teresa sobieszcsyk and lindy williams · pdf filedepartment of rural sociology warren hall...

22
Teresa Sobieszcsyk and Lindy Williams Assessing the Value of Proxy Reports: Do Spouses Really Know What They Think They Know about Their Partners' Fertility Attitudes? PSC Research Report Report No. 01-479 June 2001 P OPULATION S TUDIES C ENTER AT THE I NSTITUTE FOR S OCIAL R ESEARCH U NIVERSITY OF M ICHIGAN PSC

Upload: domien

Post on 04-Feb-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Teresa Sobieszcsyk and Lindy Williams

Assessing the Value of Proxy Reports: Do Spouses Really Know What They Think They Know about Their Partners' Fertility Attitudes?

PSC Research ReportReport No. 01-479

June 2001

P O P U L A T I O N S T U D I E S C E N T E RAT THE INSTITUTE FOR SOCIAL RESEARCH

UNIVERSITY OF MICHIGAN

PSC

Page 2: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

PSC Publications Population Studies Center, University of Michiganhttp://www.psc.isr.umich.edu/pubs/ PO Box 1248, Ann Arbor, MI 48106-1248 USA

The Population Studies Center (PSC) at the University of Michigan is one of the oldestpopulation centers in the United States. Established in 1961 with a grant from the FordFoundation, the Center has a rich history as the main workplace for an interdisciplinarycommunity of scholars in the field of population studies. Today the Center is supported by aPopulation Research Center Core Grant from the National Institute of Child Health andHuman Development (NICHD) as well as by the University of Michigan, the National Instituteon Aging, the Hewlett Foundation, and the Mellon Foundation.

PSC Research Reports are prepublication working papers that report on currentdemographic research conducted by PSC associates and affiliates. These papers are written fortimely dissemination and are often later submitted for publication in scholarly journals. ThePSC Research Report Series was begun in 1981. Copyrights for all Reports are held by theauthors. Readers may quote from, copy, and distribute this work as long as the copyright holderand PSC are properly acknowledged and the original work is not altered.

Page 3: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Assessing the Value of Proxy Reports: Do Spouses Really Know What They Think They Know

about their Partners’ Fertility Attitudes?

Teresa R. SobieszczykPostdoctoral Research Fellow

Population Studies Center426 Thompson St.

Box 1248 University of Michigan

Ann Arbor, MI [email protected]

Lindy WilliamsAssociate Professor

Department of Rural SociologyWarren Hall

Cornell UniversityIthaca, NY [email protected]

The data for this study were collected as part of a broader project studying unmet need for familyplanning in the Philippines. The principal investigators on the project were John Casterline, nowat the Population Council, and Aurora Perez, at the University of the Philippines. Substantialinput on the project was contributed by Ann Biddlecom, who is at the University of Michigan,and Joy and Florio Arguillas, now at Cornell University, and a number of others who are at theUniversity of the Philippines. We are grateful to all involved for their various roles in the projectdesign, data collection, etc. Support for the research was provided from the Evaluation Project atthe University of North Carolina (grant number 5-36701), the Ford Foundation office in Manila,and the Mellon Foundation.

Page 4: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Abstract:

The paper examines the correspondence between husbands’ and wives’ own reported attitudesabout the timing/occurrence of their most recent pregnancy and proxy reports about thoseattitudes provided by their spouses. Survey data from the Philippines are used to analyze thedegree to which agreement between proxy reports obtained from spouses and the respondents’own reported attitudes is due to chance. Potential problems are explored regarding the use ofproxy reports for assessing spousal attitudes regarding pregnancy intendedness. Several factorsthat contribute to the mismatch between male and female respondents’ reported attitudes andtheir spouses’ proxy reports of those attitudes are identified. The paper concludes with somerecommendations about the use of proxy reports when collecting data about pregnancyintendedness and related fertility issues.

Dataset used:

In-Depth Study of Unmet Need for Family Planning: Philippines, 1994

Page 5: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

1An important exception is Stycos’ research on men and women and fertility issues inLatin America (see, for instance, Stycos, 1955; Stycos, 1968).

1

Assessing the Value of Proxy Reports: Do Spouses Really Know What They Think They Know

about their Partners’ Fertility Attitudes?

In past research on fertility issues, data for married couples generally were collected fromand about the wife alone, both because researchers assumed that women are the most reliablesources of information on reproductive and contraceptive histories and because it was argued thatfertility decisions are most pertinent to women, who undergo the physical challenges ofpregnancy and childbirth. Attitudinal questions regarding contraception and fertility havelikewise been addressed mainly to women, based on the assumption either that women’s attitudesare more important or that their attitudes agree with those of their husbands.1

More recently, however, there has been growing interest in the role of men in fertilitydecisions and behaviors (see, for instance, Thomson, McDonald, and Bumpass, 1990; Bankole,1995; Stycos, 1996; Becker, 1996; Unalanl, 1998; Biddlecom, Casterline, and Perez, 1997;Bankole and Singh, 1998). While a number of recent studies have interviewed men directly, inmany cases researchers have continued to rely on proxy reports provided by wives in order toaddress the fertility attitudes and behaviors of husbands. A debate still exists about whether thebenefits of surveying both spouses separately, rather than using proxy reports, outweigh theadded expense and complexity of data collection and analysis entailed by such an approach(Williams and Thomson, 1983; Becker, 1996; Biddlecom et al., 1997; Unalanl, 1998).

Among those arguing for separate data collection strategies are those who question thevalidity of proxy reports. Possible threats to validity include the fact that respondents may haveforgotten or may not know their spouses’ views or that they may project their own views ontotheir spouses (Williams and Thomson, 1985; Becker, 1996). Moreover, people may be lesslikely to answer proxy report questions than other questions, contributing to problems withmissing data (Williams and Thomson, 1985).

