by giving subjects a series of decisions between an early, small payoff and a later, larger payoff (delay discounting tasks). These types of studies generally go further to estimate the implied discount rates of subjects’ decisions. Most studies (with a few exceptions) use monetary rewards as a payoff medium, and time spans of the tradeoffs are quite large, on the order of months or years. Using monetary rewards to study time preferences involves problems, mainly because money is not a primary reward; rather it is an opportunity set from which people can obtain a real reward. We are interested in this study in very short term impatience. Indeed, we think that short term impatience is very important to understand, as most temptation occurs when the prospect of a reward is imminent. This makes the use of monetary rewards even less appropriate for our study. We therefore introduce the paradigm of computer games as a medium of reward. Past studies have used food; we viewed this as problematic as we expect it to violate the “more is not worse” concept. We chose our computer game to be tempting and enjoyable, and contrasted it with an annoying task. We did this, so that even if subjects did not want expressly to play the game (which they were not forced to play), they would still desire to avoid the annoying task. Our treatment variable is intended to manipulate the degree to which impulsive motivations are given priority. In the less impulse-friendly environment, subjects make decisions regarding the game before they start the experiment, so before they are actually involved in the game or the task. In the more impulsive environment, they make their decisions while doing the game or task.
We link behavior in our “temptation tasks” to scores on the Barratt Impulsiveness Scale (BIS), and in experiment 3 of the series we show that behavior is additionally linked to the delay discounting task, using Amazon.de gift certificates as a reward medium. We find that scores on the BIS have more predictive power in the impulsive condition. Furthermore, we find with the delay discounting task that subjects are “present-oriented”; that is have declining discount rates.
Chapter 3 (Handing Over the Reins: On the Social Nature of the Illusion of Control) concerns the “Illusion of Control” brought to light by Ellen Langer (1975). The illusion of control appears in decisions involving risk. Risky decisions in real life not only involve risk preferences, but also skill. For example, skiing is a risky activity, but the more skill a person has, the lower the probability of an accident (holding choice of slope difficulty constant). In this situation, the amount of risk a person takes depends in part on her assessment of her own skill level. Ellen Langer argues that people conflate the skill with the risk elements, and even when they are given control over elements of a task that have no influence on the risk involved, act as if they have been given (partial) control over the lottery. This leads people to potentially take more risk in situations in which they have a feeling of control.
Recent work by economists has challenged this notion, and found conflicting results over whether the bias exists. Past studies have generally compared giving illusory control to the decision maker versus giving it to another person. In this paper, I put forth the proposition that this comparison may lead to many different conclusions about whether there is an illusion of control if there is heterogeneity in subjects’ perceptions of their own skill level versus that of another person. I therefore conduct an experiment which makes it possible to observe whether there is heterogeneity in the type of illusion subjects have, and to then categorize them according to this characteristic. I also link their observed illusion of control to scores on the Magical Ideation Scale, which has been shown to be positively related to illusion of control biases (Brugger & Graves, 1997). I find in my study what initially looks like an illusion of control. However, I argue that it is actually the result of randomization. I also observe a link between scores on the Magical Ideation Scale and the amount of variation in individual answers, which I also will argue is the result of randomization, coupled with the unusual distribution of Magical Ideation scores.
Chapter 4 (Social Decision-Making Processes) looks at the processing of different types of conflict in social decision making through observing reaction times. We look at ego conflict (conflict between selfish and social motives) and social conflict (conflict between different social motivating factors). We use traditional (1st party) dictator games, where a decision maker decides over distributions of money for herself and another person, as well as 3rd party dictator games in which a decision maker decides over distributions for two other people with no consequences for her own payoff. We record reaction times as a way of measuring the conflict in a decision and as a way to assess the automaticity or controlled nature of selfish and social motivations. Rubinstein (2007) showed that reaction times could be used as a measurement of conflict when making a decision, and that lower conflict results in quick responses and higher conflict in faster responses. Additionally, automatic processes are thought to be fast and cheap, whereas controlled ones are slow and expensive; therefore we will be able to use reaction times to examine this idea. In our context, the 1st party decisions contain more conflict than the 3rd party conditions, simply because they contain a selfish motivation. With the 3rd party condition, we can assess the individual’s personal norm; that is her attitude about what is a fair way to allocate between two people independent of any selfish motivation. Past studies have shown that selfish decisions are made more quickly than decisions in favor of another person. We introduce the 3rd party condition to assess whether it is the selfish aspect of the decision that results in faster reaction times for selfish decisions, or another, previously undetected property of the decision.
We find that an increase in social conflict (that is, conflict between social motivating factors) results in increased reaction times. We further found that though selfish decisions are made faster, this is not the result of selfish motivation, but of other aspects of the decision. We also find that the personal norm is not well characterized according to our three identified types of social motivation (efficiency, maximin preferences, and absolute inequality aversion). The personal norm predicts increases in reaction times better than ego conflict with any particular social property, showing that the personal norm captures individual heterogeneity in values.