64 Terence H. Hull and Wendy Hartanto
poses a statistical contradiction. Was there a problem with the DHS attempt to record accurate pregnancy histories, or did the two surveys suffer from unreliable
sampling? To answer this we need to look at the way the DHS selects the house- holds and individuals to be included in the survey.
In the DHS it is assumed that single women are not sexually active and have produced no children. However, to calculate standard demographic measures of
fertility the survey needs to record all women in the population irrespective of their marital status. This is because the denominator of the fertility rate is the total number
of all women of the relevant age group, while the numerator is the total number of births produced by mothers. If the DHS is justifi ed in assuming that unmarried
women do not have any births then no harm is done to the numerator by interview- ing only ever-married women. Nonetheless, it is still important to have an accurate
enumeration of all women, including single women, for the denominator.
The DHS household listing is the tool used to collect the data for the estimate of the total population of women and men in the sample households. Table 1 reveals
that the DHS household listings consistently show lower proportions of single women than do census, Supas and Susenas enumerations taken at around the
same time, particularly for the ages from 20 through 34, the peak years of repro- duction for Indonesian women. What explains the apparent lack of single women
in the DHS listings? In part it appears that there is a major difference in the type of household covered by DHS and census-type surveys. Why should this occur?
Essentially, both the Susenas and the DHS surveys are meant to include only regular households rumah tangga biasa, and interviewers should not visit ‘insti-
tutional housing’ rumah tangga khusus such as prisons, dormitories and barrack facilities or religious group housing such as convents or schools. This should have
an equal effect on both surveys if interviewers for all national surveys follow the sampling instructions carefully.
TABLE 1 Proportions Single for Women of Reproductive Ages in Successive National Surveys in Indonesia
a
Age Group Supas
DHS Census
Susenas DHS
Supas DHS
1995 1997
2000 2002
2002–03 2005
2007
Percentage of women in the age group who are single 15–19
85.7 82.1
89.3 89.7
85.4 90.8
86.9 20–24
40.1 36.1
43.1 47.0
41.2 51.4
38.3 25–29
15.2 14.1
16.7 16.3
13.8 19.7
15.4 30–34
5.5 5.3
6.9 6.5
5.9 8.1
7.0 35–39
2.8 2.4
3.5 2.9
3.0 4.3
3.6 40–44
2.1 2.9
2.4 2.1
2.1 2.6
2.6 45–49
1.9 1.7
2.0 1.4
2.0 2.0
1.9
All WRA 27.7
25.3 28.7
27.6 25.0
28.8 23.7
a
WRA = women of reproductive age. See footnote 1 for explanation of other terms. Source:
Calculated from the Measure DHS STATCompiler at http:www.statcompiler.com and unpublished data from BPS–Statistics Indonesia.
Downloaded by [Universitas Maritim Raja Ali Haji] at 19:36 18 January 2016
Resolving contradictions in Indonesian fertility estimates 65
Since the 1980s Indonesia has undergone a remarkable change in the roles young women perform in society. They are increasingly likely to pursue educa-
tion to higher levels, to work in expanding industrial and service occupations or to join the over four million Indonesian workers who are employed overseas and
sending remittances home though this group is not included in any census or survey enumeration. Single women in Indonesia often live in institutional set-
tings. These women would thus not be enumerated unless they are mistakenly listed in their parents’ households.
What is more important is that single women also form what might be regarded as ‘non-standard’ regular households even if they are not in recognised institu-
tional settings. Thus young people may be listed in a household in the BPS cen- sus listing, but they may live under very crowded or unusual circumstances, and
the household may possess none of the characteristics regarded as features of ‘normal’ family life. Perhaps the best known of these household types are the
indekos
or boarding houses that proliferate throughout urban areas. Anecdotal evidence from interviewers indicates that these households are sometimes passed
over in the DHS canvassing because fi eld-workers concentrate on units that are more likely to yield eligible respondents, and they know that the clusters of stu-
dents or young workers are unlikely to be ever-married. In contrast, the decennial census enumeration attempts to include all individuals and all households, and
the Supas makes special efforts to cover both family and non-family households, often with a particular interest in workers and students.
DHS analysts sometimes justify the lack of attention to the conditions of single women by using the assumption that unmarried women do not generally have
sexual relationships and, even if they do fall pregnant, this is likely to lead to mar- riage in a very short period of time. Sociological and anthropological research,
as well as simple observation of social trends, indicates that such an assumption has always been naïve, and recently has become completely misleading. With an
estimate of over one million induced abortions annually, of which some 20–60 occur among the unmarried Utomo et al. 2001: 8–9, 23, 27, it is vital for the DHS
to ensure that all eligible single women are included in the sample. Moreover, the rising age at marriage means that an increasing proportion of women aged 20–34
are single, and they are also moving out of their parental homes and shedding many of the strictures that apply to adolescents. It is crucial that these women be
included in any analysis of reproductive and sexual health.
