Economic Consequences of Mispredicting Utility

The misprediction of utility in general has recently been introduced in the series on anomalies in the Journal of Economic Perspectives (Kahneman and Thaler 2006). Specific suboptimal consumption choices are reviewed in Hsee et al. (2012). It is argued, for example, that the mode under which available choice options are evaluated is a source of error when people predict the utility they may accrue from the options.

Our proposition is related to a long tradition in economics arguing that individuals tend to focus too much on material goods and disregard goods providing non-material benefits (see Lebergott 1993; Lane 1991; Frank 1999). Most importantly, Scitovsky (1976) claimed that “comfort goods” are over-consumed compared to goods providing “stimulation” (for a discussion of Scitovsky’s contribution to the understanding of well-being, see Pugno 2013). Comfort goods are described as defensive activities, protecting against negative affect. They have a strong extrinsic component, consisting of the consumer goods achieved through rapid productivity growth. In contrast, stimulation comes from creative activities providing novelty, surprise, variety and complexity. These aspects accentuate the renewal of pleasurable experiences, as it is also emphasized for intrinsic attributes.

Taking commuters’ reported satisfaction with life as a proxy measure for individually experienced utility is an approach that follows a substantial recent literature on reported subjective well-being, satisfaction and happiness in economics (e.g. Frey and Stutzer 2002; Layard 2005; Di Tella and MacCulloch 2006; Stutzer and Frey 2010), as well as in psychology (e.g. Kahneman et al. 1999; Diener et al. 1999).

A characterization of the economic approach to human behavior is provided, for example, in Frey and Stutzer (2001).

This terminology follows the work of Kahneman et al. (1997) in which utility is interpreted as a hedonic experience. Both utility measures—predicted and experienced utility—diverge from traditional decision utility that is derived from individual behavior.

A theory of the role of salience in choice under risk is proposed in Bordalo et al. (2012). In their model, those attributes of choice options are salient that are dissimilar across options.

We borrow these categories from a large literature in humanistic or value psychology (e.g. Maslow 1968; Rogers 1961).

The underlying theories are manifold, and include the urge to master one’s environment for its own sake (White 1959) to be an origin (DeCharms 1968), to resist loss of control (Brehm 1966) and the reflection of perceived control in more effective behavior and higher positive affects (Bandura 1977; Seligman 1992).

Alternative classifications of attributes and characteristics might well be possible and productive. For example, in a recent application, pro-environmental attributes of choice options are linked to utility misprediction (Welsch and Kühling 2011).

When people spend time with friends because they are famous or important, the extrinsic dimension becomes more prevalent.

A complementary discrepancy is the one between the happy life filled with positive affect and the meaningful life (Baumeister et al. 2013).

Young academics might be particularly worried about life after a negative tenure decision. Gilbert et al. (1998) asked assistant professors how happy they thought they would be after a positive and a negative tenure decision. The answers were compared with the reported subjective well-being of academics affected by a tenure decision made five or less years previously. Although assistants predicted they would be less happy in the first five years after being turned down, there was no statistically significant difference between those who had and had not gotten tenure. Similarly, assistants also overestimated the positive impact of receiving tenure on their subjective well-being.

The arguments further complement the research on heuristics or “rules of thumbs”. The idea is that adopting them is leading to better decisions than evaluating expected outcomes on a case-by-case basis. The crucial challenge is the identification of conditions under which the use of rules and principles leads to favorable outcomes (for a discussion see Hsee et al. 2012).

In contrast, learning is not an option when it comes to once-in-a-lifetime choices. Biased decisions can then well affect one’s life path. We believe that misprediction of utility matters greatly when it comes to important life decisions (like career choice), but we have not studied them here.

In an empirical analysis, Simonsohn (2006) provides related evidence for the inherent complexity of commuting choice. He argues that commuting behavior can be better understood in a framework of constructed preferences. People come up with some reference level of commuting time or commuting radius that they are only prepared to give up after experiencing negative effects on their well-being. In a challenging study on people moving from one US city to another, Simonsohn finds that people coming from a city where the average commuting time of the population is high (or low) also choose to commute more (or less) than average at their new place of residence (keeping individuals’ own past commuting experience constant). In the latter model, people can end up either commuting too much or too little.

In a related study, Comerford (2011) analyzes the choice of transport mode for commuting in a framework that takes into account focalism. Focalism is a cognitive bias that leads individuals to overweigh the contribution of certain attributes of choice options to experienced utility.

It is well documented that commuting is both physically and mentally stressful (e.g. Novaco et al. 1990). The strain of commuting is associated with raised blood pressure, musculoskeletal disorders, lowered frustration tolerance, increased anxiety and hostility, being in a negative mood when going to work in the morning and coming home in the evening, increased lateness, absenteeism and turnover at work, as well as adverse effects on cognitive performance (Koslowsky et al. 1995). In a recent panel study for the UK, Roberts et al. (2011) find commuting to be related to lower psychological health in particular for women. This effect holds when statistically controlling for health status, net household income, housing quality, working hours as well as job satisfaction. As the controls include (1) variables that potentially capture compensation in the housing and labor market and (2) a measure of domain satisfaction (i.e. job satisfaction), the results cannot easily be interpreted within the framework of the current study.

The strong notion of equilibrium has only been partially tested so far. It has not been studied whether there are systematic rents: rather, derived hypotheses within the equilibrium framework have been analyzed. There is considerable evidence for capitalization of transportation infrastructure in the price of land and for compensating wage differentials due to commuting distance. However, these findings do not require an equilibrium situation, and can also be explained by the law of marginal substitution (e.g. Timothy and Wheaton 2001; van Ommeren 2000).

If income is kept constant, commuting time is expected to enter negatively into the equation.

Due to lack of data, we cannot study adaptation to luxury housing.

For the lag structure, we draw on the findings reported in Di Tella et al. (2010) on adaptation to household income. The adaptation process fades out after three lags. Moreover, the results with three lags are very similar to the results with four lags but allow us to include more observations.

Here, only ordinary least squares estimations are reported. Thus, it is implicitly assumed that the answers can be cardinally interpreted. While the ranking information in reported subjective well-being would require ordered probit or logit regressions, comparative analyses for GSOEP have shown that it makes virtually no difference whether responses are treated ordinally or cardinally (Ferrer-i-Carbonel and Frijters 2004). The 11 categories of the dependent variable seem to mitigate potential problems from assuming continuity.

Here, similar partial correlations for the covariates are estimated. The full equation in panel (A) with individual fixed effects reads as follows: u = −0.0019*** (SE = 0.0006) * commuting time −0.053 * 10⁻³ (0.078 * 10^-3) * age² +0.256*** (0.045) * single with partner +0.292*** (0.049) * married +0.340*** (0.100) * separated with partner −0.389*** (0.065) * separated no partner +0.451*** (0.064) * divorced with partner −0.024 (0.060) * divorced no partner +0.623*** (0.165) * widowed with partner −0.093 (0.090) * widowed no partner −0.446*** (0.131) * spouse living abroad +0.074*** (0.019) * one child in household +0.098*** (0.026) * two children in household +0.157*** (0.041) * three or more children in household −0.252*** (0.042) * no. of persons in the household^1/2 −0.064*** (0.024) * self-employed −0.136 (0.097) * old German Laender −0.008 (0.110) * EU foreigner +0.038 (0.079) * other foreigner +0.187*** (0.025) * first interview + year fixed effects + 7.821*** (0.114).

Several of them are empirically addressed in Stutzer and Frey (2008).

The respective estimation for real household post-government income in Di Tella et al. (2010) amounts to 65.2 %.