Cognizant of these concerns, a number of investigators have opted to assess the attitudesand behaviors of both spouses separately. Several studies have focused on the fertility desiresand behaviors of both members of married couples because of increased awareness that husbandsand wives may have different reproductive attitudes and behaviors and that husbands’ attitudesmay have an independent effect on couple behavior (Beckman 1983; Thomson et al. 1990; Ezeh,1993; Bankole, 1995; Becker, 1996; Lasee and Becker, 1997; Biddlecom et al., 1997; Bankoleand Singh, 1998; Biddlecom and Fapohunda, 1998). To date, the majority of studies includingthe desires and behaviors of both members of married couples have concentrated on assessingfamily size ideals and preferences (see, for example, Lolagene and Fernandez, 1978; Williamsand Thomson, 1985; Mason and Taj, 1987; Fapohunda and Todaro, 1988; Williams, 1990;Thomson et al., 1990; Bankole and Singh, 1998; Mason and Smith, 2000) or attitudes towardsand use of contraceptives (for example, Ezeh, 1993; Salway, 1994; Biddlecom et al., 1997; Laseeand Becker, 1997; Mason et al. 1998; Mason and Smith, 2000).

Page 6: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

2

A handful of studies have used data pertaining to both spouses to examine attitudes as towhether or not a past pregnancy was intended or wanted (for example, Williams, 1994; Williamset al., 2001). Bankole (1995) has indicated that this is important, as it is not realistic to assumethat a husband and wife will always view the addition of another child equally favorably. Further, classifying a pregnancy as unwanted when only one spouse may view the pregnancy thatway may be problematic from both a research and a policy perspective.

In this paper, we focus on pregnancy intendedness among couples in the Philippines.Using survey data from metropolitan Manila and the province of Nueva Ecija, we examine thecorrespondence between husbands’ and wives’ own reported attitudes and proxy reports providedfor them by their spouses on the issue of the intendedness of their most recent pregnancy. Forcomparison, we also look briefly at correspondence between respondents’ own reports and proxyreports from their spouses regarding attitudes towards the use of contraception and the idealnumber of children.

We then focus on potential problems arising from the use of proxy reports for assessingspousal attitudes regarding pregnancy intendedness that contribute to the mismatch between maleand female respondents’ own reported attitudes and their spouses’ perception of those attitudesas reported in their proxy reports. We conclude with some recommendations about collectingdata about pregnancy intendedness and other fertility issues from both spouses.

Potential Influences on Proxy Reports of Spousal Attitudes Regarding the Intendedness ofthe Couple’s Most Recent Pregnancy

Our analysis focuses on the accuracy of proxy reporting, looking at three main factors andcontrolling for standard demographic indicators. First, when assessing the likelihood that onespouse will accurately report the views of another, it is clear that poor communication mayimpede one spouse’s ability to accurately perceive and report the other’s attitudes (Stycos, 1955;Lasee and Becker, 1997). We examine two indicators of couple communication: respondents’reports of the frequency they discuss having children with their spouses and respondents’assessment of the difficulty they have in communicating with their spouses about sex. Weexpect that respondents who infrequently or never discuss having a(nother) child should be lesslikely to correctly report their spouses’ attitudes regarding pregnancy intendedness than thosewho discuss this issue more frequently. Similarly, we expect that respondents who report havingdifficulty communicating with their partners about sex should be less likely to correctly reporttheir spouses’ attitudes on fertility and other related topics than those who more easilycommunicate with their partners about those issues.

The second variable central to our analysis is respondents’ own views of the wantednessof the couple’s most recent pregnancy. One of the most important potential problems with theuse of proxy reports to measure fertility issues is thought to occur when respondents project theirown views onto their spouses because they either are unaware of their spouses’ true views orassume that their spouses’ views agree their own views when, in fact, couple agreement does notexist (Williams and Thomson, 1985). If respondents are projecting their own views onto theirspouses, then the respondents’ own views regarding the intendedness of the most recentpregnancy and whether or not they have achieved their own fertility goals would be expected to

Page 7: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

2 It should be noted that this classification may heighten correspondence betweenrespondents’ own reports and their spouses’ proxies over what would be observed if thewantedness variable included a third category (intended, mistimed, unwanted), as it is sometimesoperationalized. Becker (1996) argues that respondents are probably better able to provideaccurate proxy reports when there are fewer response categories than when there are more.

3

significantly influence the proxy reports they provide for their spouses (Thomson and Sprecher,1982; Williams and Thomson, 1985).

Finally, Bankole (1995) recently found that men and women may influence a couple’sfertility decisions differently at different parities and life course stages. In addition, when familybuilding is beginning, disagreement is less common (additional pregnancies are desired by bothspouses). As a result, in this analysis, we control for parity, and we add a variable that assesseswhether the respondent feels that s/he has the ideal number of children at the time of interview,whether s/he has fewer than would be ideal, or whether s/he has had more than would be ideal. We expect that in couples at lower parities, spouses will be more likely to agree that their latestpregnancy was intended and therefore will also be more likely to correctly identify their spouses’view. Similarly, respondents who feel that they have had fewer children than would be ideal maybe more likely to agree with their spouses that their latest pregnancy was intended and thus toassess correctly their partners’ preferences. Data and Methods

Our empirical analysis is based on a 1993 survey on contraception and fertility in thePhilippines, collected collaboratively by the Population Institute of the University of thePhilippines and Brown University. Survey data were gathered in two provinces on the island ofLuzon. Within barangays (neighborhoods), probability samples of currently married womenaged 25 to 44 were selected for the survey. Seven hundred and eighty married women and theirhusbands were interviewed. A supplementary sample of 420 women was drawn to add statisticalpower to analyses conducted using only the women’s data; this sample brought the total numberof women interviewed to 1,200. Insofar as was possible, couples were interviewed separatelyand simultaneously. The interviewer’s sex matched that of the respondent. For this paper, weanalyze only the data for the couples and further restrict the sample to those who had one or morepregnancies in the five years leading up to the survey (473 couples). Table 1 (attached) showsthe summary statistics for the characteristics of these respondents.

