Preferences for family size or for sex of a child reflect the values attributed to children within a given cultural setting as well as individual considerations: such preferences indicate the demand for children (United Nations, 1987). In traditional societies family size preferences are found to be greater than actual fertility, but in developing countries the family size preferences are lower than actual fertility; in developed countries the two are similar (Ware, 1974).

Although demographers have been successful in identifying direct determinants of actual fertility (Bongaarts, 1978), they have not yet been able to identify the direct determinants of reproductive preferences. Moreover, socio-economic differentials of actual fertility have long been known, but socio-economic differentials of reproductive preferences were almost unknown until recently. Using the percentage of women wanting more children as contained in World Fertility Survey data, Brackett and others (1978) were the first to dispute the hypothesis that uneducated women had unrestricted fertility desires or desired a very much higher number of children than educated women. In subsequent studies using World Fertility Survey data, the United Nations (1981) and Lightbourne (1984) examined both the percentage of women wanting more children and mean desired family size, and reported that reproductive preferences vary little by socio-economic status.

The issue of whether desired family size or actual fertility falls first is yet to be resolved. In Costa Rica, fertility preferences appeared to change very little either before or during the decade when fertility itself was falling dramatically (Stycos, 1984). In the Republic of Korea (Cho and others, 1982), Thailand (Knodel and others, 1982) and Taiwan Province of China (Sun and others, 1978), desired family sizes fell only after the fertility transition was under way.

____________

* The author of this article is Research Fellow, Population Studies Centre, International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), G.P.O Box-128, Dhaka-1000, Bangladesh. The article is based on the author's Ph.D thesis, the research for which was supported by ICDDR,B. The author would like to acknowledge with gratitude the assistance of Gordon Carmichael and Alan Gray of the Australian National University, Canberra.

In the treatment area of Matlab in Bangladesh, the total fertility rate in 1990 was 3.6, whereas it was 5.2 in the comparison area. The two areas are virtually the same with respect to social and economic setting, but differ in the availability and use of contraception. The objectives of this article are to examine the levels, motivation and differentials of reproductive preferences in the treatment and comparison areas of Matlab in 1990-1991. More specifically, it examines whether reproductive preferences and the motivation behind the preferences are similar in the two areas and whether reproductive preferences vary by socio-demographic subgroups.

The study area

The data for the present study come from Matlab upazila (sub-district) where the International Centre for Diarrhoeal Disease Research, Bangladesh, has been maintaining the Demographic Surveillance System since 1966. Matlab is about 70 km south-east of Dhaka, the national capital. The area is a low-lying deltaic plain intersected by the tidal river Gumti and its numerous canals (for details, see Ruzicka and Chowdhury, 1978; D'Souza, 1981).

The Matlab study area consists of both treatment and comparison areas. The treatment area was exposed to the Contraceptive Distribution Programme during the period 1975-1977 and has been exposed to the Family Planning and Health Services Programme since October 1977, while in the comparison area services are limited to those received through the conventional government service programme (for details, see Bhatia and others, 1980). Contraceptive use was low in both areas in 1975, but increased to 12.6 per cent in 1977 and to 31.1 per cent in 1978 in the treatment area after introduction of the Family Planning and Health Services Programme and remained at the same level until 1982. Since then contraceptive use has been increasing, reaching 60.6 per cent in 1990. In the comparison area, contraceptive use is much lower than in the treatment area although it has also been increasing, i.e. from 4.7 per cent in 1977 to 16.5 per cent in 1984 and to 27.2 per cent in 1990.

A remarkable decline in fertility has been observed in the treatment area since the introduction of the Family Planning and Health Services Programme in 1977. The total fertility rate declined from 6.9 children per woman in 1976 to 5.1 in 1980, 4.1 in 1987 and 3.6 in 1990, while the total fertility rate in the comparison area also declined, from 7.2 in 1976 to 6.7 in 1980 and 5.2 in 1987 and 1990.

