Applications
Probabilistic Population Projections by 13 World Regions

The key element of the probabilistic approach is to make projections that are not just simple numbers, but a distribution of possibilities (probabilities) at future dates. Three types of applications are briefly summarized here:

• the probabilistic analysis of age structures,
• probabilistic forecasting,
• learning through the passage of time, and
• conclusions

The probabilistic analysis of age structures

One of the main motives behind these analyses is, that the population of the world and of each of the regions will be growing significantly older. Policymakers need, for instance, to know the likelihood that their pension systems will remain viable in the future. This depends not only on the number and age distribution of future pensioners, but also on the age distribution of the contributors. Due to the uncertainties in the future fertility, mortality and migration paths, the forecasts generate patterns of uncertainty in the ratio of pensioners to contributors (in other words, the likelihood), which are impossible to forecast without probabilistic population projections.

THe following aspects of aging are discussed comprehensively in the book The End of World Population Growth in the 21st Century: New Challenges for Human Capital Formation and Sustainable Development (Lutz, Sanderson, and Scherbov, eds.):

• proportions of the population age 60 and above,
• proportions of the population age 80 and above,
• the ratio of the population 60 and above to those 20 to 59 (old dependency ratio),
• the proportion of the population less than 20 years,
• the ratios of the number of people above 60 years and above 80 years to the number of people less than 20 years (dependency ratio), and
• the average age of the voting population.

Probabilistic forecasting

In the past, demographers discussed the uncertainties inherent in their population projections by using scenarios. Scenario analysis is still common in many different disciplines. Scenarios can illustrate the laws of population dynamics but cannot provide any information to the user on the likelihood of the described path. Policymakers are interested in the answers to different potential development paths. For instance, what would be the long-term consequences of alternative fertility trends resulting from alternative policies and scenario based projections? The probabilistic population projections integrate the scenario into the broader context by focusing on certain sub-ranges of future fertility, while leaving mortality and migration unconstrained. It is a way of posing and answering those types of questions within a probabilistic framework.

The figure shows the distribution of the world's population in 2050 conditional on average fertility and mortality level over the period 2000-2050. The x-axis is divided into three ranges: "low fertility" includes all 2000 simulated futures where the average total fertility rate was below 1.6, "medium fertility" was between 1.6 and 1.8, and "high fertility" was above 1.8. In each of the panels there are three lines that have different symbols near their centers. The lines with the diamond near the center refer to life expectancy below 68 years; those with the square, life expectancy between 68 and 71 years; and those with the triangle, life expectancy above 71 years. The symbols are placed at the medians, and the circles at the end points of the lines, indicating the interdecile ranges.

Now we can answer "what-if"-type questions. We can immediately read from the figure that the difference in future fertility is very significant combined with the medium range of uncertainty for future mortality. For instance, in the middle group, the median population of the world in 2050, if we experience low fertility, would be around 7.7 billion people, with the interdecile range covering the area from 7.0 to 8.3 billion people. If we experience high fertility, the median population would be about 10.0 billion people (about 2.3 billion people higher), with an interdecile range between 9.0 and 10.9 billion people.

More examples are discussed in the book The End of World Population Growth in the 21st Century: New Challenges for Human Capital Formation and Sustainable Development (Lutz, Sanderson, and Scherbov, eds.)

Sanderson, W., Scherbov, S., O’Neill, B.C., and Lutz, W. 2003. Probabilistic Population Forecasting. Laxenburg, Austria: IIASA, IR-03-052 (October 2003).

Passive learning

...or in other words, learning with the passage of time. In the future, population foecasts may be different from those made today, because of what we observe between then and now. We can predict what our future forecasts will be, conditional on those observations. We learn through the passage of time, and what we learn will change the way we think about the future. This is important for the decision makers to decide whether to take actions now or to wait to learn more through the passage of time. To answer the questions, we need probabilistic projections with future jump-off dates. These are conditional on what happens between the beginning of the current forecast period and the future jump-off date.

As time resolves uncertainties that lead to a new vision of the future, new forecasts make the old ones appear to be wrong, because the forecasts were of particular numbers and not of distributions. It is inevitable that a probability forecast for the year 2050 that is made in 2010 will have a different median value than one made in 2000, even though the earlier projections were probabilistically correct.

More details in the book The End of World Population Growth in the 21st Century: New Challenges for Human Capital Formation and Sustainable Development (Lutz, Sanderson, and Scherbov, eds.)

Conclusions

All these are vital for decision making. Knowledge on uncertainties associated with our forecasts of age structures helps policymakers to decide on appropriate policies. The forecast helps them to understand what the potential effects of their policies could be. And learning helps to integrate the time path of policy formulation with an understanding of what new forecasts are likely to show during that interval.

Responsible for this page: Isolde y
Last updated: 27 Jul 2004