Cover Image

PAPERBACK
$36.25



View/Hide Left Panel
Click for next page ( 39


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 38
FINAL RESULTS This section first examines models for the propensity to breastfeed and for the propensity to contracept. Second, among women who breastfeed and among women who contracept, information about other experiences in the birth interval is incorporated, and models of termination rates for breastfeeding and for contraception, respectively, are examined. Finally, factors affecting conception rates are examined using information about the timing of breastfeeding, contraception, and child mortality. The scheme for the analysis ts given by the diagram below. propens ity to breastfeed~ --- how long? _ Background ~- charac terist~ - other experiences ;~ - ~~ ~:_,~ _ - ^" "~= U`~ll ' ~ fertility interval ~ ~ - propene ~ ty "v ,' contrscept~ + how longly child mortality I II The details of the variables included in each equation were given in Table 1. 38 III

OCR for page 38
39 LOGISTIC REGRESSION MODELS FOR .1~ PROPENSITY TO BREAS'1~;D AND FOR THE: PROPENSITY TO CONTRACE:PT These two equations use background characteristics to predict the probability of breastfeeding and the probability of using contraception in either the last closed or the open birth interval. Colombia and Costa Rica have quite different distributions of the two dependent variables. Nearly all women breastfeed in Colombia; in Costa Rica, although breastfeeding is very common, it is not universal. In contrast, many more women contracept in Costa Rica than in Colombia. Table 4 gives the results of the analysis. Since brea~tfeeding is nearly universal in Colombia (only 6 percent of women did not breastfeed in either of the two birth intervals sampled) , the best prediction is that all women will breastfeed. Under these circumstances, the model is not very informative. Even though there is more variability in the dependent variable in Costa Rica, the model does not perform much better. Thus, the measured character- istics indicating social and economic status do not distinguish well between a woman who will breastfeed and one who will not. The analysis of the propensity to use contraception is somewhat more successful. Although contraceptive usage is much higher in Costa Rica, the impact of the social and economic variables is much larger in Colombia. In Colombia, the probability of use increases with the education of both the woman and her husband, and is higher for urban than for rural residents. In Costa Rica, only the woman's education and urban residence have an impact, and the effects are much smaller. This confirms that contraceptive use is both more extensive and less restricted to particular social and economic groups in Costa Rica than in Colombia. It is interesting that, although breastfeeding is less widespread in Costa Rica than in Colombia, it is not any more restricted to particular social and economic groups in one country than in the other. The analysis by period in the appendix shows that these results are not significantly altered as births from progressively earlier time periods are e liminated, suggesting that these results are not solely attributable to changing practices.

OCR for page 38
40 ~ - TABL`E 4 Logistic Regression for Probability of Breastfeeding and for Probability of Using Contraception in Last Closed or Open Birth Interval Colombia I Costa Rica Breast feeding Contracept ion ~ Breast feeding Contracept ion : constant work ~ ince marriages work may f ras home worn ' ~ educat ion husband 's educat ion urban res idencea 3.626 ( .2344) .0227 (.3011) -.2051 ( .3420) -.0545 ( .0481) -.0627 ( .0402) -.5516* ( .2676) -.9816 (.1051) .1362 ( .1552) -.2262 (.1964) .2032* ( .0308) .1222* ( .0279) .7413* (.1214) 1.63? (.1248) .4303 ( .2543) -.4640 (.2848) .0241 ( .0275) - .0439 ( .0236) -.0375 (.1626) .8288 (.1221) -.0605 ( .2050) .1716 .2519) · 0900 ( .0300) .0118 t .0262) .4441* ( .1674) Mode 1 x2 d . f . prey it t ive accuracy 2 not us ing 27.35 .684 6.0 303.5 8.70 5 5 .164 .335 39.7 17.5 1643 47.24 .318 19.2 N 1643 1449 1449 Notes: Indicator variable; layer. Standard encore are in parentheses. * indicates significance at the .05 level . HAZARD MODELS FOR U:RMINATION OF BREASTFEEDING AND CONTRACEPTIVE DISCONTINUATION Figure 6 shows the estimated hazard functions for termination of breastfeeding (top panel) and contraceptive discontinuation (bottom panel). These hazards have been estimated without covariates and contain the same information as the survivor functions discussed earlier. The horizontal lines show the result of assuming constant exponential discontinuation rates; the jagged lines show the estimated hazards when they are allowed to vary over the eight subperiods defined earlier; the solid lines represent estimates for Colombia and the dashed lines those for Costa Rica. First examine the estimated hazard functions for termination of breast-

OCR for page 38
41 Breaetfeeding in(t) .150 .100 .050 in(t) .200 )~ C ontraception C ol ombis _. Costa Rica ~"~_ 3 6 1 2 18 ~4 36 48 t .050 _ 3 6 1 2 18 24 36 48 t FIGURE 6 Hazard Functions for Termination of Breastfeeding and for Contraceptive Discontinuation feeding in the top panel. The horizontal lines show that termination rates are higher in Costa Rica than in Colombia. Colombia shows a pattern of moderate discon- tinuation rates for durations of breastfeeding under 18 months, followed by a marked peak in discontinuation rates between 18 and 23 months, suggesting that the socially prescribed weaning time may fall within this

