In this study we derive a novel multi-population Lee-Carter type model. Specifically, we extend the Bayesian model in Czado et al. (Insur Math Econ 36:260–284, 2005) to allow exchangeability between parameters of a group of m populations. In a validation-based examination, the proposed model is found to be beneficial for several examined countries. Also, we examine changes in forecasting ability due to varying calibration periods. Our results suggest that mortality rates from a distant past are inferior to those from a more recent past.

中文翻译：

---

可交换死亡率预测

在这项研究中，我们得出了一个新颖的多种群Lee-Carter型模型。具体来说，我们在Czado等人中扩展了贝叶斯模型。（Insur Math Econ 36：260-284，2005），以允许一组m个总体的参数之间具有可交换性。在基于验证的检查中，发现建议的模型对几个检查过的国家都是有益的。此外，我们研究了由于校准周期的变化而导致的预测能力的变化。我们的结果表明，远古时代的死亡率低于近代。