For couples who reported having at least one pregnancy in the five years prior to thesurvey, the intendedness of the most recent pregnancy in that time period was assessed using aseries of questions on pregnancy wantedness comparable to those used in the U.S. NationalSurvey of Family Growth. Pregnancies that occurred at a time when a respondent had wanted tobecome pregnant or that occurred later than desired are classified as ‘intended.’ Pregnancies thatoccurred too soon or at a time when the respondent preferred not to have any more children areclassified as ‘unintended.’2 The survey also asked questions regarding how strongly respondentsfelt about the intendedness of the couple’s most recent pregnancy in the five years preceding thesurvey. Each spouse was then asked a similar series of questions about his or her spouse’s

Page 8: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

4

attitudes regarding the intendedness of the pregnancy. In keeping with what is noted in footnote1, we expect proxy reporting to be somewhat less accurate on the four-category intendednessvariable than on the two-category variable. Respondents’ proxy reports of their spouses’ viewson the intendedness of their most recent pregnancy had more missing cases than most othervariables we employed.

The survey instrument also asked both spouses whether they approved of contraception. Respondents were asked, “In general, do you approve of couples using ways to avoid gettingpregnant, or do you disapprove of couples using ways to avoid getting pregnant.” Each spousewas then asked a similar series of questions about his or her spouse’s attitudes towardscontraception.

In addition, the survey asked both spouses for their ideal number of children and for aproxy report of their spouse’s ideal number of children. The level of missing data for variablesregarding respondents’ proxy reports of their spouse’s ideal number of children was comparableto that for the intendedness variables.

In general, the coding of independent variables is relatively straightforward and is shownin Table 4. For our analysis of pregnancy intendedness in Table 4, we use dependent variableswith three outcomes: (1) the respondent perceived that his/her spouse felt the most recent pregnancy was unintended, but actually the spouse felt that it was intended (e.g. the respondentunderestimated the spouse’s satisfaction with the timing/occurrence of the most recentpregnancy), (2) the respondent correctly perceived that the spouse felt that their most recentpregnancy was intended or unintended, and (3) the respondent perceived that the spouse felt thepregnancy was intended, but actually the spouse felt that it was unintended (e.g. the respondentoverestimated the spouse’s satisfaction with the timing/occurrence of the most recent pregnancy). The multinomial logit models shown in Tables 5 and 6 use the same dependent variables; thereference category for the dependent variable is that the husband or wife correctly estimated thespouse’s views.

In the first part of our analysis (Tables 2 and 3), we use the Kappa index as a measure ofagreement between respondents’ own reports and the proxy reports provided for them by theirspouses. The Kappa index takes into account the level of agreement expected by chance alone. Kappa scores range from 0.0 to 1.0, with 0.0 indicating that agreement is no greater than whatwould be expected by chance and 1.0 indicating perfect agreement. A statistically significantKappa index indicates that the respondents’ own reports and the proxy reports agreed to a largerextent than would be the case by chance alone.

Results and Analysis

The level of agreement between men’s reports of their own views about variouscontraceptive and fertility issues and their wives’ perceptions of what their husbands thoughtabout those issues is shown in Table 2. The overall levels of agreement for each of the variablesshown in the table demonstrate medium to high levels of agreement between husbands’ reportsand their wives’ perceptions of what their husbands thought. The statistically significant Kappascores indicate that the levels of agreement are greater than would have occurred by chancealone; however, it should be noted that the Kappa scores are quite low.

In terms of pregnancy intendedness, women’s perceptions of whether or not theirhusbands thought their most recent pregnancy was intended or not matched their husbands’ own

Page 9: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

3 Eighty-seven percent of women’s proxy reports were within one child of their husbands’reports, however (data not shown).

5

reports in about two-thirds of the cases (66 percent). As expected, the correspondence was lower(51 percent) for the four-part variable, which also reflects how strongly the respondent felt aboutthe intention status of the pregnancy. Wives were more likely to provide a proxy report thatmatched their husbands’ own reports when their husbands considered the most recent pregnancyto be intended rather than unintended. While we expected more strongly held views tocorrespond more closely than less strongly held views, this was the case for pregnancies thatwere considered to be strongly intended but not for those considered to be strongly unintended.

Wives had a somewhat higher level of knowledge concerning their husbands’ attitudesregarding approval or disapproval of contraception. Women’s perceptions of their husbands’attitudes towards contraception agreed with the husbands’ own views in more than 80 percent ofthe cases. However, they were much more likely to have provided proxy reports that matchedtheir husbands’ own reports when their husbands approved of using contraception rather thanwhen they disapproved.

Women’s knowledge concerning their husbands’ ideal number of children was somewhatlower than their knowledge of their husbands’ approval or disapproval of contraception and theirknowledge of the wantedness of their most recent pregnancy (two-category variable). Theirperceptions of their husbands’ ideal number of children matched their husbands’ own reports inonly about half of the cases.3 Interestingly, women’s proxy reports of their husbands’ idealnumber of children were more likely to correspond with their husbands’ own reports when thelatter had an ideal of three or four children, and three was the median and the modal response onideal family size for both women and men.

Table 3 shows the level of agreement between women’s own reported views aboutvarious contraception and fertility issues and their husbands’ perceptions of what they thoughtabout those issues. The overall percentage distributions for each of the variables shown in thetable generally indicate medium to high levels of agreement between husbands’ perceptions ofwhat their wives thought and what their wives actually reported. The exception is the strength ofpreferences regarding pregnancy intendedness, which had a lower level of agreement (under 40percent). As in Table 2, the Kappa scores are statistically significant but relatively low, againsuggesting low levels of agreement beyond what would have occurred by chance alone. For the two-part pregnancy intendedness variable, men’s perceptions of their wives’views matched their wives’ actual views in about 61 percent of the cases, but for the four-partintendedness variable, they matched in fewer of the cases (only about 39 percent). Again,husbands were somewhat more likely to provide a proxy report that matched their wives’ ownreports when their wives considered the most recent pregnancy to be intended rather thanunintended. Interestingly, however, male respondents were also more apt to correctly assess theirspouses’ reports of recent unintended pregnancies than were female respondents. This wasparticularly true when their spouses’ views concerning unintended pregnancy were strongly held.

