Modelling interstate migration in México: Static Markov Chains versus Dynamic Markov Chains with Moving Average

Resumen

This paper compares the accuracy / inaccuracy of Static Markov Chains (CME) versus Dynamic Markov Chains with Moving Averages (CMD) to reproduce interstate migration flows in Mexico. This comparison has not been done in Mexico. The main contribution of this paper is the systematic way the precision of the two Markov models is compared. We detailed the metrics of comparison. Each metric is geared to a different dimension of the migration system: i. Inequality of immigration is analyzed with Gini indexes and Lorenz curves; ii. Competition of migration is compared with the Herfindahl-Hirschman index; iii. Statistical Similarity is examined with Pearson correlation coefficients and Spearman rank correlation; and iv. The Global Error is estimated with the proportional average error index. The model derived from the CMD was slightly more accurate than the static model to replicate interstate migration flows observed in Mexico and will surely be even more accurate as more information becomes available and the study period is extended.