1. Introduction

2. Selection of the Base Year Population

3. Reducing the Age Heaping

4. Selection of the Life Table

5. Estimation of Fertility Level and Trends

6. Distribution of Age Specific Fertility Rate (ASFR)

7. Estimation of Mortality Level and Trends

8. Estimation of Migration Level and Trends

9. Sex Ratio at Birth

10. Urban Population Projection

11. Annual Population

12. Sub-National Population Projection

12.1 Background

12.2 Methodology

12.2.1 Population Projection for Five Development Regions: Phase 1

12.2.2 Population Projection for Districts: Phase 2

12.2.3 Population Projection for Ecological Zones: Phase 3

Annex 1 : Rate of Change in Expectation of Life at Birth at Different Levels of Life Expectancy.

Annex 2 : Transformation of Life Expectancy at Birth from West Model of Coale and Demeny Life Table to General Model of United Nations Life Table.

1. Introduction

Population censuses have been carried out in Nepal at decennial intervals and they provide information about the size and structure of the population during census years. The latest census was taken in 2001. Now, based on the vast amount of information collected about population related variables in census 2001, projections have to be made to determine the future growth of the population. They have to be prepared by age and sex for districts, development regions as well as urban and rural areas. Projections can be made for social and economic subgroups also.

Population projections are needed for development planning and there is no doubt that they should be as accurate as possible. The estimates will be accurate if data used are accurate and assumptions involved in the projections hold true in reality. Hence, data need to be properly evaluated and adjusted for errors before using for projection and the most likely assumptions need to be made. However, the most plausible assumptions used in the projection will not be exact in reality. As such, other alternative assumptions have to be used. The various combinations of assumptions about fertility, mortality and migration will lead to a large number of population projections. In general, three scenarios of population projections are usually made. They are named as high, medium and low variants. The assumptions used in the medium variant represent the most likely assumptions in future giving plausible estimates of the population in future years. The other two variants can be regarded as the upper and lower limits of the variations in assumptions giving the corresponding limits for the estimated population.

The high variant represents declining mortality and fast decline in fertility, the medium variant represents the same trend of declining mortality and medium decline in fertility and the low variant represents the same trend of declining mortality and low decline in fertility. In all these three variants, net migration has been assumed to be insignificant. Here, for the population projection, the component method has been used. This method makes explicit assumptions about the components of population growth or in other words, separate projection of fertility, mortality, immigration and emigration are made.

Estimation of fertility and mortality levels and their trends are made based on the census 2001, Demographic Health Survey 2001 and other relevant literatures.

It is to be noted that the total population can be estimated with greater accuracy than the structure and distribution of populations. Likewise, the national population can be estimated with greater accuracy than sub-national populations.

Keeping in mind, the facts that the longer the period of projection, the greater the errors will be in the assumptions and the lesser utilization of the population projections, these are made for three variants High, Medium and Low for a period of 20 years starting from 2001 at an interval of 5 years. A computer program for making population projections spectrum system of policy models is used (Stover, J. et. al, 1997).

2. Selection of the Base Year Population Top

The reported and the final result of the population size according to census 2001 are as follows.

The final results of the population census of 2001 are obtained from the reported population and by estimating the population of disturbed areas in the census from the household listing or the observed growth of population between 1991 and 2001. Hence, this final result of the population of the census 2001 is taken as the base year population for the projection.

In fact, it has been estimated by post-enumeration sample survey that the reported population is under enumerated by 5.3 percent (Dangol, B.D.S., 2002, p.19). However, as the estimates of net-under enumeration by various methods vary in accuracy, conclusions about the relative accuracy of the national census must be drawn with great caution (Shryock, H.S. et.al. 1971, p107).

3. Reducing the Age Heaping Top

Analysis by various methods shows that the age distributions of 2001 census are disturbed by age misreporting. However, it is known that the proportion of population under age 15 years for both sexes combined is often less affected by age-misreporting in the census than other age groups (UN 1983, P. 166). Hence, among various techniques used in reducing age heaping, the technique mentioned in p. 241 of the same manual is selected because this technique gives the proportion of population under age 15 years closest to the reported population in the census.