Partly as a result of the Family Planning and Health Services Programme, the crude death rate in the treatment area declined from 12.5 per thousand in 1978 to 10.0 per thousand in 1985 and 7.6 per thousand in 1990, while the crude death rate in the comparison area declined with some fluctuations, from 13.8 per thousand in 1978 to 14.1 per thousand in 1985 and 9.4 per thousand in 1990. The general mortality decline, however, has been due in particular to improvements in infant and child mortality.

The data

As the measurement of reproductive preferences is controversial, both quantitative and qualitative data were used to examine motivation for ideal family size. In order to examine the levels and differentials in reproductive preferences, two widely used indicators, namely desired family size and desire for more children, were employed.

The study was based on three data sets: the Knowledge, Attitude and Practice (KAP) survey of 1990, Socio-economic Survey (SES) of 1982 and Qualitative Survey (QS) of 1991. The primary objective of the KAP survey was to provide updated information on contraceptive use, reproductive preferences, and maternal and child health service performance for both the treatment and comparison areas (Koenig and others, 1992). A multi-stage sampling procedure was employed: 31 villages were randomly selected from the treatment area and 36 villages from the comparison area. The number of respondents interviewed totalled 4,238 in the treatment area and 3,708 in the comparison area. In the KAP survey, women were asked: "Do you want more children in the future?" "How many more do you want to have?" and "How long do you wish to delay your next child?" Women were also asked: "If you could begin childbearing again, how many children would you like to have?" The last SES was conducted in 1982 and the present study uses information from it on religion and possession of household items.

A qualitative interview programme, lasting for five weeks, was conducted in July 1991. Collected was information on family size ideals ("How many children do you consider ideal for a newly married couple?"), reasons for wanting particular numbers of children, whether children were considered to provide old-age security, and why parents and grandparents had had many children. Four villages were chosen purposively: two in the treatment area and two in the comparison area. Forty women were interviewed, 20 in the younger age group (under 30 years of age, with one or two living children) and 20 in the older age group, 10 each from the treatment and comparison areas in each instance. The older group consisted of women who had expressed a desire to cease childbearing in 1984, but had delivered babies during the period 1984-1991. The younger women were selected purposively from the same households as the older women if available, otherwise from the same or a neighbouring bari (patrilineally related households). The interviews were conducted by a woman who had been working at the ICDDR,B as an interviewer for the previous two years, was known to the community and had some experience in conducting in-depth interviews. Each discussion lasted about two hours and was conducted during the daytime. To develop a friendly atmosphere, the interviewer was instructed to spend some time with the respondent before starting the discussion. To guide the discussion, the interviewer used a schedule and recorded answers on it. However, a tape-recorder was purposely not used because it could have attracted other household members, particularly children, to come close to the place where the interview was conducted and thus nullify the level of privacy that was desired. To gain some idea of whether reproductive preferences in the Matlab study area are similar to those in other areas of Bangladesh, seven women from the area adjoining the Matlab study area were also interviewed. These women had been visiting the upazila health complex with their children and the discussions were recorded on tape.

To examine differentials in reproductive preferences, analysis was carried out for currently married, non-pregnant women aged 15-49. Pregnant women were excluded because preferences for some of these women might depend upon the sex of the unborn baby. To minimize rationalization, desired family size was also analyzed for women under age 25. In this regard Pullum (1983) found that one of the best strategies to resolve the problem of rationalization is to restrict analysis to those at an early stage of the family-building process. For whether more children are wanted, analysis was also carried out separately for woman of parity three. These women are of particular interest since about 65 per cent of all women wanted to stop childbearing at this parity.