As was the case for women, men’s knowledge of their wives’ attitudes towardscontraception was higher than their knowledge of their wives’ views on pregnancy intendedness. Men’s perceptions of their wives’ attitudes towards contraception agreed with their wives’ ownreported views in more than 80 percent of the cases. Also like their female counterparts,

Page 10: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

4As noted earlier, the dependent variable has three outcomes: (1) the respondentunderestimated the spouse’s views regarding their most recent pregnancy (e.g. the respondentperceived that the spouse felt the most recent pregnancy was unintended, but actually the spousefelt that it was intended), (2) the respondent correctly perceived that the spouse felt that theirmost recent pregnancy was intended or unintended, and (3) the respondent overestimated thespouse’s views regarding the most recent pregnancy (e.g. the respondent perceived that thespouse felt the pregnancy was intended, but actually the spouse felt that it was unintended).

6

husbands were much more likely to have provided proxy reports that matched their wives’ ownreports when their wives approved of using contraception rather than when they disapproved.

Men’s perceptions of their wives’ ideal number of children matched their wives’ reportedideals in less than half of the cases, although the Kappa score remains statistically significant. Again, men’s proxy reports of their wives’ ideal number of children were more likely tocorrespond with their wives’ own reports when they had an ideal of three children (the modalnumber). As was true of women, a much higher percentage of men (88 percent) were able topredict their wives’ ideal number of children within one child of their wives’ own report thanthey were to provide the exact number (data not shown).

A comparison of Tables 2 and 3 indicates that women and men were about equally likelyto assess correctly their spouses’ views about contraception and ideal family size. Overall,women were slightly more likely than their husbands to correctly perceive their spouses’ attitudesabout the intendedness of the most recent pregnancy, but men were slightly more likely tocorrectly perceive an unintended pregnancy than were their wives. It may be that women aresomewhat more outspoken than men when they have negative views regarding pregnancyintendedness, perhaps because unintended pregnancies have more salient drawbacks for womenthan for men, although this cannot be determined from these data.

Table 4 shows cross-tabulations of female and male respondents’ proxy reports of theirspouses’ views regarding the intendedness of the couple’s most recent pregnancy in the five yearsprior to the survey, by selected demographic, socio-economic, and attitudinal variables. The firstset of columns displays the accuracy of women’s proxy reports of their spouses’ views onpregnancy intendedness in comparison to the men’s actual views,4 and the second set of columnsdisplays the accuracy of male respondents’ proxy reports of their wives’ views in comparison towomen’s actual views.

The findings in Table 4 indicate that a greater proportion of younger male and femalerespondents than older ones correctly estimated their spouses’ attitudes regarding theintendedness of the couple’s most recent pregnancy. No doubt related to this finding is thatwomen who had borne a smaller number of children were more apt to correctly perceive theirhusbands’ attitudes about pregnancy intendedness than were those who had borne more children. More educated male respondents were slightly more likely than their less educated counterpartsto correctly assess their spouses’ attitudes about pregnancy intendedness. Similarly, womenwhose spouses had higher levels of education were more likely to correctly perceive theirspouses’ attitudes about pregnancy intendedness. However, men whose wives were more highlyeducated were slightly less likely to correctly perceive their wives’ attitudes about pregnancyintendedness.

Page 11: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

7

Men in couples who reported discussing whether or not to have the next childoccasionally or often were more likely to correctly assess their spouses’ attitudes regardingpregnancy intendedness than were men in couples who never discussed this issue. A similarpattern occurs for female respondents, although the variable is not statistically significant. Interestingly, men in couples who discussed this issue occasionally were more likely to havecorrectly assessed their spouses’ attitudes regarding pregnancy intendedness than were those who discussed it more often. Both men and women who reported having no difficulty communicatingwith their spouse about sex were slightly more likely to have correctly assessed their spouses’attitude about pregnancy intendedness than were those who reported having difficultycommunicating about sex.

Regardless of their own views concerning the wantedness of their most recent pregnancy,both women and men correctly assessed their spouses’ views more than half of the time.However, women who viewed their most recent pregnancy as intended were somewhat morelikely to correctly assess their husbands’ views (73 percent) than women who viewed their mostrecent pregnancy as unintended (54 percent). The opposite is true for men--men who viewedtheir most recent pregnancy as unintended were more likely to correctly assess their wives’ views(72 percent) than were men who viewed their most recent pregnancy as intended (59 percent). Of those women who incorrectly assessed their husbands’ views, women who themselves viewedtheir most recent pregnancy as unintended were more likely to underestimate than tooverestimate their husbands’ views, while women who themselves viewed their most recentpregnancy as intended were somewhat more likely to overestimate rather than underestimatetheir husbands’ views. Similarly, of the men who incorrectly assessed their wives’ views, menwho themselves viewed their most recent pregnancy as unintended were more likely tounderestimate than to overestimate their wives’ views, while men who themselves viewed theirmost recent pregnancy as intended were slightly more likely to overestimate rather thanunderestimate their wives’ views. This finding is consistent with the argument that respondentsmay project their own views onto those of their spouses.

Both men and women who had had fewer than their ideal number of children were morelikely to correctly assess their spouses’ views regarding pregnancy intendedness than were thosewho had had their ideal or more than their ideal number of children. Both wives and husbandswho thought they had had more than their ideal number of children were somewhat more likelyto overestimate their spouses’ attitudes. Interestingly, the same is true of those who had fewerchildren than they considered ideal.

In Tables 5 and 6, multinomial logit models are used to estimate the accuracy of femaleand male respondents in estimating their spouses’ preferences regarding the intendedness of thecouple’s most recent pregnancy in the five years prior to the survey. We explore the ways theopenness of couple communication about reproductive matters and the respondents’ own fertilitypreferences influence the correspondence between the respondents’ own reports and thoseprovided by their spouses, net of other factors. The models in Table 5 assess the correspondencebetween women’s proxy reports of their husbands’ views in comparison to their husbands’ ownreports, and those in Table 6 assess the correspondence between men’s proxy reports of theirwives’ views in comparison to their wives’ own reports. In these tables, the first models showthe demographic and socio-economic controls only (wife’s and husband’s current age andeducational attainment, and the number of children the wife has ever born). The second modelsadd the variables concerning couple communication about fertility matters. The third model in

Page 12: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

8

each table allows us to assess whether the respondents’ own views might bias their proxy report. These models include the respondents’ own reported intendedness views and whether or not theyhad achieved their own ideal number of children.

