Decomposition of new venture growth into firm age, survey period and vintage effects

Abstract

Firm data are accumulated on a yearly basis. In view of the linear relationship of firm age + foundation year = survey year, the fluctuations of firm data classified by age and period cannot be decomposed into age, period and cohort (foundation year) effects. Three decomposition methods are briefly reviewed and applied to Japanese data on new ventures founded since 1995. Regarding sales and employment growth, the age effect is the largest with a downward trend, and the cohort effect is negligible. Regarding labour productivity, the age effect indicates upward movements, and the cohort effect is negligible. The reason of the negligible cohort effect is discussed.

Keywords:

• age–period–cohort decomposition
• Bayesian cohort model
• identification problem
• Monte Carlo simulation
• new venture growth

JEL Classification:

• L25
• M13
• C11

Acknowledgements

I am grateful to an anonymous referee for useful comments. Needless to say, any remaining errors are mine.