For the multivariate analysis, the method of multiple classification analysis was used to analyze the interval-level dependent variable (desired family size) and logistic regression to analyze the binary dependent variable (desire for more children). For desired family size, analysis was undertaken for women aged 15-49 and for those aged under 25. Used as covariates in examining socio-economic differentials of desired family size were the following: for women aged 15-49, the age of the woman and number of living children; and for women aged under 25, number of living children. For desire for more children, analysis was undertaken for women aged 15-49 and for those with family size of three children. In the case of desire for more children, the dependent variable was given the value 1 if the respondent wanted more children and 0 if she did not. The independent variables used in the logistic regression were age of woman, number of living children, education of woman, education of husband, possession of items, religion and use of contraception; age of woman and number of living children were used as continuous variables. For both multiple classification analysis and logistic regression, education of woman (0, 1-4 and 5+ years), education of husband (0, 1-5 and 6+ years) and possession of items (none, 1-2 and 3+) were grouped into three, while religion (Muslim and Hindu) were dichotomized. Education levels were obtained by asking about completed years of schooling in the modern sector; religious education, which does not lead to a certificate, was not counted. The education categories of the women were defined differently from those of their husbands to get sufficient numbers in each category. Possession of household items, such as radio, watch, bicycle or quilt, and the receiving of remittances were considered here to be indicators of a household's economic condition.

The quality of data on reproductive preferences can be assessed to some extent through non-response levels, test-retest reliability studies and inter-item consistency analysis. Over the period between 1975 and 1990, the non-response for desired family size declined dramatically, from 17.1 per cent to 1.7 per cent in the treatment area and 27.2 per cent to 3.4 per cent in the comparison area. An inter-item consistency check was made on the 1990 KAP data: desire for more children was compared with desired family size. Of those who wanted more children about 80 per cent were consistent compared with 95 per cent of those who wanted no more children.

Levels

Table 1 shows reproductive preference levels for the treatment and comparison areas in 1990. Mean desired family sizes were found to be similar, but the percentage desiring more children was slightly higher in the treatment area than in the comparison area. Two populations with identical desired family size distributions could have very different proportions of women not wanting more children if they had different tempos of childbearing, which depend upon such factors as level of contraceptive use, breastfeeding pattern and post-partum abstinence (Lightbourne and MacDonald, 1982; United Nations, 1987). When the percentage desiring more children in the comparison area was standardized, the percentages were, however, found to be closer in the two areas (46.2 compared to 47.9).

	Desired family size			Desire for more children
Area	Mean*	Non- numerical (%)	N	Percentage	Non- numerical (%)	N
Treatment	3.12	1.7	3,648	46.2	1.1	3,669
Comparison	3.24	3.4	2,814	42.7(47.9)*	2.3	2,846

* Note: Standardized for number of living children in the treatment area.

Reproductive preferences in the two areas were further examined through qualitative interviews using the ideal family size question: "How many children do you consider ideal for a newly married couple?" Examples of responses of the 20 women interviewed in each area are as follows:

A couple should have two boys and one girl.
	(35-year-old mother of four boys and a girl, treatment area)
A couple should have one boy and one girl.
	(18-year-old mother of a boy, treat- ment area)
A couple should have two boys and one girl.
	(39-year-old mother of five boys and three girls, comparison area)
A couple should have one boy and one girl.
	(36-year-old mother of six boys and a girl, comparison area)

To examine whether reproductive preferences in the treatment and comparison areas are similar to those in other areas of Bangladesh, the ideal family size question was also asked through qualitative interviews with women residing in the area adjoining the Matlab study area. Examples of responses of the seven women who were interviewed while visiting the upazila health complex are as follows:

A couple should have two boys and one girl.
	(30-year-old mother of two boys and one girl, adjoining Matlab study area)
One should be happy with two, either boys or girls.
	(25-year-old mother of two boys and one girl, adjoining Matlab study area)

The above results suggest that motivation for ideal family size is similar in the treatment and comparison areas and a similar motivation also exists in areas adjoining it; this may reflect national-level estimates.

The desired family size data collected in the 1990 KAP survey was found to be 28.0 per cent lower than those recorded in 1975 by the contraceptive distribution project (Koenig and others, 1987). As there are no other data available for the Matlab study area between 1975 and 1990, it is not known whether or not desired family size declined gradually. However, one study from another rural area of Bangladesh conducted in 1985 found a mean desired family size of 3.0 (Hoque, 1988), which suggests that the decline probably occurred during the period 1975-1985.

