Investigating well-being at work via composite indicators

The paper investigates the effect on workers’ well-being of selected risk factors existing in the workplace. Following the framework proposed by EU-OSHA (2013), the two categories of risk factors under investigation will be on the one hand the physical risk factors and on the other hand the psychosocial ones. More in particular, our objective is to display how psychosocial risk factors can affect workers’ health conditions as much as the physical risk factors, using some of the evidence of the European Working Conditions Survey (Eurofound, 2017). As a proxy of workers’ well-being, the variable of interest, that is the self-assessed health (SAH), stems from question Q75: “How is your health in general? Would you say it is (..:)” measured on a 5-point Likert scale (1 Very good; 2 Good; 3 Fair; 4 Bad; 5 Very bad). Current literature, as summarized in OECD handbook (2008), emphasizes the stages to achieve effective and consistent composite indicator. In order to build a model-based composite indicator for the SAH, the methodology chosen for this paper involved a number of steps. First, due to the ordinal variable of interest, Ordered Probit analyses were run based on explanatory variables describing the physical risks at work, the psychosocial risks and some individual characteristics. Then, a Principal Components Analysis was carried out to build a composite indicator summarising the selected variables, and the resulting Principal Components (PCs) have subsequently been used as explanatory variables in further Ordered Probit analyses. Moreover, results from the latter models are compared to measure the intensity of the relationship between SAH and variables identifying physical and psychosocial risks, highlighting those more relevant in influencing SAH. The results display that both types of risk factors do exert a significant impact on workers’ health, and in both cases the synthetic indicator (i.e. the first PC) accounts for most of the variance providing and effective synthesis of the data. When included in Ordered Probit models to measure the strength of their effect on the self-reported health, the indicators built for the two sets of risks turn out to be significant, both together and alone. The benefit of building these synthetic indicators relies on that they allow for simplifying a model-based analysis and may help in disentangling specific drivers of work-related well-being, as long as they are actually carriers of information, with the additional advantage of removing redundant information, obtaining more robust models.