In Table 5, the first model indicates that women married to men who were more highlyeducated were slightly less likely than those married to less educated men to overestimate theirhusbands’ views regarding the intendedness of their most recent pregnancy, and more likely tocorrectly perceive their views. In addition, lower parity women were somewhat more likely thanhigher parity women to correctly assess their husbands’ views on pregnancy intendedness ratherthan overestimating them. Couples at higher parities may be less adept at communicating aboutpregnancy intendedness or planning their families in general, while those who recently begantheir childbearing, or those who have had just a few children overall, are probably more likely toagree that their most recent pregnancy was intended. These findings are consistent with thecross-tabular findings discussed earlier. Once we control for other factors, no effect of age isobserved.

Model 2 adds the variables regarding couple communication about childbearing and sex,but none of the new variables have significant effects. The parity control variable remainssignificant, indicating that higher parity women are both more likely to overestimate and tounderestimate their partners’ satisfaction with the occurrence/timing of the couple’s most recentpregnancy than are lower parity women, who are more apt to provide a correct assessment, net ofother factors.

Model 3 adds the variables concerning women’s own views on pregnancy intendednessand whether or not they had achieved their own fertility goals. Compared to those who felt thattheir most recent pregnancy was unintended, women who thought their most recent pregnancywas intended were much less likely to underestimate their husbands’ views on pregnancyintendedness than to provide a correct assessment. Again, this may provide some support for theprojection argument. Also, compared to women who had had what they considered to be theirideal number of children, women who had had fewer than their ideal number were less likely tounderestimate their husbands’ views on pregnancy intendedness then to provide a correctassessment. These results also match the cross-tabular findings. As noted previously, thisfinding may support the argument that women who have not yet achieved their ideal number ofchildren are, along with their husbands, still in the family-building stage and that their level ofagreement is high on the issue of pregnancy intendedness.

The first model in Table 6 once more includes only the control variables. It shows thatolder men were slightly more likely than younger men to underestimate their wives’ views onpregnancy intendedness, and thus less likely to provide a correct assessment. Men married towomen with higher education levels were slightly more likely to overestimate their wives’pregnancy intendedness views than to correctly assess them (compared to men married to womenwith less education), while men who were themselves more highly educated were slightly lesslikely than men with less education to overestimate their wives’ pregnancy intendedness views. Again, both findings are consistent with those shown in Table 4.

Model 2 adds the variables on couple communication. As expected, the more frequentlycouples discussed having their next child, the less likely men were to overestimate their wives’views on pregnancy intendedness and the more likely they were to provide a correct assessment. In addition, compared to couples who reported having no difficulty communicating about sex,

Page 13: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

9

among those who reported difficulty, husbands were more likely to overestimate their wives’attitudes on pregnancy intendedness than to provide a matching report.

Model 3 adds the variables concerning the husbands’ own views on pregnancyintendedness and whether or not they have achieved their own fertility goals. Men who thoughtthe most recent pregnancy was intended were more likely than those who did not, to overestimatetheir wives’ views on pregnancy intendedness and less likely to provide a correct assessment ofthose views. Again, this may support the notion that there is a certain amount of projectioninvolved in proxy reporting. Additionally, compared to those who felt they had had the idealnumber of children, men who had had fewer than their ideal number were less likely tounderestimate their wives’ views on pregnancy intendedness and more likely to provide a correctassessment. This finding matches the cross-tabular result and also parallels the finding discussedfor women’s reports in Table 5. We expect that this is driven largely by family-building couples.

Summary and Conclusions

Our analysis of data from the Philippines demonstrates moderate levels of overallagreement between proxy reports provided by spouses and the spouses’ own reported viewsregarding their satisfaction with the timing/occurrence of the couple’s most recent pregnancy,and to a lesser extent, on the strength of their preferences regarding pregnancy intendedness. These levels of agreement are lower than those for spousal attitudes regarding the use ofcontraception. As expected, there were higher overall levels of agreement on the two-partpregnancy intendedness variable than on the four-part variable, which also included strength ofpreference. The Kappa statistics indicate that in general there was relatively low levels ofagreement between the two types of reports beyond what would have occurred by chance alone. In addition, proxy reports generally were much more apt to match spouses’ own reports when theattitude discussed was a positive one. This may reflect a face-saving strategy on the part of eitherthe respondent or the proxy reporter or both. Multinomial logit models regarding the accuracy with which male and femalerespondents’ assessed their spouses’ attitudes towards the intendedness of their most recentpregnancy indicate that men who reported better communication with their wives about sexualmatters and more frequent discussions about having the next child may have been better able toprovide accurate proxy reports for their wives’ views on pregnancy intendedness than those whoreported having difficulty communicating with their partners about sex or never or onlyinfrequently discussing having the next child. It is interesting that this effect was not statisticallysignificant in the analysis of women’s proxy reports. While women were generally slightly moreapt than men to provide accurate reports of their spouses’ views, our measures of communicationwere not significantly correlated with increased reporting consistency among women. We concurwith Lasee and Becker (1997) that multiple indicators are needed to capture the complexities ofthe process of communication between spouses; although we are able to include two indicators,one on frequency of communication and one on level of difficulty that is experienced incommunicating, we would suggest that future data collection efforts include more detail on theprocess of communication about fertility issues.

Page 14: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

10

In general, both men’s and women’s own attitudes towards pregnancy intendedness andachievement of their own fertility goals had a significant influence on their ability to correctlyassess their spouses’ views on pregnancy intendedness, indicating that proxy reports may havebeen influenced to some extent by respondents’ projection of their own fertility desires and goalsonto their spouses. It is also likely that there is some presumption on the part of one spouse orthe other that agreement existed where in fact it did not.

