Behavioral brokerage

School staff members’ bridge-building behavior was assessed using a short form of Obstfeld’s “Brokerage tertius iungens orientation” measurement instrument (Obstfeld, 2005). In the original version, the brokerage orientation was assessed based on six items with a 7-point Likert scale (Cronbach’s alpha = 0.88). Due to the principle of parsimony, we assessed only four of these items, covering different aspects, such as whether teachers see opportunities for collaboration between people or whether they introduce school actors to each other who might have a common strategic work interest (see Table 1). The test instrument was assessed using a 6-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree). Results revealed a high reliability of the test instrument (N = 1,114; M = 3.77, SD = 0.91; Cronbach’s α = 0.85). Further, a confirmatory factor analysis was conducted to test empirically our theoretical measurement instrument adapted from Obstfeld (2005). The measurement instrument based on four items revealed acceptable model fit indices: χ² (2) = 31.274, p < 0.001, scaling correction factor Yuan-Bentler correction (Mplus variant) = 1.456: CFI = 0.98, TLI = 0.93; RMSEA [90% CI] = 0.138 [0.098 –0.183], SRMR = 0.026. We are aware that the values for the root mean square error approximation (RMSEA) clearly deviate from the cut-off value close to.06 suggested by Hu and Bentler (1999). However, as Kenny et al. (2015) indicated, the RMSEA in properly specified models with a small number of degrees of freedom (in our case df = 2) often falsely indicates a poor fitting model. We therefore evaluated the goodness of fit for our latent construct based on all the other alternative fit indices not or less dependent on the degrees of freedom (such as CFI, TLI, and SRMR). As all of these values revealed an acceptable fit, we went on to address our research question based on our measurement instrument. Moreover, intraclass correlation coefficients (ICC₁ = –0.010; ICC₂ = –0.283) indicated that school actors did not resemble each other based on the school they belonged to Lüdtke and Trautwein (2007). Hence, behavioral brokerage, as we have theoretically outlined above, is examined most suitably on an individual level.

Measurement instrument to assess school actors’ behavioral brokerage.

M = mean and SD = standard deviation. r^it indicates item-total correlation coefficients. α-drop indicates Cronbach’s alpha of latent construct if item is dropped. α indicates Cronbach’s alpha of the latent construct.