OCR for page 38
42 interval. The pattern for Costa Rica shows much higher discontinuation rates at short durations, with one group of women discontinuing before 3 months have elapsed and a second group discontinuing after 6 to 11 months. Turning to the estimated hazard functions for contraceptive dis- continuation in the lower panel, the horizonal lines show that on average, discontinuation rates are only somewhat higher in Colombia than in Costa Rica. However, the time pattern of discontinuation is quite different. Colombia shows much higher discontinuation rates at shorter durar Lions, and the two curves tend to converge after durations of use longer than two years. The curves for Costa Rica show two peaks of discontinuation, one at short durations, the other at durations of 18 to 23 months. As discussed below, these differences are primarily accounted for by differences in the pattern of discontinuation of coitus- dependent methods in the two countries, while the pattern of discontinuation of coitus-independent methods is more . ·, similar. Tables S and 6 give the estimated coefficients, their standard errors, and the antilogs for the equation with covariates predicting termination of breastfeeding. These results are displayed graphically in Figure 7. The strategy used in this and the other analyses of duration is first to examine the impact of the background variables alone (Model 1), then to add the covariates describing other demographic characteristics (Model 2), and finally to add the covariates that describe other events or behaviors that occur in the interval (Model 3). The log-likelihood for each model is given at the bottom of the tables so that likelihood-ratio statistics can be computed to determine the statistical significance of the added covariates. First consider the effects of the background covariates alone. In Colombia, as in Costa Rica, the woman's education, her husband's education, and urban residence are all statistically significant, and the estimated coefficients are nearly of the same magnitude. In both countries, these three variables are all associated with shorter durations of breastfeeding. Model 2 adds parity, which is also statistically sig- nificant and of nearly the same order of magnitude. In both countries, higher parity is associated with longer durations of breastfeeding. Some of this may be a cohort effect since average parity at the start of the last closed interval increases by approximately one for each five-year cohort of women between ages 20 and 45. Because the number of observations is small, the model

OCR for page 38
43 TABLE 5 Coefficient Estimates for Termination of B reastf ceding: Colombia Model 1 Model 2 Model 3 coeff. antilog coeff. antilog coeff. antilog work since -.0557 .9458 -.0562 .9454 -.0616 .9403 marriages (.0629) (.0629) (.0628) work away .0021 1.002 .0012 1.001 .0084 1.008 from home a (.0765) (.0764) (.0765) women 'e education .0766* 1.080 .0722* 1.075 .0720* 1.075 (.0110) (.0112) (.0111) husband's .0111 1.011 .0074 1.007 .0093 1.009 education (.0095) (.0096) (.0096) urban residences .2343* 1.264 .2339* 1.264 .2327* 1.262 (.0516) (.0517) (.0516) parity ' -.0234* .9769 -.0245 .9758* (.0082) (.0082) child survivals -.4022* .6688 (.0922) Periot 1 (0-2 months) constant -2.854* .0576 -2.693* .0676 -2.316* .0987 (.0585) (.0808) (.1176) Periot 2 (3-5 months) constant -2.787* .0616 -2.624* .0725 -2.245* .1060 (.0617) (.0835) (.1199) Period 3 (6-11 months) constant -2.579* .0758 -2.414* .0895 -2.032* .1316 (.0524) (.0776) (.1162) Period 4 (12-17 months) constant -2.661* .0699 -2.491* .0828 -2.102* .1222 (.0739) (.0945) (.1294) Period 5 (18-23 months) constant -2.429* .0881 -2.261* .1042 -1.866* .1547 (.0910) (.1080) (.1406) Periot 6 (24+ months) constant -3.130* .0437 -2.964* .0516 -2.561* .0772 (.1452) (.1562) (.1815) loglikelihood -6594. -6590. -6582. N - 2264 Notes: a Indicator variable: 1-yes. Standard error. are in parentheses. * indicates significance at the .05 level.

OCR for page 38
44 TABLE 6 Coefficient Estimates for Termination of Breastfeeding: Costa Rica Model 1 Model 2 Model 3 coeff. antilog coe f f . ant i log coe f f . ant i log - work. since -. 0826 .9208 -.0627 .9392 -.0640 .9379 marr iage a ( .0742 ) ( .0744 ) ( . 0744 ) work away .0942 1 .099 .08102 1 .084 .0847 1 .088 f ram home a ( . 0881 ) ( . 0881 ) ( . 0881 ) woman 's . 0439* 1 .045 . 0371* 1 .038 . 0368* 1 .038 educat ion ( .0095 ) ( .0098 ) ( .0098 ) husband's .0131 1.013 .0099 1.010 .0118 1.012 educat ion ( .0088 ) ( .0088 ) ( .0088 ) urban residences .1426* 1.153 .1342* 1.144 .1412* 1.152 ( .0556 ) ( . 0557 ) ( . 0558 ) parity -. 0289* .9715 - . 0301* .9703 ( . 0089 ) ( . 0089 ) ch i Id survivals - . 5909* .5538 (. 1296) Period 1 (0-2 months) constant -2 . 371* .0934 -2 . 160* . 1153 -1 . 597* .2025 ( .0561 ) ( .0846) ( . 1484) Period 2 (3-5 months) constant -2.511* .0812 -2.297* .1005 -1. 727* .1778 (.0640) (.0908) (. 1535) Period 3 (6-11 months) . constant -2.285* .1018 -2.071* .1260 -1.496* (.0552) (.0848) (. 1513) Period 4 ( 12-17 months) constant -2.418* .0891 -2. 205* . 1102 -1 . 627* . 1966 ( .0876) (. 1086) (. 1664) Period 5 (18-23 months) constant -2 .465* .0850 -2. 238* .1066 -1 . 657* . 1908 t. 1226) (. 1403) (. 1893) Per lad 6 ( 24+ months ) constant -2 . 791* .0613 -2. 562* .0772 -1 . 973* . 1391 ( . 1719) ( . 1853) ( . 2258) 10gl ikel ihood -5524 . -5519 . -5510 . N a 1908 Notes: a Indicator variable: layer. Standard errors are in parentheses. * indicates significance at the .05 level .