Motivation

Studies conducted in Bangladesh up until the early 1980s found that familial, social, economic and religious conditions were favourable to having many rather than few children (Arthur and McNicoll, 1978; Cain, 1977, 1981, 1983). From a village study, Khuda (1977) concluded that a large family, with both boys and girls, was an asset to the household and Cain (1977) noted that children, especially sons, provided parents with significant net economic returns through their labour. In a comparison with India, Cain (1981) argued that in a rural economy, when senility, illness or environmental disaster strike, children, especially sons, are needed for long-term family security. However, Robinson (1983) argued that it is not the high benefit but the very low cost of children which induces high fertility. According to Robinson, at a very low cost, even a modest benefit would make the value of children positive. In the remainder of this section, qualitative data are presented to examine the motivation for ideal family size when fertility decline is under way: fertility decline has, however, been much steeper in the treatment area than in the comparison area.

As reproductive preferences have been found to be similar in the treatment and comparison areas, the qualitative data presented below are not shown by area. The younger women interviewed in the qualitative survey possessed on average 1.9 items compared with 1.1 items for older women. They had almost no education (mean schooling 1.8 years compared with 0.6 years), and their mean ages and mean numbers of living children were 25 and 1.2, respectively, for younger women, and 39 and 6.1 for older women.

Both younger and older women were asked the ideal family size question. As younger women had 0-2 children, it was assumed that their own desired family sizes would be reflected, while older women were expected to report their perceptions of the family size preferences of the generation just commencing family formation. The mean ideal family size was 3.2 children for younger women compared with 3.0 children for older women. The following section examines motivation behind the reported ideal family size.

Increase in cost of living

In the qualitative interviews, younger women were asked about the advantages and disadvantages of having the small family sizes which they reported as ideal. Older women were asked the same question, but with reference to the number they considered ideal for a younger couple. Almost every woman from both the younger and the older groups reported that it was the increase in the direct economic cost of children that had influenced ideal family size to decline.

One will face difficulties in providing food, education and land for cultivation, so a small family is better.

(45-year-old mother of six boys and four girls)

It is difficult to provide food, clothing and education. Many children create many problems. My husband's income is low, so few children is better.

(34-year-old mother of two boys and one girl)

Availability of family planning method

Women were asked why their fertility preferences were low compared with the number of children their parents and grandparents had had. Almost all women reported that family planning was not available in the past, and some also said that there were more resources per capita in the past.

They had plenty of property and also family planning methods were unavailable.

(36-year-old mother of six boys and one girl)

Family planning methods were not available.

(26-year-old mother of two boys and two girls).

Change in perception of old-age security

To seek insight into the contemporary link between fertility and old-age security, women were asked whether they considered the option of having a large family so that their children could provide security for them in their old age. Out of 40 women, none supported the contention that large families, of a size similar to those of their parents and grandparents, provided security for old age. Moreover, their statements favoured the quality rather than quantity of children.

I do not think many children is a strength for the family. If you have many children, you can't properly educate them. So few children is better.

(30-year-old mother of one boy)

I think many children create many problems. In these days, boys do not look after parents, so having many children is bad.

(45-year-old mother of seven boys and two girls)

Favourable attitude of in-laws

It is often argued that in societies such as Bangladesh, other family members (in-laws) have a major role in fertility decision-making for couples. On three occasions mothers-in-law were interviewed, and their reactions towards small family size were positive.

My daughter-in-law is very young (recently delivered a baby boy) and does not know about family planning, but we will tell her in the future what she needs to do to keep a small family size because times have changed.

(A mother-in-law)

The foregoing responses indicate that motivation for small family size is widespread; however, these motivations were absent even a decade ago. Motivations were found to be similar for both younger and older women.

Differentials

Desired family size

Demographic differentials: In both areas, the mean desired family size increased with the age of the woman; the difference in the number of children desired between women aged under 20 years and those 40 years and older was about half a child (table 2). When mean desired family size was adjusted for number of living children, the relationship was reversed: desired family size was higher for younger age groups and lower for older age groups. The differences between women aged under 20 and those 40 years and older were 0.3 of a child in the treatment area and 0.4 of a child in the comparison area. A similar finding has been documented for Jamaica after adjusting for the number of living children and number of living children squared (Lightbourne, 1984). Desired family size may be higher for younger than for older age groups because the older age groups are more aware of the cost of a large family than are the young ones, whose members have not yet experienced the burden of excess childbearing. The finding of minimal difference in desired family size between younger and older women suggests that when desired family size changes it changes in all generations equally.

