Johnes, G and Johnes, J (2009) Strategic Responses to Companies'' Own Past Performance: Why do some Firms Fare Better Than Others? Working Paper. The Department of Economics, Lancaster University.
Recent work on business strategy considers the evaluation of company performance using frontier methods (Devinney et al., forthcoming). The present paper builds on that work to examine the extent to which company performance in one period impacts on business practices and hence performance in subsequent periods. We investigate this using a panel of annual data on some 4280 firms over the period 1983-2003, drawn from the Osiris data set of Bureau van Dijk. A data envelopment analysis is conducted to evaluate the efficiency of firms in converting inputs - in the form of shareholders' funds, liabilities and costs - into sales. The efficiency scores are then modelled in a random parameter framework where one of the determinants of current period efficiency is the firm's own lagged efficiency. In a parsimonious model, we find that the extent to which lagged efficiency affects current efficiency varies considerably from firm to firm. Some firms maintain a relatively constant level of efficiency period after period, while the efficiency of other firms is much more variable over time. Companies with extreme values of the random parameter (either low or high) are less likely than others to have high efficiency scores. These results are used to inform a number of qualitiative case studies of companies. Our evidence suggests that firms for which the random parameter is high tend to be long established enterprises operating in narrowly and clearly defined markets, and enjoying sustained periods of market stability; firms for which the random parameter is low tend to have had a turbulent recent past involving either rapid growth (including merger activity) or decline. Meanwhile efficiency is determined in part by the industry and country with which a firm is associated, and also by the opportunities to exploit scale economies.
Actions (login required)