How Important is Methodology for the Estimates of the Determinants of Happiness?

Abstract

Psychologists and sociologists usually interpret happiness scores as cardinal and comparable across respondents, and thus run OLS regressions on happiness and changes in happiness. Economists usually assume only ordinality and have mainly used ordered latent response models, thereby not taking satisfactory account of fixed individual traits. We address this problem by developing a conditional estimator for the fixed-effect ordered logit model. We find that assuming ordinality or cardinality of happiness scores makes little difference, whilst allowing for fixed-effects does change results substantially. We call for more research into the determinants of the personality traits making up these fixed-effects.The empirical economic literature on self-reported happiness, also termed life satisfaction, seems to be taking off. Whereas in the 1970s and 1980s there was only a trickle of articles on happiness, 1 the last couple of years witnessed a spate of empirical studies on this subject. 2,3 Next to the economic literature, there are more than 3000 studies done in the last 30 years by psychologists and sociologists (Veenhoven, 1994;Veenhoven, 1997). This means that there is now a large combined literature on what causes happiness.In this paper, we investigate the robustness of all these findings. To this extent, we categorise the empirical studies on methodologies used that reflect the assumptions imposed on the meaning of satisfaction questions and on the influence of unobservables.On the meaning of satisfaction questions, psychologists have by and large interpreted the answers as cardinal, i.e. that the difference in happiness between a 4 and a 5 for any individual is the same as between an 8 and a 9 for any other individual. In the economic profession, cardinality is still considered very suspect (Ng, 1997). In studies of individual happiness therefore, we find that economic papers generally assume that satisfaction answers are only ordinally comparable, i.e. that it is unknown what the relative difference between satisfaction answers is but that all individuals do share the same interpretation of each possible answer.

Monetary valuation of informal care: the well‐being valuation method

“…The first ones assume that the answers to well-being questions are cardinal, while the second type of techniques only assume ordinality. Ferrer-i-Carbonell and Frijters (2004) show that assuming cardinality or ordinality generates similar results in terms of the trade-offs between variables, which is the information we use in this paper. The first measure of well-being (see Figure 1 in section 'Survey questions') is regressed by means of an ordered probit (OP).…”

“…Subjective questions on self-reported well-being have been used in economics to understand and explore a large range of interesting topics, such as unemployment, health, job situation, and income (Blanchflower and Oswald, 2004;Clark and Oswald, 1994;DiTella et al, 2001;Easterlin, 2001;Ferreri-Carbonell, 2005;Ferrer-i-Carbonell and Frijters, 2004;Ferrer-i-Carbonell and Van Praag, 2002;Frey and Stutzer, 2002;Frijters et al, 2004;Oswald, 1997;Van Praag and Ferrer-i-Carbonell, 2004). In doing so, economists take individuals answer to well-being questions as a proxy to measure utility (see, e.g.…”

“…Frey and Stutzer, 2002;Luttmer, 2005). In other words, one of the main assumptions in the well-being literature is that the answer to the well-being question is a monotonic positive transformation of the theoretical concept we are interested in (Ferrer-i-Carbonell and Frijters, 2004).…”

The effect of unemployment on self-reported health and mental health in Greece from 2008 to 2013: A longitudinal study before and during the financial crisis

“…[ Table 3 (Ferrer-i-Carbonell and Frijters;Schmitz, 2011). Although a random effects model is a possible alternative specification, a Hausman test revealed that this alternative is not supported by the data and that a fixed effect model is therefore preferred.…”

Can alcohol make you happy? A subjective wellbeing approach

“…The strength of FE models is that they remove unobserved, time-invariant confounding, and as a result have been recommended for studying both the impacts of alcohol (French & Popovici, 2011) and influences on wellbeing (Ferrer-i-Carbonell & Frijters, 2004). In Study 1, we use FE analyses of a conventional cohort dataset with gaps of several years between waves.…”