OCR for page 38
45 C al ombia h ( .02 .01 h (t) .02 .01 survive 1_ yes; urb a n- n o s urvival-~; urban n o ...... survival no; urban - no . . ·. . .. .. 3 6 12 1 8 24 Costs Rica 'e .,..~e "~ - ·. . t . . . ~1 1 3 6 12 18 24 t FIGURE 7 Hazard Functions with Covariates for Termination of Breastfeeding

OCR for page 38
46 could not be reestimated for each cohort of women, and the statistical package used does not permit age to be treated properly as a time-varying covariate. The issue therefore remains unresolved. Model 3 adds a dummy variable indicating whether the child survived past age two. The coefficients of this variable are of the same sign and are statistically significant in both countries, but the estimate for Costa Rica is nearly 50 percent greater than that for Colombia. The corresponding hazards are shown in Figure 7. The top panel in the figure graphs the results for Colombia, and the bottom those for Costa Rica. The scale has been adjusted for the mean parity and mean education of the woman and her husband. The sol id line gives the hazard for women whose child survives past age two--that is, well past the usual weaning time--and who do not live in an urban area. When we compare the two countries, there is a tendency among women who have not yet stopped breast- feeding to stop between 18 and 23 months in Colombia and between 6 and 11 months in Costa Rica. The dashed line shows the estimated hazard for women who live in urban areas and whose children survive. The dotted line shows the hazard for women whose children do not survive. Clearly, breastfeeding durations are much shorter for these children in Costa Rica, where durations of breast- feeding are already relatively short. It was not possible to obtain more detailed information on the interactions of child mortality and breantfeeding from these data, for reasons already cited; however, in both countries, child mortality is lower among children who are breastfed than among all children, even in Costa Rica, where mortality in general is quite low. Tables 7 and 8 give the estimated coefficients for the equation predicting contraceptive discontinuation These results are displayed graphically in Figure 8. Model 1 includes the effects of the background covariates alone. Husband's education is statistically significant in both countries, while urban residence is significant only in Colombia. This is no longer the case when other covari- ates are added (Models 2 and 3). In the larger models, h usband ' s education is no longer s ignif icant; urban residence becomes significant in Colombia, reaches borderline significance in Costa Rica, and is associated with longer durations of use. This last effect is larger in Colombia than in Costa Rica, possibly due to the greater concentration of family planning efforts in urban areas in Colombia. Model 2 adds parity, desire for an

OCR for page 38
47 additional child, and contraceptive method. All three variables are statistically significant in both countries. Higher parities are associated with longer duration of use in both countries, and the coefficients are about the same size. Women who desire an additional child and those who use a coitus-dependent method tend to use for shorter durations; these effects are somewhat larger in Costa Rica than in Colombia. When breastfeeding and child survival were added, neither var table was statistically significant in either country. Model 4 contains terms for the interaction of desire for an additional child and method of contraception with subperiod to see how the shape of the hazard is modified by each of these covariates. These results are presented graphically in Figure 8. As before, the results for Colombia are given by the top panel and those for Costa Rica by the bottom panel. The scale for Costa Rica has been expanded to show the detail; the level of discon- tinuation is therefore considerably higher in Colombia. The shape of the discontinuation curves is quite differ- ent for the two countries. The solid line gives the estimated hazard function for women who use coitus- independent contraception and who do not desire additional children. In Colombia, high discontinuation rates for these women are concentrated in the first 3 to 5 months of use. In Costa Rica, the curve is not only lower, but much flatter, although somewhat higher discontinuation rates can be found in the first year of use. The line of dashes and dots shows the estimated hazard for women who use a coitus-dependent method and who do not want an additional child. In Colombia, discontinuation rates are relatively higher for durations of use under 18 months, then drop to join those for the coitus-independent methods. In Costa Rica, the discontinuation rates are relatively high at rather short durations of use, moderate at intermediate durations, and quite high for durations longer than 18 months. The two dashed lines give the estimated hazards for each of the two classes of methods for women who desire an additional child. In Colombia, the principal effect of this variable is to shift the curve upward by about the same amount at all durations. In Costa Rica, for each set of methods, it is only after durations of use of one year or more that the discontinua- tion curve for women who desire an additional child consistently diverges from that of women who do not. This suggests that in Costa Rica, for durations of use of one year, women who want an additional child are as likely

