By Dr. Gary L. Deel, Ph.D., J.D.
Faculty Member, Wallace E. Boston School of Business
I’ve written in previous articles about the importance of competent and informed workforce management. I’ve also written about the need to collect and analyze good data in these processes in order to understand the dynamics of an organization or an industry.
However, data analysis is not necessarily intuitive or easy. Correlation is not causation, and it is very easy to draw misguided conclusions when you’re looking at many different variables and trying to assess which inputs are responsible for which outputs and how much.
Workforce Analysis Is Highly Complex
In the world of workforce management, analysis is extremely complex. Consider, for example, the list of 51 different human resource metrics offered by the Academy to Innovate HR. Each of the metrics on this list is important to organizational operations in different ways and in different proportions, depending upon the specific company being studied. It’s easy to imagine how different variables might affect one another with different dynamics.
For instance, we might expect to see that variables around employee training might influence the variables around employee productivity. We could also expect a relationship between employee training and employee satisfaction.
And in completing this one triangular dynamic, we would probably expect to observe some type of relationship between employee satisfaction and employee productivity, too. Consequently, it’s quite clear that many of these metrics are probably interrelated, but the extent and scope to which they affect one another is not at all obvious.
Furthermore, our intuition does nothing to account for the potential unexpected relationships between variables. For example, is it at all obvious that the age of an employee should have an effect on satisfaction or organizational revenue?
Not necessarily, but these kinds of correlations are nonetheless entirely possible. Such correlations only become apparent with the application of sophisticated statistics tools, wielded by well-trained mathematicians and analysts.
Using Software for Statistical Analysis
In business disciplines, academics and scholars very commonly use a software application called Statistical Package for the Social Sciences (SPSS), which is developed and sold by IBM. SPSS allows for a lot of intense, extremely nuanced statistical analysis, including correlations and significance testing, linear and multiple regressions, chi-square assessments, and even more complex structural equation modeling (SEM).
The details of these different statistical tools are far too complex to demystify in one article; they require advanced graduate-level coursework in statistical analysis to understand them and leverage their insights. However, it suffices for our purposes to say that these tools enable users to determine the relationships between variables that would otherwise be invisible to people only observing the raw data.
Regression Modeling
A very common tool used for assessing statistical correlation and causation is regression modeling. In short, regressions allow for a comparison of many different variables to determine the strength of their associations – individually and in different collections – so as to examine which variables might influence others and to what extent.
Consider a very simple example such as individual political persuasions. We could assess several different variables that could potentially be correlated with and/or causative of different political allegiances.
These variables might include family and upbringing, education level, geographic location, the influence (or not) of religion, military affiliation (if any), gender, race, and many other variables. But the key questions are: Which of these variables are the best predictors of political ideology and how much?
This type of analysis is where linear and multiple regression become useful. They allow us to input data for many different independent variables and one or more dependent variables, and then run a statistical analysis that will reveal to us the strength of relationships between the different independent variables and different dependent variables. Then, these results are at least, in principle, suggestive of some level of potential causation, although it is important to note that regression modeling is not conclusive evidence of causation.
So how might we apply this kind of analytical approach to the 51 metrics offered by the AIHR? Well, suppose we are interested in understanding the dynamics of a variable like employee satisfaction. Using satisfaction as a dependent variable, we could look at the other 50 factors in the list and hypothesize about which might be influential upon satisfaction.
For the sake of this example, we might posit that training, average hours, overtime expectations and promotion rate could all be variables that have some effect on employee satisfaction. Assuming we have collected statistically valid, reliable data on these variables, we could input the data to SPSS or some other tool and run a regression analysis to see how these independent variables affect the dependent variable, if at all. The subsequent results can help us to guide changes in business policy, management, and practice to improve upon the conditions that are the strongest predictors of employee satisfaction.
Advanced Statistical Analysis Requires Proper Study and Training to Provide Insights
To be clear, you do not necessarily need to hold a Ph.D. or a tenured professorship to run advanced statistical analyses. However, there should be no illusions that this kind of work can be done without serious study and training. It takes a lot of work to understand the mechanics of advanced statistics to do them competently and avoid wasting time through error.
The need for this intelligent, investigative approach in advanced workforce management – especially for the largest companies with the most to lose from incompetent decision making – is one of the reasons why many large businesses are investing in dedicated workforce analytics programs to steer these efforts. There is certainly an up-front cost involved in retaining the specialized talent that is needed to manage business processes, but the potential gains in terms of clear awareness of business dynamics and aligned strategies for improvement often more than offset the expenses.
Because workforce analysis is complex, advanced statistical skills and knowledge are vital to leverage the very best value from collected data. That way, it becomes less complicated to steer businesses in the right directions.
Comments are closed.