Age of woman (years)	Treatment area			Comparison area
	Unadjusted	Adjusted*	N	Unadjusted	Adjusted*	N
<20	2.90	3.33	187	3.00	3.53	145
20-24	2.92	3.22	720	3.02	3.35	592
25-29	2.97	3.08	879	3.11	3.24	673
30-34	3.15	3.08	667	3.33	3.22	488
35-39	3.28	3.08	514	3.38	3.13	378
40+	3.44	3.08	681	3.52	3.12	538

Note: * Adjusted (by regression) for number of living children.

In both areas, mean desired family size, with minor exceptions, increased with number of living children; the difference in mean desired family size between those who had no child and those who had seven or more children was 0.8 of a child in the treatment area and 0.9 of a child in the comparison area (table 3). When mean desired family size is adjusted for age of woman, an essentially positive relationship remains; however, the difference between women with no child and those with seven or more children becomes wider, i.e. 1.0 child in the treatment area and 1.1 child in the comparison area. This result replicates findings for Guyana, Jamaica and Trinidad and Tobago after adjusting for age and age squared (Lightbourne, 1984). The results indicate that parity affects women's reports on the number of children desired; implementation of preferences, rationalization of achieved family size and underestimation of eventual desired family size are operating here to produce these relationships (Razzaque, 1994).

No. of living children	Treatment area			Comparison area
	Unadjusted	Adjusted*	N	Unadjusted	Adjusted*	N
0	2.80	2.71	232	2.89	2.78	203
1	2.78	2.70	604	2.80	2.72	401
2	2.87	2.83	693	2.97	2.92	439
3	3.12	3.12	661	3.16	3.14	469
4	3.40	3.44	570	3.49	3.52	420
5	3.37	3.44	441	3.44	3.50	355
6	3.48	3.57	255	3.57	3.65	260
7+	3.60	3.71	192	3.77	3.88	267

Note: * Adjusted (by regression) for age of woman.

Socio-economic differentials: In the treatment area, the mean desired family size of women aged 15-49 was inversely related to the education of the woman, education of her husband and possession of items (table 4, column 1). Muslims had a higher mean desired family size than Hindus. For women under age 25, the patterns were similar to those of women aged 15-49 (column 3). When the mean desired family size of women aged 15-44 were adjusted for the age of the woman, number of living children and other socio-economic variables (column 2), the same patterns were found as in column 1, but within-category differences for education of husband and possession of items became insignificant. For women aged under 25, when the means were adjusted for number of living children and other socio-economic variables (column 4) the patterns were usually similar to that for women aged 15-49 (column 2), but the inter-category differences, except for education of husband, became insignificant.

In the comparison area the mean desired family size of women aged 15-49 was inversely related to education of the woman, education of her husband and possession of items (table 4, column 5). Hindus had a higher mean desired family size than Muslims. For women under age 25, the patterns were similar to those of women aged 15-49 (column 7). When the mean desired family size of women aged 15-49 was adjusted for the age of the woman, number of living children and other socio-economic variables (column 6), the same patterns were found as in column 5, but within-category differences became insignificant. For women aged under 25, when the means were adjusted for number of living children and other socio-economic variables (column 8), the patterns were usually similar to those for women aged 15-49 (column 6).

At the multivariate level, the mean desired family size did not usually vary significantly by socio-economic categories in either area for either women aged 15-49 or those aged under 25. In fact, the differences in mean desired family sizes by socio-economic categories never exceeded 0.2 of a child, even when they were significant; from a practical point of view, these differences are negligible. Thus, it can be argued that desired family sizes are homogeneous across socio-economic categories.