OCR for page 38
48 - e~ o S ·. ~: o ·_t c- ~e, 3 O ~: ~: U] a e" ~~ _ ~ :~: o U] ·_t V U) E~ b0 O _ · - ~4 ~ o x: O O _ · - C O c~ ~0 o · - c: q5 u" o b0 O _ · - C ~S o _ ~ O ~ ~ ~ ~ ~ ~ t_ O O u~ `0 ~ u~ O · ~ CS, ~ r— 0\ _ · · · · ~ ~ ~_ _ ~ ,ie. ~ .~.e 0 r~ ~ ~ l_ ~ ~ 0 ~ 0 _ ~ _ ~ `0 e~ `~0 _ ~ ~ ~ ~ ~ 0` ~_ = - mc~ 0_ 00 0_ O~ 0_ 00 00 ~~ ~0 ~ e ~ ~ ~ e — 1 — 1 — 1 — 1 — 1 u~ O ~ O ~ `0 ~ ~ 0 ~ ~ ~ 0 ~ ~ ~ 1_ ~ _ · · · · - ~ _ u~ u~ · . _ _ _ _ _ '_ a~e _ ·|' ~ c'. ~ t: a' r_ ~ ~J ~ 0 ~— 0 ~ ~ t—~ ~ c~ O ~ ~ _ ~ I_ ~ ~ c~ C~ ~ ~ ~ ~ ~ _ ~ ~ - , ~ 0 _ o~ 0 0 _ ~ _ ~ a, 0_ 0_ 00 00 c~— _0 40 - 0 — 1 ~-- 1 ~ 1 — 1 ~ 1 — ~ ~ , 0 0 - c~ ~ ~ ~ c~ ~ oo O ~ ~D ~ ~ O · ~' o' =` ~ - o' · - · · · · · - - ~ - ~ - ~ ~ ~ ~ ~ ~ ~ ^ c~ ~ o ~ o - oY oo ~ ~ _ ~ ~ u~ ~ ~ a' 0 ~ ~ u~ - - ~ ~ ~ ~ o - ~ o o - ~ - ~ ~ o - o - o o oo ~ - - o ~ o ~ o · · . · . . . . . . . ~ · · ~ . — 1 — 1 — 1 — 1 — 1 ~ — — o' c~ r~ o ~ - o ~ o ~ c~ ~ · · . ~ - - ~ ~ ~ r~ 0 ~t— _4 _~o u~= So Yo ~ ~ - c~ ~ o c~ o - o o o - o - oo oo ~ - · · · · . · · · · — ~ 1 — 1 — — 1 — 'e ~ c c o ~m ·- ~ c ~ c o .- ~- = - o~ :^ ~ ~ 0 e" ~ ~C' ~ o" ~ ~ ~ ~ o - o ~ ~ c ~ ~ · - C3O ~ 0 0- - =- - C ~ 1 oo ~ q5 ~ - ~ ~ ~ ~ :^ · - c ~ ~ ~ c~ ~ ~= ~; ~ .~ ~ ~ U D ~ ~- - · - . - ~L4 ~0 e3 0~ =0 ~ -~- ·-c O ~ O ~ 0= ~= ~ ~ ~ ~ ~ O O 3 e ~- 3 ~ ~ ou 3 ~ CL ~ U ~ t~

OCR for page 38
58 h (I) .02 ~'~ Colombia _ _ _ . C osta Rica A; 3 ~ 12 1 ~ 24 36 He t FIGURE 9 Bazard Functions for Live-Birth Conceptions other variables are added. Model 2 adds contraceptive method and parity. Use of either coitus~dependent or cottu~independent contraception lengthens the time to next conception. The "pact of coitus-independent contraception is nearly twice that of coitus-dependent contraception in both countries, but the impact of either set of methods is greater in Costa Rica than in Colombia. This result holds when other variables are added to the model, suggesting that contraceptive efficacy may be greater in Costa Rica {see Goldman et al., 1982). Bigher parities are associated with longer intervals in be ~ countries, and the estimated coefficients are close in value. Model 3 removes the background covariates and adds the full set of t~.e-varying intermediate variables. Model 4 retur N the background variables to the acde1. Likelihood ratio tests show that, although the effect of the back- ground variables is small, it is statistically signifi- c~nt. A co peridot of Models 3 And ~ reveals that few coefficients change Then the background variables are added. As already noted, along the background variables, only urban residence is artistically significant. Figure 10 displays to iapact of the ti_'rarying cavariates frog Hodel d. The results for Col~bla are in the upper panel Id those for Costa Rice in the later panel. The solid line gives the estimated hazard for men who do not breasted and do not contracept' this corresponds to

OCR for page 38
59 the Natural fertility. line in Diagram 2 with gestation subtracted from all intervals. Although the shapes of the hazards for the two countries accord moderately well, that for Costa Rica suggests an unexpected rise in the interval 3 to 5 months. As expected from previous bio- metric research reviewed earlier, the impact of breast- feeding is more marked at durations of less than one year, after which the two curves converge, while the largest impact is for durations of under 3 months. The dotted line, which shows the case where the child does not survive, is most useful for comparison. For example, if a child were breastfed for 6 months and then died, the risk of conception for its mother would be the dashed line for the first 6 months and the dotted line thereafter. The estimated effects of this variable are rather larger than expected, particularly at very short durations. Some of this effect may be due to the tendency of short birth intervals and high infant mortality to be mutually rein- forcing in a manner that is not completely captured by either breastfeeding or contraceptive behavior. The two lowest lines in each panel show the conception rates for women who contracept. The variable indicating contracep- tive method shifts the hazard to a very low level in both countries; the time-varying cover iates act principally to rearrange the shape of the hazard slightly so that the effect of contraception is not a simple proportional shift of the solid line. This is particularly noticeable in the last subperiod (24+ months), where the hazard for contracepting women levels off from its downward course. The similarity of these results for the fertility equation in the two countries is striking, particularly in light of the behavioral differences shown by the models of breastfeeding and contraception discussed earlier: not only are the signs on many of the coefficients the same but many of the coefficients are close in magnitude. Although different propensities and durations of breast- feeding and contraception are used to obtain close to the same level of fertility, the impact of a particular behavior is nearly the same. This reinforces the idea expressed earlier that this portion of the model accesses fundamental biometric aspects of fertility, whereas the other portions of the model are more behavioral in nature.

