Marriage: Distance between birthplaces and age difference between mates
ANDRE BOUCKAERT AND BRUNO BOULANGER Abstract The methodology of potential mate analysis can be used to obtain information about the determinants of marriage. This study investigates the influence of the distance between birthplaces and of age difference between mates. We use a new method of potential mate analysis that involves a sample of potential mates simulated from a sequence of actual mates. The differences between actual and potential mates are investigated using a logistic model and by direct comparison of distributions. The logistic model gives a poor fit. The comparison of distributions reveals no significant difference in distance between birthplaces for actual and potential mates or, at any rate, no difference large enough to be demonstrated by our small sample. A marked difference is observed for marriages in which the male is slightly older. The distance between birthplaces in our sample is shorter than in Great Britain.
Several studies have investigated the determinants of choice of mates. These determinants are obviously complex because they include family relationships, social habits, and ways of avoiding consanguineous matings. Even more important, marriage presupposes accessibility, which is conditioned by social, psychological, and geographic structures. The present study attempts to investigate two particular aspects of accessibility: distance between birthplaces and age difference between mates.
Distance limits accessibility because people born in places remote from one another are unlikely to meet or even to be aware of each other’s existence. Moreover, a long distance between birthplaces often goes hand in hand with cultural differences. Cultural differences can have a positive or negative impact on the probability of marriage.
Are there any other factors involved in the relationship between distance and marriage? We can easily understand why people are more likely to meet someone born relatively close to their own birthplace. The distribution of distances between acquaintances’ birthplaces is therefore positively skewed, but preferences as regards distance might also affect one’s choice of an acquaintance as a mate. The distribution of distances between birthplaces of married couples conditional on acquaintance could therefore assume a different shape.
The influence of distance between birthplaces on the probability of marriage may be more subtle than mere availability. For example, if there is traditional avoidance of marriage between people born in the same city, block, street, or village, the probability of marriage, conditional on acquaintance, will increase with distance. However, at the same time the marginal probability of marriage might increase with distance just because there are many more acquaintances in a large circle than in a small circle drawn around someone’s birthplace.
In this study we attempt to estimate the probability of marriage for given distances between birthplaces, taking into account the probability that there is no marriage between people born in the same place. Our study will thus shed light on effects of distance beyond those of accessibility. A new methodology is adopted and tested not only for the study of distances between birthplaces but also for the study of age difference between mates.
As a consequence of demographic structure, age distribution may be heterogeneous, and age availability is not necessarily the same for men and women. Our method takes such heterogeneities into account when estimating the effect of age difference. For example, we will be able to answer the following two questions: (1) Many studies have shown that the modal age difference between mates is that husbands are two years older than their wives; does this difference reflect a preference, or does it just reflect the fact that most older males are unavailable because they are already married? (2) Many studies show that most distances between birthplaces of mates lie between 5 and 10 km; does this mean that there is a clear preference for such distances or that such distances are more common for all acquaintances, whether married or not?
For example, Coleman (1979) investigated the distance between birthplaces in a sample of 660 first marriages in southeast England. The distribution of observed distances was strongly skewed: The mean was 103 km but the median was only 40 km, and 25% of the couples were born less than 10 km apart. In Cambridge, England, Mascie-Taylor (1986) found a mean of 118 km (and a median of 25 km) for 60 couples with a white collar husband and a mean of 77 km (and a mean of 25 km) for 109 couples with a blue collar husband. A stronger dependence on occupation was found in rural Oxfordshire by Jeffries et al. (1976), who studied 613 couples married only once. At extremes of the social scale the mean was 141 km for professional occupations versus SS km for agricultural workers. In all these studies positively skewed distributions of distances were found. In Cambridge and Oxfordshire different distributions were observed for different occupations, but nothing tells us whether, among available potential mates, the agricultural worker of Oxfordshire marries a person born 55 km from his own birthplace just because such persons are more numerous in his “pool” of potential mates or because he likes (or dislikes) exotic persons and that 55 km is very far (or very near) in the area covered by his pool.
This brings us to the issue of the potential mate analysis methodology reviewed by Leslie (1985). Leslie stated that “potential mate analysis (PMA) is based on the idea that mates are chosen from among a ‘pool’ of possible choices, and that it is possible to characterize this pool in terms of its size and the attributes of its members…. The characteristics of actual pairs of mates are compared with those of random pairs drawn from the pool.” For example, in a study of marriage in the Orkney Islands Brennan and Dyke (1981) defined a male and a female in their population as potential mates if (1) they were both within reproductive age limits, (2) the difference of age between the female and the male was included in an interval between -15 and + 5 years, (3) both were unmarried at the time of the study, and (4) they were not members of the same nuclear family.
Using these inclusion-exclusion criteria, Brennan and Dyke (1981) were able to compare married males with their actual and potential wives and the unmated with their potential mates. This enabled them to observe that the distance between birthplaces of actual mates was shorter than that of potential mates, indicating that closeness of birthplaces had a positive effect going beyond mere availability.
