The title of Véronique Hertrich’s (1998) recent article in Population “Are the answers of men and women to be equally trusted?” sums up the essential conclusions of past research on the gender-specific reliability of responses in demographic surveys: answers provided by women tend to be more comprehensive, detailed and relevant than men’s, especially when they concern the common family life experienced by the couple (Poulain, Riandey, and Firdion, 1992). American and English authors refer more directly to “incomplete reporting of male fertility” as opposed to the supposedly complete female reporting (Rendall, Clarke, Peters, Ranjit, and Verropoulou, 1997). They conclude that male self-reported fertility histories are reliable (i.e., as reliable as female ones) for births from the current marriage, but marred by omission for extra-marital births (a third in Britain and half in the United States) and for births from dissolved marriages (over a third in both countries). In Canada too, the consistency of reporting among surveyed married partners contrasts with 40% under-reporting after separation. The omissions mostly relate to children with whom the fathers have little contact and whom they left recently: “The picture of children not living with both biological parents drawn from retrospective surveys is likely to differ significantly by sex of the responding parent” (Juby and Le Bourdais, 1997).

These generalisations, however, do not necessarily hold true in all cases. Véronique Hertrich, for example, shows that “from the point of view of the marriage, male reporting seems better [than female reporting]”. She credits the fact that “most African populations, like the Bwa, are organised ‘around the men’”. Similarly, Poulain, Riandey and Firdion conclude with regard to migration that “spouse-specific feedback differentials are not ultimately significant (…). By confronting the spouses, it is possible to assign a ‘reasonable’ date back to events that had been significantly antedated in individual memory, but to a very similar extent by husband and wife”.

Discrepancies in children’s home-leaving dates reported by the spouses are even so extreme as to “militate against an interpretation in terms of recall error. Mothers perceive their children’s independence as a later and more formal event”. Staying within the retrospective perspective that we have adopted so far, Nadia Auriat (1996) has worked these conclusions up into a general “theory of imprinting … based on the social character of memory”, arguing that “recall error assumes non-random forms that vary with the nature and specific characteristics of the events to be remembered”.

We take a similar approach in the present study that compares the replies of divorced men and women to identically phrased questions in two Russian surveys. The interviews were in the main retrospective, since a majority of the collected data referred to previous periods of the respondents’ lives: the time of their now-dissolved marriage, the periods of separation and divorce, and finally the post-divorce period up to the survey date. The surveys generally returned factual information: dates (of birth, marriage, divorce, etc.), number of children, educational level, occupations, etc. There were few questions on opinions (e.g.: “were you and your spouse happy with the arrangements made for the children at the time of the divorce?” or, “how much influence did the father have on his children?”), but some factual questions were less specific than those exemplified above (e.g.: “who did the housework while you were married?” or: “who took the initiative for the separation?”). The questions range broadly from long-term periods of recall (e.g.: “at the time of your marriage, did your spouse already have children?”) to short-term ones (in the last twelve months….), and from precise facts to purely subjective evaluations.

In the latter case, there is no objective reality against which the male and female responses can be checked for reliability, but even on many factual questions with an important qualitative component, there is no benchmark that will show who is closest to the truth. In all these cases, it is not possible to do more than measure the consistency between male and female responses. The same will also apply to most other factual questions for lack of an independent statistical measure of the “right” response. In only one example of external check (concerning the number of children of the dissolved marriage) will it be possible to ascertain the reliability of male and female responses.

Since there is no way of conveniently pigeonholing gender-specific responses as “right” or “wrong”, we will try to determine whether the dividing line between consistent and inconsistent responses maps out homogeneous and easily identifiable data sets and in particular whether the clustering of inconsistencies suggests distinct male and female perceptions of divorce.

The male and female questionnaires had the same structure, sequencing the family life course by reviewing the marriage before its dissolution, the divorce proceedings, the period since the divorce and the current situation. There were some departures from this general scheme, however, such as linking questions on working life up to the survey date with the earlier questions on the occupation at the time of marriage.

The respondents of both sexes were asked about numerous experiences that they had shared or were sharing now with their former partner: e.g. their date of marriage or divorce, the number of their children, their relations as a couple, and the story of their divorce. In those instances, the symmetry of the questions put to men and women is the result of the shared nature of the investigated topic. The situation is different when information is sought on one spouse only: e.g., his or her birth date, occupation or opinion on a particular subject. Here, the symmetry was created by introducing parallel questions about the spouse. Thus, men were interviewed about their ex-wife’s date of birth, her occupation or her opinion on the particular subject.