OCR for page 38
60 o" ~ . - _ o ~: o (U C o .. C) - ~ so 3 O _ V dV ~ O V ~ ~ O O O ~ .- _ ~ OQ ~ _ ~ C) V 0 _ _ ~ O \0 0 0 ~ · · _ _ _ ~ ~_ m0 ``D ~ r~ ~ 0 00 00 · . . — 1 — _ _ `0 ~ o~ ~ 0 0 U~ _ U~ o' 0 · ~ ~ e _ _ iC 'D~ _0 mm _ O ~ ~ ~ ~ 0\ ~ ~ O~ ~_ U~^ _ - 00 00 _0 ·_ · · · · · · _ - ~ 1 — 1 — 1 — .^ * m~ 00 _ 0 `0 0\ 0C~ 00 U~ _ ~ O · · · ~ 1 ~ 1 ~ U~ ~ _ O O ~ O 0` · . — o - iC _ O U. ~ C O C · C~ ' 1 _ 0 - - . ~ ~ ~ ~ o~ ~ r_ _ 0 0 ~ ~ l~ O ~ ~ ~ _ - - ~ 00 ~m ._ ~ ~ ~ O · ~ _ · · · , · · ~ - 1 ~ 1 — 1 — 1 O ~ O O ~ _ _ O O ~ `0 C~J ~ O ~ · · ~ o0 ~ U~ 0\ 0 · _ _ · · · · · ~ o ~ o - ~ o' io ~ ~ 0 - o 0 ~ o' ~ ~ ~ o' ~ ~o o c, o o 8 0 80 _ o . o ~ o ° 8 ~ ~ .. o . .. .. .. _ . .. .. ~ . 1 — ·— — 1 — 1 — 1 — 1 — 1 — 1 ~o ~ o o' ~ o O ~D 0\ 0 0 0 0 o~ .0 _ o~ ~ o, ~ o · · · · · . ^ *^ _ ~ o _ 0 ~ ~ U, ~ _ o~ o ~ ~o _ ~ ~ _ ~ 0 o o ~ o 0 _ _ o _ ~ U~ ~o C~ o o o o o o o o C~ o . _ .. .. .. .. .. ~ . 1 — 1 — 1 — 1 — 1 — 1 — 0 s C C o C. ~a C ~ C =0 C 0 1 ~ ~ ~ ~ ., ~ ~ o O. ~ ~ c 0 c ~ ~ ~ ~ ~ c~ ~e 0 - 0 ~ -~ =~ ~ ~ rl o0 ~ 0 0~ - =~- C 1 ~ 1 U C o.c: ~ ~ ~ ~ 0 ~ 0 · - C ~ ~ ~ C ~ ~ ~ ~ ~ ~ 0 ~ D ~ .- - ~ ~ ~ ~ · - . - 0 ~ ~ ~ 0 e ~ 0 ~ D 0 · - C · - C ~ ~ ~ O ~ 0= ~= ~ ~ O O O O ~ ~ O xe 3- ~ -~ ~ oo oo ~ ~ ~

OCR for page 38
61 A At A o' At · · * ^ ^ * ~ ~ o c~ - 0 u~ ~ C~ O u~ ~D `O mm - ~ · - · · · · - ~ 1 _ — 1 — ~D o . I_ ~ ~D - 0 ~ ~ - ~ co · · ic ^ 1 - _ m0` C~ - 0= ~ ~ ~ ~ ~ - 0 ~ u~ ~ ~ ~ - - ~ ~ - c~ . ~ · c~ _ ~ ~D c~ · - · · ~ · . 1 — 1 — _ 1 — ~ - ~ c~ c~ c~ ~ a' ~— ~ — 0 o - o 0 ~ - - · ~ o ~ · · - · · · - ~ * - ^ ~ ^ * ^ ~ 0 ~ ~ o ~ ~ - ~ ~ o ~ u~ ~ '= ~o 0`n - ~o ~c~ 00 - ' O C~ ~ ~D _ ~D ~ 0 ~— ~ ~ ~ c~ a~ ~ —~ . ~ . ~ _ c~ _ ~ ~D ~ · - · - - · c~ · .. . . .. 1 — — 1 — 1 ~ 1 — — 1 — - 1— o . ^ o o · - - 1 — c~ c~ o . * - o ~ a' 0 · o 1 — ~c o D D t~ '_ b0 C ~ U~ C O :. 1 · - · - · - ~ _ C~ ~ 0 V ~ ~ 0 ~ ~ O '_ . - C . - ~ O D c' c~ 00 C C O . - . - C ~ C) 0 0 ~ ~ c ~ {: o ~ o D - . - 0 - . - ~C C)

