Data Resource Profile: The Human Fertility Database

The Human Fertility Database (HFD) was created in response to the growing need of freely available, high-quality, internationally comparable population-level fertility data. In the midst of discussions about the shift to later childbearing and low fertility levels in industrialized countries, the lack of proper fertility data became particularly apparent. A number of other sources of fertility data exist, but accessibility, scope and comparability of these data vary widely, and data documentation is often inadequate or absent. Officially launched in 2009, the HFD is a collaborative endeavour of the Max Planck Institute for Demographic Research (MPIDR) in Rostock (Germany) and the Vienna Institute of Demography / Wittgenstein Centre for Demography and Global Human Capital (Austria). The development of the HFD was inspired by the success of the Human Mortality Database (HMD) that had become an important resource of detailed mortality data.1 By providing free access to detailed fertility data which meet the highest quality standards, the HFD not only fills the gap in the availability of comparable fertility data, but also promotes more sophisticated analyses and methodological advances. Information about HFD history, methodology, research team and other important material can be found at [www.humanfertility.org]. The HFD is based on official data on live births and population counts originating from national statistical agencies. Because of strict data quality requirements, the HFD provides data only for developed countries with full coverage of vital statistics. Eventually the HFD is expected to cover all countries included in the HMD (currently 38 countries and 47 territories or populations), as well as several other countries with relatively high-quality fertility and population data. At present, the HFD provides detailed fertility data for 27 countries and five sub-national populations. For all countries, the HFD assembles the longest possible continuous time series. Figure 1 shows the ranges of years for which data series for all birth orders combined and by (biological) birth order of the child [considering all previous live-born children of the mother, including those she has had outside marriage or with different partner(s)] are available for each population in the database. The HFD also features selected advance data for countries that will be included in the near future. Figure 2, depicting trends in total fertility rate, mean age at first birth, completed cohort fertility and cohort childlessness, highlights the wide diversity existing among the countries in the HFD. Period fertility is measured here by the period total fertility rate (TFR), which represents the average number of births a woman would have if the age-specific fertility rates observed in a given period remained the same throughout her entire reproductive life.4 Despite the increase in the TFR witnessed in many countries in the early 2000s, Southern Europe, Central and Eastern Europe (CEE) and East Asia continue to be the regions with the lowest fertility globally. The oldest age at first birth is found in Southern Europe and East Asia, whereas the post-communist countries of CEE still have considerably younger childbearing schedules. Completed cohort fertility, measuring mean completed family size of women born in the same year,4 has been decreasing in most of the HFD countries. Among the women born in the late 1960s, the lowest levels of cohort fertility, at 1.4–1.6 children per woman, prevail in Germany, Italy and Japan, whereas the countries and territories with relatively high period fertility levels, including Iceland, Northern Ireland, the USA and France, also have higher cohort fertility at or above 2 children per woman. The lowest shares of childless women are recorded in CEE. Women born in Japan in the late 1960s have the highest level of childlessness (close to 25%), followed by women from Canada, Finland and the Netherlands. HFD period (left panel) and cohort (right panel) fertility data series for all birth orders combined and by birth order (as of April 2016). Trends in period total fertility rate and mean age at first birth and in completed cohort fertility and childlessness (data as of April 2016). Czech R., Czech Republic. For each country or territory, HFD data are provided in four blocks: ‘Summary Indicators’, ‘Age-Specific Data’, ‘Fertility Tables’ and ‘Input Data’. The first three data blocks include sets of indicators of period and cohort fertility. The age in the age-specific data ranges from ≤ 12 to 55+ and the birth order ranges from 1 to 5+. The fourth data block, ‘Input Data’, provides original data used for the calculation of HFD indicators. Data for each country and territory are thoroughly documented in a Background and Documentation file. Figure 3 illustrates a typical HFD country page. Screen shot of a typical HFD country page. The first data block, ‘Summary Indicators’, contains conventional aggregate measures of period and cohort fertility, specified also by birth order when available. These measures comprise total number of live births, crude birth rates, total fertility rates, mean age at birth and standard deviation in the mean age at birth. For period data, tempo-adjusted TFRs estimated by Bongaarts and Feeney's2 method are displayed alongside the conventional TFRs. These indicators aim to correct distortions caused by changes in the age at childbearing, which depress the conventional TFR when births are postponed to higher ages and inflate it when births are shifted to younger ages. For cohorts, parity progression ratios are available in addition to the completed cohort fertility. They show the proportion of women with a given number of children (n) who eventually give birth to another child (n + 1).4 The second block, ‘Age-Specific Data’, also consists of period and cohort data, specified by birth order when available. The period and cohort data sub-blocks cover the same array of indicators: birth counts, female population exposures and unconditional age-specific and cumulative fertility rates. Period data are presented by different configurations of calendar year and mother's year of birth and age. Cohort data are specified by mother's year of birth (birth cohort) and age in completed years. The third block, ‘Fertility Tables’, contains two sub-blocks: ‘Period fertility tables’ and ‘Cohort fertility tables’. These are increment-decrement life tables which model the process of childbearing in real or synthetic (period) female