This enables a comparison between the answers of men and women on common subjects, and between self-reported items and the ex-spouse’s answers concerning them. In neither case, however, can the comparisons be made within the couples, because the samples are not made up of former husbands and their ex-wives but of divorced men and women who are statistically matched by an identical sampling procedure in the same sampling frame. The consistency of male and female responses will thus be assessed by comparing their distributions in the male and female samples. Identical frequencies of the same item will reflect the consistency between men and women, while the discrepancies will constitute evidence of inconsistency.

We will use statistical tests to determine whether there is a “significant difference” between the distributions of responses by sex. In general we use the t test, applied to frequencies [2] or means [3]; for complex sets of possible responses [4] the chi-square test will be used. Where t is greater than 1.96 or the chi-square distribution is higher than its critical value, we will accept the hypothesis that there is a significant difference between the answers of men and women, within a 5% margin of error.

Selected questions from the first cluster are given in Table 1A. They fall into three types:

Table 1A
Large differences between “dual” information relating to the ex-spouses

The information set out in Table 1B reveals other discrepancies between the answers given by men and women, this time concerning the families of ex-spouses rather than concerning the latter. The two instances relate to assistance provided to the couple during married life. A first question about help itself is followed by a question on the helper. There is disagreement on the latter topic, although not necessarily on the former.

Table 1B
Equally large differences between “dual” information relating to parents and in-laws

Thus, men and women agree on the proportion who needed help with childcare (around 69%), but women report needing financial help more often than men (80.7% against 66.5%). In either case, on the other hand, both men and women claimed that more help was provided by their own family than by that of the ex-spouse. Men, however, claimed less predominance for their family’s help than women for theirs. Financial help is a case in point: while 43.6% of the men thought that the assistance of their own family was larger, and 29.4% that their wife’s family contributed more, the corresponding proportions for women were 64.8% and 18.2%. The same order of difference applies to childcare.

These discrepancies characterise men and women who share the experience of having been at odds with an ex-spouse. Most bear on issues that are the focus of marital discord: contributions to the household budget, the respective roles of family and in-laws, infidelity and the desire to start a new life. It cannot be inferred that these issues and the divergent points of view were contributing factors to the divorce. That would require proof that couples that stayed together give more consistent responses to the same issues. The impossibility of such an analysis prompted us instead to compare this set of questions to those that receive consistent answers between men and women, to ascertain their specificity.

Table 2A
Few differences on factual information concerning the past

The exceptions to overall consistency are shown in Table 2B. They relate to the wife’s pre-marriage condition (marital status and number of children) and the childlessness of the dissolved marriage. The three variables could well be connected.

Table 2B
Minor differences in factual information on the past, mainly child-related

A small proportion — approximately 5% — of the interviewed women reported an earlier marriage or children born before the union that ended in the observed divorce. Males reported a much higher percentage of wives with previous marriages (15.4%) or children (11.4%). There is a clear relationship between the two discrepancies, in that women who had been married before the dissolved union were also more likely to have had children.

Conversely, more women than men reported the birth of children during the dissolved marriage. This warrants closer scrutiny on several counts:

At the time of the male survey, some information was collected in the divorce decrees that constituted the sampling frame. This included the age of the spouses at divorce, and the number of under-age children. The male and female responses for the first of these two variables are consistent, and thus less interesting than those for the second variable.

Among the cases selected for inclusion in the sample, 32.8% of couples were childless at the time of divorce (Table 3). A significant percentage of these decrees could not be used for reasons explained below (the individual was absent, or had migrated from the city where the divorce took place); the proportion of childless couples was not significantly different among them than in the whole sample (34.1%). By contrast, a significantly higher proportion of males who refused to answer — nearly a quarter of usable addresses — were childless (42.4%); only 27.7% of those who responded had no under-age children at the time of divorce. The non-response bias, therefore, leads to an underestimation of childlessness by approximately 5 percentage points.

Table 3
Proportion of couples with no under-age children at the time of divorce

On the one hand, male responses given during the interviews overestimate by about three percentage points the results compiled from the divorce decrees for the same persons (30.5% against 27.7%); that the gap is so small shows the reliability of the divorced males’ responses. On the other hand, a comparison of male survey responses and the result expected on the basis of the total sample produces a gap of approximately two percentage points in the opposite direction (30.5% against 32.8%); because it is also small, that gap demonstrates the effectiveness of the survey procedure. The proportion estimated from the survey falls right in between those compiled from the divorce decrees, before and after dropouts (32.8% and 27.7% respectively).

