Globalisation, ICT and the nitty gritty of plant level datasets

The net entry contribution to aggregate productivity growth has increased dramatically in the UK over 1990s according to calculations based on data from the Annual Respondents Database (ARD). Some recent studies have tried to link this to other structural changes over the same period such as increased globalisation and usage of ICT. I argue that the increase might equally have been caused by a systematic bias that is introduced to growth decompositions through random survey sampling of the underlying plant or firm panel datasets. This bias – despite being a general problem of growth decompositions does not seem to have been noticed in the literature yet. In the 1990s the Office for National Statistics (ONS) has successively increased the share of plants in the population of the ARD that are subject to random sampling. I show that this could cause the bias to spuriously increase the net entry contribution. My results show that correcting for the bias makes a substantial difference: the net entry contribution is about 10 percentage points lower on the corrected series in the 1990s. Surprisingly however, the positive correlation between ICT and net entry share – a main result of earlier studies – becomes more significant.