Hugo Zagor?ek,Vlado Dimovski, Miha ?kerlavaj
significant positive direct effect on organizational learning (? = .21) (Kurland/Hertz-Lazarowitz 2006). The effect of transactional leadership was still positive but somewhat weaker (? = .15).
While clear theoretical arguments for the influence of transformational leadership on organizational learning exist, the role of transactional leadership is not so clear. Vera and Crossan (2004) propose a theoretical model where good leaders are those that know how to switch between a transformational and a transactional style of leadership in accordance with the situation (regarding the environment, strategy, prior firm performance, and stage of organizational life) in order to facilitate organizational learning. One might expect that transactional leaders promote the acquisition and exchange of information, but only to the extent that it clarifies role and task requirements or serves some other clear purpose. In a similar manner, information interpretation is encouraged, yet not to the same degree as with a transformational leader. Consequently, some cognitive and behavioural changes in organizational members do occur as a result of organizational learning. These arguments lead to the following hypotheses.
Figure 1. Hypothesized impacts of laissez-faire, transactional, and transformational leadership on organizational learning
Organizational learning process Information processing Organizational learning process Information processing H1a:+ Information acquisition H2a:+ Information acquisition Transformational leadership H1b:+ H1c:+ H1d:+ Information distribution Transactional leadership H2b:+ H2c:+ H2d:+ Information distribution Information interpretation Information interpretation Behavioral and Cognitive changes Behavioral and Cognitive changes
JEEMS 2/2009
149
Transactional and transformational leadership impacts on organizational learning
Hypothesis 1: Transformational leadership positively influences the information acquisition (H1a), information distribution (H1b), information interpretation (H1c), and cognitive and behavioural change (H1d) dimensions of organizational learning. Hypothesis 2: Transactional leadership positively influences the information acquisition (H2a), information distribution (H2b), information interpretation (H2c), and cognitive and behavioural change (H2d) dimensions of organizational learning. Hypothesis 3: Transformational leadership has the strongest impact on all four aspects of the organizational learning process as compared to transactional leadership.
The conceptual models in Figure 1 summarize the relationships between the three types of leadership and organizational learning.
Method
Research instrument
Using pre-tested constructs and measures allows for the validity and reliability of the data collected. For the constructs of the organizational learning process, the study used the OLIMP questionnaire (Dimovski 1994; ?kerlavaj et al. 2007; see Appendix A). The instrument has been refined and tested on several occasions (in the USA, Slovenia, Croatia, Malaysia) at various points in time (1994, 2002, 2003, 2004, 2005, 2006) always yielding adequate psychometric properties. The questionnaire uses five-point Likert scales and consists of four dimensions and nine sub-dimensions, totalling 36 items.
For the measurement of leadership styles, the study uses the Multifactor Leadership Questionnaire (MLQ), which is one of the most widely used and tested measures of transformational and transactional leadership. Antonakis et al. (2003), Avolio et al. (1995), and Lowe et al. (1996) document its sound psychometric properties. It contains 36 items representing the nine factors described above. The MLQ was translated into the Slovenian language using the translation ./ back-translation technique. Prior to the study 130 respondents filled in the Slovenian version of the MLQ questionnaire, obtaining good levels of reliability.
Research design and data collection
The study uses the organizational unit as the unit of observation and defines an organizational unit as a geographically or functionally distinct part of an organization which has its own leader. The respondents assessed the four constructs of organizational learning in their unit and as well rated their leaders (the leaders of their organizational unit) regarding the various leadership behaviours specified in the previous section.
150 JEEMS 2/2009
Hugo Zagor?ek,Vlado Dimovski, Miha ?kerlavaj
Where random sampling is problematic (as in management research), one way to enhance the generalizability of findings is to deliberately sample for heterogeneity (Mark/Cook 1984). By intentionally selecting subjects who come from diverse organizational settings, the researcher can determine whether a selected model accurately describes the actions of individuals across these divergent contexts. On the other hand, choosing a sample of firms located in a relatively homogenous geographic, cultural, legal, and political space reduces the impact of confounding variables (variables that cannot be controlled in the empirical research) (Triandis 1994). In line with these observations, the empirical context of the study is Slovenia in 2006. As such, the context of the present study differs from the predominantly North American research settings. This internationalization of the research context might contribute to the enhanced generalizability of the research findings and a better understanding of leadership styles, organizational learning processes as well as the impact of leadership on organizational learning.
Surveys were mailed to 1,914 alumni of the undergraduate programs of the Faculty of Economics of the University of Ljubljana, the premier business school in the country. In addition, electronic versions of the survey were sent to current and former students from the same school, which accounted for an additional 4,485 units.. Within the first three weeks of the mailing, 418 questionnaires had been completed and sent back, for a response rate of 22.3%. In addition, 374 usable electronic questionnaires were completed. After removing some questionnaires with a large number of missing answers, the final sample size equalled 753.
