IIASA/VID Educational Attainment Model

The World Population Program of the International Institute for Applied Systems Analysis (IIASA) in Laxenburg, Austria and the Vienna Institute of Demography, Austrian Academy of Sciences developed the IIASA/VID educational attainment model for reconstruction of educational attainment distributions by age and sex from 1970 to 2000 and projections through alternative scenarios to 2050 (the results of this last model will be available shortly).

  1. What is the central question that the model answers?
  2. What is the primary use for the model?
  3. What are the data needs?
  4. What does the model do well in comparison to other models, and what are its main limitations?
  5. What are the key assumptions?
  6. What parameters can be changed?
  7. Are the policy implications clear?
  8. How could this model link to other projection models?

a. What is the central question that the model answers?

The central question is how the composition of human capital, approximated as the levels of educational attainment of the working population by gender and five-year age groups, has developed between 1970-2000 and will further evolve from 2000 to 2050 depending on a set of scenarios on schooling achievements. The two models, Reconstruction (model A) and Projections (model B), are separate although they both start from the same empirical input data, namely, the absolute population divided by five-year age groups, by sex, and by levels of educational attainment for the most recent date.

The Projections model included in the EdStats projection database basically translates present trends in schooling achievements into future levels of educational attainment for the working-age population. This translation from schooling (flow) to human capital (stock) is important because the stock of educational attainment of a population is the determinant of many of the desired effects of education. It is also essential to two important determinants of human wellbeing, namely, population and economic growth, mostly through the influence of education on fertility, mortality, and wealth at the micro and macro levels. The model aims primarily at showing the combined momentum of education growth from the increase in school enrolment to improving educational attainment of the working-age population.

b. What is the primary use for the model (initial diagnosis and planning, detailed programme development, budget support)?

The primary use is to show the combined mechanisms of population and education momentum considering the full composition of the population by age, sex, and levels of educational attainment. The education scenarios are not to be interpreted as predictions or forecasts, but as exercises in ‘what if’ reasoning. As such they serve the important purpose of illustrating the consequences of different kinds of trends and policy environments on global human capital.

The model is designed as a scientific tool for projections but can clearly be used as an advocacy tool to show the long-term implications of improvements in schooling at the level of countries (or other geographical or administrative units). Up to now, projections have been carried out for 120 countries according to four scenarios.  IIASA/VIS plans to increase its geographical coverage (to 150 countries) as well as to apply the model to sub-national regions for large states (like China and India).

c. What are the data needs?

The projections require five types of input: four are standard to cohort-component projections and one is specific to multi-state models.

  • Base-year data on population by age, sex, and education. This is mostly available from censuses and surveys. At the moment, the Projections model is set for four broad education categories (no education, primary, secondary and tertiary) but it could be changed to include any levels of detail.
  • Base-year fertility by age and education and assumptions up to the end of the projection period (usually 2050). Fertility data are usually available through censuses and surveys (DHS and others).
  • Base-year mortality by age, sex, education and assumptions up to the end of the projection period. Although information on educational mortality differentials is not readily available for most countries and generally is much more difficult to obtain than fertility differentials, IIASA/VIS has implemented some differentials resulting from the observation of a large number of countries for which IIASA/VIS has information.
  • Base-year net number of migrants by age, sex, and education. Migration assumptions are the most difficult to estimate because migration flows tend to be volatile. However, since past data are only of very limited use in defining migration assumptions, this is not much of an additional obstacle. In practice, age-sex-education distributions of the populations of sending countries were pooled for each period. IIASA/VIS then calculated the difference in the population distribution by age at the end of the period between the UN projection and our projection (aggregated). A positive difference means positive net migration. Age-specific negative net migrants are distributed (subtracted) proportionally to the education distribution in the age group. In the event of age-specific positive net migration, the distribution from the pooled distribution is used to distribute the positive net migrants to the four education categories. In the case of 0 in a category, the net migrants are added to the next non-zero higher category.
  • The transition between education categories is one input that is specific to multi-state population projections. It is the age-specific probabilities for young men or women to move, e.g., from the category of primary educational attainment to secondary attainment. The set of necessary transition rates is very limited because education is hierarchical, e.g., persons can only move to the next higher category. Transitions also occur over a short period of time and only for a few age groups that are in school (usually from 5 to at most 24 years of age), as the phenomenon of adult education is rather minimal. Since data that would allow for the calculation of precise transition rates is not readily available, they need to be approximated from the educational attainment at each age and level, translating cohort data in age groups 5-9, 10-14, 15-19, 20-24 into period data.

d. What does the model do well in comparison to other models, and what are its main limitations?

Its main strength is that it takes full advantage of the demographic nature of improvements of education along cohort lines, i.e., people typically are educated at a young age and then maintain their highest educational attainment throughout life. It also explicitly considered the often-disregarded fact that fertility and mortality rates vary considerably with the level of education. It is very flexible in terms of scenario making, or adjusting the specificities of education categories. Its main advantage compared to other education modeling tools is that it gives the big picture in terms of educational attainment of the population.

The main drawback of the IIASA/VID model is that it does not attempt to model the schooling processes themselves but rather summarizes them in terms of their outcomes that lead to a specific educational attainment distribution.

e. What are the key assumptions?

The notion that we can avoid making assumptions about future educational attainment trends is a fallacy: Since fertility is influenced by education levels, population projections inevitably imply assumptions about the population’s future educational attainment, even if these remain unstated. In our view, it is preferable to be explicit about these assumptions.

The main assertion that the dynamics of improvements in educational attainment by age and sex follow the laws of demography is not an assumption but part of the mathematical laws of population dynamics. Assumptions only concern the right choice of fertility, mortality and migration patterns as well as the patterns of educational transitions by age and sex. If the right (and for the future still unknown) parameters are assumed for transitions that are specific for each educational group, then by definition, the projected or reconstructed distributions must be correct.

f. What parameters can be changed?

As mentioned above the model is quite flexible and all the key parameters can be changed. The number of educational attainment categories is also flexible.

g. Are the policy implications clear?

The Projections model provides levels of educational attainment for the whole population by age and sex. This provides a history and future of the consequences of specific education policies. Hence, the model usually points at the obvious necessity to invest more in education, since any lack of investment will have implications for the future that will be visible through the education/age/structure of the population.

h. How could this model link to other projection models?

IIASA/VID model can be linked to any of the other models with more detail on the schooling processes. Such joint models could translate specific aspects of schooling policies into alternative enrolment, educational transitions and ultimately attainment distributions.