OCR for page 38
62 g t_ ~ . - _ C o ~: 8 C) o" _ · - _ x: ~4 o" _ x: 8 C) _ ~ C _ 3: o 0 0 ~ 0 U~ 0 o ~ ~o _ C~ o~ ~ ~ ~ I_ o o~ O ~— ~ r— O · a' o~ · · . . . _ · ~ ^ 1— 0~ ~n c~ _ _ u~ ~o o~ I_ ~ r~ ~ c~ c~ ~o - - cr~o u~ - ~o_ ' - '_o mc~ _ ~ C~ o 0 o~ _ ~ ~ _ C~ o 0 .` C~ · ~ C~ _ o _ C~ C~ · C~ o _ o _ o C~ · · - · - · ~ C~ · · - · ~ · ~ 1 — 1 _ 1 _ 1 _ 1 _ ~ 1 _ 1 _ C~ I_ 0 ~ U~ 0 ~ ~ o' _ o o o' _ ~ o · 0 ~ ~ o · ~ 0 · · . . _ · ~ ~^ ^ - ~ ~a' __ _0 m~ - - ~o _ - o _ C~ o 0 ~ _ · C~ ~ _ o _ ~ ~ C~ · · ~ · . · ~ 1 ~ 1 _ 1 _ 1 _ o o' U~ o . ^ o o ~ o~ 0 0 · o - 1 — 1— C~ o . ^ r~ ~o o ~o ~o · o - 1 — o - - 1 ~D o ~ ·. - CO C ~ o P~ D ._. a' D D 1_ C) ~ _ 1 O · - 1 ·— - ~ ~ - " ~ 0 - ~ ~ o · - c · - o ~ ~ ~ ~ P4 m 0 r~ _ c~ r~ _ .~ O ~ c~ . . . r_ ~ C~ o~ ~— oo =_ 0 0 _ C~ 0 _ _ · . · . 1 _ 1 — ~ I 0 ~o 0 . U~ ~ ~ '_ _ C 0` r · O C~ e C 1 1— o o . ~ O C~ o, ~0 · O · r. 1~ ~ 1 0 tD ~C ~ . C ~ 8 8 D ~0 ·~4 C ~ 0 O ~ C) C o . - C~ C~ o D _ C~ ql 1 ~ 0 · - _ 0 O _ · - ., - ~ C~ cc: C U~ C o tJ

OCR for page 38
63 _ a, 0 oo ~ ~ o ~ ~ _ C~ ~ _ C~ ~ _ ~ ~ · O ~_ ~ ~ e · · _ · · _ _ _ O 0 ~ *~ ~ .~ .~ ~ .~ 1 C~— ~ ~ —~ ~ ~ a~ ~ ~ ~ ~ . _ ~ U~ O C~ ~ ~ ~ O C~ ~ ~ ~ _ 0 ~0 ~ C~ U-~ 1— o0 ~0 1—~ ~ ~ ~ ~0 O— 1—~ —~ · _ C~— _ _ ~ _ ~ ~ · · e · · · 1 — 1 — — 1 — 1 — — — =~ I_ =~ ~ ~ ~ 0 C~ ~ ~ 0 0 O 0 ~ ~ `0 _ · ~ · O t— · · _ . _ . · _ _ ~D 0o ~ .~ ~ *~ _ _ ~ 1 u~ ~D —~ ~ O O C~ 0 ~ ~ `0 0 ~ C00 =0 0~m 0~ t~'C~ =1~ =0 C~`D C~C~ C~ - 0= = - ^= =`D 0— r~ c~ _ ~ · _ c~ _ _ _ c~ _ .. .. · - d" · · - · - · - — 1 — — 1 — 1 — ~ — lt ^ 1 U~ 0 0 C~ 1 ~ C~ · O 0 Os ~ 8 c~ _ D c-) _ 00 ~ ~ + ~ C ~ ~ ~ O ~ ~ C O ~ O ._~ .,4 ·— C~ ·,4 .~4 ·— O _ ~ ~ ~ ~ .- 0 C ~ ~ 0 1D eU ~ ~ ~ ~ ~ 0 ~ ~ O ~ 0 ~ ~ .,` tt ~ _ ·- 0 C' ~ _ ~4 ~ C '- ~ C ~ C ·- 00 s" 0 ~e ~ 0 ~ 0 ~ 0 c' F~ c' D ~ c' _' _ _ C CL "= U~ C~ z e - ea o · - 0 o · - O O S" 0. C C ·~~ ·~d . 0 0 :^ :^ _ _ _ U~ ~ e O ~ ~ a' '_ _ DD .1 40 ..,4. - 0 (U ~ 0 C :^ _ ~ S~ O O =o ~ ~ C .. ~ ~ C ~ ~ C)- - ~ _ · - · - · - ~ ~ ~ ~ - ~ e e ~.- .~ - ~ C oo oo sa 0~ - C C ~ 0 .,1~ - 0 ~ :^ :^ ~ 0 0 0 ~ ~= ~ 1 1 ~ ~) .~. ~ ~ ~.~4 a~ C-~- - ~ C ~E- - C.- cn ~D ~ · ~ 0 o

