The aim of this commentary is to illuminate, for a wider audience, the essential features of the analysis described in the paper on “China’s low fertility may not hinder future prosperity” (1), to place the paper in the context of the program of work on the role of education in demography at the Wittgenstein Centre, Vienna, and to evaluate its claims. Tables 1 –4 provide an overview of the options for forecasting a country’s demographic future used in the Wittgenstein program. The paper’s authors chose from the options set out. Tables 1 –4 provide an interpretation of their work, based on reading the paper and discussion with the authors. Table 1 characterizes the model, its inputs, and its outputs. Table 2 provides details of alternative demographic models that might be used. In Marois et al. (1), a microsimulation model based on sample data is used, with the sample numbers factored up to a larger population. The demographic assumptions of the model embed the key relationships between fertility, mortality, and migration by educational attainment. The model for China focuses on the national population, with no role for internal migration. International migration into and out of China is very small in relation to its 1.4 billion 2020 population, so it does not play much part in determining future populations, in contrast to Western countries, where immigration is a vital contribution, and to developing countries, where emigration is important. Table 3 presents information on indicators of population aging. The companion paper (1) shows that very different results are obtained when education and productivity are included in the dependency ratios. Table 4 reports the implications of the work for national policy in China. There are also pointers to improvements that might be introduced, such as integrating recent estimates of educational quality (2), the skills and knowledge that pupils and students gain through education.
Table 1.
Features of a population projection model including educational attainment
| Model elements | Details |
| Inputs | Base year estimates plus forecasts of rates for demographic components and prevalence levels for education, labor force status, and productivity |
| Features | Multidimension Population Projection Model (MDPM) |
| Outputs | Forecast populations, demographic components, and prevalence |
Table 2.
Model options, relationships, and uses
| Demographic model options | Relationships | Uses |
| Single region cohort survival (SRCS) | Fertility rates fall with increasing education Mortality rates fall with increasing education Migration rates increase with increasing education |
Independent forecasts for a set of country or region or group populations |
| Multiregional cohort survival (MRCS) | Forecasts of prevalence levels for education are driven mainly by cohort replacement at ages 30 years and over, by educational investment at ages less than 30 years | Aligned forecasts for a set of regions or countries, linked through migration flows/rates |
| Biregional cohort survival (BRCS) | Forecasts of labor force status and productivity dependent on education forecasts | Reduced version of the MRCS model applied when region or country sets are large |
| Microsimulation model (MSM) based on microdata from large survey or census | Simulates changes in person and household attributes for individuals | Provides scaled-up forecasts for the whole population |
Table 3.
Indicators derived from population forecasts
| Derived indicators | Definitions | Forecast changes |
| Old age dependency ratio (OADR) | Ratios of older populations to working age populations | Serious increase in ratios in future Shifts upwards in threshold ages for retirement |
| Labor force dependency ratio (LFDR) | Ratio of older retired population to employed population at all ages or ratio of inactive population to active population at all ages | Increase in ratios diminishes Older population needs support to continue working |
| Productivity weighted labor force dependency ratio (PWLFDR) | Ratio of inactive population (all ages) to active population (all ages), weighted by productivity for education grade | This measure remains relatively constant for 2020−2070 in China |
Table 4.
Implications for use of a multidimension model incorporating education
| Indicator to use | Policy implications | Future model improvements |
| Use PWLFDR not LFDR or OADR to inform policy | Do not panic and introduce “forced measures” Invest in education and life-long learning, child health, and support for disabled workers |
Add indicators of educational quality beyond years of schooling Investigate other heterogeneity effects: subnational regions, rural—urban settlement, the Hukou registration system in China |
Population Projections with Added Education
The authors (1) apply demographic methodology developed by Wolfgang Lutz and colleagues over more than two decades (3) at the International Institute for Applied Systems Analysis (IIASA), Vienna Institute of Demography (VID), and the Wittgenstein Centre (a collaboration of IIASA, VID, and the Vienna University of Economics). Innovations include the rationale for including education as a variable in world population projections (4) and the use of education as a key population attribute in the projection model (5). Educational attainment is usually measured by years of attendance at an educational system of schools, colleges, and universities. Education attainment develops over a person’s early lifetime and, in most societies, has largely been completed by age 25 y. The assumption that educational attainment is fixed from this age makes its estimation for countries feasible using commonly available statistics.
