Ordered Response Models and Non-Random Personality Traits: Monte Carlo Simulations and a Practical Guide

Abstract

The paper compares different estimation strategies of ordered response models in the presence of non-random unobserved heterogeneity. By running Monte Carlo simulations with a range of randomly generated panel data of differing cross-sectional and longitudinal dimension sizes, we assess the consistency and efficiency of standard models such as linear fixed effects, ordered and conditional logit, and several different binary recoding procedures. Among the binary recoding procedures analyzed are the conditional ordered logit estimator proposed by Ferrer-i-Carbonell and Frijters (2004) that recently has gained some popularity in the analysis of individual well-being, as well as the new developed "Blow-Up and Cluster" (BUC) estimator of Baetschmann et al. (2011). The Ferrer-i-Carbonell and Frijters estimator (FCF) performs best if the number of observations is large and the number of categories on the ordered scale is small. However, the BUC method performs similarly well and even outperforms the FCF estimator if the number of categories on the ordered scale is large. If the researcher is only interested in the relative size of coefficients with respect to a baseline, however, the easy-to-compute linear fixed effects model delivers essentially the same results as the more elaborate binary recoding schemes.

Does parenthood increase happiness? Evidence for Poland

“…In some situations, if the scale of the dependent ordinal variable is relatively long and approximates interval data, linear regression may yield similar results as ordered probit or logit regression (Ferrer-i-Carbonell and Frijters 2004). However, in our case the dependent variable has only four categories and so using the linear specification, which requires the cardinality assumption, 3 is questionable since it might severely bias our estimates (for the discussion on the topic see Baetschmann et al 2011;Geishecker and Riedl 2010). The simplest alternative approach to this problem is to recode the ordinal 3 The cardinality assumption implies that the intervals between the categories of the dependent variable are equal.…”

Does Income Inequality Make Us Less Happy?

“…We thus utilise an estimator suitable for non‐linear models. Specifically, we adopt the ‘blow‐up and cluster’ (BUC) estimator proposed by Baetschmann, Staub and Winkelmann () which, as Geishecker and Riedl () find via simulation, outperforms the popular Ferrer‐i‐Carbonell and Frijters () estimator when the number of categories of y is large, as in our case. Our preferred model is thus a fixed‐effects ordered logit model, which does not rely on the assumption of cardinality; for comparison, we also present results from linear fixed‐effects models and pooled ordered logit models.…”

The Impact of Matching Mission Preferences on Well-Being at Work

“…The ordinal …xed e¤ects estimator used in comparison is the 'blow-up and cluster'estimator (see Baetschmann et al 2011 for an extensive discussion). Geishecker and Riedl (2012) show that this method performs as well as or better than the other available ordered …xed e¤ects methods.…”

“…Given that the ordinal …xed e¤ects methods o¤er no bene…ts in interpretation over the linear …xed e¤ects speci…cation, I use a cardinal scale of job satisfaction in the estimation. Geishecker and Riedl (2012) show that the assumption of cardinality still allows to interpret the results in ratio's of parameter estimates. A feel for the size of an e¤ect can be obtained by comparisons with other estimates from the same regression.…”