Fertility in Bangladesh is high even by the standards of developing countries. Recent evidence suggests that fertility has started to decline in Bangladesh (Amin and others, 1993). The total fertility rate has declined from nearly seven births per woman in 1975 to about five births per women in 1989 and by 1990 this rate was well below five births (Cleland and others, 1994; Caldwell and Caldwell, 1992). A number of demographers have argued that the mechanism of this steep fertility decline was the consequence of a recent increase in contraceptive prevalence within marriage (Amin and others, 1990; Cleland and others, 1994; Cleland, 1993). In Indonesia, research suggests that fertility decline resulted mainly from a high rate of increased use of contraception which was induced primarily through economic development and an increasing rate of female education as well as greater work force participation (Gertler and Molyneaux, 1994). It is of great concern to policy makers to know the reasons why other socio-economic, demographic and cultural variables do not seem to contribute to a decline in fertility in Bangladesh. Such variables are important for a study of fertility; investigations are needed in order to produce findings. Reliable information about the factors influencing fertility is indispensable in the process of planning for the overall socio-economic development of a developing country such as Bangladesh. Human fertility is the outcome of the function of a number of variables within a complex process. The mechanism of factors affecting fertility is that intermediate variables influence fertility directly, while socio-economic and environmental variables affect fertility indirectly through intermediate variables (see, for example, Davis and Blake, 1956; Bongaarts, 1978; Bongaarts and others, 1984). This study is an initial framework for the classification of variables to be analysed using the path analytical approach. In the context of Bangladesh, only a few studies, not all of them nationally representative, have been carried out to examine the effects of various factors on fertility (Ahmed, 1981; Rob and Kabir, 1988; Islam and Khan, 1991). These studies provide very useful information. Ahmed's study was based on national data of the 1975 Bangladesh Fertility Survey and two other studies based on a micro-level study.

___________

* The authors of this paper are S. M. Shafiqul Islam, Professor, Department of Statistics, University of Chittagong, and H.T. Abdullah Khan, Lecturer, Department of Statistics, University of Dhaka, Bangladesh. Correspondence should be addressed to: H.T. Abdullah Khan, Research Assistant, Department of Mathematics, Napier University, Sighthill Court, Edinburgh EH11 4BN, Scotland, United Kingdom.

The second national Bangladesh Fertility Survey was conducted in 1989 and its data are available for analysis, the results of which may provide new knowledge useful for national population policy-making. Therefore, the purpose of the present paper is to study the direct, indirect and joint influences of various socio-economic and demographic factors on fertility in Bangladesh, employing the technique of path analysis.

Data

The data used in this study are from the 1989 Bangladesh Fertility Survey (1989 BFS), which was conducted on behalf of the Government of Bangladesh by the National Institute of Population Research and Training (NIPORT) during the period December 1988 to April 1989. A two-stage probability sample design was used for the survey. At the first stage, a sample of areas was drawn and all households in each of the selected areas were listed. A sample of households in each of the selected areas was drawn at the second stage and finally a total of 11,905 ever-married women below 50 years of age belonging to the selected households were successfully interviewed. A detailed description of the methodology of data collection including sample design for the survey can be found elsewhere (Huq and Cleland, 1990).

Selected variables

In the 1989 BFS a number of socio-economic, demographic and cultural variables are available. Among them, nine variables have been taken into consideration in the present study in order to construct a path model. Table 1 gives a detailed description of the variables. Respondent's education, religion, place of residence and possession of modern objects in the family are considered as the socio-economic and background variables, while the demographic as well as intermediate variables regarded in this study are age at first marriage, contraceptive use, duration of breast-feeding, and fetal loss. Finally, the number of children ever born to each woman is used here as a measure of fertility. It has been assumed here that these selected variables affect fertility behaviour of women according to the theoretical framework of Bongaarts and others (1984:515).

Variable	Abbreviation	Measurement
Wife's education (X1)	REDU	1 = No schooling 2 = Lower primary 3 = Upper primary 4 = Higher
Religion (X2)	REL	1 = Muslim 0 = Other
Place of residence (X3)	PR	1 = Rural 0 = Other
Possession of modern objects (X4)	PMO	Number of household possessions
Age at first marriage (X5)	AFM	Completed years
Contraceptive use (X6)	CU	1 = Ever use 0 = Never use
Duration of breast-feeding in last closed birth interval (X7)	DBF	Completed months
Fetal loss (X8)	FL	Number of wasted pregnancies
Children ever born (X9)	CEB	Number of live births

Analytical method

The technique employed in this study to examine the effects of various factors on fertility is path analysis. This technique has been used by many social scientists in various fields of enquiry for quantifying and interpreting causal linear models (Ahmed, 1981; Duncan, 1966; Loebner and Driver, 1973; Shin, 1977; Balakrishnan and others, 1980; Miller and Stokes, 1985; Tan, 1987). One of the main tasks in path analysis is to construct a path diagram in which variables are arranged in a meaningful manner per se; the direction of influences is shown by arrows. The path diagrams (figures 4 and 5) represent the hypothetical causal model of relationships between fertility and some of its determinants. The model considered here is the recursive type in which each variable is assumed to be dependent upon all prior causal variables. As can be seen from the schematic diagram, variables REDU, REL, PR, PRO have effects on fertility, i.e. children ever born (CEB), through AFM, CU, DBF and FL.