Educational attainment has important effects on fertility, mortality, and migration—the drivers of population change. The higher a woman’s educational level, the lower her fertility (6). Higher education levels mean longer life expectancy (7, 8). Higher education also means higher levels of migration activity (9).
The Cohort Replacement Process or “Demographic Metabolism”
The basic process in a population projection model is cohort survival. People enter the population at birth and then, over time, survive or die from one age to the next, taking fixed attributes (educational attainment at age 25 y) with them through the rest of life. This process steadily improves the educational profile of a population, a process termed “demographic metabolism” (10). Of course, other socioeconomic indicators such as income, occupation, or health also influence demographic component rates and flows. But these other attributes fluctuate over time for individuals and populations, tracking ups and downs of economies. They are more difficult to operationalize. The overall educational attainment of the population improves over time because of the steady replacement in each age band of the older, less educated cohort by younger, more educated cohorts.
Multidimension Projections Combine Multistate and Prevalence Models
In Table 1, the overall model has been characterized as multidimension rather than multistate (11). A multistate model incorporates two-way flows of persons between attribute states, so that outcomes in one state are dependent on those in all other states. The original model used regions within a country between which people migrated (12). Data on migration between regions are widely available from periodic censuses or continuous population registers. Equivalent flows between educational grades are not as accessible, so a prevalence method was used. This links prevalence of a condition in a population by age and sex in a recent survey or census to the age and sex variables in a life table or a projection. What needs to be added is a separate model for projecting educational prevalence (13).
Measuring Population Dependency Using Improved Ratios
One of the key challenges faced by nation states in the twenty-first century is population aging. To measure this challenge, demography uses dependency indexes, principally the ratio of older age groups to working age groups (Table 3). However, crude demographic ratios are misleading, so researchers have developed ratios that account for the economic contributions of the different age groups by measuring the degree of labor force participation (14) and labor force participation by education grade (15). But education and labor force participation are insufficient to measure the potential for the working population to support dependents. Productivity-weighted labor force dependency ratios have been estimated using education and wages by education grade as productivity proxies (16). Most recently, measures of educational quality using literacy data for countries of the world have been estimated and used to generate a “skills in literacy adjusted mean years of schooling” or SLAMYS (2).
National Homogeneity versus Subnational Heterogeneity
The China paper (1) considers the country as one homogenous territory. However, there is ample evidence of strong quality differences between schooling in rural and urban areas, through a collaborative project involving American and Chinese researchers (17). They have identified both the causative factors and solutions. However, in China, the locality registration system, or Hukou, generates further inequality by denying the children of urban in-migrants to cities the same access to schooling and welfare benefits as holders of urban Hukou. If the model were rerun using either a rural/urban or rural Hukou/urban Hukou classification or a combination, would the results for China be less favorable? We don’t know. But a study of India which used Indian states, a rural/urban classification of settlements, and educational grades for a set of fixed assumption projections found a close agreement between the results of a national and a triply classified projection (18). Subnational inequality was closely linked to inequalities in educational attainment.
Policy Implications: Don’t Panic
Table 4 points to policy implications of the China study (1) for the national government in planning for demographic aging. Indicators of education combining time in school and quality of outcome should be used in projection models where possible. The main message is not to panic and impose draconian measures which affect family life, as happened under the one-child fertility policy. China should invest in improving the educational quality of rural populations and of rural Hukou migrants in cities. National projections should be supplemented by subnational projections capturing heterogeneity in population attributes to confirm the India result that inequality in education across the country largely accounts for differences between national and subnational projection outcomes. This might also resolve the conflict between top-down and bottom-up approaches to country projections. In top-down projections, the norm in national work, subnational projections are adjusted to agree with the national projections. In bottom-up projections, national projections are the sum of subnational results. This is the approach of the Population Division of the United Nations in generating successive revisions of World Population Prospects.
