Defloor, Bart; Bleys, Brent; Van Ootegem, Luc; Verhofstadt, Elsy; Schalembier, Benjamin (2014). 'Measuring preferences with different life evaluation questions: is there a common cognitive core?' Paper presented at the annual conference of the HDCA, 2-5 September 2014, Athens, Greece.

There is considerable consensus in the literature that well-being is a multidimensional concept (Stiglitz et al. (2009)). Functionings such as health and material living standards contribute to individual well-being, but also other functionings such as social contacts or the state of the environment play a role. One common observation is that the relative contribution to well-being of each of these functionings differs across individuals, i.e. their individual judgments differ. E.g. older individuals might value health more than younger people. Policymakers, if they want to avoid paternalism, should take into account these differences in individual judgments (Fleurbaey et al. (2009)). The aim of our paper is to suggest a way to measure these preferences.

The problem is that information on preferences is difficult to acquire (Beshears et al., 2008). One popular way is to ask people directly to evaluate their life, using questions about how satisfied or how happy they are and then derive information about preferences from that. One downside of this approach, which we focus on in our paper, is that the trade-offs between the functionings depend on the type of life evaluation question if the questions are treated separately (see e.g. Defloor et al. (2013) on this), which is inconvenient from a policymaker's point of view. The World Happiness Report (Helliwell et al. (2012)) makes reference of this issue as well. When several types of life evaluation questions are compared, it is concluded that there are similar patterns, but that happiness, seen as an emotional report at a point in time, has different properties than life satisfaction. It is much less correlated with major life circumstances.

There are several reasons why the answers on the different life evaluation questions differ. The first refers to the distinction between affect and cognition. When people are asked to evaluate their life, their answers are a mixture of two things: a cognitive judgment reflecting how they judge their life and an affective judgment reflecting how they feel in their life at the moment (see Fleurbaey et al. (2009)). The former reflects the individual's preferences and can be seen as a policy relevant evaluation of the life the individual is leading, the latter is related to the mood of the day or the individual's positive or negative attitude towards life. The wording of the question or the position of the question in the questionnaire also play a role, e.g. happiness has a more affective connotation than life satisfaction (Defloor et al. (2013)). Helliwell et al. (2012) distinguish furthermore between experienced and remembered well-being and whether or not the question refers to a momentary evaluation or a longer time span.

If we believe that individuals have preferences, and these preferences are reflected in the cognitive part of the life evaluation question, then it could be possible to get better information on preferences by combining a number of life evaluation questions. This is the aim of our paper.

We aim to disentangle affective and cognitive information using a questionnaire (1662 respondents) carried out in Flanders, Belgium. The questionnaire contains, apart from information on functionings achievement, individual characteristics and personality traits, a total of 10 life evaluation questions. The standard happiness and satisfaction questions are included, yet it also contains questions about how valuable people find their life or about the extent to which they have capabilities. We assume that the individuals' cognitive evaluation about their life is the common core of their answers, but that affective information influences each separate question in a different way.

In order to get this information, we estimate a system of equations with the same list of independent variables –functionings, individual characteristics, personality traits and interactions between functionings and individual characteristics— but with different life evaluation questions as dependent variable. The coefficients with the functionings and the interactions are restricted to be the same in all estimations, while the coefficients with individual characteristics and personality traits are allowed to differ across estimations. The interactions between functionings and individual characteristics are used to allow for inter-individual differences in judgments.

Our analysis will provide insight in the common cognitive core underlying all life evaluation questions on the one hand, and affective information, which influences each question in a different way, on the other hand. This information can be used to calculate equivalent income, a normatively interesting measure of well-being proposed by Fleurbaey et al. (2009).