Is recent Afghanistan survey data suitable for fertility analysis? A regional investigation based on fertility inhibiting determinants
Authors:
Jamal Abdul Nasir aff001; Sohail Akhtar aff001; Syed Arif Ahmed Zaidi aff002; Andleeb Rani aff001; Hina Bano aff003; Andrew Hinde aff004
Authors place of work:
Department of Statistics, Government College University, Lahore, Pakistan
aff001; Department of Community Medicine, Quaid-e-Azam Medical College, Bahawalpur, Pakistan
aff002; National College of Business Administration & Economics, Rahim Yar Khan, Pakistan
aff003; Department of Social Statistics and Demography, University of Southampton, Southampton, United Kingdom
aff004
Published in the journal:
PLoS ONE 14(10)
Category:
Research Article
doi:
https://doi.org/10.1371/journal.pone.0223111
Summary
Afghanistan has been a country blighted by war over the past five decades and limited research is available on its demography. This study seeks to assess the suitability of recent survey data for Afghanistan (the 2010 Afghanistan Mortality Survey (AMS)and the 2015 Afghanistan Demographic and Health Survey (ADHS)) for estimating levels and trends in fertility. As several fertility measures rely on the quality of age data, we first apply demographic tools for the identification of age misreporting, finding evidence that it is severe. We then explore the consistency of fertility reporting across the two surveys, finding that the 2015 ADHS reports higher fertility among older women than the 2010 AMS although the seasonal pattern of fertility is consistent across the two surveys. We then estimate total fertility rates in 2008–2010 and 2012–2015 and measures of Bongaarts’s key proximate determinants of fertility for Afghanistan and its provinces for urban and rural areas separately. The results show that fertility is similar in urban and rural Afghanistan. Although most of the provincial data on the proximate determinants is reasonably consistent with the fertility rates, there are anomalies in some provinces which indicate the possible under-reporting of births. Overall, we conclude that the fertility data in the two surveys can be used with care to give an indication of broad regional fertility patterns and trends in the country.
Keywords:
Age groups – Urban areas – Rural areas – Female contraception – Health surveys – Fertility rates – Afghanistan
Introduction
Many aspects of the demography of the Islamic Republic of Afghanistan remain under-researched. The exact population of the country is not known due to the lack of a recent census. According to the Central Statistics Organization of Afghanistan (CSOA) the total population of the country is estimated to be 32.2 million people [1]. The urban population (24 per cent) is only one third of the rural population (71 percent). Afghanistan’s population and socio-economic development has faced the challenge of war over the past five decades, and unrest and political instability still affect the living conditions and fertility choices of inhabitants. According to the Population Reference Bureau, among the countries of the South Asian region, Afghanistan has the highest total fertility rate (an estimated 4.8 children per woman in 2018) [2].
The fertility transition in Afghanistan has attracted limited attention in demographic research [3]. Although there have been several surveys co-ordinated by the CSOA, including the Afghanistan Multiple Indicator Cluster Surveys (1997, 2000, 2003, 2010–2011), Afghanistan Living Conditions Survey (2005, 2007–2008, 2011–2012, 2013–2014, 2016–2017), a Socio-Demographic and Economic Survey (2016) at provincial level (the province is the first level administrative unit of the country), and the Afghanistan Health Survey (2003, 2005, 2007–2008, 2015), these surveys did not collect birth history data. In addition, most surveys coordinated by the CSOA are not widely available to researchers outside the country.
Neighboring countries of the South Asian region have shown a noticeable decline in fertility. For instance, Iran’s total fertility rate is 2.0 and that of Pakistan is 3.1 [2]. Contraceptive prevalence in Iran among married women of reproductive age is 77 per cent, compared with 35 per cent in Pakistan and only 23 per cent in Afghanistan (for modern methods, the corresponding percentages are 57, 26 and 20) [2]. Contraceptive use is one of the three major proximate determinants of fertility as proposed by John Bongaarts [4,5].
In this paper we examine the quality of the data on fertility collected in the only two recent surveys that did include birth histories: the Afghanistan Mortality Survey carried out between April and December 2010 (hereafter 2010 AMS) and the Afghanistan Demographic and Health Survey conducted between June 2015 and February 2016 (hereafter 2015 ADHS).