4. Selection of the Life Table Top

It has been seen during the analysis of census 2001 that childhood mortality are highly under reported and adult mortality to different extent along with misreporting of age at death. Fertility and mortality decline being far away from the stable population and in the lack of estimation of net migration, the method of constructing life tables are highly questionable. Hence, model life tables with expected age pattern of mortality are selected.

Basically, there are two kinds of life tables i) Coale and Demeny and ii) UN model life tables for developing countries which are often used in demographic projections. The Infant Mortality Rate (IMR) and Child Mortality for various levels of mortality are plotted for both these life tables. The IMR and Child Mortality for both sexes combined based on 1976 Nepal Fertility survey are compared with plotted values. The comparisons show that UN General Model fits better for Nepal (UN 1990, p. 12). In fact, this model gives values of IMR and Child Mortality Rate closest to values given by DHS 2001 and 1996. As such, the “General” mortality pattern of United Nations Model life tables for developing countries is selected for the population projection.

5. Estimation of Fertility Level and Trends Top

Estimates of TFR based on various methods are shown below.

Table 1: Estimates of TFR based on various methods

The estimate of Total Fertility Rate (TFR) from the census 2001 is 1.8 births/woman. However, with adjustment for underreporting of births and declining fertility by Arriaga’s Method, the values range from 3.7 to 4.1 (UN 1988, p.73). This method is suitable when fertility and mortality are declining. The estimate of TFR 4.1 is based on children ever born (CEB) and age specific fertility rate (ASFR). The mean value of 3.9 of these two estimates can be regarded as the plausible estimate of TFR in 2001 based on the census. The estimate of TFR from DHS 2001 based on past 3 years preceding the survey is 4.1. In fact, this is an interval estimate with true value lying in between 3.9 to 4.3 with 95 percent confidence (MOH, 2002, p. 225).

In DHS surveys, though every effort is made to get correct number of births, it does not ensure that all births are enumerated. In fact, the omission of birth is difficult to detect from the data unless there is gross under-reporting (IRD, 1990, p100). There are other factors in sample surveys which affect in estimating true fertility level. Clearly, it will be affected by the inability of interviewing women who have died. Omission rates for birth reports for visitors or those who have moved away are to 6 times larger than for those interviewed at their usual residence (Seltzer, W. 1973, p23).

On the other hand, for the population projection, the estimates of true fertility levels are needed. These are obtained by the application of Arriaga’s Method on census 2001. The estimates of TFR by the same Arriaga Method to DHS 2001 data range from 4.6 to 5.5 (Joshi, P. L. 2003). The relation between TFR and contraceptive prevalence rate (CPR) namely TFR = -.06895 X CPR + 7.43671 (IRD 1991, p.31) shows TFR to be 4.7 when CPR = 39.3 percent in 2001.

However, data on live births during one year preceding the survey (September, 2000) gives TFR of 4.04 as shown in the table above. Therefore, for the base year 2001 June even at the slowest decline a TFR of 4.0 is seen to be more realistic for the projection.

With these various estimates in background, the TFR value of 4.0 in 2001 corresponding to the value given by the nation’s tenth five year plan is selected for the base year. Three kinds of fertility trends are assumed. They are as follows.

i. Fast decline in TFR

To reach the replacement level of fertility in 2017 that is by the end of 12 th plan, it is feasible to decrease the TFR by a slightly higher amount in initial years and then by 0.12 per year up to 2016 and then after maintain the fertility level constant (UN, 1992, p.13).