The system of equations for the model considered may be written as below, which will provide information on the estimation of influences of variables:

x5 = p54x4+p53x3+p52x2+p51x1+p5ueu

x6 = p65x5+p64x4+p63x3+p62x2+p61x1+p6vev

x7 = p75x5+p74x4+p73x3+p72x2+p71x1+p7wew

x8 = p85x5+p84x4+p83x3+p82x2+p81x1+p8xex

x9 = p98x8+p97x7+p96x6+p95x5+p94x4+p93x3+p92x2+p91x1+p9yxy

where pji values are path coefficients from standardised variables xi to xj, and eu, ev, ew, ex and ey are random disturbance terms.

Path analysis permits one to estimate the direct, indirect and joint effects of predetermined variables on the dependent variable by decomposing the correlation coefficient between the dependent variable and any predetermined variable according to the following fundamental theorem:

where j and i are two variables and q runs over all the variables that have direct paths to variable j (e.g. Duncan, 1966; Shin, 1977).

It is well documented that the current age of women has a high correlation with their fertility, and fertility variations may be observed due to characteristics of women between these two age cohorts. Figure 1 reveals that fertility increases with higher order cohorts of women. On the other hand, split data for women aged below 30 years and women 30 years or older show different types of distribution. The former (figure 2) shows a skewed distribution; however, the later (figure 3) shows more or less symmetric distribution. The average parity is seen to be merely two per woman aged below 30 years, whereas it is about six per woman aged 30 years or older (see table 2). Thus, the path model has been applied separately to two age cohorts of women: (a) the cohort aged below 30 years at the time of interview (younger cohort) and (b) the cohort aged 30 years or older at the time of interview (older cohort). Table 2 presents the means (X) and standard deviation (S.D.) of the variables included in this study for the two age cohorts of women.

_

Variable	Women aged below 30 years			Women aged 30 years or older
	N	X	S.D.	N	X	S.D.
REDU	6,536	1.846	1.167	5,369	1.604	1.02
REL	6,536	0.874	0.331	5,369	0.847	0.359
PR	6,536	0.716	0.450	5,369	0.705	0.456
PMO	6,536	3.595	2.225	5,369	3.657	2.216
AFM	6,536	15.053	2.324	5,369	14.501	2.535
CU	6,536	0.453	0.498	5,369	0.534	0.499
DBF	2,670	17.587	10.381	1,289	18.097	9.699
FL	6,536	0.219	0.634	5,369	0.520	1.044
CEB	6,536	1.965	1.604	5,369	5.691	2.633

Note: See table 1 for abbreviations.

Various path coefficients are shown in figure 4 for the cohorts of women aged below 30 years and in figure 5 for the cohorts of women aged 30 years and older. Out of 23 hypothesised paths, 12 and 10 paths are found to be statistically significant for the younger and older cohorts, respectively. The direct, implied, indirect and joint effects of each of the selected explanatory variables are presented in table 3 for the cohort aged less than 30 years and in table 4 for the cohort aged 30 years or older. For both the cohorts, age at first marriage and duration of breast-feeding are found to have significant direct negative effects while the number of fetal losses has a significant direct positive effect on the number of children ever born. Compared with the younger cohort, both the direct influence of age at marriage and duration of breast-feeding on fertility are observed to be higher for the older cohort. The direct influence of the number of wasted pregnancies on the number of children ever born is found to be higher and significant for women aged below 30 years than for those aged 30 years or older; however, their signs are in expected directions.

Total effects of respondents' education and possession of modern objects on fertility are negative while those of religion and place of residence are positive for both cohorts. This indicates that women with more education and possessing more modern objects in the family to which they belong have fewer children ever born to them as compared with those having less education and fewer modern objects, whereas Muslim women and rural women have more children ever born to them as compared with their non-Muslim and urban counterparts.