There exists an enormous literature on the types of error (examples include omission, duplication, misreporting, telescoping due to memory lapse) affecting estimates of fertility [6,7]. In this paper we focus on the extent of age misreporting before examining the consistency of fertility estimates from the two surveys. By comparing two surveys conducted only five years apart, we hope to be able to identify where the fertility outcomes are inconsistent and therefore indicate errors or omissions in at least one of the surveys, and also to emphasise topics where the surveys produce consistent results, which increases our confidence that the data are accurate. We then estimate total fertility rates in 2008–2010 and 2012–2015 and measures of Bongaarts’s key proximate determinants of fertility for Afghanistan and its provinces for urban and rural areas separately.
Data
The 2010 AMS used a sampling frame provided by the CSOA. It aimed to provide estimates of demographic variables which were representative of the whole country, and of urban and rural areas within three domains [8]. The domains were (1) North (the Northern and North Eastern regions), (2) Central (the Western, central highland and Capital regions), and (3) South (the Southern, South Eastern and Eastern regions). A two-stage sampling process was used. The country was divided into strata based on urban and rural residence and the three domains listed above (additional strata were used for rural areas within each of the domains). Within each stratum enumeration areas were selected in a first stage, and then households were selected within each enumeration area in a second stage. Interviews were conducted with all women aged 12–49 years in the selected households. In all, 23,897 households were selected, and 22,381 were successfully contacted [8]. A total of 47,848 women were interviewed. A small number of areas of the country were not covered in the survey for security reasons [8]. The population sampled represents 87 per cent of the total population, and most of the omitted areas were in the Southern region [8]. The urban areas of Kabul were heavily over-sampled.
The 2015 ADHS used an updated version of a sampling frame provided by the CSOA. The sampling frame used information about 34 provinces, control areas, districts and urban or rural residence. In the 2015 ADHS a two-stage stratified sample design was used [9]. The first stage involved selecting 950 clusters (260 in urban areas and 690 in rural areas). Due to security issues in some areas of Afghanistan 101 reserve clusters were preselected, giving a total of 1,051 clusters for the survey. Of these, 75 clusters were classified as insecure during the household listing operation; for the 976 remaining clusters household listing was successfully completed. Eventually the survey was carried out in 956 clusters. The second stage used systematic random sampling of households. A total of 25,741 households were selected for the sample, and from these households 30,434 ever-married women age 15–49 years were identified for individual interviews and 29,461 interviews were successfully completed [9]. Some areas, notably Nooristan province, were over-sampled,
The 2015 ADHS only interviewed ever-married women aged 15–49 years. For comparative purposes, we compare these women with the 26,730 ever-married women in the same age range in the 2010 AMS. The distribution of these women in both surveys by province and urban-rural residence is presented in Table 1. Although the sampling fractions varied by province and from survey to survey, both surveys provided weights, by applying which they may be rendered representative of the national population. When comparing national estimates, therefore, we use weighted data.
Methods
This paper examines the quality of the birth history data in the two surveys using the following approaches. First, we measure age misreporting using two well-known methods: Whipple’s Index and Myers’s Blended Index. Whipple’s Index measures age heaping on ages with digits ending in 0 and 5 in the adult age range [10]. Its normal application uses the age range 23–62 years and the formula
Because we only have data on women aged 15–49 years, we calculate Whipple’s Index using the age range 18–47 years.
Modifications and extensions to Whipple’s Index have been suggested in the literature [11,12]. Here, however, we use the original version designed to measure heaping on ages ending in the digits 0 and 5. This is the dominant form of age heaping in Afghanistan. Myers’s Blended Index is a more general measure, which takes into account preferences for (or antipathy towards) ages ending in any digit [10,13].
