Year Census, Supas, Susenas CPS and DHS pre-2000 DHS Retro DHS 2007 Retro

Transcription

1 levels and trends in Indonesia Over the last four decades Indonesia, like most countries in Asia, has undergone a major transition from high to low fertility. Where up to the 1970s had long born an average over five children, the pace childbearing has slowed since then through a combination delayed marriage and the increased use contraception to prolong the time between births and ultimately end childbearing with fewer children. The transformation in behaviours surrounding family formation is reflected in the sequence blue diamonds in Figure 1. This gives us a long term macroscopic view fertility trends from the beginning the family planning program through the most recent national surveys. Each blue diamond is an estimate fertility spanning a three to five year period centred on the point. This calculation is drawn from a comparison the number children between the ages 0 to 3 or 4 enumerated in the population, and the number childbearing ages (the Own Child method). The line is not linear but falls continuously through the turn the century, at which point the decline slowed near the so-called replacement level 2.1 children per woman. Figure 1 Indonesian fertility trends, TFR Year 1995 Census, Supas, Susenas CPS and pre-2000 Retro 2007 Retro

2 The fertility calculations the more detailed Contraceptive Prevalence Survey and the Demographic and Health Surveys (CPS and ) since the late 1980s do not follow the census trendlines. In early years they were below the own child estimates. Since 1995 the estimates total fertility have increasingly exceeded the fertility levels found in the census-type national surveys. Since the year 2000 the published estimates current fertility (three years prior to the survey) have been stagnant at 2.6 children per woman, well above replacement level and nearly half a child higher than the census estimates. enumerations focus on the experience ever-married between the ages 15 and 49, from whom complete histories pregnancies and births are obtained. The total number births in discrete periods time prior to the survey is obtained from these histories. Thus in Figure 1 the two most recent produce estimates fertility both for the four years prior to the survey and five to nine years retrospectively. In each case they show declining fertility, not a plateau. Table 1. Marital status distributions for reproductive ages in successive national surveys in Indonesia Age Group SUPAS Census SUPAS age in the age who are All WRA age in the age who are ever-married All WRA *Calculated from the Measure STATCompiler: and 2007 data provided by Statistics Indonesia. In the it is assumed that are not sexually active and have not produced any children. However, to calculate fertility rates the survey needs to record all in the population irrespective their marital status. This number is obtained from the household census compiled by interviewers when they first arrive at selected sample households. The census listing is the tool used to assess the number household members who are eligible for the various interview forms collected during the survey. Table 1 reveals that the household listings consistently show lower proportions compared to Census, SUPAS or 2

3 enumerations taken at around the same time, particularly for the ages from 20 through 29. What explains the apparent lack in the listings? In part it appears that there is a major difference in the type household covered by and census type surveys. Essentially, the interviewers are on the lookout for ever married and given the nature the survey they are particularly attuned to households with families. Since the 1980s Indonesia has undergone a remarkable change in the roles young perform in society. They are increasingly likely to pursue education to higher levels, to work in expanding industrial and service occupations, or join the over four million Indonesian workers who are employed overseas sending remittances home. Single ten live in institutional settings dormitories, industrial barracks, and boarding houses. Anecdotal evidence from interviewers indicates that these places are ten passed over in the canvassing because fieldworkers concentrate on households that are more likely to yield eligible respondents. In contrast census type enumerations are designed to include both family and non-family households, ten with particular interest in workers and students. Assessing the coverage listing The faces the challenge working out some way estimating the number missing from the household census and using that to recalculate the denominators fertility rates. Table 2 shows total number actually in the survey in column B. Column D shows the number who would have been listed if the had obtained the same proportion as found by the 2000 Population Census (column C). This calculation implies that there could have been as many as 4492 missing from the household listings. Table 2. Estimation total number in if reflecting 2000 Census marriage patterns by age percent in age 2000 Census in age Estimate sample if had Census prile all Difference -- missing A B C D=A*C/B A-D All WRA Of course there are some strong reasons to question the use a full count census carried out over two years earlier to test the coverage the While the census did not have to contend with sampling errors, it did have problems with non- 3

4 sampling issues. It was a mammoth operation with an army interviewers, many whom lacked experience and supervision. Thus the two years individual aging and population mobility plus the potential lower quality data collection in the census could undermine the confidence we have in applying the marital status the full count census data to the Importantly, though, the census enumeration served as the basis for the sampling frame for subsequent surveys. Both the and the National Social and Economic Survey () carried out a few months earlier in 2002 were constructed from identical sampling procedures, as indicated in the Final Report for the : The sample developed for the 2002 National Socioeconomic Survey (Susenas) was used as a frame for the selection the I sample. Household listing was done in all CBs covered in the 2002 Susenas. This eliminates the need to conduct a separate household listing for the I. (I Final Report, 2003: 267) The annual followes a census-type household enumeration including questions on the age, sex and marital status each resident. There should be no difference in the proportion who were in the two surveys. As Table 3 shows, there is still a very large difference nearly 1499 though in three age s, 35-39, and there were more in the than would have been expected from the marital status pattern the. Table 3. Estimation total number in if reflecting 2002 marriage patterns A B C D=A*C/B A-D by age percent in age 2002 percent in age Estimate sample if had prile all Difference - - missing All WRA Adjusting the fertility rates for missing These two calculations missing allow the reconstruction age specific and total fertility rates for the. In the Main Report the method used for calculating fertility rates indicates that: 4

