Dynamic econometric models for cohort and panel data: Methods and applications to life-cycle consumption.

The purpose of this research is to analyze dynamic models for cohort and panel data, with special emphasis in the applications to life-cycle consumption. In the second chapter of the thesis we analyze the estimation of dynamic models from time-series of independent cross-sections. The population is divided in groups with fixed membership (cohorts) and the cohort sample means are used as a panel subject to measurement errors. We propose measurement error corrected estimators and we analyze their asymptotic properties. We also calculate the asymptotic biases of the non-corrected estimators to check up to what extent the measurement error correction is needed. Finally, we carry out Monte Carlo simulations to get an idea of the performance of our estimators in finite samples. The purpose of the second part is to test the life-cycle permanent income hypothesis using an unbalanced panel from the Spanish family expenditure survey. The model accounts for aggregate shocks and within period non-separability in the Euler equation among consumption goods, contrary to most of the literature in this area. The results do not indicate excess sensitivity of consumption growth to income. In the last chapter, we specify a system of nonlinear intertemporal (or Frisch) demands. Our choice of specification is based on seven criteria for such systems. These criteria are in terms of consistency with the theory, flexibility and econometric tractability. Our specification allows us to estimate a system of exact Euler equations in contrast to the usual practice in the literature. We then estimate the system on Spanish panel data. This is the first time that a Frisch demand system has been estimated on panel data. We do not reject any of the restrictions derived from theory. Our results suggest strongly that the intertemporal substitution elasticity is well determined