OCR for page 38
64 - o ~: u - ~; _ ·. c o - ~3 C~ ~ _ _ ~ _ o ~4 ~C ~4 o U) — 0 — ~ o #) ~: - v o ~; E~ o _ · - o o - . - c - 4 o so o _ . - 4 e os 0 ~o - . - 0 0 ~ _ 0 ~ _ 0 0 ~ ~ ~ 0 0 0 ~ 0 0 0 ~ ~ _ `o o, · o~ ~ ~ ~ ~ o _ · _ · _ _ ~ ~ ~ ^ ~ ~ ^ .^ ~ ~ ^m ~D~ _~ —~ ~o ~o~ ~r~ _— —< U~ 0 ~ C~ _ _ 0` U~ ~ ~ O O _ ~ O 0 0 ~ o, 0 _ _ 0 ~ `0 ~ ~ ~ ~ ~ _ ~ ~ oo oo oo oo ~o ·_ 0_ oo ·L .. .. .. .. . . _ . .. .. ~ — 1 — —~ 1 — 1 — 1 — 1 ~ 1 — 1 ' ~ 0 ~ O ~o 0 ~ o' · · ~ o o, U~ o . — ~ ~ ~ ~ ~ ~ 0C~ 4 - o~ o. `0 0 _ 0 _ o~ r~ r 4m 0~ 00 0C · _ o~ _ 0 0 U _ . · · · ~ C~ 1 — 1 ~ 1 ~ 1 ~ ur, c~ o~ 0 ~o _ 0 0 U~ ~o 0 ~ _ , 0` ~ O 0 ~ 0 0 o' U~ 0 _ · 0` · C~ r~ t~ ~ ~ O _ · _ · · · ~ ~ - ~ ~ ~ '_ ~ ~ ~ ~ ~ _ ~ _ r~ _ ~ ~o O c~ ~ ur~ ~o t_ ~ _ ~ ~ ~ ~ ~ O ~ ~ ~ O u~ 0 ~ o' 0 _ —0 ~ ~ ~ a, —~D C~ _ ~ ~ O O O O O O O O ~ O · O r~ O O O · - .. .. . . .. . . _ . .. .. — ~ — — 1 — ~ — 1 — 1 — 1 — 1 O ~ ~ 0 ~ 0 0 0 ~ _ ~ O ~ 0 0` 0 _ · C~ · 0` 0 0 _ · _ · · _ ~ ~ ~ ~ ~ ' U~ O _ 0 C~ ~ ~ ~ `0 C] oY ~ U~ 0 ~ O ~ 0` _ C~ ur~ ~; _ 0 0 ~ O _ O O ~ `0 ~ ~ 00 00 00 00 C~O ._ · · · · · · · · · · -;t — 1 — ~ 1 ~ 1 ~ 1 ~ C C O C C. ~. C ~ C =0 C 0 1 .~4 ~ - O ~ q5 ~ ~ = - _ : - ~ `: ~ C ~ ~ ~ ~ ~ ~ e o - O ~ ·- ~ ~ ~ _ 3 0 0- - =~ - ~ 1 ~ 1 c, .C - ~ C ~ ~ 0 ~ ~ ~ :^ ~ ~ a, ~ ~ ~ ~:5 ~ ~ ~ ~ ~ 0 :~ ~ ~ ~ D ~ ~ - ~ ~ ~ ~ . - . - ~ 0 e ~ 0 ~ D 0 ·— C ·— C kd k4 O ~ O ~ ~= ~ ~ O O O O ~ - 3 C q' 0 C o u

OCR for page 38
65 ~ - ~ 0 ~ o' 0 ~ o · ~ o o' ~ ~ ~ - ~ 0 u~ e~ t_ ~_ ~ e * ~ ~ ~ ~ ~ * ~ ~ _% 0\ e~ C*' ~ e;' 0` ~ ~ _ ~ ~ a ~ O - . u~ ~ ~ 0 0 ~ r~ 1— 0 ~ ~— 0 0 _ c~ ~ a' m~ - - 0^ ·m ~D— C~~ C~'— · e ~ C~ · e · e e e — 1 ~ 1 — — 1 — 1 _ 1 _ 1 _ `0 `0 ~ 0 _ ~ ~ O ~ ~ 0 C~ ~0 · O U~ e _ ^ _~ 0 `0 C~0 0 0` ~ 0 C~J C~ C~ ~ ~ ~ C~ ~ ~ ~ 0 ~ · e e ~ ~ ~ 1 — 1 — _ . . C o · - C o C) ~ ^ * ^ ^ _ a, _ c~ _ c~ 0 0 ~ _4 m~ _~ - 0 r~e~ _a) - 4 == · ~ `o _ ~ ~ ~ ~ 1 ~ 1 _ 1 _ 1 _ U~ o . * 0 · C~ 1 o C~ o . * ^ U~ · O r~ ~ 1 — D c) _ 00 C ~ O · e" ~ - 4 · ,e" e~, ~ ~ C~ ~ dJ ~ (IJ e, ~e ~ C., 0 e" 0 hd ~ ~ ~ ~ _ C ~ ~ · - O ~ C} {l. ) eC) ~ C)