This study has shown that in a large number of cases, spouses could not accurately reporteach others’ attitudes by proxy, either regarding views on the intendedness of past pregnancies oron their ideal family size. In the Philippines, as in the rest of the world, decision-makingregarding family planning and fertility issues clearly involves a complex process of negotiationwithin marital couples; decisions may be influenced by the attitudes and intentions of one orboth spouses and may represent a compromise made by both members of a couple. Therefore,collecting data regarding the fertility attitudes and intentions directly from both male and femalemembers of couples remains vital to improved understanding of attitudes towards contraceptiveuse, ideal family size, and the intendedness of past pregnancies. While such a strategy isadmittedly much more costly in monetary terms than are single-sex studies, omitting one spouseor the other from the research process is extremely costly in other very important ways.

Page 15: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

11

References

Bankole, Akinrinola. “Desired Fertility and Fertility Behavior among the Yoruba of Nigeria: A Study ofCouple Preferences and Subsequent Fertility.” Population Studies 49 (1995): 317-328.

Bankole, Akinrinola, and Susheela Singh. “Couples’ Fertility and Contraceptive Decision-Making inDeveloping Countries: Hearing the Man’s Voice.” International Family Planning Perspectives 24, 1(March 1998): 15-24.

Becker, Stan. “Couples and Reproductive Health: A Review of Couple Studies.” Studies in FamilyPlanning 27, 6 (November/December 1996): 291-306.

Beckman, Linda J. “Communication, Power, and the Influence of Social Networks in Couple Decisionon Fertility.” In Determinants of Fertility in Developing Countries, Vol. 2, ed. Rudolfo A. Bulatao,Ronald L. Lee, with Paula E. Hollerbach and John Bongaarts, 415-443. New York: Academic Press,1983.

__________. “Husbands’ and Wives’ Relative Influence on Fertility Decisions and Outcomes.” Population and Environment: Behavioral and Social Issues 7, 3 (1984): 182-197.

Biddlecom, Ann E., John B. Casterline, and Aurora E. Perez.. “Spouses’ Views of Contraception in thePhilippines.” International Family Planning Perspectives 23, 3 (September 1997): 108-115.

Biddlecom, Ann E., and Bolaji M. Fapohunda. “Covert Contraceptive Use: Prevalence, Motivations, andConsequences.” Studies in Family Planning 29, 4 (December 1998): 359-372.

Coombs, Lolagene C., and Dorothy Fernandez. “Husband-Wife Agreement about Reproductive Goals.” Demography 15, 1 (February 1978): 57-73.

Ezeh, Alex Chika. “The Influence of Spouses Over Each Other’s Contraceptive Attitudes in Ghana.” Studies in Family Planning 24, 3 (May/June 1993): 163-174.

Fapohunda, Eleanor R., and Michael Todaro. “Family Structure and Demand for Children in SouthernNigeria.” Population and Development Review 14, 4 (1988): 571-594.

Lasee, Ashraf, and Stan Becker. “Husband-Wife Communication about Family Planning andContraceptive Use in Kenya.” International Family Planning Perspectives 23, 1 (March 1997): 15-20,33.

Mason, Karen Oppenheim, and Herbert L. Smith. “Husbands’ Versus Wives’ Fertility Goals and Use ofContraception: The Influence of Gender Context in Five Asian Countries.” Demography 37, 3 (August2000): 299-311.

Mason, Karen O., Herbert L. Smith, and S. Philip Morgan. “The Husband’s Role in DeterminingWhether Contraception is Used: The Influence of Gender Context in Five Asian Countries.” In Men,Family Formation, and Reproduction. Liege, Belgium: IUSSP, 1998.

Page 16: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

12

Mason, Karen O., and Anju Taj. “Women’s and Men’s Reproductive Goals,” Population andDevelopment Review 13, 4 (December, 1987): 611-638.

Morgan, S. Philip. “Individual and Couple Intentions for More Children: A Research Note.” Demography 22 (1985): 125-132.

Salway, Sarah. “How Attitudes Toward Family Planning and Discussion between Wives and HusbandsAffect Contraceptive Use in Ghana.” International Family Planning Perspectives 28, 2 (June 1994): 44-47, 74.

Stycos, J. Mayone. Family and Fertility in Puerto Rico. Westport, CT: Greenwood Press, 1955.

__________. Human Fertility in Latin America: Sociological Perspectives. Ithaca, NY: CornellUniversity Press, 1968.

__________. “Men, Couples. And Family Planning: A Retrospective Look.” Cornell UniversityPopulation and Development Program Working Paper Series No. 96.12. (1996). Ithaca, NY: CornellUniversity.

Thomson, Elizabeth, Elaine McDonald, and Larry L. Bumpass. “Fertility Desires and Fertility: His,Hers, and Theirs.” Demography 27, 4 (November 1990): 579-588.

Thomson, E., and S. Sprecher. “Perceptions of Marital Partner’s Desire for Children.” Centerfor Demography and Ecology Working Paper 82-13. Madison: University of Wisconsin.

Unalanl, T. “Problems of Collecting Information from Men in Demographic Surveys: Experience fromthe 1988 Turkish Population and Health Survey.” In Men, Family Formation, and Reproduction. Liege,Belgium: IUSSP, 1998.

Williams, Lindy B. “Variations in Reproductive Goals among Indonesian Spouses.” Working Paper onWomen in International Development Series No. 211, Michigan State University, East Lansing, MI, (July1990).

__________. “Determinants of Couple Agreement in U.S. Fertility Decisions.” Family PlanningPerspectives 26, 4 (July/August 1994): 169-173.

Williams, Lindy, Teresa Sobieszczyk, and Aurora Perez. “Couples’ Views about Planning Fertility in thePhilippines.” Rural Sociology 65, 3 (September 2000): 484-514.

____________. “Correspondence between Survey and In-depth Interview Data on the Subject ofPregnancy Wantedness in the Philippines.” Studies in Family Planning, forthcoming.

Williams, Richard, and Elizabeth Thomson. “Can Spouses be Trusted: A Look at Husband/Wife ProxyReports.” Demography 22, 1 (February 1985): 115-123.