ii. Medium decline in TFR

The decline is in accordance with normal decline in the tenth five year plan (.10 per year) up to 2011 and then after slower decline (UN, 1992, p.13). However, once the level of TFR reaches 3.0, the further decline in TFR has to be accompanied by the increase in the proportion of married women with schooling. For example, the proportion of married women with no schooling is 42 percent in Egypt and 45 percent in Bangladesh and the TFR of this group exceeds 4 and the fertility transition stalled at slightly above 3 (Bongaarts, J. 2003, p.22). Infact, in Bangladesh, fertility stalled at 3.3 inspite of increasing CPR from 44.6 percent in 1993/94 to 53.8 percent in 1999/00 (Kamal, N and Chaudhary, R.H. 2003, p. 157). In Taiwan, where the overall TFR was 1.7 in 1991, fertility estimates for primary school graduates was 3.6 births per woman (Bongaarts, J. 2003, p.20). In Nepal, the proportion of married women with no schooling was 80 percent in 1996, which decreased to 72 percent in 2001 (MOH, 1997, p. 23 and MOH, 2001, p. 25). Hence, with the sluggish decrease of non-schooling among married women, it is very likely that TFR might be stalled at around 3 in spite of vigorous effort in increasing the CPR and decreasing the proportion of married women with no schooling. As such, TFR is estimated to be 2.8 in 2021. This is also compatible with the mean desired number of children of 2.6 as shown by DHS, 2001, p.123.

iii. Slow decline in TFR

This is in accordance with G. Feeney’s trend of slow decline (MOPE 1998).

6. Distribution of Age Specific Fertility Rate (ASFR) Top

Distribution of ASFR in the base year is taken from DHS 2001. For future years, the fertility is assumed to decrease in the younger and older age groups. Distribution of ASFR in percentages is shown below.

7. Estimation of Mortality Level and Trend Top

As has been noted in the selection of the life table, the estimation of expectation of life based on constructing life tables are highly questionable. Moreover, most of the methods give valid results if the population is stable. But as the population in 2001 is far away from stable, the expectation of life at birth is obtained by the method based on the proportion of population under age 15 years and the probability of surviving to age five years (UN, 1983, p. 167). The expectation of life at birth for males and females are 60.1 and 60.7 years respectively in 2001.

The mortality trend for males is assumed according to G. Feeney’s scale in MOPE, 1998. The difference between expectation of life at birth for females and males in 2021 is assumed to be 1.4 years which is in accordance with medium variant population projection made by United Nations Population Division for Nepal (UN, 2000, p.338). The level in 2001 and the trend are shown below.

8. Estimation of Migration Level and Trend Top

Migration data in census 2001 are analyzed based on citizenship, persons absent and living in other countries, place of birth, duration of stay in the current place and place of residence 5 years before (Joshi, P.L. 2003). However, the adjustment of net migration are not made as the estimate of net migration based on survival ratios are -492000 during the period 1991 to 2001 which are not consistent with the estimate based on foreign born population and absentee population. Survivorship ratio method applied to 2001 and 1991 age distributions also shows inconsistent estimate of net migration.

The estimates of net migration by various methods vary in accuracy and the adjustment in the population projection needs great caution. Moreover, the effects of net migration have been estimated to be about 2 percent and as such, have been assumed to be insignificant.

9. Sex Ratio at Birth Top

The sex ratio at birth is estimated to be 1.05 male births per female birth (Joshi, P. L. 2003). This is assumed to be constant throughout the period of the projection.

10. Urban Population Projection Top

Urbanization is taking place rapidly in Nepal since the last several years. In 1991 the number of urban areas was only 33. This number increased to 58 in 2001. For the population projection, urban population is defined as population residing in urban areas. As such urban populations are affected by several factors like natural growth, migration and reclassification of areas. Urban – Rural population projections are made by United Nations method of urban and rural population projections which are based on the following equation:

Where, T, U and R are the total, urban and rural population respectively for the year t and T’, U’ and R’ are corresponding values for the year t+1; d = u – r where, u and r are the urban and rural rate of increase respectively (UN 1974, p. 36).

