1The Kingdom of Bhutan is a small, landlocked country of South Asia, located in the eastern part of the Himalaya and surrounded by the Indian States of Sikkim, Assam and Arunachal Pradesh to the west, south and east, respectively, and the Chinese province of Tibet to the north. Due to its geographic location, its historical isolation from the international community and the small number of available data sources, very little is known about the country’s demographic development (Véron, 2008). Its territory extends over about 38 thousand square kilometers and was home to a population of 635,000 according to the most recent census conducted on 30-31 May 2005 (Office of the Census Commissioner, Royal Government of Bhutan, 2006). Bhutan’s population is young, with a median age of 22 years. The country’s population growth has slowed down considerably, from a high of 3.1% per annum in the 1990s to 1.8% in 2005. Mortality rates have improved greatly, with a significant decrease in the infant mortality rate (IMR) from a very high level of 102.8 deaths per 1,000 live births in 1984 to 30 in 2012, while the under-five mortality rate fell to 37 per 1,000 live births in 2012 compared with 162.4 in 1984. [1] Although the rate of growth has declined, the population of Bhutan is expected to continue increasing in the coming decades as the population momentum will keep the growth rate positive for some time to come. Most notably known for its unique approach to development, Bhutan is the only country in the world to use Gross National Happiness to measure quality of life or social progress among its population.

2The focus of this short paper is to review and examine the evidence documenting changes in fertility levels and trends in Bhutan over the last 50 years. Based on the available data sources, we study the fertility transition in the country by presenting a quasi-exhaustive list of estimates of total fertility obtained with both direct and indirect methods, and using census and sample survey data. The reconstruction of fertility levels and trends in Bhutan shows that total fertility was around 6 children per woman until the mid-1980s and fell rapidly from about 5.5 children per woman in the 1990s to close to replacement level today. We discuss the consistency of the various results in the light of the potential issues liable to affect the estimates given by each method. This short paper can be considered as laying an empirical foundation for further research on fertility change in Bhutan and, more broadly, on the demography of the country.

I – A large downward adjustment in the Bhutan population count

3The figures for the total population of Bhutan diverge significantly between data sources. According to civil records for 1969, [2] Bhutan had a population of 1,035,000 persons on 1 December of that year. This figure remains difficult to assess, but is probably a gross exaggeration. Indeed, at that time, it was believed in Bhutan that unless a country had a population of more than one million, it was not eligible to join the United Nations. The 1969 population figure should therefore be understood within this context and considered as a very crude estimate of the country’s population, established with a view to applying for membership of the United Nations, which Bhutan joined in 1971.

4In 1990, the population of Bhutan was officially close to 1.5 million. Since then, this figure has been significantly revised, starting in the early 1990s, when a first drastic revision was made. According to the 1992 Statistical Yearbook (Central Statistical Office, 1994), the country was home to only 624,000 people. Later on, a further downward adjustment was made; only 502,000 persons were recorded as living in the country in 1992 (United Nations Statistics Division, n.d.). No official explanation is available for these successive revisions.

5Bhutan’s first modern census was conducted in 2005. Before that, in response to large immigration flows, the country took citizenship censuses to check the nationality of its populations (Véron, 2008). The first revision in the early 1990s coincides with changes in the definition of the legal population of Bhutan following the entry into force on 10 June 1985 of a new citizenship act (Ministry of Home Affairs, n.d.). Following implementation of the act, many people of Nepalese descent moved out of the country in the late 1980s and early 1990s (Hutt, 2003). While no authoritative figure exists on the number who actually left, these departures certainly cannot fully explain the drastic reduction in the official population count in the early 1990s.

6Despite the successive revisions of Bhutan’s population count, they are unlikely to affect our estimates. As we shall see below, no sudden and noticeable change in the fertility levels and trends are observed in the late 1980s and early 1990s.

II – Data and methods

7Our analysis is based on period fertility estimates drawn from a variety of data sources. Several different estimation methods were used. As the first (modern) population census was not conducted in the country until May 2005, our study relied mostly on nationally representative sample surveys. We used as many estimates of total fertility as possible that could be drawn from the available sources.

