Catching Gazelles with a Lasso: Big data techniques for the prediction of high-growth firms

We investigate whether our limited ability to predict high-growth firms (HGF) is because previous research has used a restricted set of explanatory variables, and in particular because there is a need for explanatory variables with high variation within firms over time. To this end, we apply “big data” techniques (i.e., LASSO; Least Absolute Shrinkage and Selection Operator) to predict HGFs in comprehensive datasets on Croatian and Slovenian firms. Firms with low inventories, higher previous employment growth, and higher short-term liabilities are more likely to become HGFs. Pseudo-R2 statistics of around 10% indicate that HGF prediction remains a challenging exercise.