Funciones cóncavas y cuasicóncavas
5.1 Optimización libre Criterio del Hessiano
5.1.2 Condiciones suficientes de segundo orden
In this section analysis was undertaken to examine the net effects o f demographic and sociocultural characteristic on use of contraception separately for women who
wanted no more children and those who wanted to delay the next birth for at least
one year. The former group is very important because they are motivated to cease
childbearing. Moreover, this group is also used in calculating 'unmet need' for
contraception. Current use of contraception was used as the dependent variable while
age of woman, number of living children, whether has a son, education of woman,
possession of items and religion were used as independent variables. For theoretical
reasons age of woman and number of living children were kept in all the models,
whether they were significant or not.
The dependent variable took the value 0 if the respondent was not currently using
contraception and 1 if using it. As the dependent variable is binary, logit regression
was more appropriate than the ordinary least squares model. Cox and Snell (1980)
mention at least two approaches for model selection, 'forward' and 'backward'. Both
approaches are found in the literature, but forward selection is adopted here because
it required less computer memory. In forward selection the variable which produces
the maximum reduction in deviance is incorporated in the model first and the
procedure continues until no further significant reduction in deviance is recorded. A
positive log odds (coefficient) for any category of a predictor indicates that women
in that category were more likely to be using contraception than those in other
categories of the same predictor. Parameter estimates with negative signs indicate the
opposite. Odds ratios show the relative change, compared to the value for the
reference category.
5.5.1 Those who wanted no more children
The results of the final models are shown in Table 5.9 (for full models see Table
B5.3). In the treatment area in 1984 and in 1990, use of contraception among those
who wanted no more children did not vary significantly by demographic and
1990. Use of contraception also varied by religion in the comparison area in 1984,
while in 1990 it varied by age of woman, number of living children and education of
woman.
Table 5.9
Results of logistic regression models of contraceptive use of currently married, non pregnant women aged under 40 who wanted no more children, treatment and
comparison areas 1984 and 1990
V ariable Treatm ent Com parison 1984 1990 1984 1990 B SE Odds ratio B SE O dds ratio B SE Odds ratio B SE O dds ratio C onstant -0.516 0.431 -0.962 0.499 -0.470 0.558 -2.266 0.458 A ge o f w om an 0.030 0.016 1.03 0.077 0.018 1.08 -0.001 0.020 1.00 0.079 0.016 1.08 Living children -0.056 0.052 0.94 -0.086 0.063 0.91 -0.096 0.063 0.91 -0.162 0.051 0.85 E ducation o f woman (yrs)
0 1.00 1-4 0.463 0.165 1.59 5+ 0.727 0.170 2.07 Religion M uslim 1.00 1.00 1.00 H indu 0.534 0.190 1.70 0.456 0.197 1.57 0.841 0.257 2.32 C om parison of models (1) Full m odel+ L og-likelihood (df) 1345.9 (991) 1347.3 (1290) 790.4 (656) 1447.6 (1076) (2) N ull m odel L og-likelihood (df) 1364.3 (999) 1374.5 (1298) 8 1 1 .6 (6 6 4 ) 1495.4 (1084) (3) Final M odel L og-likelihood (df) 1352.0 (996) 1349.7 (1295) 797.9 (661) 1449.6 (1080) 2-3 12.3 (3) 24.8 (3) 13.7 (3) 45.8 (4) 3-1 6.1 (5) 2.4 (5) 7.5 (5) 2.0 (4)
Sources: SES 1982, IDS 1984 and KAP 1990 Note: + See Table B5.3
Differentials in contraceptive use by demographic and sociocultural characteristics
depend on the length and nature of the family planning program that is operating. In
the Matlab study area Bhatia (1983) reported that when contraceptive use was low,
all occupation categories accepted contraception equally, but as contraceptive use
Thailand, the sociocultural differential was found to be small in Indonesia (Cleland,
Little & Pitaktepsombati 1979; Freedman, Khoo & Supraptilah 1981; Streatfield
1986) because pressures to adopt contraception were different in the two programs.
In the treatment area over time contraception was become equally acceptable to all
sociocultural subgroups except religious categories, while in the comparison area a
differential by education of woman appeared in 1990, but that by religion
disappeared. Similar levels of contraceptive use by sociocultural categories in the
treatment area are an indication that the FPHSP has been able to motivate all
subgroups equally.
