|
Article Excerpt
1. Introduction
Throughout our lives, we face many choices between things we know we should do and things we want to do: whether or not to visit the gym, whether or not to smoke, whether to order a greasy pizza or a healthy salad for lunch, and whether to watch an action-packed blockbuster or a history documentary on Saturday night. In this paper, we investigate the effects of this type of internal conflict between the desire to do what will provide more short-term utility and the knowledge that it is in our long-term interest to do something else. In particular, we focus on the way this type of conflict leads individuals to make systematically different decisions in the domain of film rentals when they make choices in the present about what to watch versus choices for the future about what to rent.
A number of authors have discussed the distinction between goods that provide primarily long-term benefits, which we call should goods, and goods that provide primarily short-term value, which we call want goods. Options conceptually similar to shoulds have also been called "cognitive," "utilitarian," "virtue," "affect-poor," and "necessity" options, while options that are conceptually similar to wants have also been called "affective," "hedonic," "vice," "affect-rich," and "luxury" options (see Khan et al. 2005 for a review). We rely on the terms should and want to convey the internal tension produced by these competing options. The distinction between these different types of goods is important because evidence suggests that the context in which a decision is made may affect which types of goods, should goods or want goods, a person prefers.
The tendency to put off options preferred by our should selves (e.g., saving, eating vegetables) in favor of options preferred by our want selves (e.g., spending, eating ice cream) is stronger for decisions that will take effect immediately than decisions that will take effect in the future (Loewenstein 1996, Bazerman et al. 1998). Economists have modeled this phenomenon by proposing that people dramatically discount future utility relative to present utility (see, for example, Phelps and Pollak 1968, Ainslie 1992, Loewenstein and Prelec 1992, Laibson 1996, O'Donoghue and Rabin 1999), and with "multiple-selves" models in which individuals' decisions are controlled by multiple internal agents with competing preferences, one of which optimizes over a longer time horizon than the other and is more likely to control choices that are made for the future than the present (see, for example, Thaler and Shefrin 1981, Read 2001, Fudenberg and Levine 2006). In this paper, we empirically test for the time-inconsistent preferences that these models of present bias predict people will demonstrate when making repeated choices over the same set of goods, ranging from extreme want goods (items with only short-term benefits) to extreme should goods (items with only long-term benefits), when some decisions will take effect in the present and some will take effect in the future.
Evidence that people prefer want options over should options more frequently when making choices about the short run rather than the long run has been found in numerous domains (Oster and Scott Morton 2005, Wertenbroch 1998, Rogers and Bazerman 2008, Read and van Leeuwen 1998, Milkman et al. 2008b), including that of film rentals in a laboratory setting (Read et al. 1999, Khan 2007). To extend the study of the impact of present bias on people's preference rankings of want and should options beyond the laboratory, we obtained a novel panel data set containing individual-level information about consumption decisions over a period of four months from Quickflix[R], an Australian online DVD rental company. This data set comes from a domain in which individuals make rental choices for the future and consumption choices for the present from a set of goods that range from extreme should items (highbrow films) to extreme want items (lowbrow films). Repeated observations of the same individuals over time allow us to investigate both whether customers exhibit present bias and whether they learn to reduce their present bias with experience.
To test the theory that people exhibit present bias in the domain of film rentals, we begin by scoring the films in our data set on the spectrum from should to want items. We then use our rental data to test and confirm the hypothesis that the same Quickflix customer holds films longer the closer the films fall to the should end of the want/should spectrum. We also test and confirm the hypothesis that when customers rent two sequential films, the first of which has more should and fewer want characteristics than the second film, they are more likely to reverse their preferences (watching and returning the films out of order) than when they rent a movie with more want and fewer should characteristics first. Both of these hypotheses stem from the combination of a model of consumers as present biased and our definition of relative should and want goods. We thus interpret our findings as evidence that people exhibit present bias in the field when making decisions about film rentals. Finally, we address and attempt to rule out a number of alternative explanations for our findings. One set of analyses designed to rule out alternative explanations for our results reveals that consumers reduce the extent to which they hold films that fall closer to the should end of the want/should spectrum longer than other DVDs as they gain rental experience. This suggests that people learn about their present bias over time and that the effects we detect are unlikely the result of "optimal" decision-making strategies.
The remainder of this paper is organized as follows. In [section]2, we review the relevant literature on time-inconsistent preferences and clarify the origins of our hypotheses. In [section]3, we describe our data set and methods for rating films along the spectrum from should to want. We present the results of our analyses and discuss alternative explanations for our findings in [section]4, and present our conclusions in [section]5.
2. Past Research on Time-Inconsistent Preferences
A considerable literature on time-inconsistent preferences has developed since Strotz (1956) pointed out that people exhibit more impatience when making decisions that will take effect in the short run rather than the long run. Loewenstein and Thaler (1989),, Ainslie (1992), O'Donoghue and Rabin (1999), and Frederick et al. (2002) provide partial reviews of the literature on intertemporal choice, and Milkman et al. (2008a) review the literature on the context effects that have been shown to alter people's preferences for should versus want options.
Evidence from numerous laboratory studies indicates that consumers exhibit present bias when making choices about money (McClure et al. 2004, Thaler 1981, Kirby and Herrnstein 1995, Kirby and Marakovic 1996, Kirby 1997), lottery tickets (Read et al. 1999), relief from pain and irritation (Solnick et al. 1980, Navarick 1982, Trope and Fishbach 2000), films (Read et al. 1999, Khan 2007), and foods (Wertenbroch 1998, Khan 2007, Read and van Leeuwen 1998), among other things. Models of present bias have also been tested and confirmed in the field in the domains of gym attendance (Malmendier and Delia Vigna 2006), magazine newsstand and subscription pricing (Oster and Scott Morton 2005, Wertenbroch 1998), savings behavior (Angeletos et al. 2001, Ashraf et al. 2006), and supermarket quantity discounts (Wertenbroch 1998). Past field studies, however, have not directly tested whether people's preference rankings over a set of goods in advance of consumption are systematically different from those at the time of consumption, as predicted by a combination of a model in which consumers dramatically discount utility from future periods and our definition of want and should options.
For a number of reasons, it is empirically difficult to test models of consumers as present biased outside the laboratory. A direct test of any such model requires a data set containing information about the consumption decisions of the same consumers over time, where different decisions take effect at different points in the future. Past field studies have overcome the hurdle of obtaining individual-level consumption data over time in the domains of savings behavior and gym attendance. Partnering with a bank in the Philippines, Ashraf et al. (2006) offered commitment savings products to a subset of the bank's former clients. They confirm (for female subjects) the prediction that consumers who exhibit more present bias on hypothetical questions are more likely to take up commitment devices. Ashraf et al. (2006) track individuals' take up of a savings commitment device as well as the amount those individuals save in their bank accounts over a 12-month period. Malmendier and Delia Vigna (2006) employ a panel data set to examine individual-level gym attendance and contract types over a three-year period at several health clubs. They find that present bias explains the popularity of flat-fee contracts among gym customers who could have saved money by paying per visit. Although neither of these studies employs data that would permit the identification of explicit reversals in preferences at the within-subject level, both test predictions of models of present bias in the field at the within-subject level. Both studies also examine the choice of whether or not to engage...
|