The total effect of the respondent's education on fertility is found to be more pronounced among women belonging to the older cohort than to the younger cohort. Of the total effect, -0.177 of respondent's education on fertility (X9) for women aged 30 or older, -0.102 (57 per cent) is transmitted through its implied effect (p91), -0.0793 (45 per cent) through its indirect effect via age at marriage (X5) in the same direction, while 0.0092 (5 per cent) is transmitted through joint association in the opposite direction. Other indirect effects of the respondent's education via X6, X7 and X8 and via second order interactions X5X6, X5X7, X5X8 on X9 are negligible. Although the indirect effect of REDU on CEB transmitted through contraceptive use is negative (-0.0080) in the case of women belonging to the older cohort, the indirect effect (0.0085) of REDU acts through increasing contraceptive use that consequently decreases fertility among women belonging to the younger cohort.

Higher total positive influences of religion on fertility and of place of residence on fertility are observed among women aged 30 years or older than those below 30 years. This indicates that fertility among older Muslim women and older rural women is higher than among their non-Muslim and urban counterparts, respectively. It should be noted that the implied effect (p92) of religion has contributed about 58 per cent of its total effect on fertility, while implied effect (p93) of place of residence has contributed about 69 per cent of its total influence on fertility among women aged 30-49 years.

For both age cohorts, fertility, as expected, is found to be lower among women belonging to the households possessing more modern objects. However, the implied effect of the possession of modern objects on fertility is not in the expected direction in the case of both cohorts; the bivariate analysis shows its significant negative effects on fertility. Explanations can be given likewise to multicollinearity among variables, although for the regression, it is not found to be serious. Further research is needed to investigate this situation in order to clarify this interpretation.

Ideally, a zero-order correlation coefficient between CEB and any predetermined variable should be the same as the total effect of that predetermined variable on CEB. A closer look at tables 3 and 4 reveal that, with few exceptions, the zero-order correlation coefficients between CEB and each of the selected predetermined variables generally do not differ much from their corresponding total effects. It should be pointed out that about 10 per cent and 22 per cent of variances of fertility for women belonging to the younger and older cohorts, respectively, have explained the selected predetermined variables which are significant at the 0.001 level. These low proportions of explained variance may be attributed to the following: (a) there might be more explanatory variables that could not be included in the model and (b) individual rather than aggregate data have been employed in this study.

Some findings of this study deserve consideration from the viewpoint of their policy implications. It has been found that female age at marriage has a significant direct negative influence on fertility. Thus, raising the age at marriage by implementing a minimum-age marriage law is likely to lower fertility on a national scale. Duration of breast-feeding is also found to have a significant direct negative effect on fertility. Encouraging women to breast-feed their children for a relatively longer duration may also contribute to a reduction in fertility. Fetal loss appears to have a significant direct positive effect on fertility in Bangladesh, which means that mothers who have experienced fetal loss are found to have higher fertility. Mothers always try to replace their dead child as early as possible. Such behaviour is a result of social fear about the survival of children. Maternal mortality is also high in Bangladesh. Therefore, it is essential to provide primary health care, particularly maternal and child health care, for surviving children.

The total effect of female education on fertility is found to be negative. Education may provide better employment opportunities outside the home and age at marriage can be raised by providing education to females, especially at the secondary and higher levels. Based on the findings of this study, it may be suggested that attention should be focused on the need for providing educational facilities, particularly for Muslim women in rural areas in order to depress the level of fertility in Bangladesh.

Household possession of modern objects has a negative influence on fertility. This finding reveals that, if women have access to modern objects in the house such as radio and television, their fertility is likely to be lower than those who do not have these modern objects. Most women in Bangladesh work at home as housewives, for example, cooking, maintaining the home, taking care of children and so on. They rarely go outside the home. If they have at least a radio, they can hear population programmes, gain knowledge of population issues and learn about family planning, since it is through this mass medium that many family planning messages and related health information are transmitted. Women can gain great advantage from such modern objects. In a study of five African countries, Kojima (1993) suggested that socio-economic development policies and family planning programmes giving special emphasis to the mass media, especially radio, may produce significant desired effects with regard to fertility control.

Ahmed, B. (1981), "Differential fertility in Bangladesh: a path analysis", Social Biology, 28(1-2):102-110.

Amin, R., S. Becker and J. Chowdhury (1990). "Recent evidence on trends and differentials in Bangladesh fertility", Journal of Biosocial Science, 22:225-230.

Amin, R., J. Chowdhury, A.U. Ahmed, R.B. Hill and M. Kabir (1993). "Reproductive change in Bangladesh: evidence from recent data", Asia-Pacific Population Journal, 8(4):39-58.

Balakrishnan, T.R., G.E. Ebanks and C.F. Grindstaff (1980). "A multivariate analysis of the 1971 Canadian Census Fertility Data", Canadian Studies in Population, 7:81-98.

