|
One
of the basic but vital information provided by a census is the age
profile of the population. It is therefore important to evaluate the
accuracy of the age distribution and correct for deficiencies wherever
necessary. Digit
preference or age heaping is a common error that occurs in age
reporting. It is the result of deliberate mis-statement or ignorance of
age on the part of the respondent. The figures 2.1, 2.2 and 2.3 give the
age and sex distribution of the Republic of Mauritius, Island of
Mauritius and the Island of Rodrigues. From the graph, it appears that
there is a slight preference for the digit 0 among both males and
females in the Island of Mauritius as well as in the Republic. For the
Island of Rodrigues, however, there seems to be a preference for the
digit 9. To
quantify the magnitude of digit preference, two indices have been used
and these are the Myer's Index and the Preference Pattern Index(PPI).
Both indices involve the calculation of the percentage of the
population, P(i) having age ending in i where i =0,1,2……9. In the
ideal situation where there is no digit preference, p(i) equals 10 for
each i. The
index is thus computed as the sum of absolute deviations of P(i) from
10, i.e
The
indices can therefore lie between 0 (no digit preference) and 180
(absolute digit preference). Table 2.10 gives the indices by sex for the
Republic of Mauritius, Island of Mauritius and the Island of Rodrigues.
The figures show no marked digit preference in age reporting at the
Census. Age seems to be more accurately reported among males than among
females in the Republic of Mauritius and the Island of Mauritius. A
comparison with the digit preference indices computed for the 1990
Census indicates that age reporting was better at the 2000 Census. Table
2.11 gives the most preferred digits at the 1990 and 2000 Censuses. The
table indicates that the preferred digit has shifted from 5 in 1990 to 2
in 2000 while the preference for 0 has persisted. Preference for digit 0
may be due to the tendency to round off ages to years ending in 0. The
preference for the digit 2 may be due to a combination of factors
namely: · Increases in the number of births in the periods 1947/48, 1957/58
and 1967/68. · The year 1968, which was the year of independence being used as a
reference year for calculating age for those born around that year. For
the Island of Rodrigues, preference for the digit 9 was recorded in 2000
as well as in 1990. Preference for digit 0 was also noted. Fig 2.1 - Population by age and sex, 2000 Census
Republic of Mauritius
Fig 2.2 - Population by age and sex, 2000 Census Island of Mauritius
Fig 2.3 - Population by age and sex, 2000 Census Island of Rodrigues
Fig 2.4 - Population pyramid by single year of age - Republic of Mauritius 2000 Census
Table 2.10 - Index for digit preference in age data - 2000 Census
Table
2.11 - Most preferred digits by sex - 1990 & 2000 Censuses
2.4.2 Vertical consistency checks for age-reporting error In
a population with no drastic changes in fertility, mortality and
migration and where
migration is either negligible or does not occur at selective age
groups, the percentage change in population in given ages is expected to
decrease smoothly with age. The
population pyramid at figure 2.4 shows that for ages 35 and above, there
is a smooth decrease with age. For the ages below 35 years, the changes
with age are irregular. This is mostly due to variations in number of
births over the last few decades. The bulge at around age 20 is the
result of high births around the year 1980 while the bulge around the
age 8 is due to high births recorded around year 1992.
Table
2.12 - Age ratio, sex ratio and U.N joint scores, 1990 and 2000 Censuses
- Republic of Mauritius
To
quantify the degree of age -sex reporting errors, various indices such
as the age ratio score, the sex ratio score and the UN joint score have
been calculated for the population aged under 70 years. The
age-ratio score is defined as the average of the sum of absolute
deviations of age-ratios from the expected 100. The age ratio is itself
defined as the ratio of the population of a given age to half the sum of
population in the preceding and following groups and expressed per 100. The
sex ratio score is defined as the average of the sum of absolute
deviations of consecutive sex ratios. The
UN age-sex accuracy index or UN joint scores combines indices on
accuracy of the age distribution and accuracy in sex ratio to come up
with a single index measuring the accuracy of the age-sex distribution
of a Census. The census data is considered to be "accurate",
"inaccurate " or "highly inaccurate" depending on
whether the index is under 20, 20 to 40 or over 40 respectively. For
the 2000 Census, the UN joint score works out to 18.3 thus indicating
that the Census data can be considered as accurate. A comparison of the
score for the 2000 Census with that for the 1990 Census shows that the
age and sex reporting was slightly better in 1990. 2.4.3 Diagonal consistency checks The quality of census data can also be assessed by the analysis of survival ratios. Cohort survival ratio is the ratio of the population of a particular age group at a given census to the population ten years younger in the preceding census (assuming that the two censuses are 10 years apart). The survival ratio is usually below unity unless there is massive in-migration. In a country with no migration, the survival ratio is expected to decrease smoothly with increasing age. However, the survival ratio for age group 5-9 years is normally higher than that for age group 0-4 years, due to higher mortality prevailing at the youngest ages. Chapter 2Table
2.13 - Intercensal cohort and overall survival ratios 1990-2000 by sex -
Republic of Mauritius
The
data in table 2.13 shows a decreasing pattern for ages 30 and above
while for younger age groups, the pattern is irregular. One plausible
explanation would be the effect of migration at these ages. The
overall survival ratios however decline continuously with age indicating
that the census data is of good quality. A comparison of overall
survival ratio by sex indicates higher survival ratios among females at
all age groups. This is expected since mortality is lower among females.
2.5
Adjustment for under-enumeration of young children In
order to assess the degree of under-enumeration at the 2000 Census, the
enumerated population has been compared with the expected population in
2000. As explained in section 2.3.3, the expected population has been
derived by surviving the 1990 census figures on the basis of data on
live births, deaths and migration for the inter-censal period July 1990
to June 2000. The
comparison exercise for the Island of Mauritius reveals that the
enumerated population in the older age groups is generally lower than
the expected population among males (table 2.9(a)) whereas the reverse
is true among females (table 2.9(b)). This may be due to the fact that
the sex ratio of migrants used in the population estimation was too low.
As regards the age bracket 0-9 years, the census figures are
systematically lower than expected, indicating possible
under-enumeration of young children. Given
the good quality of the data as assessed in the previous sections of the
report, it is not necessary to adjust the age data except for the age
group 0-9 years. The actual adjustments for that age group (table 2.14)
have been obtained as the difference between the expected and the
enumerated population; the rationale for this method being that vital
registration is complete in Mauritius and that migration has little
impact at the youngest ages since it is concentrated in the working age
group. Table
2.14 - Adjustment for under-enumeration of young children, 2000 census -
Island of Mauritius
As regards Rodrigues, an independent estimate of the population by single year of age and sex as at mid-2000 is not available. However, in order to assess the degree of under-enumeration of young children for Rodrigues the population aged under 10 years has been estimated by surviving live births occurred in Rodrigues during the period July 1990 to June 2000. It has been found that these estimates do not differ significantly from the census figures. Thus no adjustment has been made. |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