cohorts, focusing on the age and parity (number of children a woman already has) dimension. Along with the period fertility tables, information on age- and parity-specific female population exposures, conditional fertility rates by age and parity, parity- and age-adjusted total fertility rates (PATFR)3 and table mean ages at birth are provided. The fourth block, ‘Input Data’, includes input data files containing original birth and population data organized in standardized HFD format and often covering wider ranges of ages, birth orders or years of coverage than the standardized data in the first three HFD blocks. References to data sources as well as notes relevant to the input data are also available in this data block. In addition to the detailed main data set, users are offered an ‘HFD lite’ option, which includes easily accessible Excel tables with the most commonly used aggregate fertility indicators for all HFD countries: total fertility rate, mean age at birth, mean age at first birth, completed cohort fertility and cohort childlessness. The HFD follows precise data processing procedures, which are described in detail in the HFD Methods Protocol.4 The entire data processing can be described as a sequence of six steps as follows. Births. HFD input birth counts vary significantly with respect to age and birth order ranges, Lexis shapes, etc. When necessary, specific splits and adjustments are performed to transform these varying data formats into the uniform Lexis format where birth counts are classified by calendar year, age of the mother and mother's birth cohort. In these data, the age varies from 12 and younger to 55+, and the birth order ranges from 1 to 5+. The Lexis data by year-age-cohort cells form the basis for all further calculations. Population denominators. The HFD uses data on female population by age except for one indicator, the crude birth rate. There are two broad data types used for the female population. The first type specifies female population exposure by age, which is needed for the computation of fertility rates by age. For most countries, the female population exposure is estimated using data from the HMD. If the country is not included in the HMD, the HMD methodology is used to produce necessary data from additional population statistics collected. The second type of population data specifies women by age and parity; these estimates are needed for the computation of conditional (parity-specific) fertility rates, which serve as the major input for period fertility tables. These data are usually available from population censuses or registers. Fertility rates. The HFD provides both unconditional and conditional fertility rates. Unconditional age-specific fertility rates relate births by age of the mother and (when available) child's birth order to the entire female population of a given age. Conditional age- and parity-specific fertility rates measure childbearing intensity among women of specific ages and parities (e.g. second births are related to women of parity 1 only). Cumulative fertility rates are based on a summation of fertility rates across ages up to the indicated age limit, shown for each single age category. Cohort fertility rates are computed for every combination of birth cohort and age in the available data. Period fertility rates are computed for every combination of calendar year and age in the available data. Summary measures. The summary measures are crude birth rates, total fertility rates (including the completed cohort fertility), tempo-adjusted total fertility rates, mean ages at birth, cohort parity progression ratios and standard deviations in the mean age at birth, computed for all birth orders combined and by birth order. The simplest summary measure is the crude birth rate, which is the ratio of total live births to total population in a given year. The other summary measures are based on unconditional age-specific fertility rates. Cohort fertility tables by age and parity. These increment-decrement life tables model childbearing in female birth cohorts by age and parity. In the HFD their computation is possible if the available fertility data cover time periods long enough to observe cohorts from age 15 or younger until age 25 or older. Life table functions for each cohort are computed from unconditional age- and order-specific fertility rates (horizontal parallelograms) as the major input data. Period fertility tables by age and parity. Period fertility tables describe the fertility progression in a ‘synthetic cohort’ of women. Each period table is built on the basis of conditional age- and parity-specific fertility rates of 1 calendar year. The key input in period tables is the parity-specific distribution of female population by age. These distributions are obtained either from the cohort fertility tables or from census data that provide the initial parity distribution in 1 base year, or directly from repeated population censuses or annual register data. In the latter case, the period fertility tables are called ‘census- or register-based’. All input and output data files undergo extensive checks by HFD country specialists (i.e. HFD team members responsible for a specific country or countries in the HFD). Since the launch of the HFD in 2009, more than 4000 unique users have accessed it. As of April 2016, a (non-exhaustive) list of publications that used the HFD includes 106 journal articles, 45 books, book chapters or dissertations and 89 other publications: see the full list at [http://www.humanfertility.org/Docs/publications.pdf]. Cross-national studies discussing fertility trends and aiming to provide comparable data for different countries are a typical example of research benefiting from the HFD. Using HFD data supplemented with other data sources, Goldstein et al.5 examined the effect of the late 2000s financial and economic crisis on fertility trends in Europe. Rindfuss et al.6 used HFD data to pursue an argument that economically developed countries experienced a ‘bifurcation’ in their fertility rates into two distinct fertility regimes. Frejka and Zakharov7 used HFD data to analyse the effects of pro-natalist policies, adopted in Russia in 1981 and 2007, on the timing and the level of fertility by birth order. The World Fertility Report8 by the United Nations shows that this global agency also used the HFD as an important source of fertility data. The availability of birth-order-specific fertility data is another advantage of the HFD. Using HFD data for 15 countries of Western and Eastern Europe, Burkimsher9 looked into demographic