Acknowledgments
My research is supported by a Visiting Fellowship at the University of Leeds, which provides access to library electronic resources.
Footnotes
The author declares no competing interest.
See companion article, “China's low fertility may not hinder future prosperity,” 10.1073/pnas.2108900118.
References
- 1. Marois G., Gietel-Basten S., Lutz W., China’s low fertility may not hinder future prospects. Proc. Natl. Acad. Sci. U.S.A. 118, e2108900118 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Lutz W., et al., Skills-adjusted human capital shows rising global gap. Proc. Natl. Acad. Sci. U.S.A. 118, e2015826118 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Lutz W., Sanderson W., Scherbov S., Eds., The End of World Population Growth in the 21st Century: New Challenges for Human Capital Formation & Sustainable Development (Earthscan, London, United Kingdom, 2004). [Google Scholar]
- 4. Lutz W., Skirbeck V., “How education drives demography and knowledge informs projections” in World Population and Human Capital in the Twenty-First Century, Lutz W., Butz W., KC S., Eds. (Oxford University Press, Oxford, United Kingdom, 2014), pp. 14–38. [Google Scholar]
- 5. Lutz W., Butz W., KC S., Eds. World Population and Human Capital in the Twenty-First Century (Oxford University Press, Oxford, United Kingdom, 2014). [Google Scholar]
- 6. Basten S., Sobotka T., Zeman K., “Future fertility in low fertility countries” in World Population and Human Capital in the Twenty-First Century, Lutz W., Butz W., KC S., Eds. (Oxford University Press, Oxford, United Kingdom, 2014), pp. 39−146. [Google Scholar]
- 7. Caselli G., Drefahl S., Wegner-Siegmundt C., Luy M., “ Future mortality in low mortality countries ” in World Population and Human Capital in the Twenty-First Century, Lutz W., Butz W., KC S., Eds. (Oxford University Press, Oxford, United Kingdom, 2014), pp. 226−272. [Google Scholar]
- 8. Garberro A., Pamur E., “Future mortality in high mortality countries” in World Population and Human Capital in the Twenty-first Century, Lutz W., Butz W., KC S., Eds. (Oxford University Press, Oxford, United Kingdom, 2014), pp. 273−332. [Google Scholar]
- 9. Bernard A., Bell M., Educational selectivity of internal migrants: A global assessment. Demogr. Res. 39, 835–854. (2018). [Google Scholar]
- 10. Lutz W., Demographic metabolism: A predictive theory of socioeconomic change. Popul. Dev. Rev. 38, 283–301 (2013). [Google Scholar]
- 11. Rees P., Education and demography: A review of world population and human capital in the 21st century. Vienna Yearb. Popul. Res. 16, 37–53 (2018). [Google Scholar]
- 12. Rogers A., Introduction to Multiregional Mathematical Demography (Wiley-Interscience, New York, NY, 1975). [Google Scholar]
- 13. Goujon A., et al., A harmonized dataset on global educational attainment between 1970 and 2060 – An analytical window into recent trends and future prospects in human capital development. J. Demogr. Economics 82, 315–363 (2016). [Google Scholar]
- 14. Kupiszewski M., Ed., International Migration and the Future of Populations and Labour Force Resources in Europe (Springer, Dordrecht, The Netherlands, 2013). [Google Scholar]
- 15. Loichinger E., Labor force projections up to 2053 for 26 EU countries, by age, sex, and highest level of educational attainment. Demogr. Res. 32, 443–486 (2014). [Google Scholar]
- 16. Marois G., Bélanger A., Lutz W., Population aging, migration, and productivity in Europe. Proc. Natl. Acad. Sci. U.S.A. 117, 7690–7695 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Rozelle S., Hell N., Invisible China: How the Urban-Rural Divide Threatens China’s Rise (The University of Chicago Press, Chicago, IL, 2020). [Google Scholar]
- 18. KC S., Wurzer M., Speringer M., Lutz W., Future population and human capital in heterogeneous India. Proc. Natl. Acad. Sci. U.S.A. 115, 8328–8333 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]