2. Data quality

The data on deaths have been subjected to a number of tests on their plausibility and internal consistency. While age heaping is easily observed in age-specific death rates, evidence of age overstatement can be detected by examining age distributions of deaths. Measurement of the slope of mortality is also apt to reveal dubious patterns caused by overstatement of age. As this error is almost invariably greater among old men than women, it usually leads to an abnormal sex ratio. Countries with data of proven high quality provide a set of indicators against which less certain data can be compared. At the opposite end, indicators found in notoriously unreliable data show the direction and magnitude of alarm signals. Another diagnostic tool are time series of data quality indicators which often reveal improvement over time towards more plausible levels. As a whole the data quality indicators for the 30-odd countries present a consistent and logical picture with a range of values in which it is possible to discern various degrees of reliability with some assurance when due regard is paid to the degree of aging and the general level of mortality in the country in question. A report on quality assessment of the Kannisto-Thatcher database will be published in a forthcoming volume in this series.

        On the basis of these findings the data have been classified into the following four quality categories.

 

A. Good quality

Austria Iceland
Belgium Italy
Czechoslovakia Japan
Denmark Luxemburg
England and Wales1) Netherlands
Finland Norway
France Scotland1)
Germany, Ease2) Sweden
Germany, West2) Switzerland
Hungary

        Although quality differences can be noted within this group of 19, it has not been found easy nor important to divide it into sub-groups. For the following five countries the data go in single years to age 99 only: Czechoslovakia, Germany (East), Hungary, Luxemburg, Scotland. The three first-mentioned among them form an analytically very useful group of the former East bloc. The data for the remaining 14 countries will be frequently used as the source of the most reliable information available on the human life span. It should be noted, however, that the Italian data for 1950s and the early Japanese data on centenarians were not fully reliable and that pre- 1960 data on Iceland and the Netherlands are missing.


B. Acceptable quality

Australia
New Zealand, non-Maor 3)
Portugal
Singapore, Chinese 4)

        The Australian data show an apparent increase in mortality at ages 100-104 which is probably due to data improvement while it is not known whether this has already run its course. The New Zealand data are affected by uncertain age information on South Pacific immigrants who are classified as non-Maoris. Both countries show an unusually high proportion of deaths over age 100 which may or may not be true. 'Ilie quality indicators for Portugal show notable improvement over time resulting in approximately correct data for the last two decades except. at ages over 105. The series for Singapore is as yet too short for conclusive evaluation.


C. Conditionally acceptable quality

Estonia
Ireland
Latvia
Poland
Spain

        These data give probably a roughly correct description of the mortality trend though at a level artificially lowered by age overstatement. The Estonian data are considerably better than those for Latvia and Poland.


D. Weak quality

Canada
Chile
New Zealand, Maori
United States

        Used with due caution, these data may give approximate information on the size and development of the old age population below age 90 or 100 but estimation of mortality by the method applied in this report would be too uncertain. The data on Chile, New Zealand Maoris and United States non-whites are clearly the least reliable.


1) England and Wales on one hand and Scotland on the other will be treated as separate countries because their vital statistics are prepared and published separately.
2) The two parts of Germany before reunification will be called East and West Germany, as this may be more convenient than acronyms because different readers have been used to acronyms based on different languages.
3) Data for New Zealand, unless otherwise stated, refer to the non-Maori population only. The Maori population is small and the data for it unreliable.
4) Data for Singapore refer to the Chinese population only. Data for other ethnic groups are available but are too small to be significant.

Updated by V. Castanova, 1 March 1999