It is generally accepted, unfortunately without convincing evidence, that the proportion of unusable records in the sample and the refusal rate were lower in the female than in the male survey. Hence, the range of values derived from the female sampling frame should not be larger than for the males, and the extreme values in the range should be close — say, 28-32% as an order of magnitude.

The result from the female survey — 19% — falls well outside of the range. The more than eleven-plus percentage points discrepancy between the proportions of childless couples calculated from male and female reports would seem to stem from marked underestimation by women compared to a more adequate assessment of the situation by men [6].

There may, therefore, be an underlying logic to the inconsistencies in Table 2B: women would be reporting that their older children are born from the dissolved marriage, while they were in fact born from a previous marriage. In doing so, they would underestimate the fertility of these earlier unions, and perhaps even deny that they ever existed, and overestimate the fertility of the divorced couple. Male reports are presumably more balanced, although neither sex gives an incorrect picture of women’s completed fertility. This is a radically different conclusion than that derived from the British and North American surveys.

First, the two operations were not simultaneous: the female survey was conducted in 1993, the male survey in 1998. Second, the number of divorce cohorts was not the same: twenty in the female survey, and only twelve in the male survey. Finally, the fieldwork was limited to the six cities where the samples of divorced persons were taken, excluding from the surveys those who had migrated since the decree.

The five-year time lag distorts the comparisons between male and female reporting of time-dependent variables. The unequal cohort numbers implies that the divorces in the female survey had occurred a longer time before the survey; this affects the comparisons of time-dependent magnitudes. The effects of limiting the study to non-migrants are more complex. Because men’s spatial mobility is higher than women’s, more men are “missing” from the 1998 survey than women from the 1993 survey, but the surveyed females (in the town of their divorce) live more often far from their ex-husband than the surveyed males do from their ex-wives. The comparisons between male and female reports are therefore biased when they involve migration-dependent variables or variables related to the geographic distance between ex-spouses.

Table 4A provides examples of conflicting information obtained from men and women that cease to be so when the comparison is restricted to years of divorce common to the two samples (Table 4B).

Table 4A
Examples of inconsistencies in total samples

Table 4B
Consistency over the period common to the samples (divorces in 1986-1993)

The last item in the table is most characteristic of a time trend that masks the consistency of responses between men and women. Over time, an ever-growing proportion of agreements are concluded between divorced spouses over child support payments, pre-empting court orders to the same effect. Because the divorces in the male survey were more recent than those in the female survey, the share of these agreements between ex-spouses appears higher (and court orders less frequent) than in the female reports. The agreement is very close, however, on the period of divorce common to both surveys. The same is true for the incidence of premarital cohabitation and for women’s dissatisfaction with their ex-husband’s career development — two rising trends over time.

The range of topics on which the spouses agree is thus extended to include the various questions about the spouses’ occupations (at different stages of their careers), but also a wide range of information on informal situations (consensual union, out-of-court agreements) as well as personal opinions.

Table 5A features two variables known to be strongly affected by the remoteness of the divorce in time: the frequency of contacts between ex-spouses, and of the visits of the father to his children. Analysis of the survey results confirms that, in Russia as elsewhere, the frequency of these contacts is inversely related to the time elapsed since the divorce decree. Accordingly, the table had to restrict the female survey to the same number of years after divorce as the male survey. Even with this precaution, however, the male and female responses can only be validly compared if the effect of geographic distance between ex-spouses is controlled, since the frequency of contact or visits is significantly higher when the ex-spouses still live in the same city than when one has migrated.

Table 5A
Effect of distance on contacts (samples limited to the last twelve years of divorce)

When only those cases are considered where both ex-spouses still live in the city of the divorce, the two samples give very similar results. The frequency of contact between ex-spouses also matches closely when the ex-wife stayed and the ex-husband left, or when the ex-husband stayed and the ex-wife left. That is not true of visits to the children: they are less frequent when the father is away than when the mother is.

Examples of more directly migration-related variables associated with the last marital home are shown in Table 5B.

Table 5B
Selective effect of migration (samples limited to the last twelve years of divorce). Percentage distribution of responses

Because the proportion of migrants rises with the time elapsed since divorce, the analysis was restricted, as before, to the decrees of the last twelve years in both samples. The probability of the departing spouse having been the ex-husband (wife) is almost automatically markedly higher where he (she) has also migrated from the town. But the reports of both spouses concur on who took the initiative to leave when both still live in the locality of divorce. They also agree on the type of housing under the same conditions.

