*3.2. Cluster Analyses: Different Operationalizations of the Wellbeing-Performance Patterns.*

As mentioned above, we used cluster analysis to find different patterns of relationships between performance and wellbeing, taking into account different operationalizations of wellbeing (i.e., hedonic vs. eudaimonic) and performance (i.e., self-rated vs. supervisor ratings). The results are shown below. Models 1 and 2 consider self-rated performance by the employee (hedonic wellbeing in Model 1 and eudaimonic wellbeing in Model 2). Models 3 and 4 consider performance evaluated by the supervisor (hedonic wellbeing in Model 3 and eudaimonic wellbeing in Model 4).

When performance is evaluated by the employee, there are four clusters: (1) employees who are high in both wellbeing and high performance; (2) employees who are medium low in wellbeing and medium high in performance; (3) employees who are medium high in wellbeing and medium low in performance; and (4) employees who are low in both wellbeing and performance.

When performance is evaluated by the supervisor, there are three clusters: (1) employees who are high in both wellbeing and performance; (2) employees who are high in wellbeing and low in performance; and (3) employees who are low both in both wellbeing and performance.

The results show that there are antagonistic patterns of wellbeing and performance (i.e., happy-unproductive, and in some cases, unhappy-productive). In fact, the results indicate that, on average, over 50% of the respondents belong to these clusters.

#### 3.2.1. Model 1: Hedonic Wellbeing vs. Self-Rated Performance (H-PE).

In Model 1, we consider two variables: hedonic wellbeing and self-rated composite performance rated by the employee. The auto-clustering algorithm indicated a four–cluster solution as the best model because it minimized the BIC value (BIC = 1060.892, BIC change from the previous cluster = −228.184). The average silhouette measure of cohesion and separation was 0.5, indicating fair to good cluster quality. The importance of both predictors was 1.00.

Four clusters emerged (see Figure 1): (1) employees high in hedonic wellbeing (*M* = 6.17, SD *=* 0.35) and high in self-reported performance (*M* = 6.29, SD = 0.36), i.e., "hH-hPE" (*n* = 411; 24,95%); (2) employees medium low in hedonic wellbeing (*M* = 4.97, SD = 0.49) and medium high in self-reported performance (*M* = 6.10, SD = 0.31), i.e., "mlH-mhPE" (*n* = 383; 23,25%); (3) employees medium high in

hedonic wellbeing (*M* = 5.45, SD = 0.46) and medium low in self-reported performance (*M* = 5.26, SD = 0.34), i.e., "mhH-mlPE" (*n* = 578; 35,09%); and (4) employees low in hedonic wellbeing (*M* = 3.82, SD = 0.71) and low in self-reported performance (*M* = 4.88, SD = 0.69), i.e., "lH-lPE" (*n* = 274; 16,67%).

**Figure 1.** Four cluster analyses of different combinations of well-being dimensions and performance from two sources. h stands for high level; mH stands for medium high level; ml stands for medium low level; l stands for low level. H-Pe stands for Hedonic-Performance (self-rated by the Employee); E-Pe stands for Eudaimonic-Performance (self-rated by the Employee); H-Ps stands for Hedonic-Performance (evaluated by the Supervisor); E-Ps stands for Eudaimonic-Performance (evaluated by the Supervisor); A and B inside the arrows denote different types of comparisons that can be made among the different operationalizations of well-being and performance within the "happy-productive" worker framework.

3.2.2. Model 2: Eudaimonic Wellbeing vs. Self-Rated Performance (E-PE).

In Model 2, we consider the following variables: eudaimonic wellbeing and self-rated performance. Although the auto-clustering algorithm indicated a two-cluster solution as the best model, we decided to opt for a four-cluster solution to maintain a similar cluster structure to Operationalization 1, and because the four-cluster solution also presented fair to good quality (BIC = 1067.114, BIC change from the previous cluster = −197.159, average silhouette measure of cohesion and separation = 0.5). The importance of both predictors was 1.00.

Four clusters emerged (see Figure 1): (1) employees high in eudaimonic wellbeing (*M* = 6.39, SD = 0.41) and high in self-reported performance (*M* = 6.27, SD = 0.34), i.e., "hE-hPE" (*n* = 596, 36%); (2) employees medium low in eudaimonic wellbeing (*M* = 5.10, SD = 0.40) and medium high in self-reported performance (*M* = 5.63, SD = 0.43), i.e., "mlE-mhPE" (*n* = 425, 26%); (3) employees medium high in eudaimonic wellbeing (*M* = 6.02, SD = 0.35) and medium low in self-reported performance (*M* = 5.28, SD = 0.36), i.e., "mhE-mlPE" (*n* = 474, 29%); and (4) employees low in eudaimonic wellbeing

(*M* = 4.60, SD = 0.61) and low in self-reported performance (*M* = 4.38, SD = 0.45), i.e., "lE-lPE" (*n* = 152, 9%).

3.2.3. Model 3: Hedonic Wellbeing vs. Performance Evaluated by the Supervisor (H-PS).

In Model 3, we consider two variables: hedonic wellbeing and performance assessed by the supervisor. The auto-clustering algorithm indicated a three-cluster solution as the best model because it minimized the BIC value (807.301, BIC change from the previous cluster = −172.428). The average silhouette measure of cohesion and separation was 0.5, indicating fair to good cluster quality. The importance of the predictors of hedonic wellbeing and performance evaluated by the supervisor is 1.00 and 0.91, respectively.

Three clusters emerged (see Figure 1): (1) employees high in hedonic wellbeing (*M* = 5.76, SD = 0.57) and high performance evaluated by the supervisor (*M* = 4.80, SD = 0.26), i.e., "hH-hPS" (*n* = 334, 37%); (2) employees high in hedonic wellbeing (*M* = 5.46, SD = 0.56) and low in performance evaluated by the supervisor (*M* = 3.86, SD = 0.36), i.e., "hH-lPS" (*n* = 402, 44%); and (3) employees low in hedonic wellbeing (*M* = 3.91, SD = 0.83) and low in performance evaluated by the supervisor (*M* = 3.67, SD = 0.86), i.e., "lH-lPS" (*n* = 179, 20%).

3.2.4. Model 4: Eudaimonic Wellbeing vs. Performance Evaluated by the Supervisor (E-PS).

In Model 4, we consider two variables: eudaimonic wellbeing and performance evaluated by the supervisor. Although the auto-clustering algorithm indicated a four–cluster solution as the best model, we decided to opt for a three-cluster solution to maintain a similar cluster structure to operationalization 3, and because the three-cluster solution also presented fair to good quality (BIC = 786.235, BIC change from the previous cluster = −242.320, average silhouette measure of cohesion and separation = 0.5). The importance of the predictors of eudaimonic wellbeing and performance evaluated by the supervisor was 1.00 and 0.81, respectively.

The three clusters identified are (see Figure 1): (1) employees high in eudaimonic wellbeing (*M* = 6.14, SD = 0.52) and high performance evaluated by the supervisor (*M* = 4.92, SD = 0.14), i.e., "hE-hPS" (*n* = 240, 26%); (2) employees high in eudaimonic wellbeing (*M* = 6.19, SD = 0.45) and low in performance evaluated by the supervisor (*M* = 3.75, SD = 0.56), i.e., "hE-lPS" (*n* = 416, 46%); and (3) employees low in eudaimonic wellbeing (*M* = 4.93, SD = 0.52) and low in performance evaluated by the supervisor (*M* = 4.14, SD = 0.55), i.e., "lE-lPS" (*n* = 259, 28%).