Second, we look at the consistency of the estimates of fertility produced by the birth histories in the two surveys. We do this in two ways, first by looking at the seasonal pattern of births and, second, by computing age-specific fertility rates (ASFRs) and the total fertility rate for four-year periods from 1984–1987 to 2012–2015 using data from the two surveys. Because both surveys are designed to be representative of the majority of the population, ASFRs for the whole country estimated for the same four-year period using weighted data should be similar for both surveys. We compute the ASFRS using weighted data to maximise comparability. The ASFRs are computed using the exact exposure method. The numerators are the total numbers of births reported by the women in each survey in each four-year period when they were in each five-year age group at the time of giving birth. The denominators are obtained by working out the exact exposure (in years and fractions of a year) for each woman in each age group in each four-year period and summing these over all women.
Third, we use the ASFRs in the most recent period for each survey to estimate total fertility rates for the urban and rural populations in each province. To maximise comparability, we have amalgamated strata with fewer than about 200 women in either survey to produce a set of sub-samples defined on the basis of province and urban-rural residence which are comparable across the two surveys and which include approximately 200 women or more. We estimate the total fertility rate for each of these sub-samples for the period 2008–2010 from the 2010 AMS, and for the period 2012–2015 from the 2015 ADHS.
We then examine some key determinants of fertility in the same set of sub-samples. John Bongaarts, in his classic work, built on the ideas of Kingsley Davis and Judith Blake, and identified the four most important proximate determinants of the fertility outcome in any population: the prevalence of marriage, the use of contraception, the use of abortion and the impact of breastfeeding [4,5,14]. Here, we focus on marriage, contraception and breastfeeding. For each of the provincial and urban/rural sub-samples we compute a measure of the likely impact of each of these determinants on fertility. For marriage, we calculate the percentage of 20–24 year old women who are currently married, using the entire sample from 2010 AMS (which interviewed both ever-married and never married women). For contraception, we compute the percentage of ever-married women aged 15–49 years who are using a modern contraceptive method at the time of the 2015 ADHS. Finally, for breastfeeding we measure the percentage of the most recent births to women in the 2015 ADHS born 12–23 months before the survey who were still being breast fed on the survey date. We expect high fertility to be associated with a high proportion of 20–24 year olds being married, a low prevalence of modern contraception and a low proportion of babies aged 12–23 months being breast fed.
Results
Fig 1 presents the distribution by single years of age of the samples from the two surveys, distinguishing women living in urban and rural areas. The age distribution is highly irregular, with clear peaks on ages ending in the digits 0 and 5 (and to a lesser extent 2 and 8). This indicates substantial age misreporting by respondents. The degree of age misreporting is not obviously different in the two surveys, neither is it clearly greater in rural than in urban areas.
Table 2 presents the values of two indices of age misreporting at the national and provincial level in Afghanistan for urban and rural samples in 2010 and 2015. Whipple’s Index measures age heaping on ages with digits ending in 0 and 5 [10]. According to United Nations criteria, values of the index above 175 indicate ‘very rough’ age reporting and values between 125 and 175 indicate ‘rough’ age reporting. It is clear that in almost all of Afghanistan’s provinces age reporting is ‘rough’, and in most it is ‘very rough’. The highest levels of age misreporting occur in the Central Highlands region and, in 2010, in Eastern region.
The pattern revealed by Myers’s Blended Index is similar to that for Whipple’s Index; in particular, in no province is the value of Myers’s Blended Index below 10, indicating substantial levels of preference for certain terminal digits in all areas of the country.
An important feature of age reporting in Afghanistan is that it is not obviously better in urban areas than in rural areas in either 2010 or 2015 (Table 2). It is, however, slightly better among women with some education than among women with no education. In 2010, Whipple’s Index was 158 for women with some education compared with 207 for women with no education; the corresponding values in 2015 were 155 and 186. Myers’s Blended Index took the values 14.3 and 23.8 among women with some education and no education respectively in 2010; by 2015 the corresponding values were 13.1 and 21.3. Taken together, these results suggest a slight improvement in age reporting in rural areas between 2010 and 2015, but very little improvement in urban areas.