These controls, however, are not sufficient to erase the discrepancies when the attention focuses on variables that weigh the respective roles of the two spouses and their families (“Who owned the marital home occupied on the eve of the separation?”) or that evaluate the merits of either ex-spouse in subjective terms (“What influence did the father have on his children?”). Table 6A brings us back to the variables in Table 1A, but this time at a later stage of the family life cycle.

Table 6A
Persistent differences on “dual” information (samples limited to the last twelve years of divorce). Percentage distribution of responses

Women assign their parents and themselves a significantly greater role than men do. Men are less assertive and give more weight to the sharing of responsibilities. Women attribute substantially less influence to the father than do men.

Added to these discrepancies are those resulting from the respondents’ failure to respond when they are asked to evaluate the subjective opinion of their partner after the divorce on an issue that the ex-spouses may no longer have the opportunity to discuss (Table 6B).

Table 6B
Answers with large proportions of “don’t know” (samples limited to the last twelve years of divorce). Percentage distribution of responses

The share of non-responses increases with the geographic distance between the ex-spouses, but they remain important when they are still living in the same city. Over a third of responses remain non-committal, when either women or men are asked about their ex-spouse’s satisfaction with the outcome of the divorce decree. Assuming the distribution of male non-responses identical to that of male responses, the result would come very close to the degree of satisfaction expressed by the women. But the same procedure applied to female non-responses would lead women to greatly overestimate male satisfaction. The symmetry of appreciation between the sexes is thus more apparent than real.

Table 7
Consistency (+) and inconsistency (–) of male and female responses, by sample type

In the first group, union and family formation can be described through a series of variables. These are objective characteristics summarised by dates, numbers, or simple classifications (marital status). Male and female reporting is generally in agreement, with one exception: when there is more than one marriage, women erroneously include some events from the first marriage — especially children’s births — in the second.

Another series of variables address the conditions of married life. There is a high degree of consistency in the reporting of basic characteristics such as educational level and spouses’ occupations at the beginning and end of the union. This makes it even possible to use either the male or the female surveys to study the occupational mobility of the spouses before their divorce. On a purely subjective level, however, women pass judgement on their ex-husband’s careers, and the men seem to be aware of this. Mid-way between fact and value judgement, men and women agree on their housing conditions (especially the frequent co-residence with parents) and the help received to raise their children. They differ only on the amount of financial help received, which women rate higher than men.

By contrast, the responses on living conditions diverge systematically as soon as the questions address the respective contributions of the two spouses or their families. Each sex rates its own or its parents’ contribution as predominant or, at the very least, as more significant than the respondents of the opposite sex do.

In the second group of questions — concerning the time from separation decision to survey date — the spouses exhibit a common pattern of inconsistencies in identifying the initiator of the decision to divorce or in assessing intentions to remarry after the divorce. Both overestimate their own role and underestimate their own intentions compared to the other’s assessment.

By contrast, both sexes describe the material conditions of the divorce — who left the marital home, who won custody of the children, what are the child support arrangements? — in similar terms. All these questions are answered consistently in the male and female surveys. The same is true in reference to the period after the divorce, for the frequency of contacts between ex-spouses and the father’s visits to his children, where the ex-spouses still live close by.

Several years after they separated, it would be unrealistic to try to match the responses of ex-spouses, since they may not know what has become of their former partner. Answers that involve subjective impressions, however, for example on the father’s influence over his children or on the satisfaction of the ex-spouses about the decisions taken at the time of the divorce are, or appear, contradictory.

Overall, consistent items are much more numerous than inconsistent ones. Men and women give very similar descriptions of their marriage and their divorce. There is no gender-specific view of the family.

Yet although the discrepancies are few in number, they are significant because most of them have a common theme: the place occupied by each spouse (and their parents) within the couple and in family life. Men and women attribute to themselves the dominant role. The contrast between husbands and wives is more marked, however, in the women’s than in the men’s responses. For example, whereas only 23% of women report receiving financial help from their in-laws against 80% from their own parents (a difference of 57 percentage points), 44% of men acknowledge help from their in-laws against 65% from their own family (a difference of 21 percentage points); and whereas 14% of women report having intended to form a new relationship and ascribe the same intention to 45% of their ex-husbands (+ 31 points), 27% of men report the same for themselves and 33% for their ex-wives (+ 6 points).

The main focus of disagreement is on the marital relationship, or even more, on the marital contributions. The imbalance of these contributions is also far more pronounced in the female than in the male responses. Women describe themselves as more involved than men in the marriage. The survey does not tell us whether this disparity is a factor in the divorce, or a fact of life that is also true in stable marriages [7]. But this disagreement does not prevent men and women separated by divorce from displaying a remarkable knowledge of one another and great consistency in the description of the life they shared.