Firm versus Sector Effects on Profitability and Growth: The Importance of Size and Interaction

Abstract

In this study we contribute to the long‐standing debate on the impact of firm versus industry effects on firm performance in three distinct ways; firstly by testing the firm, industry and their interaction effects on performance, secondly by examining the impact of each effect for different size groups, and lastly by measuring performance in terms of sales growth in addition to profitability. We use data of 71,750 UK firms, between 2002 and 2004, and employ moderated regression analysis for three sub‐samples namely micro, SMEs and large firms. With regards to profitability, we find the interaction effect to be significant in all sub‐samples for broad level of industrial aggregation (SIC4). For narrow industrial aggregation (SIC2), the interaction effect is only significant for micro firms. Neither of the above effects is significant for sales growth.

Key Words:

• Firm Effect
• Industry Effect
• Interaction
• Profitability
• Growth

JEL classifications:

• D21
• L11
• M21

Notes

1. To demonstrate this we cannot improve on Eriksen and Knudsen’s explanation (2003, 194) ‘To see this, note that the measure used to capture industry effects (in all previous models) is an aggregate of the firm‐level ROA for a particular industry. The model then includes firm‐level ROA’s and the aggregated industry‐level ROA’s as main effects. If we further add a firm‐industry interaction term, the interaction term will capture all variance and the main effects will be completely empty and insignificant (since all effects are composed of firm‐level ROA). In other words, it is impossible to use a variance decomposition model to test the interaction hypothesis we are interested in’.

2. Commission Recommendation of 6 May 2003 concerning the definition of micro, small, and medium‐sized enterprises (2003/361/EC).

3.

Where R² _F refers to the R² of the model with the interaction term, R² _R refers to the R² of the initial model (without the interaction term included), N_F is the number of variables in the new model and N_R the number of variables in the initial model. Last S stands for the number of observations.