Finding the right fuel for the analytical engine: Expanding the leader trait paradigm through machine learning?

Using self-report personality data and 360-degree performance evaluations of 973 managers across various contexts, we investigated the leader trait paradigm using a range of machine learning methods. We found that a relatively simple linear ordinary least squares model incorporating direct effects of traits and context performed equally as well as our best performing complex machine learning alternatives (e.g., lasso and random forests) at predicting leader effectiveness under low-dimension conditions (i.e., a small number of predictors). We then increased dimensionality and found that newer machine learning methods excelled. Overall, our computationally intensive approach supports the argument that (a) direct effects (not interactions) of traits and context are important predictors of leader effectiveness and (b) appropriately matching combinations of methods, models, and data (from simple and conventional to complex and novel) creates a powerful machine learning engine for investigating leadership. We end with opportunities for future research, discuss practical implications, and provide a list of resources for those interested in learning more about this analytical future.