Incorporating big microdata in life table construction: A hypothesis-free estimator

The IT revolution, now more than ever, offers a cheaper and faster way to collect, store, transmit and process data. Detailed microdata of dates of death, migration and birth are already becoming available for general populations. In this paper, we develop within the family of period-based estimators a new, assumption-free estimator for constructing life tables. The estimator proposed exploits all the detailed data available and is free of the theoretical inconsistencies that the estimators currently used by most official statistical agencies have. We compute the proposed estimator for a real database and test the suitability of the hypotheses on which the estimators used so far rely. The hypothesis of uniform distribution of birthdays is proven to be inadequate and the one having the largest impact on the estimated probabilities. Given its influence on public pension systems and life insurances, we advocate for adopting the more efficient approaches proposed in this paper.