Aims & objectives
The broad aim of this project was to explain the heterogeneity of firm performance in industries. In the period under review there have been three substantive contributions.
One strand of work has contributed to the analysis of the relative importance of “firm” and “industry” effects in determining variations in corporate profitability. A number of analyses (primarily for the US) have used a technique of decomposing firm level profits into their firm and industry specific components to explain variations in corporate returns in terms of firm and industry level components.
Statistical inference has been somewhat ad-hoc in these studies, which drew their conclusions based on the sizes of estimated firm effects and industry effects within a random effects model framework. In particular, none of these studies has attempted the computationally difficult task of testing whether any of the variances are statistically significant, and whether one or more of the variance components are significantly larger than the others. Our work contributed by redressing this methodological shortcoming in inference. In this programme of work, we applied the method to data on a number of economies, specifically, the UK, India, and the US. The completed analysis for India, demonstrates an interesting relative shift of firm and industry effects as the economy moved from a regulated regime (pre-1985) to one of partial liberalisation (1985-1991) and finally, more comprehensive liberalisation (1991 onwards).
We find that surprisingly, firm effects dominated not only in the comprehensive liberalisation, but also in the regulated period. Managerial efforts at attending to procedural norms and playing by the rules of the political game competently can lead to adept firms doing well under the regime; and the development of such competencies ensure that firm effects were important in this period. In contrast, in the partially liberalised regime, firm managements came up against constraints in their pursuit of competitive strategies; targeted industries had differential benefits of liberalisation, and thus industry effects came to prevail.
Another strand of work has made a contribution to the assessment of statistical assessment of market structure. Here we determined the precise relationship between a commonly used structural measure of market structure and a standard dynamic model of firm growth. Starting from the well known model of firm growth (Gibrat’s law) we derived the asymptotic probability distribution for the concentration ratio. Empirical applications for the US shows that only in a few industries did small firms significantly outgrow the large; in most industries, large firms significantly outgrew the small.
The third strand of work addressed the economic processes that underlie the evolution of market structure. It has been noted that even in periods of great economic change, observed market structure (concentration) changes little. We show the precise way in which changes in market structure are underpinned by two dynamic processes: systematic patterns in the growth of small firms as against large, and increased market share volatility. We show that in periods of change, the degree of turbulence in market shares, and the relationship between growth and size, change quite dramatically, but offset each other leaving summary measures of market structure relatively unchanged. Thus a more structural approach analysing underlying processes of size related growth and market share volatility are important to understand changes in market structure. The empirical application was to India under liberalisation.
Kattuman, P. and Chirmiciu, A. (2003) ‘Significant feedbacks in firm growth and market structure’, ESRC Centre for Business Research Working Paper No. 270.
Kattuman, P and Kambhampati, U. (2003) ‘Growth response to competitive shocks: market structure dynamics under liberalisation’, ESRC Centre for Business Research Working Paper No. 263.
Growth and performance datasets created from company accounts of UK private and public limited companies for the period from 1990 to 2000; for five industries automobiles, pharamceuticals, steel, chemicals, office equipment covering UK private and public limited companies for the period from 1990 to 2000. Density (both univariate and bivariate) estimation software was coded in the past year.
Kernel density estimation program in GAUSS.
Stochastic Kernel Density estimation program in GAUSS.