Abstract:
This report assesses the quality and
consistency of age and date reports in DHS
surveys conducted since 2000 in 67 countries.
It is the most recent of several reports on
various aspects of DHS data quality. The
first chapter describes the steps of editing
and imputing during fieldwork and data
processing. Great
care is taken to train and supervise the
interviewers to obtain the best possible
estimates of ages and dates. The eligibility
of adults for the surveys of women and men
depends on obtaining accurate values of age
near the lower and upper age boundaries
within the household survey. The eligibility
of young children for the detailed health
questions depends on obtaining accurate
estimates of when they were born within the
surveys of women. Age-specific fertility
rates, under-five mortality rates,
immunization rates, anthropometry scores, and
many other DHS indicators depend on accurate
estimates of age.
An appendix provides an inventory of all the
locations in DHS surveys where the
respondents are asked for ages and dates.
This assessment focuses on just a few of
those locations: the ages of all household
members, provided by the household respondent
during the household survey; the self-
reported ages and birthdates of women and men
in the surveys of women and men; women’s
self-reports of age and date of first union
in the survey of women; the birthdates (and
ages, if living) of children in the birth
histories, provided by the mother; and the
women’s and men’s estimates of their
respective spouses’ ages in the surveys of
women
and men.
The second chapter assesses the ages listed
above, other than spousal estimates, in terms
of three types of measures: incompleteness,
heaping, and transfers. A total of 11
indicators are used. For each indicator, the
distribution across all surveys is described
and the surveys with the most extreme levels
are identified. All of these measures vary
substantially. There are many surveys with
values close to zero on all measures, and
others with very high values. There are some
surveys in which month of birth is hardly
ever given. Age/date transfers are sometimes
large but in a direction opposite to what we
would expect, particularly around age
15 or around the date for the health
questions, clearly as a result of over-
correction during training and supervision.
Surveys with extreme values are listed.
Summary indices of incompleteness, heaping,
and
transfers are constructed and tracked over
time. The indicators fluctuated substantially
from 2000 to 2015 and did not show a
systematic trend. A single composite index is
constructed for each of the 67 countries.
The countries in the highest quintile (with
the most problems) and the lowest quintile
(with the fewest problems) are identified.
The third chapter investigates the quality
and consistency of spousal age estimates
compared to self-reports in the surveys that
included interviews of men and where women
and/or men were asked to estimate the age of
their spouse(s). In the absence of an age gap
between spouses, women tend to estimate that
their husbands are older than their self-
reported age and men tend to estimate that
their wives are younger than their self-
reported age. Evidence indicates that where
there is an age difference between spouses,
women tend to estimate in a way that reduces
the gap: they underestimate the age of older
husbands and overestimate the age of younger
husbands. Men underestimate the age of wives
who are older than they are, which reduces
the gap, but they also tend to underestimate
the age of wives who are younger, which
increases the gap. In the vast majority of
countries, there was more heaping for
estimates of spouse’s age
than for self-reported age. There is evidence
that displacement and heaping, in particular,
can be reduced, through training and
supervision, but there is also evidence that
too much focus on displacement of children or
on heaping at final digit 0 can lead to over-
correction. The biggest determinant of good
age reporting is probably the
value, to the respondents, in everyday life,
of knowing their ages or the ages of their
children. This component of data quality
varies from one setting to another and is
outside the control of a survey operation.
An Assessment of the Quality and Consistency of Age and Date Reporting in DHS Surveys, 2000-2015 (PDF, 2733K)