Autori
Espa, Giuseppe
Cefis, Elena

Titolo
Assessing accuracy in transition probability matrices
Periodico
Università degli studi di Trento. Dipartimento di Economia. Discussion papers
Anno: 1998 - Fascicolo: 1 - Pagina iniziale: 1 - Pagina finale: 25

Transition probability matrices have been used in empirical studies to investigate different topics. When transition matrices are calculated from empirical data, the resulting probabilities are estimates of the transition probabilities of underlying Markov chains. The purpose of the paper is to access the accuracy of these estimates. The suggested approach consists in applying the bootstrap methodology to the transition matrices to obtain standard errors associated with transition probabilities. In order to illustrate the approach, we choose different kinds of transition matrices computed by Cefis (1996) and by Cefis and Orsenigo (1998) to investigate persistence in innovative activities. The validity of the approach suggests that transition probability matrices can be applied not only as a pure qualitative and descriptive instrument (as suggested in the social mobility studies), but they can be treated as estimates in an inferential context.




Esportazione dati in Refworks (solo per utenti abilitati)