*2.2. Measures*

Hedonic wellbeing. Hedonic wellbeing was conceptualized as the employee's job satisfaction, and it was measured by a 10-item reduced version of the Job Satisfaction Scale (IJSS) by Cooper, Rout and Faragher [52], referring to intrinsic job satisfaction and extrinsic job satisfaction, and one additional item measuring general job satisfaction. The score for hedonic wellbeing was the global mean score for the three types of job satisfaction. It includes items such as "Opportunity to use your skills". The items have a seven-point Likert response format, ranging from 1 (quite dissatisfied) to 7 (very satisfied). Cronbach's alpha for the global score of Hedonic Wellbeing was 0.87.

Eudaimonic wellbeing. Eudaimonic wellbeing was conceptualized as a feeling of meaning and purpose at work, and it was measured by an 8-item reduced version of the scale constructed by Ryff [53], with two subscales: purpose at work and personal growth. The score for eudaimonic wellbeing was obtained by computing the global mean score for the two dimensions of the scale. It includes items such as "For me, life has been a continuous process of learning, changing, and growth". The items have a seven-point Likert response format, ranging from 1 (strongly disagree) to 7 (strongly agree). Cronbach's alpha for the global score of eudaimonic wellbeing was 0.72.

Performance—rated by the employee. Employees' self-rated work performance was operationalized as in-role performance (carrying out tasks required by the job), extra-role performance (carrying out tasks that are not required in the job description, e.g., helping others), and creative performance (carrying out tasks that are both creative and useful at work). In-role performance was measured by 3 items from a scale constructed by Williams and Anderson [54], extra-role performance was measured by 3 items from a scale by Mackenzie and colleagues [55], and creative performance was measured by a 3-item method constructed by Oldham and Cummings [36]. The composite score for performance was obtained by calculating the global mean score for the in-role, extra-role, and creative performance scales. It includes items such as: "I adequately complete assigned duties" (in-role performance); "I do not hesitate to challenge the opinions of others who I feel are leading the store/company in the wrong direction" (extra-role performance); and "How original and practical am I in my work?" The items have a seven-point Likert response format, ranging from 1 (strongly disagree) to 7 (strongly agree). Cronbach's alpha for the global work performance score was 0.71.

Performance—rated by the supervisor. Employee work performance evaluated by the supervisor was also operationalized as a general measure of performance quality. We measured these three aspects using three items: "What is his/her performance like?"; "What is the quality of his/her work?"; and "What was his/her level of goal achievement in the past year?" The items have a five-point Likert response format, ranging from 1 (very bad) to 5 (very good). Cronbach's alpha for the global work Performance score was 0.89.

Demographic variables included. Organization's sector: dummy variable (0 service, 1 production/construction). Gender: dummy variable (0 female, 1 male). Age: under 35 years old, between 35 and 50, and over 50 years old. The highest educational level achieved: no education or compulsory education, professional training or high school, advanced university degree. Occupational category: unqualified manual work, technician or administrative work, highly qualified professional, manager. Type of contract: dummy variable (0 = temporary, 1 = permanent). Seniority in the position: dummy variable (0 = less than 5 years, 1 = more than 5 years).

#### *2.3. Statistical Analysis*

The sample was divided into clusters using the two–step cluster analysis method developed by Chiu and colleagues [56] in SPSS v.22 (IBM Corp., Armonk, NY, USA). The SPSS two-step cluster method is a scalable cluster analysis algorithm designed to handle large datasets, such as those analyzed in the present study. The algorithm is based on a two–stage approach: in the first stage, it undertakes a similar procedure to the k-means algorithm. In the second step, based on these results, a modified hierarchical agglomerative clustering procedure is carried out that combines the objects sequentially to form homogenous clusters [57].

The two-step clustering algorithm output offers fit information, such as the Bayesian Information Criterion (BIC), as well as information about the importance of each variable for the construction of a specific cluster [57], which is an additional attractive feature of the two-step cluster method in comparison with traditional clustering methods. Empirical results indicate that the two-step clustering method shows a near-perfect ability to detect known subgroups and correctly classify individuals into these subgroups [58]. Based on these analyses, the sample was classified into groups reflecting different configurations of wellbeing and performance dimensions.

After finding cluster solutions for each of the combinations of variables of interest, we applied multidimensional logistic regression to explain cluster membership based on the demographic covariates described. Multinomial logistic regression is a statistical technique that specifies the dependent variable as a categorical variable that can take more than two values (in our case, the number of clusters). In multinomial logistic regression, one of the responses is chosen to serve as reference. Switching the reference group allowed us to compare the effects on all the groups. The independent variables are also categorical, with K categories. They are introduced in the model coded as k-1 binary variables. When the variables have two categories, they have been introduced as a dummy variable with a value of 0 or 1. In this case, the exponential beta coefficient represents the change in the odds of the dependent variable, associated with a one-unit change in the corresponding independent variable. When the variables have more than two categories, the coding system used is deviation coding. In this case, because there is no clear reference category, the reference category is coded as −1. This coding system compares the mean of the dependent variable for a given level to the mean of the dependent variable for the other levels of the variable. The exponential beta coefficient estimates the magnitude at which the probability of the occurrence of the event varies, comparing that category to the average of all the subjects in the study. Because the analysis does not show results for the reference group, we have repeated the analysis using the coding system with a different group as reference. With this system, we can obtain the coefficients for all the categories, which are presented in the results tables.

#### **3. Results**

#### *3.1. Descriptive Analysis*

The descriptive results are shown in Tables 1 and 2.


**Table 1.** Descriptive statistics (demographic variables).


**Table 2.** Descriptive statistics.