Bongaarts, J. (1978). "A framework for analyzing the proximate determinants of fertility", Population and Development Review, 4(1):105-132.

__________, O. Frank and R. Lesthaeghe (1984). "The proximate determinants of fertility in sub-Saharan Africa", Population and Development Review, 10(3):511-537.

Caldwell, J.C. and P. Caldwell (1992). "What does the Matlab fertility experience really show?", Studies in Family Planning, 23(5):292-310.

Cleland, J. (1993). "Equality, security and fertility: a reaction to Thomas", Population Studies, 47(2):345-352.

__________, J.F. Phillips, S. Amin and G.M. Kamal (1994). The Determinants of Reproductive Change in Bangladesh: Success in a Challenging Environment (Washington, D.C.: World Bank).

Davis, K. and J. Blake (1956). "Social structure and fertility: an analytical framework", Economic Development and Cultural Change, 4(3):211-235.

Duncan, O. D. (1966). "Path analysis: sociological examples", The American Journal of Sociology, 72(1):1-16.

Gertler, P.J. and J.W. Molyneaux (1994). "How economic development and family planning programs combined to reduce Indonesian fertility", Demography, 31(1):33-63.

Huq, M.N. and J. Cleland (1990). Bangladesh Fertility Survey 1989, Main Report (Dhaka: NIPORT).

Islam, S.M.S. and H.T.A. Khan (1991). "Effects of selected socio-economic and demographic factors on fertility: a path analysis", Asian Profile, 19(6):561-574.

Kojima, H. (1993). The effects of mass media on contraceptive and fertility in African countries. Paper presented at the IUSSP International Population Conference, Montreal, 24 August-1 September.

Loebner, H. and E.D. Driver (1973). "Differential fertility in central India: a path analysis", Demography, 10(3):329-350.

Miller, M.K. and C.S. Stokes (1985). "Teenage fertility, socioeconomic status and infant mortality", Journal of Biosocial Science, 17(2):147-155.

Rob, A.K.U. and M. Kabir (1988). "A path model for analyzing sequential fertility in Bangladesh", Journal of Statistical Studies, 8:8-21.

Shin, E.H. (1977). "Socioeconomic development, infant mortality and fertility: a cross-sectional and longitudinal analysis of 63 selected countries", Journal of Development Studies, 13(4):398-412.

Tan, Boon Ann (1987). "Multivariate areal analysis of the impact and efficiency of the family planning programme in Peninsular Malaysia", Asia-Pacific Population Journal, 2(2):45-66.

					Indirect effects through
Dependent variable	Predetermined variable	Direct effect	Implied effect	Joint association	X5	X6	X7	X8	X5X6	X5X7	X5X8	Total effect	Zero order correlation
X9	X1		-0.078	-0.0157	-0.0392	0.0085	0.0007	-0.0013	-0.0000	0.0008	-0.0003	-0.124	-0.118
	X2		0.031	0.0034	0.0208	-0.0010	0.0029	0.0010	0.0000	-0.0004	0.0001	0.058	0.159
	X3		0.067	0.018	0.0034	-0.0075	-0.0033	0.0014	0.0000	0.0000	0.0000	0.079	0.044
	X4		0.027	-0.0395	-0.0237	0.0075	0.0024	0.0016	0.0000	0.0005	-0.0002	-0.024	-0.081
	X5	-0.245				-0.0001	0.0055	-0.0022				-0.242	-0.291
	X6	0.057										0.057	0.181
	X7	-0.123										-0.123	-0.109
	X8	0.053										0.053	0.124

Table 4: Effects of variables used in the path model for explaining fertility of ever married wormen (30+ years)

					Indirect effects through
Dependent variable	Predetermined variable	Direct effect	Implied effect	Joint association	X5	X6	X7	X8	X5X6	X5X7	X5X8	Total effect	Zero order correlation
X9	X1		-0.102	0.0092	-0.0793	-0.0080	0.0017	0.0003	-.0008	0.0024	-0.0004	-0.177	-0.198
	X2		0.049	0.0004	0.0301	0.0014	0.0025	0.0012	0.0003	-0.0009	0.0001	0.084	0.115
	X3		0.089	0.0191	0.0172	0.0054	-0.0009	-0.0004	0.0001	-0.0005	0.0000	0.129	0.137
	X4		0.039	-0.0895	0.0205	-0.0069	0.0086	0.0005	-.0002	0.0006	-0.0001	-0.027	-0.078
	X5	-0.367				-0.0039	0.0113	-0.0018				-0.361	-0.269
	X6	-0.056										0.056	-0.047
	X7	-0.195										-0.195	-0.164
	X8	0.036										0.036	0.052

For further information on this material please contact: loftus.unescap@un.org