8Table 1 summarizes the data sources and methods we used to estimate the fertility changes in Bhutan over the last 50 years. The number of data sources available to estimate total fertility is unevenly distributed. Out of the eight data collection operations conducted in Bhutan, six have taken place since 2000. This distribution reflects the government’s recent focus on population in the monitoring and planning of national development policies (United Nations, 2013). It was only in 1995, after the publication of the 1994 National Health Survey results revealing poor reproductive health indicators, that a strong political commitment to curbing population growth was first expressed. As a result, five nationally representative sample surveys have been conducted since 2000 and a first modern census in line with international recommendations took place on 30-31 May 2005. A second census is scheduled for 2016. In comparison, fewer data – mostly based on indirect methods – document the changes before the 1990s.

Table 1

Data sources and methods to estimate total fertility, Bhutan

source estimation methods direct cohort fertility reverse survival 1984 demographic survey (ds) x x x 1994 national health survey (nhs) x 2000 national health survey (nhs) x 2003 bhutan living standard survey (blss) x 2005 population and housing census x x 2007 bhutan living standard survey x x 2010 bhutan multiple indicator survey (bmis) x x x 2012 bhutan living standard survey (blss) x x x 2012 national health survey (nhs) x x

Data sources and methods to estimate total fertility, Bhutan

Note: Direct estimates were taken from published reports.

9Different methods were applied to these data sources to derive estimates of total fertility in Bhutan. A first series consists of direct estimates taken from various published census and survey reports. These estimates are based on information on the number of births over a given period preceding the data collection operation (usually 12 months). This information is usually collected at the household level in the case of a census or among women aged 15-49 in sample surveys. A second set of estimates are based on data on the number of children ever born (CEB) for women aged 40-44 and 45-49. These data were used to approximate cohort fertility. Using Ryder’s correspondence between period and cohort measures (Ryder, 1964, 1983), the mean number of children ever born to a cohort can be used to approximate the period total fertility rate at the time this cohort was at its mean age at childbearing (see Feeney (2014) for further details about time translation of mean CEB for women aged 40 and over). The mean age at childbearing was computed from the number of births in the household during the last 12 months. Finally, the last series of fertility estimates was obtained by applying the reverse survival method of fertility to the population by single age and sex. Fertility was estimated using the Coale-Demeny West model life table, using mortality figures (₅q₀ and ₄₅q₁₅) and age patterns of fertility drawn from from United Nations estimates (2015), and using the Excel template “FE_reverse.xlsx” provided with Timæus and Moultrie (2013).

III – Quality of fertility estimates

10Several factors can ultimately affect the quality of the fertility estimates derived from the different methods used in reconstructing fertility levels and trends in Bhutan (Table 1).

11The reverse survival method of fertility estimation relies on the population structure by age and sex collected in population and housing censuses or in sample surveys. These data are often affected by data quality issues such as population under-enumeration at certain ages and age heaping on selective age digits. Such patterns can potentially affect fertility estimates based on a population age structure. A typical pattern found both in population censuses and sample surveys is the under-representation of young children due to both omission and bias related to age misstatement. This leads to under-estimation of fertility in the years directly preceding the data collection. The misstatement of children’s ages results in a transfer to older ages and age heaping on ages 5, 10 and 15. The transfer to older ages may lead to under-estimation of fertility in the one or two years preceding data collection, whereas age heaping on ages 5 and 10 results in over-estimated fertility around the fifth and tenth preceding years. Age heaping on age 15 contributes to under-estimation of fertility at the end of the 14-year period preceding data collection. These expected effects are often not clearly delineated because the quality of age-reporting of women aged 10-64 also affects the estimation of fertility in the reverse survival method.

12Mortality is another factor that can affect fertility estimation based on the population structure. The reverse survival method requires an estimate of mortality to derive fertility estimates. The use of unrealistically low mortality levels would result in adding too few births in the population and therefore under-estimating the level of fertility that is ultimately derived. As described above, Coale-Demeny model life tables were used. Yet, even though these models may not account fully for the peculiarities of the age patterns of mortality in Bhutan, such effects have only a minimal impact on the estimation of fertility (Spoorenberg, 2014a). In fact, the under-five mortality rate differs only slightly between the four families of the Coale-Demeny model life tables and the wrong selection of the level and age patterns of mortality only marginally affects the reverse survival fertility estimates (ibid.).

