A Neural Approach to Improve the Lee-Carter Mortality Density Forecasts

Marino, Mario; Levantesi, Susanna; Nigri, Andrea

doi:10.1080/10920277.2022.2050260

Cited by 19 publications

(9 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, we strongly recommend the careful selection of the countries' subgroup on which the DNN model will be trained. Finally, although some studies have attempted to provide a viable alternative (see, e.g., Marino, Levantesi, and Nigri (2022); Richman (2021)), how the uncertainty in the prediction could be derived when using deep-learning techniques is still considered a big challenge. Indeed, the resulting DNN estimate is a point estimation that does not provide any information on the uncertainty given by Ŵ , since DNN suffers from different uncertainty sources that affect the learning process.…”

Section: Discussionmentioning

confidence: 99%

Leveraging deep neural networks to estimate age-specific mortality from life expectancy at birth

Nigri¹,

Levantesi²,

Aburto³

2022

DemRes

Self Cite

View full text Add to dashboard Cite

BACKGROUNDLife expectancy is one of the most informative indicators of population health and development. Its stability, which has been observed over time, has made the prediction and forecasting of life expectancy an appealing area of study. However, predicted or estimated values of life expectancy do not tell us about age-specific mortality. OBJECTIVEReliable estimates of age-specific mortality are essential in the study of health inequalities, well-being and to calculate other demographic indicators. This task comes with several difficulties, including a lack of reliable data in many populations. Models that relate levels of life expectancy to a full age-specific mortality profile are therefore important but scarce. METHODSWe propose a deep neural networks (DNN) model to derive age-specific mortality from observed or predicted life expectancy by leveraging deep-learning algorithms akin to demography's indirect estimation techniques. RESULTSOut-of-sample validation was used to validate the model, and the predictive performance of the DNN model was compared with two state-of-the-art models. The DNN model

show abstract

Section: Discussionmentioning

confidence: 99%

Leveraging deep neural networks to estimate age-specific mortality from life expectancy at birth

Nigri¹,

Levantesi²,

Aburto³

2022

DemRes

Self Cite

View full text Add to dashboard Cite

show abstract

“…Nigri et al (2019) used data by sex for six countries and fixed fitting and forecast periods to evaluate the performance of their RNN method and the best fitting ARIMA model (Hyndman and Khandakar, 2008) in forecasting the LC k t series, finding the RNN to be superior. Marino et al (2022) show that, for three countries by sex and two fitting periods with with fixed forecast period, their RNN LC approach improves point and interval forecast accuracy compared with the Poisson-LC.…”

Section: Forecasting Evaluations and Comparisonsmentioning

confidence: 95%

“…Moreover, these authors introduce an LC model enhanced by machine learning, whereby an additional set of LC parameters, derived from machine learning, are included in the LC model Nigri et al (2019). introduced a Recurrent Neural Network (RNN) approach to model and forecast k t ; andMarino et al (2022) extend this approach to derive prediction intervals Richman and Wüthrich (2021). employ neural networks to estimate the parameters of theLi and Lee (2005) method.…”

mentioning

confidence: 99%

Thirty years on: A review of the Lee-Carter method for forecasting mortality

Basellini¹,

Camarda²,

Booth³

2022

Preprint

View full text Add to dashboard Cite

The introduction of the Lee-Carter (LC) method marked a breakthrough in mortality forecasting, providing a simple yet powerful data-driven stochastic approach. The method has the merit of capturing the dynamics of mortality change by a single time index that is invariably linear. This 30th anniversary review of its 1992 publication examines the LC method and the large body of research that it has since spawned. We first describe the method and present a 30-year ex-post evaluation of the original LC forecast for U.S.~mortality. We then review the most prominent extensions of the LC method in relation to the limitations that they sought to address. With a focus on the efficacy of the various extensions, we review existing evaluations and comparisons. To conclude, we juxtapose the two main statistical approaches used, discuss further issues, and identify several potential avenues for future research.

show abstract

“…In clusters 3 and 6, which display strong cointegration evidence, VECM has a slightly better overall performance than ARIMA. The coverage indicator refers to the number of forecasted values lying within prediction intervals (see, e.g., Marino et al, 2022, which derived the prediction intervals for future mortality rates and their coverage probability). It is defined on 95% prediction intervals, which are in turn based on the assumption that the residuals follow a Normal distribution.…”

Section: Models' Performancementioning

confidence: 99%

Modeling Gender Life Expectancy Ratio in a Multi-population Framework

2023

Self Cite

View full text Add to dashboard Cite

This paper aims to assess whether the male-to-female ratio in life expectancy is driven by cross-national long-period common trends. If a common trend is detected across a group of countries, then a model taking it into account should provide a more reliable description of the process in scope. We model the gender life expectancy ratio of a set of countries as a multivariate time series. Since our study includes data from 25 countries that are characterized by different longevity patterns, we preliminarily aggregate populations in clusters. Once we verify that all these series are non-stationary, we apply Johansen’s cointegration rank test to each cluster. The idea behind this methodology is to detect any stationary linear combination out of nonstationary variables, which implies that the series share some underlying long-period equilibrium relationships. If the cointegration rank test provides evidence for any cointegrating relationship, we further check whether estimated linear combinations of the single series - namely the error correction terms - are stationary. Finally, we measure forecasting errors with respect to the 10-year testing set. At least for two clusters over six, we find enough evidence to outline some cointegrating relationships in the male-to-female life expectancy ratio. This result leads us to assume that an underlying long-period equilibrium in life expectancy gender differentials does exist among the single countries included in these clusters.

show abstract

A Neural Approach to Improve the Lee-Carter Mortality Density Forecasts

Cited by 19 publications

References 48 publications

Leveraging deep neural networks to estimate age-specific mortality from life expectancy at birth

Leveraging deep neural networks to estimate age-specific mortality from life expectancy at birth

Thirty years on: A review of the Lee-Carter method for forecasting mortality

Modeling Gender Life Expectancy Ratio in a Multi-population Framework

Contact Info

Product

Resources

About