5 Numerators the ASFRs are calculated by summing the number live births that occurred in the period 1 to 36 months preceding the survey (determined by the date interview and the date birth the child) and classifying them by the age (in five-year s) the mother at the time birth (determined by the mother s date birth). The denominators the rates are the number womanyears lived in each the specified five-year s during the 1 to 36 months preceding the survey. Since only who had ever married were interviewed in the I, the numbers in the denominators the rates were inflated by factors calculated from information in the Household Questionnaire on populations ever married in order to produce a count all. Nevermarried are presumed not to have given birth. (I Main Report, 2003:43) Ideally the analysis carried out by the would adjust the number in the Household Questionnaire to take account the missing and use that number to calculate the years lived in the period from That calculation requires more information than is currently available, so in this paper a number simple assumptions are made to generate a close approximation to the adjustment needed to calculate fertility. In Table 4 the published age specific fertility rates and the calculated numbers in the Household Questionnaire are used to estimate the annual number births for all in 2002, assuming no decline in fertility over the period. Then the annual fertility rates are recalculated using the adjusted numbers who should have been listed in the Household Questionnaire if the 2002 or the 2000 Census marriage patterns had prevailed for the. Table 4. Adjustment ASFR and Total Rate for the rates Final Report ( ) in Annual births implied by fertility rates and number in 2002 adjusted for 2002 marital status Rates with 2002 based estimate adjusted for 2000 Census marital status Rates with 2000 Census based estimate Age mothers Total TFR Where the Main Report showed a TFR 2.6 (rounding f the calculation 2.57), the adjustment based on both the Census and the marriage patterns implies that the true fertility level was closer to 2.2 only a small amount above the commonly accepted target replacement level 2.1. A similar procedure has been applied to the 2007 in the Annex table, yielding a TFR 5

6 2.18. Figure 2 takes a microscopic view the fertility estimates since the advent decentralization in As recent estimates show, the adjusted results are slightly under the census-type survey trend line and both indicate a slowing in fertility decline but substantially below the unadjusted levels. Figure 2 Unadjusted trend lines from and 2007 and three census type surveys compared with adjusted fertility estimates for and TFR Year 2000 Census, , 2005 SUPAS retro estimate for retro estimate for Adjusted estimate for Current fertility Adjusted estimate for 2007 current fertility Adjusting the fertility rate for missing in the produces a TFR 2.25 for the three year period (centred on 2002). This is below the trend line for census-type enumerations. A similar adjustment applied to the 2007 using the marital status distribution from the 2005 SUPAS produces a TFR 2.18 for the period from , or a point estimate for The gap nearly half a child between and census-type estimates produces understandable disquiet among policymakers. Some prefer to rely on the higher number through the application a precautionary principle it is better to respond to the implications the higher fertility than to assume the lower fertility is real, but then be found to have been complacent. This creates a credibility gap undermining policy and confusing planning. It represents one the most serious problems revitalizing the population and family planning program in Indonesia today. 6

7 Conclusions There are serious problems with the methods used to estimate fertility in Indonesia. The own-child method fertility calculation used in the decennial census and the annual produces a steady downward trend fertility from the 1970s through to the present day. In contrast the pregnancy history techniques used by the yielded lower fertility than the census prior to 2000, and higher levels since then. The discrepancy appears to be caused by the failure the sample and interview methods to capture a true prile all reproductive ages, and in particular missing a substantial number. Once adjusted for these missing the fertility rates for the three years prior to the and 2007 surveys are around 2.2, slightly above the long term BKKBN goal replacement level fertility. ANNEX tables: Estimation total number in 2007 if reflecting 2005 SUPAS marriage patterns numbers by age in age 2005 Supas in age Estimate sample if had all Difference -- missing A B C D=A*C/B A-D All WRA

8 Estimation total number in if reflecting 2007 marriage patterns A B C D=A*C/B A-D numbers by age in age 2007 Susenas in age Estimate sample if had all Difference -- missing All WRA Adjustment ASFR and Total Rates -- Annual rates births 2007 in 2007 implied by Final fertility Report rates (2005- and 2007) number adjusted for 2005 SUPAS marital status Rates with 2005 SUPAS based estimate adjusted for 2007 marital status Rates with 2007 based estimate Age mothers in #DIV/0! #DIV/0! #DIV/0! #DIV/0! #DIV/0! #DIV/0! #DIV/0! Total TFR #DIV/0! 8