Using this research design, we were able to obtain data for organizational units of different sizes, from different functional backgrounds, and from different levels within the organizations. In addition, the units belonged to a wide range of companies in terms of size and industry. In this way the influence of confounding and background variables was randomized and cancelled out, increasing the validity of the empirical findings (Van de Vijver 2003).
Characteristics of the sample
The average age of the respondents was 34 years; 40% of them were female. Most (48%) have spent between 1 and 5 years in their present position, while 38% of them have worked in the same organizational unit for more than 5 years. The majority of the leaders rated by the respondents were male (67%). Most of them were between 30 and 50 years old (70%) and only 4% were younger than 30. 63% have spent more than 3 years as the leaders of their present organizational units, while only 11% have been in their present leadership position for less than one year. More than half of the leaders rated were in top management (52%), while the rest were in middle management (37%) and line
JEEMS 2/2009 151
Transactional and transformational leadership impacts on organizational learning
management (11%). The median number of subordinates that these leaders supervised was 17.
Statistical methods used and model specifications
The method framework for data analysis is structural equation modelling. The relationships between the constructs were estimated using LISREL 8.7 with the correlation matrix (Appendix B) and asymptotic covariance matrix as inputs. We also conducted tests of multivariate normality and found non-normal data (both in terms of skewness and kurtosis) which yielded a need to report Satorra- Bentler (SCALED) Chi square fit indices (Sattora/Bentler 1988). In large samples the ?2 test becomes highly problematic because even trivial differences between theoretical and empirical covariance matrices may result in a large value of this statistic (Joreskog 1993). Therefore, researchers typically provide several measures of model fit and use the Comparative Fit Index (CFI), which appears to be the most accurate in a wide variety of situations (Hu/Bentler 1995), as the primary criterion of model fit.
Given the fact that transformational leadership and transactional contingent reward style overlap (Avolio et al 1999; Vera/Crossan 2004), the study used two separate models. Each model tests the influence of a particular type of leadership (transformational and transactional) on the four constructs of organizational learning. Two or more indicators, consisting of several items, measure all latent constructs.
Results
Validity and reliability
We used a confirmatory factor analysis (CFA) to assess the reliability and validity of the constructs used in the study. The construct validity measures show how well the indicators represent the corresponding latent variables. Table 1 presents non-standardized and completely standardized factor loadings together with the corresponding t-values for each indicator and construct in the measurement model.
The results show that the factor loadings for all indicators are statistically significant and exceed the threshold of .50 for convergent validity (Hair et al. 1998). The only exceptions are active and passive management by exception (MBE(a) and MBE(p)), which are a part of transactional leadership. This finding indicates that transformational leadership is a problematic construct. In addition, Cronbach?s alpha for the transformational leadership as a whole only equals .57. A large divide between the contingent reward leadership dimension and the two management-by-exception dimensions is evident. Passive management by
152 JEEMS 2/2009
Hugo Zagor?ek,Vlado Dimovski, Miha ?kerlavaj
exception even has a negative correlation with the overall transactional leadership construct.
It seems that the management-by-exception dimensions relate more to laissez- faire leadership than to contingent reward leadership, which epitomizes transactional leadership. Several empirical studies have obtained similar results. In a meta-analysis by Lowe et al. (1996), an MBE scale was the only scale that shows evidence of low reliability (mean Cronbach?s ? = 65). A meta analysis by Dumdum et al. (2002) obtained similar results for the MBE(p) dimensions. In addition, the correlation between MBE(p) and leader effectiveness was negative (-.28) and similar to the correlation between laissez-faire leadership and effectiveness (-.29).
On the other hand, the correlation between MBE(a) and effectiveness was low (.08), while the correlation between contingent reward and effectiveness was significantly higher (.45). On the basis of these results, we removed the two management-by-exception dimensions (MBE(a) / MBE(p)) from further analysis. The contingent reward dimension serves as a proxy for transactional leadership in subsequent analyses. We have measured it with four MLQ items serving as indicators.
Table 2 shows the values of the Cronbach ?, composite reliability index (CRI), as well as the average variance extracted (AVE) for all latent variables in the final measurement. Diamantopoulos and Siguaw (2000) suggest that the threshold for CRI should be set at .60. Constructs exceeding that value have good composite reliability, which is the case with all latent variables. The cut- off value for AVE is .50 (Hair et al. 1998). For the Cronbach alpha it is .70 for studies in advanced phases (?1), and from .50 to .60 (?2) for studies in exploratory stages. (Nunnaly 1978; Van de Ven 1979). All of the constructs attain the recommended cut-off values using all three measures of construct reliability. The only exception is the information acquisition construct, which slightly fails the AVE internal consistency test but meets the required Cronbach alpha value for exploratory studies (?2) and, moreover, satisfies the CRI criteria, which is considered to be the most robust of all three reliability criteria (Diamantopoulos/Siguaw 2000).
JEEMS 2/2009 153