Non-causality in Bivariate Binary Panel Data

Non-Causality in Bivariate Binary Panel Data

Abstract

In this paper we develop a suitable dynamic discrete time bivariate
probit model, in which the conditions for Granger поп-causality may be
represented and tested. The conditions for simultaneous independence are
also worked out. The model is extended in order to allow for covariates,
representing individual as well as time heterogeneity. The proposed model
may be estimated by Maximum Likelihood; Granger поп-causality, as
well as simultaneous independence may be tested by Likelihood Ratio
tests. A specialized version of the model, aimed at testing Granger non-
causality with bivariate survival data is also discussed. The proposed
tests are illustrated using data concerning the relation between marriage
and fertility timing in a sample of 266 American women and the adoption
of two interrelated technological innovations by 552 Italian metalworking
plants.

Acknowledgments This paper has benefited at different stages by discus-
sions with Per Kragh Andersen, Clelia Di Serio, Clive Granger, Philip
Hougaard, Spren Johansen and Niels Keiding. The authors are particu-
larly indebted to Fabio Sartori, who significantly contributed to the de-
velopment of the main ideas presented in the paper. The usual disclaimer
applies.

1 Introduction

The epistemological status of the statistical-probabilistic notion of causality
based on predictability is still a matter of profound controversy among philoso-
phers and methodologists (see Geweke, 1984). This notion fits, in a probabilistic
sense, two key aspects of causation: the systematic conjunction of cause and
effect, and the time precedence of the cause with respect to the effect. Nonethe-
less, it fails to account for what probably is the deepest, though empirically less
helpful, aspect, i.e. the idea that the cause “forces” or “produces” the effect.
Despite these limitations, the notion of causality based on predictability proved
to be a valuable tool for applied research thanks to its operational usefulness
—

_a⅛ * Dipartimento di Economia e Produzione₅ Politecnico di Milano₅ P.za Leonardo da Vinci
32, 20133 MILANO, ITALY, email: rocco.mosconiSpolimi.it

tCRESt-LFA, Timbre J320, 15 bd Gabriel Péri, 92245 MALAKOFF CEDEX, FRANCE,
email: seriQensae.fr

1   2   3   4   >>

More intriguing information