Migration correlation: Definition and efficient estimation

The aim of this paper is to explain why cross-sectional estimated migration correlations displayed in the academic and professional literature can be either not consistent, or inefficient, and to discuss alternative approaches. The analysis relies on a model with stochastic migration in which the parameters of interest, that are migration correlations, are precisely defined. The impossibility of estimating consistently the migration correlations from cross-sectional data only is emphasized. We explain how to handle with individual rating histories, how to weight appropriately the cross-sectional estimators and how to estimate efficiently the joint migration probabilities at longer horizons.