Hence, if the values of T, U and R are known for the base year t, then the total urban population for the year t+1 can be obtained by putting the value of d. In this projection, the value of d is obtained as the difference between urban and rural population growth rate during 1991 and 2001. Proceeding successively, the total urban population is obtained for each year from 2001 to 2021.

Here for the purpose of projection, urban-rural growth rate difference is set at 4.73% (for urban, it is 6.44% and 1.71% for rural areas). The projection is made by using software package called ‘Spectrum’ developed by the policy project of policy modeling system (Stover, J. et. al, 1997).

Annual total population by sex has been computed by the procedure know as the “Central difference method” using first and second differences (UN 1974, p. 38).

12. Sub-National Population Projection

12.1 Background Top

Nepal is pursuing decentralized development planning. Population information, as explained earlier is one of the main foundations for development planning. Since in addition to national level development planning, district /village level or local level planning also has been placed as the basic development planning units in the country. Therefore at least district level population information is indispensable for all planning activities. Similarly, other sub-national level projections like development region/ecological zone level are needed for various developmental activities.

This section presents the sub-national projections including district level projections based on the national projections. Other sub-national projections included in this volume are development regions and ecological zones. Definition of ecological zones and development regions are in accordance with the standard definition used by His Majesty’s Government of Nepal.

The projection methodology used for sub-national projections is the changing ratio method. This method was used due to unavailability of reliable estimates of population parameters at the sub-national level. In the following sections, the projection methodology is described in detail.

12.2 Methodology Top

Analysis of the trends on three components of population change i.e. fertility, mortality and migration in the development regions/districts/ecological zones are constrained by the availability of reliable data. This has precluded the possibility of using component method for projecting the sub-national populations as we did on national level population projection. For this reason the changing ratio method has been used for projection of the population at sub-national levels.

The sub-national projections were completed in three phases. In the first phase, projections were made for five development regions of the country. In the second phase district level population projection were made based on the population projection of each development region. Population for three ecological zones was obtained by adding the projected populations of districts within the respective zones.

12.2.1 Population Projection for Five Development Regions: Phase 1 Top

The ratio method is mainly based on the relationship between the populations of the development regions with that of the nation. The age and sex distributions of projected population of development regions were obtained in three steps.

Step 1: Projection of total populations in five development regions Top

Following three procedures were used in the first step to obtain the total projected population in the development regions in the years 2001, 2006, 2011, 2016 and 2021.

1. Ratios of the population of development regions to the total population of the country were calculated from 1991 and 2001 censuses.

2. Based on the observed ratios of five development regions in 1991 and 2001 future ratios were projected using changing ratio methods as described below.

The average annual rate of change ‘r’ in the population proportion (ratio) between the 1991 and 2001 censuses is calculated using the following formula.

r = [1/10] * Ln [R 2001/R 1991]

Where,

R 2001= Proportion of population of the development region to the national population in the 2001 census.

R 1991 = Proportion of population of the development region to the national population in the 1991 census.

Ln = Natural log.

Half of the r-value is taken as annual rate of change ‘. r b ‘ at the base year, i.e. r b = [r/2]

Theoretically, two options can be considered for the projection of observed population ratios of development regions to the national population. Firstly, the ratios of five development regions may be held constant at the last observed level. Or, secondly, the ratios may be modified in some way to take account of the past trend. Therefore, ratios can be modified on the basis of following two alternative assumptions:

Alternative 1: The observed trend in ratios will continue for many years to come.

Alternative 2: The observed ratios will approach a stable condition after a certain period of time. In other words, there will be no change in the ratios after that period of time. The ratio may be assumed to approach a stable condition after certain number of years on the ground that differences in fertility and mortality will have disappeared and net migration will have fallen off to zero for each area (Shryock and Siegel. 1976. Page 454).