Note: Adj = adjusted; SES = socio-economic status; NLC = No. of living children; AGW = age of woman.

Desire for more children

Demographic differentials: In both areas, the percentages and adjusted percentages (for number of living children) wanting more children decreased with increases in the age of the women (table 5). A similar finding has been documented for Jamaica, Guyana, and Trinidad and Tobago (Lightbourne, 1984). Several reasons can be offered to explain the higher percentages of older women not wanting more children after the number of living children has been adjusted. The first is the selection effect: by age 35, most women who have, for example, only two children may have wanted only two, whereas proportionately more younger women with only two children may have wanted more, but have not yet had time to achieve their desired family size. Secondly, older women may want to avoid births once they have children old enough to marry and bear children themselves; this has often been called the "grandmaternal" effect (Tan, 1983). Thirdly, women who have fewer children by a given age may be more aware of childbearing costs as they become older and so wish to cease childbearing.

In both areas, the percentages and adjusted percentages (for age of the women) wanting more children decreased with increases in the number of living children (table 6). The fall in percentages wanting more children with increases in the number of living children occurs because, once desired family size has been attained, the desire for more children ceases.

Note: * Adjusted (by regression) for number of living children.

Note: * Adjusted (by regression) for age of woman.

Socio-economic differentials: In the treatment area, the percentage of women aged 15-49 who wanted more children was positively related to education of the woman and possession of items, and it exhibited a U-shaped relationship with education of her husband (table 7). This was similar for Muslims and Hindus. For family size of three children, the pattern was different from that for women aged 15-49: desire for more children was negatively related to education of the woman, education of her husband and possession of items; also, it was higher for Muslims than for Hindus. After all variables were controlled, better educated women (5+ years of schooling) in either group (aged 15-49 or having a family size of three children) were less likely than illiterate women to desire more children (table 8).

In the comparison area, the percentage of women aged 15-49 who wanted more children was positively related to education of the woman, and it exhibited a U-shaped relationship with education of her husband and an inverted U-shaped relationship with the possession of items (table 7). The percentage wanting more children was higher for Muslims than for Hindus. For a family size of three children, the pattern was similar to that for women aged 15-49 for possession of items only. It had an inverted U-shaped relationship, with education of the woman and the percentage wanting more children being lower for those with illiterate than literate husbands; and it was higher for Hindus than for Muslims. After all variables were controlled, desire for more children was found to be largely insignificant in either group (aged 15-49 or having a family size of three children) for the socio-economic variables considered here (table 8).

Characteristics	Treatment area						Comparison area
	Aged 15-49			Family size: 3 children			Aged 15-49			Family size: 3 children
	B	SE	Odds ratio	B	SE	Odds ratio	B	SE	Odds ratio	B	SE	Odds ratio
Constant	6.950	0.292		3.669	0.617		6.449	0.308		3.392	0.623
Age of woman	-0.141	0.011	0.87	-0.141	0.020	0.87	-0.131	0.011	0.88	-0.132	0.021	0.87
No. of living children	-0.998	0.052	0.36				-0.959	0.051	0.38
Education of woman (years)
0			1.00			1.00
1-4	0.072	0.074	1.07	0.153	0.126	1.16
5+	-0.232	0.072	0.79	-0.411	0.127	0.66
-2 Log-likehood(df)	2560.5 (3705)			797.7 (666)			1948.4 (2911)			587.3 (474)

The issue of the meaningfulness of reproductive preference data has not yet been resolved. Desired family size data, particularly from developing countries, have been criticized for fatalistic responses and forced responses to questions for which the respondent really has no answer. But in the 1990 KAP survey in Matlab, the rate of non-response was found to be very low. Observations from the field during the qualitative data collection process were also convincing; respondents were able to give spontaneous responses to the ideal family size question and provided valid reasons for choosing their ideal family sizes.