13Commonly, data on the number of children ever born are affected by some kind of recall error, i.e. the women or the informed household member providing this information tend to omit some of their children born many years previously (United Nations, 2004). Furthermore, information on the number of children ever born is collected for surviving women only. Due to mortality selection, fertility can be under-estimated when using cohort fertility because the reproductive experiences of women with a higher number of children, and therefore exposed to higher mortality risks delivering higher-order births, are not accounted for. These two factors contribute to under-estimation of total fertility when cohort fertility estimates are based on information on the number of children ever born to women aged 45 and older.

14Migration can also distort the quality of fertility estimates given by each method. For methods based on the population by age and sex, the departure or arrival of migrant women can potentially result in over-estimating or under-estimating, respectively, the level of fertility. Yet, migration will affect fertility estimation only if the female migrants leave their children behind (thus affecting the denominator only) or if female migrants have a significantly higher or lower number of children than native women (thus affecting the numerator only). Finally, migration could potentially affect the estimation of fertility based on the number of children ever born if the number of children ever born works as a selection mechanism to migration. However, for all estimation methods, only the migration of large portions of the female population would affect the estimation of fertility.

IV – Fertility levels and trends in Bhutan

15While a certain difference is to be expected between fertility estimates based on different data sets and estimation methods, a fair amount of consistency nonetheless exists between direct and indirect series.

16Figure 1 presents the 18 series of fertility estimates that could be extracted from the existing data sources in Bhutan. Contextual information in the form of important historical periods and political decisions are also displayed to facilitate interpretation of the national fertility trends and levels since the mid-1960s. Total fertility in Bhutan stood at about 6 children per woman until the mid-1980s when total fertility started its decline. Because the fertility estimates for the 1960s are based on completed cohort fertility (i.e. children ever born among women aged 40-44 and 45-49), it is likely that the levels given by these estimates are under-estimated – a habitual pattern (see above and below). Yet, the fertility decline revealed by these estimates is certainly real.

Figure 1

Fertility in Bhutan, 1965-2012

Fertility in Bhutan, 1965-2012

17From a level of about 5.5 children per woman in the early 1990s, total fertility decreased swiftly and reached a level close to 2.2 children per women in the early 2010s. In 15-20 years, the number of children per woman fell by more than half in Bhutan. The latest estimates indicate a levelling-off of the decline as fertility approaches the replacement level of 2.1 children per woman.

18Very few other countries (Algeria, Iran, Mongolia, Vietnam) have experienced a fertility decline of this magnitude over such a short time. Bhutan’s fertility decline is comparable to that of Iran, where fertility plummeted from about 5.6 children per woman in 1988 to 1.9 by the end of the first decade of the 21^st century (McDonald et al., 2015).

19Figure 2 shows the changes in the age-specific fertility rates between 1984 and 2012 in Bhutan. In the first half of the 1980s, fertility was concentrated in four age groups: 20-24, 25-29, 30-34 and 35-39 years. Ten years later, in 1994, the age distribution started to change. Fertility was lower in all age groups starting from age 25, but the reduction was larger in the older age groups. This is frequently observed over the course of the fertility transition and is the result of a stopping mechanism at older ages. The distinct profile of the age distribution between 1994 and 2000 suggests that women aged 30 and above started using the contraceptive methods that became available in Bhutan in the 1990s. In the year 2000, the 20-24 and 25-29 age groups accounted for more than half of all births. While the age pattern of fertility has changed very little since the early 2000s, its level has continued to fall, a possible sign that women of all ages are now controlling their reproduction using means of contraception that have become widely accessible over the last two decades. [3]

Figure 2

Age-specific fertility rates, Bhutan, diverse sources

Age-specific fertility rates, Bhutan, diverse sources

20While it is true that the distinct series of fertility estimates give a rather consistent picture of what has happened to fertility over the last decades in Bhutan, some variations need to be further considered.

21First, as expected (see Section III), the fertility estimates based on cohort fertility are systematically under-estimated due to omissions and selection through mortality (see the diamond series in Figure 1). This is especially evident in the 1980s and 1990s and does not depend on the type of data source. The cohort fertility estimated from the 2005 census and the 2010 BMIS, 2012 NHS and 2012 BLSS are systematically lower than the other series. It is likely that the cohort fertility based on the 1984 DS indicates too low a level, but without other estimated series for this period (late 1960s) it is difficult to draw any firm conclusion.