causes of the transition from declining to increasing first-birth rates in the period 1995–2002. Sobotka's10 comparative study on childlessness in Europe made extensive use of HFD data, including input data from different censuses. Detailed and reliable data available for a number of countries are indispensable for researchers developing new fertility indicators as well as new methodology. Bongaarts and Sobotka11 and Goldstein and Cassidy12 used HFD data in combination with other data sources to test the usefulness of their newly developed tempo-adjusted fertility measures. Schmertmann13 developed a new method of data-splitting by age, called the calibrated spline (CS) estimator. For the identification of empirical regularities in the actual (known) fertility schedules, this study uses data also from the HFD. The development of new fertility projection methods and generation of fertility projections constitute another important research area where HFD data have been used. For this purpose, HFD data were employed in the studies by Myrskylä et al.,14 Schmertmann et al.15 and others. A short survey of data users conducted electronically in 2012 revealed that HFD data are also widely used for teaching and studying purposes. High level of detail in data and documentation, uniformity of methods and data design, availability of birth order-specific fertility indicators and free access to all data upon registration are the features that make the HFD particularly attractive to users. The HFD adheres to the same four guiding principles as the HMD:1 comparability, flexibility, accessibility and reproducibility of the data. The data series are regularly updated. Data are provided in text files which can be easily downloaded into any statistical package. The scripts used for calculations are published as MPIDR Technical Reports.16–18 The HFD also introduced a streamlined version of the database (‘HFD lite’) which provides easy and quick access to the major summary indicators of period and cohort fertility, without registration. The weaknesses of the HFD stem partially from its strengths. By aspiring for a high level of quality and detail of collected data, the HFD restrains itself to only a limited number of countries, historical periods and data dimensions. Also, because the procedure of acquiring, documenting and processing the data is time consuming, the latest HFD data often lag behind the official statistics published by statistical agencies. The HFD aims for the inclusion of all countries featured in the HMD and a few additional countries that are not part of the HMD but which have fertility data as well as population data at reproductive ages that are of sufficient quality. Access to HFD data is free of charge upon registration, which is quick and simple. New users can register by accessing the following URL: [http://www.humanfertility.org/cgi-bin/registration.php]. No registration is required for accessing the ‘HFD lite’ data. The HMD [www.mortality.org] inspired the development of the HFD. Moreover, the HFD uses HMD data on population and death counts for the calculation of population exposures for most of the countries. The Human Fertility Collection (HFC), available at [www.fertilitydata.org], is a companion to the HFD project. The project, covering 100 countries as of April 2016, serves as a repository of historical and recent fertility data by age and birth order from various sources. Often, these data do not meet quality standards of the HFD. The HFD has also inspired the creation of other related databases, especially: the Russian Fertility and Mortality database (RusFMD) based at the Center for Demographic Research at the New Economic School, Moscow ([http://demogr.nes.ru/index.php/en/demogr_indicat/data]); and the Cohort Fertility and Education database (CFE), developed by the Vienna Institute of Demography ([www.cfe-database.org]). HFD in a nutshell The HFD was created to provide free access to high quality, internationally comparable fertility data and to facilitate fertility research and informed policy making. Launched in 2009, the HFD currently features data for 27 countries, plus five sub-national populations. The database provides detailed and uniform data on cohort and period fertility for all birth orders combined and by birth order, including birth counts, age-specific fertility rates, period and cohort fertility tables and aggregate fertility indicators. HFD output is generated by applying standardized methods. The HFD methodology includes procedures for transformation of input data into a uniform and highly detailed structure, and ensures high consistency of output data across time and countries. Great emphasis is placed on data quality checking and documentation. Data documentation, usually prepared by external country experts together with HFD team members, is available for each country. HFD data series are regularly updated and can be accessed upon registration at [www.humanfertility.org]. This project is supported by the Max Planck Institute for Demographic Research, Germany and the Vienna Institute of Demography (VID), Austria. The European Commission funded the initial work by the VID as part of the project ‘Reproductive Decision-Making in a Macro-Micro Perspective’ (REPRO) in the Seventh Framework Programme (Grant Agreement: SSH-CT-2008-217173). Since February 2012, the work of the VID team (Kryštof Zeman and Tomáš Sobotka) has been funded by the European Research Council under the European Union´s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement 284238 (EURREP project). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the institutions to which the authors are affiliated (the MPIDR and the VID) or those of the other funders. We acknowledge the key contribution of the HFD Founding Director Joshua R. Goldstein. The HFD project greatly relies on the support of national statistical offices, other data holders and country experts who assist us in acquiring and documenting the data. We are grateful for their contributions. Our gratitude for their valuable comments on the HFD methodology and general advice to the project is due to the members of the HFD Advisory Board: John Wilmoth, Wolfgang Lutz, Laurent Toulemon and Jean-Paul Sardon. We also thank Edward Nash, Dimiter Philipov and Evgeny Soroko, who have importantly contributed to the development of the HFD. Finally, we are grateful to Karolin Kubisch and Sigrid Gellers-Barkmann for assistance in data searches and formatting. For language editing, we thank Renée Flibotte-Lüskow. Conflict of interest: None declared.