We turn now to the comparison of fertility patterns revealed by the birth history data in the two surveys. Fig 2 shows the percentage of births in each calendar month according to the 2010 AMS and the 2015 ADHS. The pattern is very similar in the two surveys. There is a pronounced seasonal pattern, with births being fewest in February and March, and most in April, May and June. This implies that conceptions reach a low point in May and June, and a peak in July, August and September. the difference in the number of births between the trough in February and March and the peak in April, May and June, is very substantial, and among the highest reported from national populations [15]. The peak is not an artefact of the imputation of the month of birth for some of the births: the proportion of births for which imputation was carried out was much too small in either survey to account for anything more than a tiny fraction of the seasonal pattern.
Fig 3 shows the age-specific fertility rates (ASFRs) revealed by the birth history data for the six five-year age groups from 15–19 to 40–44 years in four-year periods from 1984–1987 to 2012–2015. There is evidence of a decline in fertility at ages 25–29 years and over since around 2000, suggesting that Afghanistan has entered the fertility transition as first observed by Thomas Spoorenberg using the 2010 AMS [3]. However, a second pattern in Fig 3 is that the 2015 ADHS generates lower ASFRs than does the 2010 AMS for the age group 15–19 years and higher ASFRs in all periods than does the 2010 AMS for age groups 25–29 years and older. The difference is substantial for age groups 35–39 and 40–44 years. The lower reported fertility at older ages in the 2010 AMS merits further investigation. One possible reason, a lower proportion of older women currently married in the 2010 AMS, can be ruled out, as typically more than 90 per cent of women were currently married at all ages up to 49 years in both surveys. Another reason is a tendency to under-report female births, which was noted in the report to the 2010 AMS [8]. There is some evidence that this tendency was greater among older women in 2010 AMS than in 2015 ADHS. The sex ratio of births to women aged 40–44 years was 115 boys per 100 girls in the 2010 AMS and 110 boys per 100 girls in the 2015 ADHS. Finally, some of the difference is the result of the different distribution of women by age within each the age groups arising from the fact that we are estimating period fertility using data from a cohort aged 15–49 years at the date of interview.
The total fertility rate for the period 2008–2011 based on the 2010 AMS was 6.46 and that for the same period based on the 2015 ADHS was 7.05. This also represents a substantial difference. The total fertility rate for the period 2012–2015 based on the 2015 ADHS is 6.61, still higher than that reported for the earlier period by the 2010 AMS, but marking a decline in fertility since 2008–2011.
The ASFRs reported in Fig 3 relate to ever-married women. As such they cannot be compared with the rates reported by the report into the 2010 AMS, which are based on all women in the survey and are consequently considerably lower [8]. The all-women total fertility rate reported in the 2010 AMS and based on births in the three years preceding the survey was around 5.0 births per woman. In the 2015 ADHS the reported total fertility rate in the three years preceding the survey was 5.3. This was based on an ever-married sample, but the denominators were inflated before calculating the ASFRs to account for never-married women [6]. It is interesting to observe that the total fertility rate for all women reported for the three years preceding the 2015 ADHS was higher than that reported for the three years preceding the 2010 AMS, consistent with the pattern we find.
Finally, Tables 3 and 4 present the total fertility rate by province and urban or rural residence, along with selected indicators of the proximate determinants of fertility. Table 3 deals with urban areas, and Table 4 deals with rural areas. The first point to note is that, among ever-married women, fertility in Afghanistan is almost the same in urban and rural areas. The lower overall fertility reported from urban areas in the report to the 2015 ADHS, for example, derives from the lower prevalence of marriage in urban areas [6]. The proportion of 20–24 year olds in the 2010 AMS who were married was more than ten percentage points lower in urban areas than in rural areas (Tables 3 and 4).
A comparison of the reported total fertility rates by province among urban areas suggests either that fertility has increased between 2008–2010 and 2012–2015 in several areas (for example Faryab, Kunduz and Urozgan), or that fertility was under-reported in these areas 2010 AMS relative to the 2015 ADHS. Regional patterns are hard to discern, save perhaps especially high fertility in urban areas in Eastern province. The low fertility in the city of Herat is consistent with a very high (by Afghan standards) prevalence of modern contraception.