Studies conducted until the early 1980s documented motivation for many rather than few children (Cain, 1977; Khuda, 1977; Arthur and McNicoll, 1978), but the present study found the opposite to be the case. Mean desired family size was found to be slightly over three children and women said that they preferred usually to have two sons. In the qualitative interviews it was reported that an increase in the direct economic cost of children and knowledge of family planning were mainly responsible for the small family-size desires. These days, women in Bangladesh do not believe that having many children provides strength for the family and security for their old age; rather the quality of their children is believed to be important. Moreover, it was reported that the intergenerational relationship between parents and children is changing; the costs of having many children are beginning to outweigh the benefits. Regarding decline in reproductive preference, Nag and Duza (1988) reported that the new small-family norm in the Matlab study area is mainly aspiration-induced rather than poverty-induced, but qualitative data from the present study did suggest that both aspiration and poverty are responsible for the decline in reproductive preference.

Desired family size, ideal family size and desire for more children were found to be similar in the two areas. Moreover, except for education of husband (for wanting more children) in the treatment area, neither mean desired family size nor percentage wanting more children varied remarkably by socio-economic categories for any group of women (aged 15-49 and either under age 25 or with a family size of three children). Similar reproductive preference in the two areas indicates that the Family Planning and Health Services Programme has had no effect on preference. According to Freedman and Freedman (1992), the norms for family size ideals are too deeply rooted to be changed simply by mass media messages; however, they reported that it is yet to be known whether an aggressive family planning programme has had desired impacts. Similar preferences across categories of socio-economic variables suggest that factors responsible for small family-size preferences work at the societal level and that preferences are influenced by broader socio-economic change. This has implications for both policy and programme purposes.

Further, when parities are controlled, there is no strong evidence that the younger women reported smaller family-size desires than older women. One possible explanation is that women's preferences have changed little over time; however, the argument was not supported by the data: desired family size in the Matlab area in 1990 was about 30 per cent lower than that reported in 1975. The other explanation is that women's preferences do change, but they change almost equally in all age groups. In fact, within culturally homogenous populations, change in reproductive preferences spread to all sectors over a short time (Knodel and others, 1987; Freedman and Freedman, 1992), resulting in similar preferences across age groups.

Our study found that the Family Planning and Health Services Programme in the treatment area of Matlab has been able to reduce fertility significantly compared with the comparison area, but it has not been able to reduce desired family size. Perhaps broader socio-economic change, such as improvements in educational and employment opportunities for women and raising their status, is needed to bring about a rapid decline in desired family size. Thus, policy and programme interventions are required that are in line with the new population and development paradigm adopted by the 1994 International Conference on Population and Development as part of the Programme of Action. In fact, there is a need for the Government to develop policies that better support individual members of the basic unit of society - the family - and to promote equality of opportunity for family members, especially the rights of women and children in the family.

Arthur, W. B. and G. McNicoll (1978). "An analytical survey of population and development in Bangladesh", Population and Development Review, 4(1):23-80.

Bhatia, S., W. H. Mosley, A. S. G. Faruque and J. Chakraborty (1980). "The Matlab family planning-health services project", Studies in Family Planning, 11(6):202-212.

Bongaarts, John (1978). "A framework for analyzing the proximate determinants of fertility", Population and Development Review, 4(1):105-32.

Brackett, J. W., R. T. Ravenholt and J. C. Chao (1978). "The role of family planning in recent rapid fertility declines in developing countries", Studies in Family Planning, 9(12): 314-323.

Cain, M. (1977). "The economic activities of children in a village in Bangladesh", Population and Development Review, 3(3):201-227.

__________ (1981). "Risk and insurance: perspectives on fertility and agrarian change in India and Bangladesh", Population and Development Review, 7(3):435-474.

__________ (1983) "Fertility as an adjustment to risk", Population and Development Review, 9(4):688-702.

Cho, L. J., F. Arnold and T. H. Kwon (1982). "The determinants of fertility in the Republic of Korea", Committee on Population and Demography Report, No. 14 (Washington, D.C.: National Academy Press.)

D'Souza, S. (1981). "A population laboratory for studying disease process and mortality - the demographic surveillance system, Matlab, Bangladesh", Special Publication, No. 13 (Dhaka: International Centre for Diarrhoeal Disease Research, Bangladesh).