Sample. The following information on 198 marriages recorded in Wavre, Belgium, in 1950 was copied from the official registries: day and month of marriage and day, month, year, and place of birth of mates. Wavre is a small town located near the geographic center of Belgium and has rail and road links with the whole country. It lies on the boundary between Belgium’s two main linguistic areas (French and Dutch), not far from the capital of Brussels. Belgium, particularly the central area, is and was already in 1950 characterized by a high population density. Previous studies of the population structure have shown that the consanguinity level is low but that distinct and sometimes small intermarriage areas can be identified, related in some cases to actual or past linguistic differences (Bouckaert 1982). Of the 198 marriages, 6 involved a person born in a foreign country. These marriages were excluded from the sample.
For each marriage the age difference a^sub ij^ and the straight-line distance between birthplaces d^sub ij^ were computed. The distances were estimated using the decametric coordinates of parishes recorded by the National Geographic Institute. Parishes are the finest administrative subdivisions of Belgium and were in use before the administrative reforms of the French Revolution. Simulated Populations. In addition to our sample of marriages, we simulated a sample of potential marriages as follows: 36,864 potential couples were simulated by pairing every male in our sample of 192 marriages with every female in our sample of 192 marriages. The age difference a^sub ij^ and the distance between birthplaces doj were computed for each pair. This was called the permutation sample. Another simulated sample was obtained by resampling with replacement males and females from the observed sample. For the purpose of symmetry 36,864 pairs were sampled. This second sample was called the resampled sample.
Most studies of the determinants of marriage have estimated only the conditional probabilities P(phi ^sub ij^/m). It was sometimes assumed that finding a nonuniform distribution of this conditional probability could be taken as a demonstration of a preference for certain values of bj. Because this would be true only if the marginal distribution was flat, uncontrolled studies actually fail to demonstrate that distance between birthplaces influences the probability of marriage.
If such controls are pairs of individuals of different sex drawn at random from the whole human population, they will include people from different parts of the world, people who are already married, and people of different ages. For such people marriage is impossible and they cannot be regarded as potential mates. To define potential mates one can use a priori rules, as was done for the population of the Orkney Islands (Brennan and Pyke 1981). However, there is a risk that the a priori rules do not exactly reflect the opportunities of the marriage marketplace. Furthermore, if clearcut boundaries are unavailable, as happens with nonisland populations, one is left with the task of finding a distribution of distances for potential mates. In this study the approach was to use the sample itself to generate distributions not only for geographic constraints but also for all other constraints (language, age, celibacy). All these constraints were presumed to affect all couples marrying in the same town hall during the same year.
There are, however, some drawbacks to not introducing explicit constraints. For example, a brother and a sister marrying in the same town hall during the same time interval would be considered potential mates by our method unless specific limitations were introduced.
Our findings about age differences are not surprising. Our negative findings about distance between birthplaces do not imply that the likelihood of marriage is not narrowly limited by geographic constraints. What our study does show is that, once these constraints are taken into account, there is no further influence of distance on choice of partner.
We were not able to demonstrate a significant drop in P(m/do) for small distances. Such a drop could reflect unconscious psychological mechanisms of consanguinity avoidance. But to investigate this, we need more data and more precise data, including distances within towns and villages, streets, and even buildings. This was completely out of the scope of the present study.
The permutational and resampling approaches used here led to similar results. This was expected because the proportion of samples in which a given pair is selected during the resampling process should converge asymptotically to the permutational probability. The logistic model does not seem to give a good fit to the data: The odds ratio decreases too fast for distances and too slowly for age differences as a result of the assumption of a linear effect of these variables.
Finally, our data suggest that the average distance between birthplaces of married persons is smaller in Belgium than in England. This could reflect a difference in population density. Cavalli-Sforza (1971) showed that the average distance between birthplaces of mates decreases as population density increases. Assuming that 0 is the probability for a given person A to meet the criteria for marriage to another person B, the mean number of people A has to meet before marrying will be 1/0. This amounts to a circle of radius (1/pi eta theta)^sup 1/2^, where eta is the population density. The average displacement of A is thus proportional to the inverse of the square root of the product of eta and theta. Because the population density of central Belgium is higher than that of southeastern England, this could explain the difference, although a difference in theta (i.e., the stringency of psychological criteria) cannot be ruled out.
1 Institut de Statistique, Universite Catholique de Louvain, 72 Avenue Mounier Louvain-la-Neuve, Belgium.
2 Eli Lilly, Louvain-la-Neuve, Belgium.
Bouckaert, A. 1982. La sante des Belges. Louvain-la-Neuve, Belgium: Cabay. Brennan, E.R., and B. Dyke. 1981. Assortative mate choice and mating opportunity on Sanday, Orkney Islands. Soc. Biol. 27:199-210.
Cavalli-Sforza, L.L. 1971. The Genetics of Human Populations. San Francisco, CA: Freeman.
Coleman, D.A. 1979. A study of the spatial aspects of partner choice from a human biological viewpoint. Man 14:414-435.
Jeffries, D.J., G.A. Harrison, R.W. Hiorns et al. 1976. A note on marital distances and movement and age at marriage in a group of Oxfordshire villages. J. Biosoc. Sci. 8:155-160. Leslie, P.W. 1985. Potential mate analysis and the study of human population structure. Yrbk. Phys. Anthropol. 28:53-78.
Mascie-Taylor, C.G.N. 1986. Marital distances, age at marriage, and husband’s social group in a contemporary Cambridge sample. Ann. Hum. Biol. 13:411-415.
Copyright Wayne State University Press Apr 1997
Provided by ProQuest Information and Learning Company. All rights Reserved