Method

An online survey was conducted as commonly used in meeting research (for advantages and disadvantages, see Wagner-Schelewsky & Hering, 2019). Previous studies on meeting lateness suggest that results from experimental settings (e.g., Allen et al., 2018) and online surveys (e.g., Allen et al., 2021) are equally valid. The survey was conducted using the tool soSci-Survey (Leiner, 2019) and followed a 2x2 between-subjects design (meeting: on time vs. late; country: Germany vs. Italy). Participants could choose between a German, English, or Italian language version of the survey. Participation in the survey was possible over a two months period from March to May 2022. Data collection and data analysis procedures in this study were approved by the Ethics Committee [details omitted for double-anonymized peer review] (proposal 297). The study was preregistered [link omitted for double-anonymized peer review]. All materials, raw data files, as well as the used analysis script can be found in our OpenScienceFramework project: https://s.gwdg.de/FEIU7k [anonymized link for peer-review].

Sample

According to an a priori power analysis calculated using G*Power (Faul et al., 2009) for a two-factor analysis of variance, at least a total ofN = 128 participants were required (f = .25, 1-ß = .80, α = .05). We expected a moderate effect size and approximately equal group sizes in each design cluster (cf. Allen et al., 2021).
Participants were recruited online via posts published on web portals. After opening the survey link, participants selected their preferred language (German, English, Italian). Participation was not compensated. Completion of the survey was estimated to take five minutes, though participants took about 2.5 minutes on average (M = 165.48 seconds; SD = 63.82).
The link to the questionnaire was accessed a total of 709 times during the survey period of which 482 surveys were started. Overall, 303 participants fully completed the survey (37.14 % drop out), of which 279 participants (77.42 % women, 20.07 % men, 1.43 % diverse, 1.08 % no gender; age: M = 25.1 years, SD = 4.82, range: 19-57 years) were included in the analysis. Participants were excluded if no written consent was given (n = 2) or if participants did not study in either Italy or Germany (n = 22). All included participants gave written consent to participate in the survey. Of the 279 participants, 141 had studied in Germany and 138 in Italy during the last six months; 141 participants (50.54 %) chose the German version of the survey, 129 participants (46.24 %) chose Italian, and 9 participants chose the English version. Participants studied law, economics, and social sciences (29.03 %), humanities (25.81 %), mathematics and natural sciences (11.11 %), human medicine and health sciences (9.68 %), engineering (2.51 %), arts and art science (1.79 %), agriculture, forestry, nutrition and veterinary medicine (1.43 %), sports (0.72 %), and other disciplines (17.92 %).

Operationalization and measurement

To reflect the diversity of meeting formats, meetings were broadly defined as the ”most recent online meeting in an educational context that lasted longer than 15 minutes and involved at least three people.” A similar description can be found in prior research (e. g., Crowe et al., 2019; Allen et al., 2021).
Participants’ meeting satisfaction was assessed using Briggs et al.’s (2003) twelve-item comprehensive scale. For the analyses, the four-item subscales of SP, SO and PGA were treated as individual constructs. This validated scale has already been used across countries (cf. Briggs et al., 2006; Reinig et al., 2009). Items were answered using a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree).
Meeting lateness was operationalized as a dichotomous item (”Did the meeting start late?”). This is also found in other studies (e.g., Rogelberg et al., 2014), where the duration of the delay was to be entered as free text rather than given as a cut-off (e.g., five- or ten-minute delay in Allen et al., 2018). Expanding on Allen et al.’s (2021) operationalization, participants were asked to name reasons for the delay as an exploratory variable that was not presumed to be person-specific. The subjective perception of lateness was exploratively addressed (”A lateness of … minutes in an online meeting in an educational context represents unpunctuality for me.”). Similar approaches can be found in Rogelberg et al. (2014) and van Eerde and Azar (2020).
Cultural context was operationalized in line with Allen et al. (2021) study via the following item: ”In which country did you mainly study in the last six months?”. As control variables , the number of participants, meeting duration, and number of meetings per week were recorded as control variables. Demographic variables (i.e., age, gender, study major) were also collected.
The questionnaire was initially prepared in English and translated to German and Italian by the first author. Translations were validated using a simple back-translation (Wendt-Hildebrand et al., 1983) by independent translators who had the linguistic qualifications and a required cultural understanding. A pilot test was carried out to test the functionality and comprehensibility of the questionnaire.

Analysis

Statistical analysis was performed using the open-source software R (R Core Team, 2017). The detected statistical outliers (n = 11 cases considering an interquartile range of 1.5) were not excluded for the main analysis, as indicated in the preregistration (even when excluded, the results do not change notably).
Assumptions of the two-factor ANOVA (Hays, 1978) were checked. A left skewed distribution is present for all constructs of meeting satisfaction. The data is not normally distributed according to the Shapiro-Wilk test (for SP, SO and PGA: p < .001). The variances are also not equal according to Levene’s test (exemplary for PGA: interaction F (3, 275) = 4.622, p = .004; PE landF (1, 277) = .005, p = .942; PE delay F (1, 277) = 13.089, p = .0004). Further challenges arise from the unbalanced design that could not be influenced due to the correlative nature of our study. Consequently, the planned analysis was discarded and a non-parametric procedure was applied with the correction of degrees of freedom according to Box (1953), in which no such distributional assumptions apply (Seistock et al., 2021). An overview of the procedure and the functions is given by Lüpsen (2021a; 2021b).