# Standards relationnelles utilisés

## Introduction

A number of methods presented here rely on relational models of fertility or mortality for their implementation. To enable others to reproduce the standards used in these methods, we describe briefly the standards used, and their construction.

## Mortality standards

In order to harmonise and standardise the mortality bases used in the preparation of this manual, all methods presented that rely on tabulated values from standard life tables draw on a common subset of these life tables.

In particular, the standard life tables used are those from the Princeton Regional Model Life Tables (North, South, East and West) and the UN Model Life Tables for Developing Countries (General, Latin America, Chile, South Asia, Far East), by sex, all with an expectation of life at birth of 60 years. The original life tables have been modified, extended and enhanced over time to extend them to older ages. We make use of the updated tables provided by the United Nations Population Division, and used by them in their population projections.

Standard life tables based on the 18 June 2010 revision, by sex with an expectation of life at birth of 60 years, for each of the nine standards above were used. These life tables provide values of lx and Lx (amongst other quantities) for ages 0, 1, 5, 10, ..., 130.

Values of lx for ages 2, 3 and 4 were derived as follows:

• For the Princeton Regional Model Life Tables, the proportionality factors presented by Coale, Demeny and Vaughan (1983: 21) were applied to l1 and l5 to generate l2 , l3 and l4. For the UN Model Life Tables for Developing Countries, deaths between the ages of 1 and 5 were distributed by single years of age in the same proportion as those deaths in the original sex- and region-specific life tables.

Joint-sex life tables (that is, for males and females combined) are required by some methods of child and adult mortality estimation. As these life tables (or their implementation) is not sensitive to the sex ratio at birth, a sex ratio at birth of 105 (boys per 100 girls) was used. Joint-sex life tables were then derived by appropriate weighting of the sex-specific life tables:

$l x c = (1.05) l x m + l x f 2.05$

where

$\text{\hspace{0.17em}}{l}_{x}^{c}\text{\hspace{0.17em}}$

represents the number of survivors at age x in the joint-sex life table and

$\text{\hspace{0.17em}}{l}_{x}^{m}\text{\hspace{0.17em}}$

and

$\text{\hspace{0.17em}}{l}_{x}^{f}\text{\hspace{0.17em}}$

are the equivalent life table values for men and women respectively.

As these life tables are used almost exclusively in a relational context (as originally set out by Brass (1971)), standard logits of the lx were calculated for all ages above zero by means of the formula

$Y x s =0.5ln( 1− l x l x )$

## Fertility standard

Methods based on the relational Gompertz model draw on a slightly modified version of the standard produced by Zaba (1981). Zaba's standard is the same at all ages older than 15 to that produced by Booth (1984). At younger ages, however, Zaba observed that the rate of change in the fertility schedule did not match that found in empirical observations. The standard applied here uses Zaba's tabulated values, but makes a small correction for how cumulated fertility, F, is derived at half-ages, which are required to accommodate the usual shift in tabulations of fertility data by age of mother. Zaba's approach averaged values of F(x) and F(x+1) to approximate F(x+½). However, since the purpose and the effect of the gompit transform, Y(x) = -ln(-ln(F(x)), is to linearise a curvilinear function, it makes greater sense to interpolate Y(x+½) from successive values of Y(x) and Y(x+1) and to then take the antigompit of Y(x+½) to derive F(x+½).