Last week we analyzed the survey question validity and construction provided by an external survey provider, now we are ready to test if the team engagement level is influenced by the work location and the group function.
The logic behind the location is because we tend to think people at the regional offices are less engaged than those based in “dynamic” metropolitan cities like London. The reason to include group function for testing is because we already learned that it has an impact on the workplace diversity in the organization. We want to test if it has influence on the employee engagement level as well.
The dependent variable in both tests is EMPsurvEngagement (the average team score for the nine engagement questions mentioned in this post). The independent categorical variables are “LondonorNot” (London or non-London locations) and “Function” (sales or professional services). We applied t-test on both scenarios, and the results are as follows:
Location vs Engagement:
Welch Two Sample t-test data: ldn.eng and noLdn.eng t = 1.3201, df = 856.94, p-value = 0.1872 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -0.4260878 2.1765949 sample estimates: mean of x mean of y 81.54784 80.67259
Group function vs Engagement:
Welch Two Sample t-test data: sales.eng and prof.eng t = -6.9432, df = 923.21, p-value = 7.22e-12 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -5.680489 -3.176910 sample estimates: mean of x mean of y 79.14542 83.57412
In terms of the geographical location, the t-test was not significant: t( 856.94) = 1.32, p > 0.1. There is no significant difference in engagement between teams based in London and those that are based in non-London locations. Therefore, we shouldn’t take it seriously as the engagement levels could easily occurred by random chance of fluctuations.
However, it’s a different story for group function. The t-test was significant: t( 923.21) = -6.943, p < 0.001 because the chance of finding such a pattern of team engagement differences by chance is less than 1 in 1,000 (p < 0.001). Furthermore, the team engagement means for the sales group is 79.15, and that of the professional services is 83.57. We can conclude that people who work in professional services (non-customer facing) are significantly more engaged than those in the sales team (customer facing). Giving that it is the sales team who deal with directly with customers, this highlights a key challenge to be addressed in any engagement strategy.
Besides the survey data provided by a third-party vendor, we can combine with our in-house demographic data to predict team-level engagement, and we will use multiple linear regression to do it. The dependent variable is EMPsurvEngagement, and the independent variables are below:
- Function: Sales vs professional services (HR data)
- GroupSize: The number of employees in the team (HR data)
- PercentMale: Percentage of males in the team (HR data)
- BAME: Percentage of BAME in the team (HR data)
- EmpSurvSupervisor: Supervisor aggregate score (survey)
- EmpSurvOrgIntegrity: Company integrity aggregate score (survey)
After applying the the linear regression model, we received the following result:
Call: lm(formula = EMPsurvEngagement ~ Function + GroupSize + PercentMale + BAME + EmpSurvOrgIntegrity + EmpSurvSupervisor, data = empSurvey) Residuals: Min 1Q Median 3Q Max -38.527 -3.698 1.109 5.128 25.100 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 52.610475 3.301428 15.936 < 2e-16 *** Function -1.272071 0.747138 -1.703 0.0891 . GroupSize -0.032016 0.018086 -1.770 0.0771 . PercentMale -0.139287 0.016974 -8.206 1.02e-15 *** BAME 2.043652 2.653315 0.770 0.4414 EmpSurvOrgIntegrity -0.002245 0.005425 -0.414 0.6791 EmpSurvSupervisor 0.485203 0.029835 16.263 < 2e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 8.008 on 739 degrees of freedom (181 observations deleted due to missingness) Multiple R-squared: 0.3773, Adjusted R-squared: 0.3722 F-statistic: 74.61 on 6 and 739 DF, p-value: < 2.2e-16
The regression model tests the significance of how well the model predicts variations in the team-engagement index. The result shows the statistical significance as p< 0.001. It means that there is less than 1 in 1000 chance that we would find this pattern of shared variance randomly by chance alone, so this model is significant.
Next we will look at the significance of each independent variable. It helps us understand the degree in which they uniquely contribute to the prediction of variation in team engagement (and in which direction they seem to be related to the variation in engagement).
The coefficients table tells us that only two seem to come out as the significant predictors: supervisor aggregate score and the percentage of the male in the team. Why does function no longer come out as a significant factor? A potential interpretation may be that it loses its importance when we take into account the other factors.
Team perception of how supportive their supervisors are is the most important driver of team engagement. Better supervisor rating results in higher level of team engagement. Gender diversity is also found to be the key potential driver of team engagement. The higher male percentage in a team, the less engaged the team tends to be.