Are There Differences in Health Inequality in Germany and across German Regions? Evidence for Retired Men Aged 65+ from the German Pension Fund Data.
Georg Wenau, Max Planck Institute for Demographic Research
Pavel Grigoriev, Max Planck Institute for Demographic Research
Vladimir M. Shkolnikov, MPIDR
Data: A full administrative sample of all retirees for the year 2015 was available from the German Statutory Insurance Scheme. These individual-level data contain precise information about the working life history for about six million men as well as reliable information about the end of individual annuity payment due to death.
Methods: Average Pension Points per year serve as a proxy for the relative socioeconomic position of a person over the course of working life. Pension Points reflect the amount of income which was subject to social insurance contributions and are the basis for subsequent calculations of individual pension payments. Based on Poisson Regression, mortality rate ratios (MRR) are estimated for population subgroups according to quintiles of average Pension Points, type of health insurance and place of living.
Results: Preliminary results reveal substantial socioeconomic and regional differences in mortality among the sample population. Compared to the most advantaged group, the MRR for the most disadvantaged group is about twice as high in almost all German federal states. Further results show that the mortality level of the advantaged group is relatively equally distributed across German regions, whereas the mortality level of the disadvantaged groups differ substantially between German regions.
Socioeconomicdifferences in mortality have been identified by a large number of studies for nearlyall Western societies by now. These studies reveal substantial healthdisadvantages for groups with lower education, income or occupational status.The reduction of these inequalities is a major issue of public health research.Studies which rely on health-related surveys can make use of a broad variety ofhealth indicators and socioeconomic variables while being restricted in termsof representativity and panel attrition. Census-linked mortality studies aresuitable to overcome these restrictions. Unfortunately, German populationstatistics do not offer the opportunity to link mortality register and censusdata. Therefore, there is still a lack of knowledge about how socioeconomicdifferences affect German mortality and almost nothing is known to what extendthese inequalities differ across German regions. This study is the first toinvestigate socioeconomic differences in old-age mortality among retired Germanmen on the level of German federal states.
Multivariate analysesare based on individual-level datasets from the German Statutory InsuranceScheme (Deutsche Rentenversicherung, DRV) which cover about 90% of the German malepopulation aged 65. The data are collected primarily for the calculationof individual pension payments according to the amount of earnedincome. To be more precise, it refers to income which was subject to socialinsurance contributions. Hence, the data include detailed informationabout the individual earnings over the course of working life. So calledpersonal Pension Points (PP) reflect this information for each individual. Tocalculate this indicator the DRV first ascertain the average income of thewhole population for the respective calendar year. Then, the individual incomeis valued with respect to the average income. If a person earned exactly theaverage within the respective period he or she receives 1 PP for this year. So,PP can be seen as an expression of the relative position of a person within thewhole society in terms of income for the respective period of time. Thesummation of PP for all relevant periods pools the information and serves as aproxy for the socioeconomic position over the course of working life. For thisstudy, we calculated average personal PP per year by dividing the sum of PP by therespective length of the individual contribution period to avoid potential biasthrough short contribution periods.
We consider averagePP as a very promising and representative indicator of socioeconomic inequalityfor men but not for women. This is because of lower labour force participation amongthe female cohorts under study (especially in western Germany). Therefore womenwere excluded from the analyses.
The final sample includes all menaged 65+ who are living in Germany and receive an old-agepension. Persons with non-German citizenship and migrants are excluded.The observation period is the year 2015. The datasets provide the respectivenumber of retirees based on the year-end pension portfolio of2014 and 2015, respectively. Exposure-torisk population was determined by collapsing these cross-sectional individualdatasets by population subgroups and then taking the arithmetic mean of populationcounts. The number of deaths within 2015 was obtained from another datasetwhich includes all cases for which pension payments have been stopped due todeath of the insured person. The combination of these datasets made it possibleto run Poisson Regression Models and estimate mortality rate ratios (MRR). Sociodemographicvariables under consideration are: age, quintiles of average PP; place ofresidence (at the level of federal states); type of health insurance(compulsorily or privately/voluntarily insured). Because the distribution of PPdiffers substantially by type of health insurance, the quintiles of average PPhave been calculated separately for these subgroups. For place of residence wechose to assign the same limits to make the estimation results comparablebetween the federal states.
Preliminaryresults and prospect
First results revealsubstantial socioeconomic and regional differences in mortality among thesample population. Compared to the most advantaged group (5thquintile of average PP), the MRR for the most disadvantaged group (1stquintile) is about twice as high in almost all German federal states (Fig. 1).Further results show that the mortality level of the more advantaged groups isrelatively equally distributed across German federal states, whereas themortality of the disadvantaged groups differs substantially across regions (Fig.2).
To facilitatethe comparison of the mortality differences across German regions, we intend touse the Slope Index of Inequality (SII). Standardized death rates can be usedto depict absolute health differences between subpopulations. Moreover, thedatasets offer additional, more specific income indicators, which might also beuseful to operationalize socioeconomic inequality in other ways. We would liketo test and compare these indicators with the already established ones. Theanalysis of additional observation periods is planned to be able to makestatements about how inequality developed over time.
Presented in Session 1194: Mortality and Longevity