In the treatment area about 50 per cent of contraceptors were using the injection and
a large proportion had discontinued injection use because of side effects such as
bleeding. Discontinuation of injections was probably higher for Muslims than for
Hindus because such side effects prevented Muslim women from performing their
daily prayers. Stewart et al. (1991) reported that 56 per cent of acceptors of injection
could not carry out their religious rituals as before because of side effects. In the
comparison area contraceptive use was low in 1984 and probably all sociocultural
groups accepted it equally, except for religious groups, while with the increase in
contraceptive use as in 1990, the educated women accepted more than the illiterate;
the religious differential disappeared in the comparison area, probably because there
was negligible use of injections. In recent studies (Nag & Duza 1988; Bemhart &
Uddin 1990), it has been reported that religious opposition is not a barrier to success
of family planning programs. However, religious opposition was widespread,
5.5.2 Those who wanted a child after one year
The results of the final models are shown in Table 5.10 (for full models see Table
B5.4). In the treatment area in 1984, use of contraception among those who wanted a
birth after one year varied significantly by all the sociocultural characteristics
considered; it also varied by whether the respondent had a son. However, in 1990 the
sociocultural differentials had disappeared, indicating that the program had been able
to eliminate differences in use over time. Use of contraception was relatively high
among those who had one or more sons, indicating a link between son preference
and contraception. In the comparison area in 1984, use of contraception varied only
by possession of items, while in 1990 it varied by whether has a son and education of
woman; the son preference and contraception association emerged once
contraceptive use to space births became reasonably common.
The reasons for non-use of contraception among these two groups of women (those
who wanted no more children and those who wanted a child after one year) could be
related to side effects, religion, the objections of husbands or relatives and non
availability of contraception. Some women were not contracepting because they
were not currently exposed to the risk of pregnancy as their husbands were away
from home. During the qualitative interviews women who wanted no more children
in the 1984 IDS but actually had children during 1984-90 were asked 'For what
reason did you have the additional child?'; however, no information is available for
those who wanted a child after one year but had it sooner. In the treatment area
several women who wanted no more children reported that they discontinued
contraception because of side effects, while in the comparison area several reported
that they did not want the child, but that it came naturally. Four responses to the
'I was using the injection but stopped because of excessive bleeding' (36 years old,
mother of four sons and one daughter, treatment area)
Table 5.10
Results of logistic regression models of contraceptive use of currently married, non pregnant women aged under 40 who wanted next birth after one year, treatment and
comparison areas 1984 and 1990
V ariable T reatm ent Com parison 1984 1990 1984 1990 B SE Odds B SE O dds B SE Odds B SE Odds
ratio ratio ratio ratio C onstant -0.142 0.464 -0.100 0.356 -3.675 1.083 -1.472 0.530 A ge o f w om an -0.066 0.022 0.93 -0.009 0.017 0.99 0.002 0.052 1.00 -0.030 0.024 0.97 Living children 0.240 0.082 1.27 0.117 0.070 1.12 0.225 0.186 1.25 0.048 0.081 1.05 W hether has a son
0 1.00 1.00 1.00 1+ 0.442 0.163 1.55 0.448 0.122 1.56 0.593 0.189 1.81 E ducation of w om an (yrs) 0 1.00 1.00 1-4 -0.081 0.165 0.92 0.039 0.234 1.04 5+ 0.366 0.184 1.44 1.157 0.187 3.18 P ossession o f item s N one 1.00 1.00 1-2 0.140 0.157 1.15 0.314 0.432 1.37 3+ 0.620 0.209 1.85 1.428 0.427 4.17 Religion M uslim 1.00 H indu -0.494 0.212 0.61 C om parison o f m odels (1) Full m odel+ L og-likelihood (df) 1247.7 (1027) 1821.0 (1333) 245.7 (533) 929.8 (995) (2) N ull m odel L og-likelihood (df) 1296.6 (1035) 1 8 5 1 .0(1341) 264.8 (541) 982.8 (1003) (3) Final m odel Log-likelihood (df) 1247.7 (1027) 1827.5 (1338) 250.2 (537) 935.5 (998) 2-3 48.9 (8) 23.5 (3) 14.6 (4) 47.3 (5) 3-1 0.0 (0) 6.5 (5) 4.5 (4) 5.7 (3)
Sources: SES 1982, IDS 1984 and KAP 1990
'I was taking the pill but through a mistake in taking it I became pregnant' (45 years
old, mother of six sons and four daughters, treatment area)
'I did not want the child but it was God gave it' (45 years old, mother of seven sons
and two daughters, comparison area)
'I did not want it but I was also scared of using contraception' (43 years old, mother
of seven sons and two daughters, comparison area).
On the basis of the qualitative data it can be argued that a further increase in
contraception among women who want no more children in the treatment area will
depend to some extent on either motivating them to continue in spite of side effects,
or providing alternative contraception which causes fewer side effects. In the
comparison area, suitable contraceptive methods need to be made available at the
door.