The ratio of the population of the development region to national population in practice can increase or decrease. Therefore, the first option is not plausible. In this projection, the ratio of the population of the development regions to national population was projected on the assumption that ratios will approach stable condition after 30 years from the base year of projection, i.e., 2001. In other words, it is assumed that ratios will stabilize by the year 2031. This assumption implies that annual rate of change ‘r b’, will drop to zero, linearly, in 30 years from the base year of projection. With these rates of change, the development region ratios are projected by geometric law at five-year intervals. The formula used for this is:

R U 2006 = R 2001 [1 + r b(29/30)] [1 + r b(28/30)] [1 + r b(27/30)] [1 + r b(26/30)][1 + r b(25/30)]

Where,

R U 2006 = Unadjusted ratio of a development region.

The R U 2006 for each development regions was adjusted on the pro-rata basis in such a way that the sum of adjusted ratio is unity. The adjusted ratios are denoted by R 2006.

Similar formulas are used to calculate the values of R 2011, R 2016 and R 2021.

3. The projected ratios were applied to the independently derived projection of the national population to get the total population in the five development regions. For this projection, the medium variant projection of population had been used. The formula used can be put as:

P D 2006 = R 2006 X P 2006

Where,

P D 2006 = Population of the development region in 2006

P 2006 = Projected national population under the medium variant for 2006.

Similar formulas are used to calculate the values of P D 2011, P D 2016 and P D 2021.

Step 2: Calculation of total population of development regions by sex. Top

1. First, the sex ratios in the future years were obtained by projecting the observed sex ratios in the base year in each development region as explained below.

The sex ratios of five development regions in 2001 census were observed for each development region. The observed sex ratio of 2001 would not be the same throughout the projection period. On the other hand, the ratio would not change indefinitely. Therefore we have to assume the sex ratio of each development region observed in 2001 will approach the projected national sex ratio in 2021. Thus under this assumption the sex ratio of each development region observed in 2001 was projected linearly for future years so that in each development region sex ratio will approach the projected national sex ratio in the year 2021.

SR 2006, SR 2011, SR 2016 and SR 2021 denotes the projected sex ratios for five projection years respectively.

2. In the second stage, applying the projected sex ratios for each projection years, the projected population by sex in each development region was obtained by using the following formula.

For example,

P DFU 2006 = P D 2006 X [1/(1+SR 2006)]

Where,

P DFU 2006 = Unadjusted female population in a particular development region in the year 2006.

P DMU 2006 = Unadjusted male population in the development region in the year 2006.

(The male population in a particular development region was obtained by subtracting the female population from the total population of that region.)

These unadjusted projected male and female populations were adjusted so that the sum over all the development regions was equal to the projected national male and female populations.

Therefore,

P DF 2006 = Adjusted female population in a particular development region in the year 2006.

P DM 2006 = Adjusted male population in the same development region in the year 2006.

Step 3: Distribution of male and female population in five-year age groups in each development region.

Finally, in the third step, total male and female populations of each development region of the nation was distributed in five-year age groups by applying the observed age structure of the concerned development regions in the base year, i.e., 2001 by the method developed by Deming so that the result on the sum of age distribution of population by sex of five development regions exactly tally with the national population by age and sex for the years, 2001 to 2021.

(The Deming row-column procedure is an iterative process that forces the cells of a two- way tabulation to sum to prescribed marginal values. The initial row sums are forced to the given row sums by multiplying each row by a suitable constant. The column sums in the resulting table are then forced to the given column sums by multiplying each column by a suitable constant. This procedure is then repeated until the prescribed row and column marginal is matched to the desired degree).

12.2.2 Population Projection for Districts: Phase 2 Top

Exactly the same methodology as explained step by step above was applied to project the population by the districts under each development region separately.

12.2.3 Population Projection for Ecological Zones: Phase 3

Each development region was divided into three ecological zones. Projected populations for each ecological zone are obtained by adding the projected population for all the districts in that zone.

References Top

Arriaga, E.E. et.al, 1992, Population Analysis with Micro Computers, Centre for International Research, US Bureau of Census.

