Do Affective States Influence Risk Preferences?

Affect is often employed as an umbrella term that refers to the current moods and emotions of individuals. Moods are generally described as low-intensity, diffuse, and relatively enduring affective states that often arise for no particularly salient reason. Compared with moods, emotions are considered to be more intense and short-lived affective states that generally have a definite cause and clear cognitive content (Davidson 1994; Gray and Watson 2001). The distinctions between moods and emotions are largely theoretical rather than empirical in nature. In research practice, identical methods are often used to induce both moods and emotions (Fredrickson 2002). In this paper, we use the terms “affect”, “affective state”, and “mood” interchangeably. Nevertheless, we believe that the definition of mood – low-intensity, diffuse, and not object-directed – is more appropriate to our empirical examination than the definition of emotion – intense, object-directed, and with clear cognitive content.

Pham (2007) presents a critical review on the empirical evidence on the relationship between emotion and rationality.

According to Thomson Reuter’s web of science, 188 articles have been published on the relationship between affect (emotion, mood) and risk references in the social sciences since 1980 until 2014.

The Journal of Business & Psychology edited a special issue addressing null results and several articles in this special issue served as excellent examples when null results can be meaningful.

Overall, the internal reliabilities for extroversion (Cronbach’s α = 0.74), agreeableness (α = 0.23), conscientiousness (α = 0.71), emotional stability (α = 0.69), and openness (α = 0.52) were not satisfactory. Therefore, we conducted a rotated factor analysis and extracted four components with eigenvalues that were greater than 1, that is, extroversion, emotional stability, conscientiousness, and openness. Agreeableness was not found to be a separate factor and was not included as a control variable in the analysis. The principal component analysis (PCA) considers all available information, whereas sum scores of the items for one personality dimension would ignore the personality differences among participants. The factor scores and the sum scores were nearly perfectly correlated for four of the five personality dimensions examined (rs > 0.90, ps < 0.001).

We pre-tested film clips from “Silence of the Lambs” and “Halloween” as suggested by Hewig et al. (2005) to induce fear in a sample of 50 students. These film clips did not significantly induce a fearful state in our participants. Thus, we chose to use a film clip from “Paranormal Activity” that successfully induced the desired fearful state.

We used factor scores for the analysis because they consider all the available information, whereas sum scores of the items for one mood type would ignore the mood differences among participants that arose from changes in fear or sadness. We included the same PANAS-X items in the principal component analysis (PCA) factor scores as in the sum scores. Varimax rotation is applied. The factor scores and sum scores were nearly perfectly correlated for all three of the examined moods (r > 0.95, p < 0.001). The results are robust to using sum scores instead of factor scores.

d = 0.20 denotes a small effect, d = 0.50 a medium effect, and d = 0.80 a large effect (Cohen 1988).

Detailed correlation results available upon request.

No evidence against poolability was produced by a poolability analysis across the factor of gender.

We only consider α = 0.05 given that none of our results survives the conservative Bonferroni correction. Statistical power was calculated using formulas derived from first principals, which yields results that are identical to those obtained from using standard software packages such as GPOWER (Erdfelder et al. 1996).

Ibid.