Limitations and future research directions
The findings should be considered in light of some limitations of the survey method
used for them. First, our study used student samples, which may call into question the
external validity of the results. However, the vast majority of the respondents were
graduate students who were enrolled in part-time academic programs while working
full-time in various business, government, and not for profit organizations. The second
limitation is the possibility that same-source biases might have inflated the research
findings. To address this potential problem, we collected data in two time periods and
followed recommendations by Podsakoff et al. (2003) for avoiding or detecting common
method bias. Finally, Doty and Glick (1998) and Spector (2006) have concluded that
problems caused by common method variance are somewhat overstated and seldom
serious enough to invalidate research findings based on the well-designed measures.
The third limitation pertains to the use of cross-sectional data. The use of
cross-sectional data limits the extent to which causality can be inferred. While our
research results suggest that ethical and empowering leadership relate positively with
LMX, which subsequently relates to the two outcomes, the direction of causality
between these variables cannot be determined in a survey study such as ours.
Based in part on these limitations, the proposed model should be tested in future
research using stronger methods. Survey studies in an organizational setting should
collect data from multiple sources in order to avoid common method bias. To assess
causal relationships a good research method is a longitudinal experiment using real
groups or students in a multi-week group simulation. The two types of leadership
behavior could be manipulated independently in a crossed design, and LMX and
commitment could be measured at an intermediate time as well as in the final time
period, along with objective measures of group performance. Finally, despite the focus
on LMX in this research, subsequent studies might examine mediators other than LMX
in the model. For example, the relationships between leadership behavior and
indicators of subordinate commitment and leader effectiveness may be mediated by
other perceptual variables such as psychological empowerment, personal identification
with the leader, acceptance of decisions, quality of decisions, and internalization of
ethical values.