22Second, the fertility estimates obtained by applying the reverse survival method reveal a systematic dip in the most recent years in each series. This is due to the under-representation of very young children in the population by age and sex (see the circle series in Figure 1). Such a pattern is found in many countries, both in population censuses and sample surveys. Information on the age and sex of members of a household is usually provided by the household head and this often results in the omission of some young children or the misstatement of their age.

23Third, the direct fertility estimates, based either on the reported number of births in the household during the 12 months preceding the data collection operation or on the information collected in full birth histories, also present some data quality issues (see the square series in Figure 1). The number of births in the household during the previous 12 months is recorded in the census, but as this information is often provided by the household head, some births may be omitted or misplaced in time, thus artificially lowering total fertility based on these data. Sample surveys are thought to provide better quality information on fertility, not least because they collect full birth histories using an individual questionnaire completed directly with the person concerned (in the present case, women of reproductive ages). However, fertility figures estimated from full birth histories collected in a sample survey can also be affected by different kinds of bias related to the respondent, the interviewer or the length of the questionnaires (Arnold, 1990; Schoumaker, 2011). Sample selection issues can also affect the estimation of fertility. The non-inclusion of some categories of women (i.e. young single women) in survey samples may lead to over-estimation of fertility levels (Hull and Hartanto, 2009; Spoorenberg, 2014b). Figure 1 indicates that the direct estimates from the 1994 NHS and 2000 NHS are higher than those of the other series. Such patterns could be due to sample selection issues. It remains difficult to reach any firm conclusion as only limited information is available in these two survey reports and the survey micro data are not available to conduct further analysis. However, comparison of the proportion of never married women by age collected in the 2005 census, 2007 BLSS, 2010 BMIS and 2012 BLSS gives telling results (not shown). Compared to the other data sources, the proportions of never married women in the 2010 BMIS are lower, indicating that not enough single women were included in the survey sample. Fertility levels may be over-estimated as a consequence of this bias, as married women have a higher number of children.

Conclusion

24Very little is known about the demography of Bhutan. This short paper has assessed changes in the fertility levels and trends in the country based on a quasi-exhaustive list of total fertility estimates using both direct and indirect methods and the available census and sample survey data. The different series of fertility estimates indicate that Bhutan has experienced an impressive fertility decline that started in the late 1980s. From a level of about 5.5 children per woman in the mid-1990s, total fertility declined swiftly to reach a current level of about 2.2 children per woman – a level close to replacement fertility. Although all series corroborate the fertility decline, variations remain, depending on the method of fertility estimation and the data source. Only a detailed, critical examination of each data source can ultimately shed light on their respective biases affecting the estimation of fertility.

25The objective of this short paper was to lay empirical foundations for further research on fertility change in Bhutan. Using the available census and sample surveys, the fertility levels and trends can be reconstructed quite consistently. Yet, a series of questions remain unanswered, for which detailed future studies are needed. Based on the absence of sudden and noticeable breaks in fertility levels and trends, we have concluded – possibly prematurely – that the changes in population size in the late 1980s and early 1990s had little effect on the fertility levels and trends in the country. The veracity of this statement should indeed be examined in more detail by reconstructing fertility for different population groups, including those that have left Bhutan. Greater knowledge of the intermediate and social factors that have driven the fertility transition in Bhutan would also enhance our understanding of the trends observed. Finally, compared to the other Himalayan regions or countries (Nepal, Sikkim and Tibet), [4] the onset of the fertility transition in Bhutan seems to have started slightly later. Bhutan’s fertility rates have fallen faster than those of its Himalayan neighbours, and are now comparable to those observed in Sikkim [5] and Tibet [6], and are lower than in Nepal. Further comparative research is needed to investigate more fully the country-specific factors behind this spectacular decrease. Bhutan has experienced fundamental socioeconomic and political changes over the last 50 years and a critical, detailed examination of the effect of these changes would provide useful knowledge for understanding one of the world’s fastest fertility transitions to date.