That elusive elasticity and the ubiquitous bias: Is panel data a panacea?

There is often assumed to be a unit elasticity of substitution between capital and labour. But estimates based on neoclassical capital demand equations frequently find a smaller value. Recent time-series work for the United States and Canada has suggested that, once the biases inherent in estimating cointegrating vectors are properly accounted for, the elasticity could indeed be close to 1. We investigate this possibility for the United Kingdom. First we use aggregate data and find that the estimated elasticity is in the neighbourhood of 0.4. We then exploit a unique industry-level dataset for the United Kingdom to try and further pinpoint our estimates. Estimates using dynamic panel data methods are close to our benchmark estimate using aggregate data, providing a robust statistical rejection of a unit elasticity in UK data.