Page 17: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Table 1. Description of Sample

Percentage N

Variable Women Men Women Men

Age 25-29 30-39 40-44

35.154.610.4

22.856.720.5

16625849

108268

97

Age at Current Marriage 10-19 29-24 25-29 30-42

22.445.523.58.7

7.240.036.816.1

10621511141

34189174

76

Monthly Household Income (Pesos)a

0-1,999 2,000-3,999 4,000-7,999 8,000+

18.942.525.912.7

––––

8920012260

––––

Monthly Individual Income (Pesos)a

0 1-1,999 2,000-3,999 4,000-5,999 6,000+

61.513.312.77.614.9

3.025.842.619.39.3

29163603623

14119197

8943

Education (Years) 0-7 8-11 12 or more

24.537.637.8

22.638.139.3

116178179

107180186

Ethnicity Tagala Ilocana Other

49.533.417.1

54.531.614.0

23415881

257149

66

Religion Roman Catholic Other

83.116.9

82.517.6

39380

39083

Place of Residence City Poblacion Rural

13.333.053.7

13.432.654.0

63156254

63154255

Total 100 100 473 473

a In July, 1993, U.S. $1 was equal to 27.7 pesos.

Page 18: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Table 2. Percentage Distribution of Husband’s Stated Views on the Intendedness of the Couple’sMost Recent Pregnancy in the Five Years Preceding Survey, Use of Contraception, and IdealNumber of Children, by Wife’s Perceptions of Husband’s Views

Husband’s OwnStated Views Wife’s Perception of Husbands’ Viewsa Total N

OverallAgree.

KappaIndex

Husband’s preferencesregarding intendednessof most recentpregnancy in 5 yearsprior to survey He intended He did not intend

Intended77.365.8

Unintended22.734.2

100.0100.0

322111

66.3 0.12**

Husband’s strength ofpreferences regardingintendedness of mostrecent pregnancy in 5years prior to survey He intended Strongly Medium/weak He did not intend Strongly Medium/weak

Intended-Strongly

72.127.3

62.423.1

Intended-Med/weak

12.727.3

8.226.9

Unintended-Strongly

13.527.3

22.442.3

Unintended-Med/weak

1.618.2

7.17.7

100.0100.0

100.0100.0

24477

8526

50.5 0.17****

Husband’s attitudetoward contraception He approves He disapproves Does not know

Approves90.473.366.7

Disapproves7.5

17.816.7

Does notknowb

2.28.9

16.7

100.0100.0100.0

4164512

81.6 0.15****

Husband’s idealnumber of children 1-2 3 4 5 or more

1-234.315.6

8.55.2

343.854.822.519.0

419.226.355.831.0

5+2.73.2

13.244.8

100.0100.0100.0100.0

73186129

58

50.5 0.29****

*Significant at p < 0.05 **Significant at p < 0.01 ***Significant at p < 0.001 ****Significant at p < 0.0001aThe variables for women’s proxy reports of their husbands’ views on pregnancy intendedness and strength of preferencesregarding pregnancy intendedness each had 26 missing cases, and the variable for women’s proxy reports of their husbands’ idealnumber of children had 23 missing cases. bThe category ‘does not know’ includes instances in which wives did not know what their husbands thought and those in whichwives perceived that their husbands did not know.

Page 19: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Table 3. Percentage Distribution of Wife’s Stated Views on the Intendedness of the Couple’s MostRecent Pregnancy in the Five Years Preceding Survey, Use of Contraception, and Ideal Number ofChildren, by Husband’s Perceptions of Wife’s Views

Wife’s Own Stated View Husband’s Perceptions of Wife’s Viewsa Total N

OverallAgree.

KappaIndex

Wife’s preferencesregarding intendednessof most recentpregnancy in 5 yearsprior to survey She intended She did not intend

Intended68.448.1

Unintended31.751.9

100.0100.0

237183

61.2 0.20****

Wife’s strength ofpreferences regardingintendedness of mostrecent pregnancy in 5years prior to survey She intended Strongly Medium/weak She did not intend Strongly Medium/weak

Intended-Strongly

53.255.1

28.643.7

Intended-Med/weak

15.412.2

12.411.3

Unintended-Strongly

19.222.5

42.932.4

Unintended-Med/weak

12.210.2

16.212.7

100.0100.0

100.0100.0

18849

10571

38.7 0.10***

Wife’s attitude towardcontraception She approves She disapproves Does not know

Approves87.572.550.0

Disapproves7.1

20.037.5

Does notknow b

5.47.5

12.5

100.0100.0100.0

42540

880.6 0.13***

Wife’s ideal number ofchildren 1-2 3 4 5 or more

1-239.313.4

8.74.7

345.257.033.018.6

411.925.148.734.9

5+3.64.5

9.641.9

100.0100.0100.0100.0

84179115

43

49.6 0.27****

*Significant at p < 0.05 **Significant at p < 0.01 ***Significant at p < 0.001 ****Significant at p < 0.0001aThe variables for men’s proxy reports of their wives’ views on pregnancy intendedness, strength of preferences regardingpregnancy intendedness, and ideal number of children each had 53 missing cases.bThe category ‘does not know’ includes instances in which husbands did not know what their wives thought and those in whichhusbands perceived that their wives did not know.

