Total factor productivity - a misleading concept

Banca Nazionale del Lavoro Quarterly Review, Sep 2001 by Reati, Angelo

Total factor productivity - a misleading concept*

1. Introduction

In real business cycle models, growth accounting, as in empirical research on the link between R&D spending and economic performance, the indicator usually chosen for productivity is total factor productivity - a concept that derives from a neoclassical production function (a Cobb-Douglas in the overwhelming majority of cases). In this paper I criticise this concept as a measure of technical change and economic performance on two grounds: i) theoretical; ii) relevance to an understanding of present technological change. Criticisms of the sort can be found here and there in the literature, but the problem is that in mainstream research they are simply ignored or receive bare mention, without drawing the conclusion that this notion of productivity must be abandoned - as should also be the case for its mother concept, namely the neoclassical aggregate production function. In any case, there is a striking contrast between the few notes of criticism and the thousands of studies estimating these production functions. Hence the need to go back to the subject and reinforce the argument by showing that the alternative concept of labour productivity is the most appropriate, particularly if taken at the macroeconomic level.

The paper is organised as follows: in Section 2 I recall the definition of the concept (this will be useful to outline the unrealistic hypotheses on which it is based - hypotheses that indeed form the foundations of the neoclassical production function); in Section 3 1 go on to point out the theoretical and practical weaknesses of the notion of total factor productivity; in Section 4 I discuss the concept of labour productivity, emphasising its superiority over the rival notion; Section 5 addresses the problem of measurement.

6. Conclusions

In this paper I have shown that total factor productivity (the `Solow residual is not the appropriate measure of technical change, particularly if one considers the present technological revolution, in which technical change is embodied in computers and other capital goods relating to information technologies.

a) At the theoretical level, the main weakness of the concept of total factor productivity lies in its derivation from a neoclassical production function - a device that makes nice mathematics but has no connection with reality. I am referring here to the mathematical assumptions of convexity and linear homogeneity - which allow for application of Euler's theorem and justify the hypothesis of constant returns to scale - and to the additional assumption of pure and perfect competition in both product and factors markets.

If, for the sake of realism, one drops these heroic assumptions, total factor productivity becomes a spurious magnitude that is more a measure of noise than a quantification of real-world phenomena. In fact, if we admit increasing returns to scale - something that seems to characterise the innovations in computer and information technologies - we then go on to estimate a positive Solow residual even though there is no (disembodied) technical change. Moreover, if we adopt the realistic framework of non-competitive market structures, observed total factor productivity is seriously biased because it includes an element related to market power. Empirically, this component is far from being negligible.

b) The second type of insufficiency derives from the fact that total factor productivity encompasses exclusively disembodied technical change, i.e. organisational innovations, learning by doing/using, spillover effects and such like. Now, the essential characteristic of present technological change is that it is first and foremost embodied in capital goods. And it is precisely because the labour force enjoys the use of new machines that enterprises benefit from an impressive increase in productivity of labour. Obviously, there is no doubt that disembodied technical change is important, but it is certainly not the main part of the story.

The alternative concept of productivity of labour escapes the above criticisms. In the first place, to define labour productivity we do not need a particular type of production function and, secondly, such a measure of productivity encompasses all kinds of technical change - embodied and disembodied. The relative influence of these two types of technical change can be measured separately by splitting the formula of productivity of labour into its components - the degree of mechanisation (the capital/labour ratio), and the `productivity of capital'. It appears that the productivity of capital is usually an indicator of disembodied technical change.

European Commission, Brussels (Belgium); e-mail: Angelo.Reati@cec.eu.int.

* I thank T. Antonucci, E. Bellino, G. Carone, J. Felipe, G. Hodgson, G. Nico&me, M. Pianta, W. Roeger and B. ter Weel for commenting on a previous version of the paper. I also thank two anonymous referees for their criticisms and suggestions, which afforded me great help. Finally, I am also grateful to E. Bellino, G. Mariutti, L. Pasinetti, N. Salvadori and A. Tylecote for very useful discussions, thanks to which I have been able to improve and clarify some points. The usual disclaimers apply.