Bongaarts, J., 2003, Completing the Fertility Transition in the Developing World: The Rate of Educational Differences and Fertility Preference Population Council No. 177.

Census Bureau of Statitics (CBS) 2002, Population Census 2001, National Report, Kathmandu, Nepal.

Dangol, B.D.S, 2002, Post Enumeration Survey Report, 2001. Submitted to UNFPA/CBS, Kathmandu.

Institute for Resource Development Inc. 1990, Methodological Reports – I, An Assessment of DHS – I Data Quality, Maryland, USA.

Institute for Resource Development Inc. 1991, DHS Comparative Studies, Maryland, USA.

Joshi, P.L, 2003, The Quality of Census Data 2001 : An Evaluation in Population Monograph of Nepal 2003 (Forth coming).

Kamal, N. and Chaudhary, R.H. 2003, Plateauing of Total Fertility Rate in Bangladesh : An Exploratory Analysis in Asian Profile, Vol 31, No. 2, Canada.

Ministry of Health (MOH), New ERA, Nepal and Macro International Inc, USA 1997, Family Health Survey 1996, Kathmandu, Nepal and Maryland, USA.

Ministry of Health (MOH), New ERA, Nepal and ORC Macro, USA 2002, Nepal Demographic and Health Survey (DHS) 2001, Kathmandu, Nepal and Maryland, USA.

Ministry of Population and Environment (MOPE) 1998, Population Projections for Nepal 1996 – 2016, National and Urban Projection, Vol I, Kathmandu.

Population Reference Bureau 2002, Population Data Sheet, Washington D.C.

Seltzer, W. 1973, Demographic Data Collection, An Occasional Paper of the Population Council, Connecticut

Shryock, H.S., Siegel, J.S. and Associates, 1971. The Methods and Materials of Demography, Washington, D.C.

Shryock, S. and Siegel, S. 1976. The Methods and Materials of Demography. Condensed Edition by Edward G. Stockwell. Academic Press, Inc., London, New York.

Stover, J and Kirneyer, S. 1997, Demproj Version 4, A Computer Program for Making Population Projections Spectrum System of Policy Models, The Futures Grup International and Research Triangle Institute.

United Nations (UN) 1974, Manual viii Methods for Projections of Urban and Rural Population, New York.

United Nations (UN) 1983, Manual X Indirect Techniques for Demographic Estimation, New York.

United Nations (UN) 1988, Mortpak-Lite, The United Nations Software Package for Mortality Measurement. New York.

United Nations (UN) 1990, Step by Step Guide to the Estimation of Child Mortality. Population Studies 107.

United Nations (UN) 1992, Patterns of Fertility in Low Fertility Settings , New York

United Nations (UN) 2000, World Population Prospects : The 2000 Revision, Vol. 1 : Comprehensive Tables, New York.

Annex 1 : Rate of Change in Expectation of Life at Birth at Different Levels of Life Expectancy.

Notes : Estimated form United Nations Population Division estimates of life expectancy for 181 countries for five year periods from 1950-55 through 1985-90. See text for further explanation. The two columns show the relationship between level of expectation of life at birth in a given five year period and the change in expectation of life at birth between this and the following period. To illustrate using the first scaling, if expectation of life at birth for 1986-91 is 50 years, expectation of life at birth for 1991-96 would be 50 + 2.40 = 52.40 years. Alternative scalings allow for the very different rates of increase observed in different countries. The second scaling is used for extrapolating expectation of life at birth in Nepal for future pierods based on the rate of increase estimated in previous sections.

Source: Griffith Feeney, 1998. Population Projection for Nepal 1996 2016, Volume 1, Ministry of Population and Environment HMG, June 1998. Top

Annex 2 : Transformation of Life Expectancy at Birth from West Model of Coale and Demeny Life Table to General Model of United Nations Life Table.

* Transformation from West Model compatible to General Model

** General Model.

Computed by B.D.S. Dongol