Causation provides the foundation for action. If we know what event results in another, we can increase the chance of realizing the outcome that we desire. Put simply, causation demonstrates that one event is the result of another event or that a cause and effect relationship is present. When we use analytics for this purpose, the value of the data and its utilization become considerably more valuable. Through analytics, we can test:
• if event a causes event b
• when event a causes event b
• how much event a causes event b
In order to demonstrate causality, we have to meet certain requirements: empirical association, temporal ordering, and non-spuriousness. All three are required to substantiate a causal relationship. In addition, two other factors strength the validity of the conclusion of causality: identifying the causal mechanism and context for occurrence. As discussed in our last post, association occurs when two variables change together. Temporal ordering occurs when the independent variable (cause or factors that influence what we want to explain) changes before the dependent variable (effect of what we are trying to explain). In essence, the cause has to be present to impact the effect. A relationship is spurious when third or confounding variable actually changes the dependent variable. Think back to the shark and ice cream example from our last post. Spuriousness poses a real threat to our analysis since it leads us to identifying the wrong cause. The causal mechanism and context require us to describe the story while identifying other factors.
The most powerful tool for testing causality relates to the experimental design. Most of us probably remember doing experiments in school. The experimental design requires us to come up with a hypothesis or a testable statement of how two variables relate and use the following approach:
• Two comparison groups (experimental group and a control group)
• Variation in the independent variable before assessment of change in the dependent variable
• Random assignment to the two comparison groups
We determine if an association exists between the independent and dependent variables in an experiment when we alter the independent variable.Although an experiment might be ideal, it is rare we have the luxury of human experiments. There are a variety of statistical and mathematical tools that we can use to meet some, if not most of the criteria for causality.
Causal questions come in two primary varieties: effects of causes and causes of effects. For example, if you examine the effects of causes, you might ask if taking aspirin will help your headache. Similarly, if you are concerned with the cause of an effect, you might wonder if aspirin helped your headache when it is gone. Most research utilizes the effects of cause approach or poses a question to address a specific issue or concern.
Causality provides the basis for understanding, modeling, optimizing, and diagnosing the events important to our success. It grants us the ability to:
• Determine the factors that have the greatest impact on performance;
• Predict what will happen when certain conditions are met;
• Analyze why certain outcomes occur and others do not;
• Isolate that factors beyond our influence
• Alter future outcomes
Consequently, within human resources, the ability to determine causation provides the basis for maximizing our resources in our major processes: recruitment, development, performance, and retention. By analyzing how one event impacts another, we are able to isolate the best candidate, environment, and process.