Page 20: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Table 4. Percentage of Female and Male Respondents who Knew their Spouses’ AttitudesRegarding the Intendedness of the Couples’ Most Recent Pregnancy in the Five Years Precedingthe Survey, by Selected Demographic, Socio-economic, and Attitudinal Variables

Female Male

Variable

Under-estimatedSpouse’sAttitudes

Correctabout

Spouse’sAttitudes

Over-estimatedSpouse’sAttitudes

Under-estimatedSpouse’sAttitudes

Correctabout

Spouse’sAttitudes

Over-estimatedSpouse’sAttitudes

Respondent’s Age Group 22-29 30-39 40-44

9.620.821.7

*71.864.556.5

18.614.721.7

10.018.126.5

+66.061.255.4

24.020.718.1

Spouse’s Age Group 22-29 30-39 40-44

11.018.219.8

72.065.761.5

17.016.118.7

15.820.411.6

63.858.267.4

20.421.320.9

Respondent’s Education (Years) 0-7 8-11 12 or more

18.216.616.3

57.368.770.0

24.614.713.8

24.720.311.5

*57.762.062.4

17.517.726.1

Spouse’s Education (Years) 0-7 8-11 12 or more

16.219.514.7

*59.662.274.1

24.218.311.2

24.819.511.8

**62.961.060.3

12.419.528.0

Number of Children Female inCouple has Ever Born 0-1 2 3 4 5+

6.710.522.922.121.3

**81.778.161.055.854.1

11.711.416.122.124.6

13.115.820.426.014.3

65.659.463.750.764.3

21.324.815.923.421.4

Frequency Couple DiscussesHaving Next Child Never Occasionally Often

19.315.010.9

61.567.870.9

19.317.218.2

18.015.3

5.6

*55.568.461.1

26.616.333.3

Difficulty Couple Has inCommunicating about Sex No Difficulty Difficulty

14.226.6

*67.661.7

18.311.7

18.514.7

+62.554.4

19.030.9

Respondent’s Own AttitudesRegarding Intendedness of MostRecent Pregnancy in 5 YearsPreceding Survey Unintended Intended

29.010.8

***53.872.5

17.216.7

17.618.0

*71.658.8

10.823.2

Respondent’s Views on Number ofChildren Couple Has Had Fewer than ideal Ideal More than ideal

9.523.922.4

***75.861.352.6

14.714.825.0

12.426.320.3

*64.957.955.1

22.815.824.6

Total Number of Respondents 73 287 73 75 257 88+ Significant at p < 0.10 * Significant at p < 0.05 **Significant at p < 0.01 ***Significant at p < 0.001

Page 21: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Table 5. Multinomial Logit Regression Coefficients Estimating Whether Wives CorrectlyPerceived that their Husbands Considered their Most Recent Pregnancy in the Five YearsPreceding the Survey to be Intended or Unintended

Model 1 Model 2 Model 3

WifeUnderest.Husband

WifeOverest.

Husband

WifeUnderest.Husband

WifeOverest.

Husband

WifeUnderest.Husband

WifeOverest.

Husband

Age of woman 0.038 0.008 0.019 -0.004 0.021 0.0001

Age of man -0.0004 -0.010 0.023 0.004 0.012 -0.006

Education of woman 0.003 -0.015 0.051 0.012 0.008 -0.021

Education of man -0.016 -0.082+ -0.015 -0.075 -0.042 -0.087+

Parity of woman 0.177 0.245* 0.262* 0.285* -0.022 0.173

Wife’s report of frequency couplediscussed having the next child Never (reference category) Occasionally Often

– – –

– – –

– 0.487 0.173

– 0.050

-0.157

– – –

– – –

Wife reports that couple hasdifficulty talking about sex – – 0.518 -0.324 – –

Wife thought most recent pregnancyin 5 years preceding survey wasintended

– – – – -1.124*** -0.263

Wife’s achievement of her fertilitygoals (her own report) Fewer children than desired Has desired # of children (reference category) More children than desired

– –

– –

– –

– –

-0.809*

-0.160

0.143 –

0.379

Constant -2.988** -1.050 -4.426*** -1.379 -1.049 -0.569

Number of casesX2 (df)Pseudo R2

40426.12 (10)**

0.037

34533.10 (16)**

0.055

40349.49 (16)***

0.070+ Significant at p < 0.10 * Significant at p < 0.05 **Significant at p < 0.01 ***Significant at p < 0.001

Note: The sample is married women who had at least one pregnancy within the five years prior to the survey. The referencecategory for the dependent variable is that the wife correctly estimated her husband’s preferences regarding the intendedness ofthe couple’s most recent pregnancy in the five years prior to the survey.

Page 22: Teresa Sobieszcsyk and Lindy Williams · PDF fileDepartment of Rural Sociology Warren Hall Cornell University ... Philippines, collected collaboratively by the Population Institute

Table 6. Multinomial Logit Regression Coefficients Estimating Whether Husbands CorrectlyPerceived that their Wives Considered their Most Recent Pregnancy in the Five Years Precedingthe Survey to be Intended or Unintended

Model 1 Model 2 Model 3

HusbandUnderest.

Wife

HusbandOverest.

Wife

HusbandUnderest.

Wife

HusbandOverest.

Wife

HusbandUnderest.

Wife

HusbandOverest.

Wife

Age of woman -0.048 0.027 -0.0002 0.037 -0.053 0.028

Age of man 0.056* -0.046 0.029 -0.045 0.057+ -0.060+

Education of woman -0.007 0.094+ -0.018 0.080 0.002 0.090+

Education of man -0.065 -0.0004+ -0.069 -0.004 -0.072 -0.005

Parity of woman -0.026 0.084 -0.108 -0.004 -0.130 0.099

Wife’s report of frequency couplediscussed having the next child Never (reference category) Occasionally Often

– – –

– – –

– 1.154 0.784

– -0.203 -0.833+

– – –

– – –

Wife reports that couple hasdifficulty talking about sex – – 0.140 0.939** – –

Husband thought most recentpregnancy in 5 years precedingsurvey was intended

– – – – 0.443 1.018*

Husband’s achievement of hisfertility goals (his own report) Fewer children than desired Has desired # of children (reference category) More children than desired

– –

– –

– –

– –

-1.00**

-0.196

0.090 –

0.607

Constant -0.798 -1.673 -2.042 -1.243 -0.259 -2.238+

Number of casesX2 (df)Pseudo R2

39122.07 (10)*

0.030

32829.47 (16)*

0.049

38640.88 (16)***

0.057+ Significant at p < 0.10 * Significant at p < 0.05 **Significant at p < 0.01 ***Significant at p < 0.001

Note: The sample is married men whose wives had at least one pregnancy within the five years prior to the survey. The referencecategory for the dependent variable is that the husband correctly estimated his wife’s preferences regarding the intendedness ofthe couple’s most recent pregnancy in the five years prior to the survey.