Using Birth Cohorts to Improve Long-Term Mortality Forecasts

Summary

The note distinguishes period mortality tables, which summarize mortality at a point in time, from cohort models, which include birth year as an explanatory variable. Although the question frames cohort and population-based models as competing approaches, the response clarifies that both concern populations; the key distinction is whether birth cohort is modeled explicitly. Cohort membership can carry predictive information about mortality patterns.

For longevity risk management, the response argues that cohort effects are often relevant and that cohort models are generally more suitable for long-range mortality projections. Period tables describe a snapshot of current mortality and may be less useful for projecting the experience of people as they age. The guidance is brief: it offers a broad modeling distinction rather than a selection procedure, empirical comparison, or discussion of data and model uncertainty. It does not quantify the predictive value of cohort effects or establish that one model is best in every application.

Key ideas

• A cohort mortality model includes birth year as an explanatory variable; both cohort and period tables describe populations.
• Birth cohorts can contain predictive information about mortality.
• Period tables summarize mortality at a given time and are less suited to long-term projections.
• The response favors cohort models for many long-horizon mortality forecasts but gives no formal selection rule.

Tags

# Cohort-based model vs. population-based model for mortality


# Cohort-based model vs. population-based model for mortality












A cohort-based model groups individuals with at least one common characteristic over a period of time through a state-transition process. A population-based model reflects as much information as possible about a demographic of the target population. Based on these definitions, a population-based model should be preferred because (1.) the population model does not focus on a specific representative demographic and (2.) the population model is dynamic in the sense that it captures the entry of new groups to the target population.

My question is what the advantages of are using a cohort-based model over a population-based model are in the context of modelling mortality specifically in the management of longevity risk. Furthermore, how should a person choose between cohort and population model when designing a mortality model.

## Answer by g g (score 2)

https://quant.stackexchange.com/a/74253

Your definitions are unusual. Where did you find them?

There are period and cohort life tables (models) - both refer to populations. A cohort model is a model where the cohort (i.e. year of birth) is an explanatory variable.

Including cohorts in mortality models makes a lot of sense since cohorts often carry predictive power. See for example here or this press article on the "golden cohort" in the UK. In most cases only cohort models are suitable for long term predictions of mortality. Period tables are only a snapshot of current mortality and less useful for predictions.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.