Testing the Pooling Assumption in an Industry Panel of R&D Investment: What Do We Learn?

In this paper we test the pooling assumption for the determinants of industry R&D investment. Most studies consider pooled estimates, but if the parameters differ across industries, pooled coefficients will not provide reliable estimates of individual industry effects. Moreover, pooled coefficients may not even be consistent estimates of the average. For a panel of UK manufacturing industries we find that pooling is only valid for output fluctuations, lagged R&D and real interest rates. Implementing the test results into our model, we find government funding is only significant for low-tech R&D. Skilled labour and foreign R&D matter only in high-tech industries.