Antonakis and Atwater (2002) define perceived leader-follower interaction as the “degree to which leaders interact with their followers” (p. 686). Leader-follower interaction frequency does not imply a well-established quality of relationship among both, it rather relates to followers seeking guidance and feedback. Despite taking different media channels into account when assessing correlations, the use of an overall interaction frequency index, summarizing all channels, potentially introduces biases. Media richness has not been evaluated in depth and therefore assigning different weights to face-to-face interaction and chat could have been more accurate.
Problems with generalizability arise with the data collection taking place in only one industry. Cases were gathered in business units of international corporations in the technology industry with more than 10,000 employees. The study followed a cross-sectional design as, in this fairly new field of interest, the design is rather exploratory. As a result of cross-sectional research, ambiguity of causal direction might be an issue (Cole et al., 2009). Yet, fundamental theory suggests building on given directions, following empirical publications. Due to this research design, participants were asked for perceptions of leadership behavior and organizational outcomes in the same survey. This procedure raises issues of common-method variance which can yield in inflation of observations (Cole et al., 2009, p. 1723; Davis & Bryant, 2010, p. 523, Podsakoff & Organ, 1986; Podsakoff et al., 2003). Method biases may cause measurement errors in different ways. Podsakoff et al. (2003, p. 881) report various sources for the existence of method biases. Relevant for the present study, same source or rater bias might apply as respondents answering to different variables would likely be consistent in their answers. Referring to previous distance leadership research, Howell and Hall-Merenda (1999) still followed this procedure. The design of the research would have been appropriate for structural equation modeling (SEM), yet the model revealed to be too complex for the number of observations. For this reason, only the confirmatory factor analysis was pursued using SEM. Interpretation of mediation analysis is further subject to bias as there is, to date, no accepted form of interpretation (Hayes, 2009, p. 417).