Page 94 - IJPS-10-4

International Journal of
Population Studies Mortality shapes population age structures

aging. Here we... consider this possibility, but reject it.” has started to emerge, occasionally becoming prevalent,
(Lee & Zhou, 2017, p.285-287) especially where the survival rate is high (Murphy, 2017;
Ignoring migration, which is indeed a minor force in Preston et al., 1989; Preston & Stokes, 2012). In all cases,
this case, population aging may depend primarily on low migration was consistently found to be irrelevant.
fertility or low mortality. To assess which is the case, two We join the debate on the determinants of age
main types of analysis have been proposed in the literature: structures advancing the hypothesis that survival is the
(i) simulations/counterfactuals and (ii) empirical main driving force at play, not just in recent times or
(decomposition) analysis. Simulations and counterfactuals, developed countries, and that recent survival suffices to
introduced by Coale in 1956 and subsequently adopted by “explain” most of the dependent variable, the observed age
various scholars such as Bengtsson & Scott (2005; 2010), distribution, at any point in time. To test our hypothesis,
remain a valid, possibly even the best approach to the topic we compare two age structures, the observed and the
(Lee & Zhou, 2017). “standard” one, the latter being the age structure of the
Simulations are typically done in four steps: (i) A stationary population associated with recent survival
population of the past is selected as a starting point conditions. The comparison is conducted systematically
(e.g., Sweden in 1860); (ii) one of its two demographic for all the countries and periods for which reliable data
“behaviors” (e.g., fertility) is artificially kept constant, are available. In Section 1.1, we outline our hypotheses
while the other (mortality) is, also artificially, forced to in detail and the indicators used to corroborate them.
follow its observed historical path; (iii) after some time, In Section 2, we specify our sources and describe how
three age structures are compared: the two observed ones, we treated the data therein. In Section 3, we present our
at the start and the end of the simulation period, and the results, beginning with the case of China, considering
hypothetical final one, obtained through simulation; and all world countries over a relatively short period (1951
(iv) based on the distance between these age structures, – 2021), and subsequently, we focus on a small group
conclusions are drawn on the relative structural impact of of countries with long time series data dating back to
the two processes, fertility and mortality. The conclusion 1860. Section 4 discusses the theoretical and practical
invariably drawn is that “equivalent changes” in fertility implications of our findings. Section 5 is devoted to the
and mortality (equivalent in terms of the resulting growth conclusions that can be derived from this study.
rate, for instance) do not affect age structures in the same
way: the former weighs considerably more than the latter. 1.1. Hypothesis and indicators
While this line of reasoning is correct to answer questions Our hypothesis is that mortality (its level, not its variation)
of the type “if … then …”, it does not necessarily lead to is the prevailing force that shapes population age structures.
valid conclusions when it comes to interpreting reality This hypothesis derives from the classical interpretation of
because it assumes that the conditions grouped in the the demographic transition, according to which the old
premise (the “if”) may in fact materialize, which may be demographic equilibrium (ancien régime), characterized
questionable (e.g., Fernandes et al., 2023; Murphy, 2017; by high birth and death rates, crumbled under the pressure
2021). In addition, the outcomes of counterfactuals and of declining mortality, a process that started towards the
simulations depend heavily on the initial conditions, that end of the 18 century in Europe (Davis, 1963; Dyson,
th
is, on the starting date, which is arbitrary (Murphy, 2017). 2010; Kirk, 1996). In this interpretation, mortality is the
One alternative to simulations is the decomposition leading force in the complex interplay of demographic
analysis launched by Preston et al. (1989) and later applied variables; everything else adapts more or less rapidly to
by several other scholars (e.g., Caselli & Vallin, 1990; the constraints imposed by changing survival conditions.
Fernandes et al., 2023; Murphy, 2017; Preston & Stokes, If this is true, mortality should also “explain” the age
2012). In more recent studies, the typical steps are as follows: structure.
(i) selecting a synthetic indicator of the age structure, often To test this expectation, (i) we ignore fertility and
the average age, A; (ii) observing its changes over time migration; (ii) we focus exclusively on mortality (or
t
alongside fertility, mortality, and migration indicators better, on its complement, survival, and in particular on
(birth, death, and mortality rates, respectively); and (iii) the years of life lived, or L series, of a life table); (iii) we
x
evaluating factors (fertility, mortality, or migration) with further restrict our attention to recent survival (i.e., we
the greatest influence on the variation of A. The typical consider only a life table calculated either in the same
t
conclusion of these studies is nuanced (Fernandes et al., year t, in which the population age structure is observed
2023): the influence of declining fertility was stronger in the or in a preceding year, as close to t as possible); and
past but lesser in recent times, when the role of mortality (iv) we measure what share of the observed age structure

Volume 10 Issue 4 (2024) 88 https://doi.org/10.36922/ijps.377