In rural areas the figures suggest fertility increases between 2008–2010 and 2012–2015 in several provinces (Balkh, Faryab, Takhar, Kunduz, Badghis, Daykundi, Nimroz, Khost, and all the provinces in Eastern region). However, some of the reported total fertility rates are suspiciously low. In 2008–2010 those in Kunduz, Daykundi and Nimroz provinces seem unreasonably low when compared with the contraceptive prevalence rates among ever-married women, which in Kunduz and Daykundi provinces are also some of the lowest in the country. In 2012–2015 the total fertility rate of 3.5 in Ghazni province seems unreasonably low when the contraceptive prevalence rate of 12.1 per cent is taken into account.
There are also some dramatic changes in the total fertility rate between 2008–2010 and 2012–2015 in provinces such as Logar (a decline from 9.2 to 6.3 in rural areas, but this is based on a sample of fewer than 200 women (Table 1)). In parts of Eastern region, specifically Nangahar and Nooristan provinces, there were increases from 6.9 to 8.6 and from 7.3 to 9.5 respectively, and here the numbers of women are much larger. One possible reason for this is the under-reporting of girl babies in the 2010 AMS. Reported sex ratios at birth in southern Afghanistan in the 2010 AMS were very skewed (125 boys per 100 girls) [8]. The extremely high fertility in Urozgan province is a feature of both surveys, suggesting that it is a real phenomenon. The very high fertility in rural Nooristan province in 2012–2015 is consistent with the near absence of modern contraception.
Breastfeeding for between one and two years is common in all areas of Afghanistan. In all provinces, and in urban and rural areas, at least half the children aged 12–23 months at the time of the 2015 ASHS were being breast fed (Tables 3 and 4). It seems unlikely that geographical variations in fertility are explained to any great extent with geographical differences in breastfeeding behavior.
Discussion
We focus this discussion eventually on the quality of the birth history data but, first, let us make some remarks about recent fertility trends in Afghanistan. Our results confirm the observations of Thomas Spoorenberg based only on the 2010 AMS that Afghanistan entered its fertility transition around the turn of the century [3]. The fertility transition also looks to be following the classic Asian pattern of an initial decline in fertility among older women, which gradually spreads to include all women aged 25 years and older. There is little evidence of a rural-urban differential in fertility within marriage, though the lower prevalence of marriage in urban areas means that fertility is lower there. There are some regional patterns in fertility, with especially high level being seen in Eastern region and in some other individual provinces, such as Urozgan.
The fertility data in both the 2010 AMS and the 2015 ADHS have their shortcomings. In common with other neighbouring south Asian countries, age heaping on ages ending in the digits 0 and 5 is very pronounced. This has the potential to affect reported fertility rates but the precise effects will depend on the nature of the misreporting (for example whether the tendency is to round ages up to the nearest age ending in the digits 0 or 5, or to round ages down). The age heaping is a feature of both urban and rural populations, but is slightly less prevalent among women with some education than those with no education. Trends in age-specific fertility revealed by the two surveys are broadly consistent, especially at ages 15–19 and 20–24 years. At older ages, the 2015 ADHS reports higher fertility than the 2010 AMS, and there are probably several factors contributing to this, all tending in the same direction.
Regional patterns of fertility are difficult to discern in either survey, and in some provinces there is reason to believe that the birth history data in one or both surveys are deficient. A few provinces have very low reported fertility, and in others there are dramatic increases or decreases in the current and recent fertility levels reported by the 2010 AMS and the 2015 ADHS. Nevertheless, some regional patterns seem robust. High fertility seems characteristic of Eastern region and rural areas of Urozgan province. The city of Herat has low fertility and a high contraceptive prevalence rate.
Conclusion
Afghanistan has only conducted two nationally representative surveys which collected birth history data: the 2010 Afghanistan Mortality Survey (AMS) and the 2015 Afghanistan Demographic and Health Survey (ADHS). In this paper we have examined the accuracy of age reporting in the two surveys, the consistency between the estimates of the trends in age-specific fertility they generate, and the plausibility of the provincial fertility estimates when set alongside estimates of the proximate determinants of fertility.
