We apply a panel vector autoregression model to a firm-level longitudinal database to observe the co-evolution of sales growth, employment growth, profits growth and the growth of research and development (R&D) expenditure. Contrary to expectations, profit growth seems to have little detectable association with subsequent R&D investment. Instead, firms appear to increase their total R&D expenditure following growth in sales and employment. In a sense, firms behave 'as if' they aim for a roughly constant ratio of R&D to employment (or sales). We observe heterogeneous effects for growing or shrinking firms, however, suggesting that firms are less willing to reduce their R&D levels following a negative growth shock than they are willing to increase R&D after a positive shock. © 2010 Taylor & Francis.