Synthetic indicators to analyze work-related physical and psychosocial risk factors: evidence from the European Working Conditions Survey.

In modern workplaces, alongside physical, chemical, and biological hazards, other risks are linked to the organisation of work and to the nature of the work itself. This paper investigates the association between workers' well-being and both psychosocial and physical risk factors at work proposing a synthetic measure suitable to generate insights on well-being at work and on individual risk factors. Exploiting data from the European Working Conditions Survey, we select as response variable the "self-assessed health". As this proxy of well-being is measured on a Likert scale, Ordered Probit analyses are run, and respondents' profiles are illustrated. Then, a Principal Component Analysis is carried out to build two synthetic measures summarising the selected risk determinants. The resulting first principal components are subsequently used as synthetic indicators in further, simplified, Ordered Probit models to explain the impact of different sets of risks on perceived health. Such a methodology allows for a straightforward interpretation of the results since many different risk drivers are replaced by two continuous synthetic indicators. Our findings, in line with existing research, confirm that both types of risk factors do exert a substantial impact on workers' health, although the psychosocial determinants seem to be more prominent.

A mixture model to assess perception of discrimination on grounds of sexual orientation for policy considerations.

The paper explores the relationships between subjective and contextual covariates as determinants of discrimination perception in a large sample survey. A modelling approach is implemented to detect the perception of discrimination on the ground of sexual orientation through expressed ratings, in a cross-country perspective. Exploiting the capabilities of a mixture model, which interprets ordinal evaluations as the combined results of two latent components, the cognitive process of selection among discrete ordered alternatives is discussed in terms of attractiveness towards the item and uncertainty in the response pattern. Gender, age and political orientation turn out to be significant variables at the individual level. Furthermore, the ILGA country score and education level, as contextual variables, exert prominent effects at country level. Empirical evidence from the Special Eurobarometer 2015 on Opinions as well as a simulation experiment support the usefulness of the approach.