OCR for page 38
66 - at: - Cal o ~: - - ~: - ~: o t) - o m E~ so 0 ·- 0 tlo 0 ~rl C 0 0 c) 00 0 _ ·- C ~4 0 C) 00 0 _ .- C o 0 _ r~ _ ~ ~ _ 0` _ O ~ ~ 0 ~ — o, ~ 0~ _ 0~ C~ 0~ _ 0 a~ 0 ~ 0 · ~o a~ ~ · _ · a ~ ^ ~ _ _ ~ ^ U~ ~ ~ ~ 0 ~ U~ ~ C~ ~ ~ ~ ~ ~ —~ 0 `0 _ ~ U~ ~ ~ ~ ~ O ~ ~ ~ `0 ~ 0` ~) l~ ~o ~ —I~ ~ ·^ ~ a, co ~ ~o 0 0 1— ·~ __ __ m~ ·~ __ m_ ~~ C~ · e ~ · e · · 1 — 1 ~ 1 ~ ~ 1 ~ 1 — — 1 — 1 ~ ~ _ ~ I~ O U~ t~ _ ~ 0\ 0 U~ 0 0` 0 t~ ~ ~ O 0 0 o' ~ 0 . ~ r~ · ~ · · . _ ~ ~ ~ ^ ^ ~ _ _' m' ~~ - 0 `0 ~ 0 ~ ~ o0 ~ ~ 4^ 0 - 0 r~ 01^ ·~ __ __ - ~ C~ · ~ ~ · · · ~ 1 — 1 _ 1 _ _ ~D o . _ O · _ 1 C~ C~ o ~ 0\ U~ O ~ C~ ~ 0` ~ ~ 0~ ~D 0 a~ ^~_ - - - - =` ·~ ~_ m_ ~~ . · . · . · . 1 ~ — 1 _ 1 _ 0` ~0 o . _ O · _ 1 — 0 C~ o . . .^ * ~ ~ O C~ m~ '1_ t_ r~ I_ · O · O · C - ' - 1 — ~ 1 — 0 "C c O ~ D r~ D _ D ~ _ _ ~ c) _ _ 01) C a~ 1 01) {: ~1 1 C O ~ ~ ~ O . - . - . - _ . - . - . - S" ·d ~ ~ ~ ~ ~ ~V (V 0 ~ - 4 C) q5 V C' O ~ oo ~ ~ O ~ 0 ~ ~ ·." a, ~ ~ _ · - 0 c' ~ _ C ~ C: · - ~ C ~ C · - ~ O ~ O .`C ~ O ~ O P~ ~ D c~ c) ~4 ~ D c~

OCR for page 38
67 C~ o C~ ~o ~o _ ~o O · ~ cr · _ · - 0 ~D O C~ ~ ~ u~ ~D — o o~ . · · _ _ _ _ * ~ ~ ~ ~ _ ~ ~ ~ _ ~n ~ ~ _ ~ _ c~ t_ u~ ~ ~ _ a~— _— a~ ^ ~ ~ ~o _ _ ~ o~ u~ ~ ~o~ - - mo, - 4 ~o - ` 0 - ~o' · ~ _ ~ o _ o C~ · ~ o _ ~ _ _ _ ~ . . . . ~ . p ~ 1 — — 1 — 1 — 1 ~ 1 — _ _ _ ~ U~ ~o _ C~ o . o, · _ · - m~ ol— C~_ C~ co ~ o' . - _ C~ · · ~ 1 — _ ~D o ~D o . ^ C~ o ~ 0 _ · _ C~ 1 ~ ~0 U~ ~ ~ o_ _ - · . · ~ 1 _ 1 — ^ o o 0 0 · o - 1 — 0 a, o V e c °e ~J D ~ _ 1 t3O ~ ~ + 0 ~ O _ ._, . - . - ~ _ U~ C ~ m ps ~ a' O ~ a' ~ ~ O . - ~ ~ ~ _ e" C · - O ~ O CJ - o · _ .^ ~ _oo ~ O - 0 0~ oi 0 ~ 1— ·~ O ~ . . . 1 _ ~ U~ - o ~t ^ 0 0 U~ C~ ;t ~ · O C~ - 1 — o - o . ~ ^ —1— ~ 0 U~ · O - 1 ~ _ ~D ~D 0 `0 0 · ~ _ _ *^ ~= 0r~ ~ 0` `0— 0` ~ ~ 0\ _ O— · . · _ D C) _ 00 C O .,- . - . - CL 3 C.} 0 {D ~ ~ _ · - O ~ O ~C C~ D ~ CJ t-" `0L =- O _ 1 ~_ r~ 1 r_ O l . o o ~C .- _ so n 0 _ Z o . - - . _ _ C ~ ~ P. 0 0 .,-. - ~_ 0 0 :^ 1' n ~ _ _ .... ·o ~ ~ 0 - _ DD 0 ~1 .,4-~.C · ~ ~ - 0 ~ ~ C n ~ _ 0 0 =o ~ ~ C .. ~ e. C ~ C) . - ~rl . - ~ ~ C ~- - .,4~— ~ C o. ·rl 0 s" os so 0 C ~ _ .,. ~ 0 ~ :^ :^ ~ 0 0 ~ ~ ~ ~ C) 1 1 ~ C) ' - ~ ~ ~.r4 ~ ~ ~ a~ C · - ~ ~ C f~ — C~ U, ~ ·0 Z

OCR for page 38
68 Col ombia h (t) .06 .04 .O2 h (tJ ·06 .04 .02 . . i. bra astf eedi ng ~ o ~ c on trace pt i or`: n bre If Bed i n 9 :~; c ontraceptior' . n 0 b re~tfee di n g . n o; c ontre ceptio n: coitus independent broastfeedi n 9, n o; C ontracep~tion coi t us dependent c hild survival: n 0 in; 3 6 12 18 24 Costs Rica t An. a. 3 6 12 18 24 FIGURE 10 Hazard Functions with Covariates for Live~Birth Conceptions t