We analyze the performance of firms in the German business-related services sector. A quarterly business survey provides the panel data base of our study. Firm performance is measured by the survey respondents? ordinal indication of their changes in total sales. We use a firstorder Markov chain and a multinomial logit specification to model the transition probabilitites. Three variants of the model are estimated: a linear index model with and without unobserved firm heterogeneity and a semiparametric model. Main results are that firm size has a positive effect on firm performance, that young firms outperform older competitors, that a bank-relationship with a single creditor has a stabilizing effect and that the degree of diversification has a negative impact on firm performance. The legal status appears to have no significant effect.
Markov chain service sector business survey firm performance multinomial logit model generalized additive model