Relational Coordination
Guidelines for Theory, Measurement and Analysis
Chapter 4: Analyzing Relational Coordination (continued)
Analyzing the Predictors of Relational Coordination
The organizational practices that are expected to predict relational coordination are typically measured at the site-level of analysis. To assess their impact on relational coordination, you can use a model in which the unit of analysis is the individual respondent to the relational coordination survey. In this multi-level model, the organizational practice or practices are the independent variables, the control variables or covariates are the functional identity of the individual respondents (and any other individual-level predictors you want to include), and the dependent variables are the individual-level measure of relational coordination. This model allows the effects of organizational practices on relational coordination to be tested at the level of the individual participant, controlling for his or her functional identity.
As you did when evaluating the effect of relational coordination on performance, multi-level regression analysis should again be used to adjust coefficients and standard errors for the multi-level nature of the data (individual observations within multiple sites). The unit of analysis for this model is the individual participant within the site. The random effect is the site. As before, the analysis will produce both a within-site R square, and a between-site R square. Within-site R square indicates the percent of within-site variation that is explained by the variables in the models. Between-site R square indicates the percent of between-site variation that is explained by the variables in the model. Either or both can be reported, but should be labeled and explained to readers. e A. Summing the RC scores for all 20 respondents and dividing by 20, we get an un-weighted average RC score of 3.24 for Site A.
Analyzing Mediation
If you have been able to measure relational coordination, performance outcomes, and some of the organizational practices you expect might influence relational coordination, you may be interested in articulating and testing a mediation hypothesis. This hypothesis will take the form: "Organizational practice X is expected to affect performance measure Y through its effect on relational coordination." In other words, relational coordination is expected to mediate (at least partially) the effect of certain organizational practices on performance. Relational coordination is a multi-level theory that operates across multiple levels of analysis, and mediation can be tested across these multiple levels of analysis, consistent with previous studies of relational coordination. xxxvii
Following the method developed by Reuben Baron and David Kenny, evaluating the mediation hypothesis requires three equations and a test of the path's overall significance. First, the organizational practice must have a significant effect on relational coordination. Second, the organizational practice must have a significant effect on the performance outcomes of interest. Third, if in addition the coefficient on the organizational practice becomes insignificant when relational coordination is added to the outcomes equation, this result can be taken to suggest that relational coordination mediates between the organizational practice and outcomes, or in other words that the organizational practice influences outcomes through its effect on relational coordination. xxxviii
Finally, the overall path must be significant. The Sobel test can be used to determine whether the association between organizational practices and performance is reduced significantly when controlling for the mediator of relational coordination, drawing upon the critical values identified by MacKinnon and colleagues to determine whether the results are supportive of mediation. A recent paper reported the results of a Sobel test of the theory of relational coordination: "Results of the Sobel test suggest that the association between high performance work practices and quality of care is significantly mediated by relational coordination (z' = 1.87, p<0.01). Together, these results suggest that high performance work practices predict quality outcomes, and that they do so by strengthening relational coordination among employees in different functions (Hypothesis 2)." And later: "Results of the Sobel test suggest that the association between high performance work practices and length of stay is significantly mediated by relational coordination (z' = 2.40, p<0.01). Together, these results suggest that high performance work practices predict efficiency outcomes, and that they do so by strengthening relational coordination among employees in different functions (Hypothesis 3)." xxxix
For an example of the mediation model that was tested in "A Relational Model of How High Performance Work Systems Work, please see Exhibit 22 below.
EXHIBIT 22: Example of a Mediation Model
Analyzing Moderation
If you have been able to measure relational coordination, performance outcomes, and some of the factors that are expected to increase the impact of relational coordination on performance, you may be interested in articulating and testing a moderation hypothesis. This hypothesis will take the form: "Factor X (task interdependence, uncertainty or time constraints) is expected to increase (or decrease) the impact of relational coordination on performance measure Y." In other words, factor X is expected to moderate the effect of relational coordination on performance.
Again following the method developed by Reuben Baron and David Kenny, evaluating the moderation hypothesis requires testing two equations. First, relational coordination must have a significant effect on the performance outcome of interest, controlling for factor X. In a second equation, the product of relational coordination and factor X (RC*factor X) must have significant effect on performance, controlling for both relational coordination and factor X. This approach is consistent with the recommendation of organizational theorist Claudia Schoonhoven for operationalizing contingency hypotheses. xl
An example from "Coordinating Mechanisms in Care Provider Groups" can be used to illustrate the use of this method for testing the theory of relational coordination. First, a random effects regression equation showed that relational coordination was associated with increased quality of care (r = 0.23, p<0.01), and with reduced hospital lengths of stay (r=-0.31, p<0.01). In addition, the product of relational coordination and input uncertainty was associated with increased quality of care (r = 0.14, p<0.05) and reduced hospital lengths of stay (r = -0.20, p<0.01), suggesting that input uncertainty increased impact of relational coordination on performance outcomes of interest. xli
For an example of the moderation model that was tested in "Coordinating Mechanisms in Care Provider Groups: Relational Coordination as a Mediator and Input Uncertainty as a Moderator of Performance Effects," please see Exhibit 23.
EXHIBIT 23: Example of a Moderation Model (that also includes mediation)