In short, it is evident that the DHS has missed many young single women dur- ing the household listing, with the result that the denominators used to calculate
fertility are under-estimated and fertility is over-stated. The problem of listing single women and men began in earnest in the 1990s, as massive transforma-
tions of education, occupation and living arrangements took place. This problem has grown steadily through to the present day. It is evident from comparison of
census-type surveys with the DHS, but that does not mean that the census is abso- lutely correct. In fact it is likely that Indonesia, in common even with Australia
and the US, has an under-count of young men and women because of the social and geographic mobility of this age group. Nonetheless, if we assume that the
DHS should at least include those young people who were counted in the census- type surveys, then we have a chance of adjusting the fertility rates to approach a
more realistic level. Downloaded by [Universitas Maritim Raja Ali Haji] at 19:36 18 January 2016
66 Terence H. Hull and Wendy Hartanto
RETURNING THE MISSING WOMEN TO THE DHS SAMPLE POPULATION
The adjustment of DHS fertility rates is a two-step process. First, the data in table 1 can be used to estimate the number of single women missing from the
DHS sample compared to the expected number if the DHS had the same propor- tion single as was found by recent census-type surveys. Second, once those single
women are added to the total number of women in the DHS households, the fertility rates can be re-calculated with new denominators. The calculation of the
number of missing single women is based on the following logic. The proportion single in the DHS ds in each age group can be represented as DsDw single
women in the DHS divided by all women in the DHS for each age group. Then we calculate the proportion single in the DHS population if all the missing single
women represented as M in the equations are restored to both the numerator and the denominator to achieve the same proportion single as was found in the
recent census enumeration cs. In order to solve for M the missing women, we derive the following equations:
cs = Ds + MDw + M Ds + M = cs x Dw+ cs x M
M – cs x M = cs x Dw – Ds M1 – cs = cs x Dw – Ds
M = [cs Dw – Ds]1 – cs
This calculation is shown for the two most recent DHS surveys in tables 2 and 3. TABLE 2 Estimation of Total Number of Women Missing
from the 2002–03 DHS Sample
Age Group 2002–03
DHS Numbers
Recorded by Age
Group 2002–03
DHS Single
Recorded by Age
Group 2002–03
DHS Proportion
Single in Age
Group 2000
Census Proportion
Single in Age
Group
Estimate of
Missing Women
Adjusted Total
2002–03 DHS
Women Dw
Ds ds
cs M
Dw’ 15–19
6,715 5,735
0.8540 0.8927
2,423 9,138
20–24 6,738
2,776 0.4120
0.4312 227
6,965 25–29
6,302 870
0.1380 0.1667
217 6,519
30–34 5,844
345 0.0590
0.0695 66
5,910 35–39
5,349 160
0.0300 0.0349
27 5,376
40–44 4,704
99 0.0210
0.0241 15
4,719 45–49
4,170 83
0.0200 0.0198
–1 4,169
All WRA 39,822
10,068 0.2500
0.2870 2,974
42,796
Source: Authors’ calculations.
Downloaded by [Universitas Maritim Raja Ali Haji] at 19:36 18 January 2016
Resolving contradictions in Indonesian fertility estimates 67
Table 2 shows that the 2002–03 DHS failed to record about 23 of the single women who should have been in the sample the 10,068 recorded plus the 2,974
missing mean the sample should have contained 13,042 single women. Thus 7 of the total sample was missed. In 2007 over one-third of single women were skipped,
representing 11 of the total sample who should have been listed in the DHS households table 3. In both surveys the bulk of the missing women were in the age
groups related to senior high school and university. Coincidentally, this is the age at which many young women are working in manufacturing and service industries.
ADJUSTING THE FERTILITY RATES FOR MISSING SINGLE WOMEN The census-based estimate of missing women allows the reconstruction of age-
specifi c and total fertility rates for the 2002–03 and 2007 DHS. According to the DHS main report for 2002–03 BPS and ORC Macro 2003, the method used for
calculating fertility rates indicates that:
Numerators of the ASFRs [age-specifi c fertility rates] are calculated by summing the number of live births that occurred in the period 1 to 36 months preceding the
survey determined by the date of interview and the date of birth of the child and classifying them by the age in fi ve-year groups of the mother at the time of birth
determined by the mother’s date of birth. The denominators of the rates are the number of woman-years lived in each of the specifi ed fi ve-year groups during the 1
to 36 months preceding the survey. Since only women who had ever married were interviewed in the IDHS [Indonesian DHS], the numbers of women in the denominators
of the rates were infl ated by factors calculated from information in the Household Question- naire on populations ever married in order to produce a count of all women. Never-married
women are presumed not to have given birth
BPS and ORC Macro 2003: 43 [emphasis added].
TABLE 3 Estimation of Total Number of Women Missing from the 2007 DHS Sample
Age Group 2007 DHS
Numbers Recorded
by Age Group
2007 DHS Single
Recorded by Age
Group 2007 DHS
Proportion Single
in Age Group
2005 Supas Proportion
Single in Age
Group
Estimate of
Missing Women
Adjusted Total
2007 DHS Women
Dw Ds
ds cs
M Dw’
15–19 6,849
5,949 0.8686
0.9080 2,936
9,786 20–24
7,040 2,693
0.3825 0.5142
1,908 8,948
25–29 7,156
1,099 0.1535
0.1974 391
7,548 30–34
6,730 468
0.0695 0.0810
84 6,814
35–39 6,473
235 0.0364
0.0431 45
6,518 40–44
5,722 148
0.0259 0.0255
–2 5,720
45–49 5,127
96 0.0188
0.0197 5
5,132
All WRA 45,098
10,689 0.2370
0.2879 5,368
50,466
Source: Authors’ calculations.
Downloaded by [Universitas Maritim Raja Ali Haji] at 19:36 18 January 2016
68 Terence H. Hull and Wendy Hartanto
In tables 4 and 5 the published ASFRs age-specifi c fertility rates and the cal- culated numbers of women recorded in the DHS household questionnaire are
used to estimate the annual number of births for all women in 2002, assuming no decline in fertility over the period 2000–02. Then the annual fertility rates are
re-calculated using the adjusted numbers of women who should have been listed in the DHS household questionnaire if the 2000 census marriage patterns had
prevailed for the 2002–03 DHS.
Where the 2002–03 DHS Main Report showed a TFR of 2.57, adjusting the fer- tility rate for missing single women produces a TFR of 2.35 for the three-year
