Understanding causal relationships among key economic variables is crucial for policy makers, who wish to e.g. stimulate private R&D growth. To this end, we applied a technique recently imported from the Machine Learning community (Structural Vector Autoregressions (SVARs) identified using Independent Components Analysis (ICA)) to a data-set of the world's largest R&D investors. Our analysis highlights the key role of firm growth in the areas of employment and sales, rather than growth of profits or market capitalization, in stimulating R&D growth. R&D growth appears toward the end of the causal ordering of the growth process. Our results suggest that policies to increase private R&D would do better to target growth of sales and employment rather than market capitalization or profits.