Our conclusion is that both surveys have weaknesses, notably in the reporting of women’s ages. The 2010 AMS suffers from the omission of female births in the south of the country, and perhaps among older women, but this was rectified to some extent in the 2015 ADHS. Some provinces have unusually low fertility rates given the reported rates of use of modern contraception. Despite this, the two surveys both reveal the onset of the fertility transition in Afghanistan around the turn of the century, and there is a broad consistency in the reported trends in age-specific fertility. Overall, we conclude that the fertility data in the two surveys can be used with care to give an indication of broad regional fertility patterns and trends in the country.
Zdroje
1. Central Statistics Organization, Islamic Republic of Afghanistan (2019) Estimated Population of Afghanistan 2019–2010. Kabul: Central Statistics Organization.
2. Population Reference Bureau (2018) World Population Data Sheet 2018. Washington DC: Population Reference Bureau.
3. Spoorenberg T (2013) An evaluation of the recent fertility changes in Afghanistan: a period-specific analysis. Journal of Population Research 30: 133–149.
4. Bongaarts J (2015) Modeling the fertility impact of the proximate determinants: Time for a tune-up. Demographic Research 33: 535–560.
5. Bongaarts J (1978) A Framework for Analyzing the Proximate Determinants of Fertility. Population and Development Review 4: 105–132.
6. Schoumaker B (2014) Quality and consistency of DHS fertility estimates, 1990 to 2012 (DHS Methodological Reports No.12). Rockville: ICF International.
7. Cleland J (1996) Demographic Data Collection in Less Developed Countries 1946–1996. Population Studies 50: 433–450. doi: 10.1080/0032472031000149556 11618376
8. Afghan Public Health Institute, Ministry of Public Health (APHI/MoPH) [Afghanistan], Central Statistics Organization (CSO) [Afghanistan], ICF Macro, Indian Institute of Health Management Research (IIHMR) [India], et al. (2011) Afghanistan Mortality Survey 2010. Calverton, Maryland, USA: APHI/MoPH, CSO, ICF Macro, IIHMR and WHO/EMRO.
9. Central Statistics Organization (CSO), Ministry of Public Health (MoPH), ICF (2017) Afghanistan Demographic and Health Survey 2015. Kabul, Afghanistan: Central Statistics Organization.
10. Siegel J, Swanson D (2008) The Methods and Materials of Demography. Bingley. United Kingdom: Emerald Group Publishing Ltd.
11. Nasir JA, Hinde A (2014) An extension of modified whipple index-Further modified whipple index. Pakistan Journal of Statistics 30: 265–272.
12. Spoorenberg T, Dutreuilh C (2007) Quality of age reporting: extension and application of the modified Whipple’s Index. Population (English edition) 4: 729–741.
13. Myers RJ (1940) Errors and Bias in the Reporting of Ages in Census Data. Transactions of the Actuarial Society of America 41: 395–415.
14. Davis K, Blake J (1956) Social Structure and Fertility: An Analytic Framework. Economic Development and Cultural Change 4: 211–235.
15. Becker S (1991) Seasonal pattern of births and conceptions throughout the world. Temperature and enviornmental effects on the testis. Boston, MA: Springer.
Článok vyšiel v časopise
PLOS One
2019 Číslo 10
- Metamizol jako analgetikum první volby: kdy, pro koho, jak a proč?
- Parazitičtí červi v terapii Crohnovy choroby a dalších zánětlivých autoimunitních onemocnění
- Nejasný stín na plicích – kazuistika
- Masturbační chování žen v ČR − dotazníková studie
- Eliquis (apixaban) nově hrazen ze zdravotního pojištění
Najčítanejšie v tomto čísle
- Correction: Low dose naltrexone: Effects on medication in rheumatoid and seropositive arthritis. A nationwide register-based controlled quasi-experimental before-after study
- Combining CDK4/6 inhibitors ribociclib and palbociclib with cytotoxic agents does not enhance cytotoxicity
- Risk factors associated with IgA vasculitis with nephritis (Henoch–Schönlein purpura nephritis) progressing to unfavorable outcomes: A meta-analysis
- Prevalence of pectus excavatum (PE), pectus carinatum (PC), tracheal hypoplasia, thoracic spine deformities and lateral heart displacement in thoracic radiographs of screw-tailed brachycephalic dogs