*2.1. Data Collection and Survey*

A stratified random sample composed of 300 surveyed students, 100 from each university, was collected during the period 2017–2018. The initial data started from a database for each university, made up of 200–1000 data each one. Incomplete surveys and those that showed logical inconsistencies were deleted. Finally, a group of 100 surveys from each university was randomly selected with the random function of the spreadsheet software, making up the complete database with 300 surveys. The survey included 33 items: 4 socioeconomic (age, gender, character public/private, size), 6 of students' perceived satisfaction and 23 related to RC. The survey's reliability was verified by means of Cronbach's alpha, with values greater than 0.7, acceptable to confirm internal consistency: communication dimension (0.703), relationship dimension (0.831) and satisfaction (0.793) (Table 2). The complete survey showed a Cronbach's alpha of 0.87 [20,41].

The 23 items of the RC model focused on the mechanisms involved in organizational practices are shown in Table 2. 11 variables of the communication dimension, 12 of relationship dimension and 6 related to the level of student satisfaction were used. The students answered each question of the survey (Table S1) as many times as profiles were observed at the university. Then, each relational coordination variable was disaggregated into the following profiles: lecturers, administrative officers, classmates, student representatives and me (myself), as a control variable. A Likert scale metric was used, from 1 (infrequent) to 5 (very frequent). In this case, the intervals between the points on the scale corresponded to empirical observations in the metric sense [42]. A visual analog scale was displayed on each survey question presented to the students.


**Table 2.** Relational coordination and satisfaction variables.

The proposed indicator of satisfaction was based on the student's satisfaction level [7,16,43]. This indicator was obtained from variables 24–29, related to profiles of conferences, student representatives, administrative officers, materials, communication channels, training con-

tent. The descriptive statistics of trend, dispersion and position of the satisfaction variable were calculated (Figure S1). In each university the median ranges between 18–20 points and for the total sample of 19 points. Therefore, two levels were determined: 19 points was used as border: less than 19, "Low satisfaction" and more than 19, "High satisfaction" [19,20]. Later, satisfaction level was understood as fixed or independent variable.

#### *2.2. Statistical Analysis*

The normality of the data distribution was evaluated using the Kolmogorov-Smirnov test (with the Lilliefords correction) and a Levene test was used to evaluate the homogeneity of variance. For those variables that did not show a normal distribution, the Bartlett test was applied to assess if the data had equal variances.

In the first stage, to answer RQ1 and RQ2, the RC variables influenced by the university (socioeconomic context) and the level of satisfaction were identified. 23 variables of RC were compared using the general linear model (GLM). The three universities (ARCADA, ESPAM and UCO) and two satisfaction levels (Low and High) were used as fixed factors. The interactions between both factors were also considered [43]. Three levels: \* *p*-value < 0.05; \*\* *p*-value < 0.01 and \*\*\* *p*-value < 0.001 were considered. The test allowed determining which pairs of means differed significantly and to study data whose error did not conform to the normal distribution and non-constant variances. The test allowed determining which pairs of means differed significantly and to study data whose error did not conform to the normal distribution and non-constant variances [43].

Secondly, an organizational model was built using a canonical discriminant analysis to answer to RQ3. This analysis allows studying the concrete relationships that exist among discriminated groups (universities) and their degree of association [44]. The coefficients of the discriminant model show the relative contribution of the variables to the model. The higher the value of the F-remove coefficient, the greater the contribution to group discrimination [45,46]. Therefore, variables with a *p*-value < 0.05 were accepted and a model with highest percentage of correctly classified cases was selected. The most discriminant variables were calculated applying the F of Snedecor, Wilks' Lambda and the 1-Tolerance. High values of F for each variable indicated that the means of each group were widely separated. Small Lambda values showed that the variable discriminates well among groups. Variables with a high percentage of tolerance (1-Toler) were selected [26]. Statistical analyses were performed using the STATISTICA 12.

#### **3. Results**

The three universities showed an average age of students less than 25 years in 86% of the sample (*p*-value < 0.05). Regarding gender, the distribution was uniform in ARCADA. However, in UCO most of the students were women (*p*-value < 0.001) and in ESPAM most of the students were men (*p*-value < 0.05). Regarding the field of knowledge, significant differences were found among the three universities. In ARCADA 100% of the data corresponded to the Social Sciences (*p*-value < 0.001), in UCO the Health Sciences predominated (90%; *p*-value < 0.001) and in ESPAM the Engineering Areas obtained the highest percentage (72%; *p*-value < 0.05). The sociodemographic indicators of the sample are shown in Table 3.


**Table 3.** Sociodemographic distribution of the sample (%).

\* *p*-value < 0.05; \*\* *p*-value < 0.01; \*\*\* *p*-value < 0.001; ns = not significantly different.

The three universities reached relational coordination values close to the average (69.87 ± 0.78; CV = 0.19). Regarding satisfaction, UCO obtained the lowest level (18.25 ± 0.44; CV = 0.24) and ARCADA (19.44 ± 0.53; CV = 0.27), the highest one. The dispersion coefficient was low in the three universities (data not presented).

#### *3.1. Identification of Organizational Differences*

GLM results are shown in Table 4. Significant differences were found in most of the variables of RC by university and satisfaction (*p*-value < 0.05). 82.61% of the RC variables showed significant differences by university. The highest RC values were observed at ESPAM and UCO, while ARCADA showed lower values. ARCADA showed significant differences in the variables related to solving problem communication and shared knowledge with the administrative officers. In UCO significant differences were found, highlighting the frequent communication, shared knowledge and mutual respect related to classmates. Lastly, ESPAM showed significant differences in accurate and frequent communication variables, and shared knowledge and goals were the variables that stood out, relating to lectures.

**Table 4.** Relational coordination by university and satisfaction level (Mean ± SE).


\* *p*-value < 0.05; \*\* *p*-value < 0.01; \*\*\* *p*-value < 0.001; ns = not significantly different. a, <sup>b</sup> Within row, averages with different superscript differ significantly.

> Significant differences by level of student satisfaction were found in 65.21% of the organizational variables. 26.09% of the variables showed differences according to both criteria (Table 4). The non-significant variables were those related to the classmates and representatives of the students in the two dimensions of RC. The interactions between university and satisfaction were found in six RC variables. Most of the variables were

related to the profile of administrative officers. The interactions found between both factors in these six significant variables are shown in Figure 1.

**Figure 1.** Interactions between university and level of satisfaction.
