Building a New Testable Model to Estimate Total Factor Productivity.

Andrew Smithers
World Economics, Vol 18. No. 2 April-June 2017, 20th July 2017

It is reasonable to assume that growth depends on changes in the supply of capital, labour and technology. Total Factor Productivity (“TFP”) models, which are based on this assumption, aim to attribute the growth of the economy to changes in the volume of the labour and capital stocks and a residual known as TFP (or Multi Factor) productivity, which represents the contribution from technology. It seems generally agreed that attempts to develop such a model have so far proved to be unsatisfactory. One reason is the practical one that “Fairly innocuous differences in assumptions can lead to very different estimates of TFP growth”.* This comment raises the more substantial issue of whether these models have any scientific validity. If even one version of these models were testable and robust, it would be accepted as the valid one superseding all others, but no such model seems yet available. The challenge is therefore to produce a model which is robust when tested and so strongly data based that it is not sensitive to any assumptions made.

While it is possible to imagine ways in which the existing stock of capital can become more efficient, the scope for such developments seems limited. The efficiency of comparable plants varies from country to country, but so do attitudes, regulations and skills, which are difficult to change. In practice, most growth requires investment either to replicate existing equipment to meet increased demand or to install new equipment in which improved technology is embedded. We therefore need to distinguish between investment which responds to changes in technology and investment not driven by such changes. Investment is profit driven. Improved technology raises the output of a given amount of investment and, unless offset by other changes, it raises the profitability of new investment; equally there are non-technology variables (“NTV”), which can stimulate or deter investment in the absence of any changes in technology. Models need to distinguish between these two, i.e. to differentiate between TFP and NTV. Improvements in technology can change the quality of labour or of capital. I shall treat both as part of TFP. This is simpler and avoids the difficulty of measuring improvements in the quality of labour. It should nonetheless be noted that rising levels of education and other improvements in the quality of labour may be essential for allowing improvements in capital efficiency to be implemented.
*Measuring growth in total factor productivity by Swati R. Ghosh and Aart Kraay, published by the World Bank PREM notes No. 42 September, 2000.

Please click on the Download link below for the pdf of the complete paper

This paper and other papers are also available on World Economics website, once you have registered.
Building a New Testable Model to Estimate Total Factor Productivity