The future of mortality – mortality forecasting by extrapolation of deaths curve evolution patterns
- A variety of mortality models can be used to project future mortality. However, the parameters of most of these models lack a clear demographic interpretation. Hence, the resulting projections may be demographically implausible in the sense that trends in key demographic statistics are not extrapolated in a reasonable way. When demographers make predictions on future mortality, they typically focus on one or few relevant demographic statistics related to certain aspects of the mortality evolution. However, they do not derive comprehensive mortality forecasts as required for actuarial purposes. This article aims to close the gap between these forecasting approaches.
To this end, we establish a new deterministic mortality model which can be used for best estimate and scenario forecasts. We model the deaths curve, i.e. the age-at-death distribution, and derive forecasts based on the extrapolation of statistics that have a clear demographic interpretation. The four key statistics of the model are those from the classification frameworkA variety of mortality models can be used to project future mortality. However, the parameters of most of these models lack a clear demographic interpretation. Hence, the resulting projections may be demographically implausible in the sense that trends in key demographic statistics are not extrapolated in a reasonable way. When demographers make predictions on future mortality, they typically focus on one or few relevant demographic statistics related to certain aspects of the mortality evolution. However, they do not derive comprehensive mortality forecasts as required for actuarial purposes. This article aims to close the gap between these forecasting approaches.
To this end, we establish a new deterministic mortality model which can be used for best estimate and scenario forecasts. We model the deaths curve, i.e. the age-at-death distribution, and derive forecasts based on the extrapolation of statistics that have a clear demographic interpretation. The four key statistics of the model are those from the classification framework of Börger et al. (2018). The design of our model makes sure that forecasts for the immediate future of the deaths curve are consistent with the most recent trends of all demographically relevant statistics. Moreover, expert opinions with respect to the future trends of certain demographically interpretable statistics can easily be incorporated – in particularly for the farther future where a pure extrapolation of historic trends might lead to implausible results. We present a possible implementation of the model and provide case studies that illustrate how the model can be applied.…