Freedman, R. and D. Freedman (1992). "The role of family planning programmes as fertility determinants", In: J. F. Phillips and J. A. Ross (eds.), Family Planning Programmes and Fertility (Oxford: Oxford University Press) pp. 10-27.

Hoque, M. N. (1988). "Rural electrification and its impact on fertility: evidence from Bangladesh". Unpublished Ph.D. thesis, Pennsylvania State University.

Khuda, Barkat-E (1977). "Value of children in a Bangladesh village", In: J. C. Caldwell (ed.), The Persistence of High Fertility (Canberra: Australian National University), pp. 681-728.

Knodel, J., A. Chamratrithirong, N. Chayovan and N. Debavalya (1982). "Fertility in Thailand: trends, differentials and proximate determinants", Committee on Population and Demography Report, No. 13 (Washington D.C.: National Academy Press).

Knodel, J., A. Chamratrithirong and N. Debavalya (1987). Thailand's Reproductive Revolution: Rapid Fertility Decline in a Third-World Setting (Madison: University of Wisconsin).

Koenig, M. A., J. F. Phillips, R. S. Simmons and M. A. Khan (1987). "Trends in family size preferences and contraceptive use in Matlab, Bangladesh", Studies in Family Planning, 18(3):117-127.

Koenig, M. A., U. Rob, M. A. Khan, J. Chakraborty and V. Fauveau (1992). "Contraceptive use in Matlab, Bangladesh in 1990: levels, trends and explanations", Studies in Family Planning, 23(6):352-364.

Lightbourne, R. E. (1984). "Fertility preferences in Guyana, Jamaica and Trinidad and Tobago, from World Fertility Survey, 1975-77: a multiple indicator approach", WFS Scientific Reports, No. 68 (Netherlands: International Statistical Institute).

__________ and A. L. MacDonald (1982). "Family size preferences", Comparative Studies: Cross-National Summaries, No. 14 (London: World Fertility Survey).

Nag, M. and M. B. Duza (1988). "Explaining high contraceptive prevalence in Matlab, Bangladesh: a qualitative study". Report submitted to the Swedish International Development Agency.

Pullum, T. W. (1983). "Correlates of family size desires", In: R. A. Bulatao and R. D. Lee (eds.), Determinants of Fertility in Developing Countries, Vol. 1 (New York: Academic Press) pp. 344-368.

Razzaque, A. (1994). "Reproductive preferences and subsequent behaviour in a rural area of Bangladesh". Ph.D. thesis, Australian National University.

Ruzicka, L.T. and A.K.M. Chowdhury (1978). "Demographic surveillance system-Matlab: vital events, migration and marriages-1976", Special Report, No. 13, (Dhaka: International Centre for Diarrhoeal Disease Research, Bangladesh).

Robinson, W.C. (1983). "Public policy, risk and fertility in Bangladesh: a comment", Bangladesh Development Studies, 11(3):39-58.

Stycos, J. M. (1984). "Putting back the K and A in KAP: a study of the implications of knowledge and attitudes for fertility in Costa Rica", WFS Scientific Reports, No. 48 (London: World Fertility Survey).

Sun, T. H., H. S. Lin and R. Freedman (1978). "Trends in fertility, family size preferences, and family planning practice: Taiwan, 1961-76", Studies in Family Planning, 9(4):54-70.

Tan, J. P. (1983). "Marital fertility at older ages in Nepal, Bangladesh and Sri Lanka", Population Studies, 37(3):433-444.

United Nations (1981). "Selected factors affecting fertility and fertility preferences in developing countries: evidence from the first fifteen WFS country reports", In: World Fertility Survey Conference 1980: Record of Proceedings, Vol. 2 (Voorburg: International Statistical Institute) pp. 141-227.

__________ (1987). Fertility Behaviour in the Context of Development: Evidence from the World Fertility Survey, Population Studies, No. 100 (New York: United Nations).

Ware, H. (1974). "Ideal family size", WFS Occasional Papers, No. 13 (London: World Fertility Survey).

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