The Rise of Age Homogamy in 19th Century Western Europe

Age Homogamy as an Indicator of Partner Selection Criteria

There is not a single source of direct information that allows us to study the rise of less instrumental and more egalitarian partner selection strategies. To study the criteria applied in selecting a partner, use has been made of court proceedings diaries, letters, paintings, and advice literature, but these sources mostly cover earlier periods and only refer to very specific time periods, social classes, and regions. We propose the use of information on the age differences between spouses as an indirect measure of the instrumentality or equality of the marital relationship.

A first reason is that there is a widespread consensus among sociologists, anthropologists, and historians that the magnitude of the age difference between spouses is an indicator of the egalitarian nature of the relationship between men and women (Atkinson & Glass, 1985; Cain,1993; Mitterauer & Sieder, 1983; Wheeler & Gunter, 1987). Men in traditional societies could maintain their control over women by ensuring that husbands were generally considerably older than wives and thus could add the advantage of superiority of age to that of superiority of sex. The age difference gave the husband a considerable advantage in status, experience, and power. In most societies in which grooms are much older than brides, a patrilineal kinship structure and patrilocal residence pattern is dominant, whereas a small difference is observed in societies in which bilateral kinship and greater flexibility in the residence pattern of newly married couples is the rule.

A second reason is that age homogamy has also been viewed as a prerequisite for the emergence of romantic love. Romantic love, which can be defined as “the capacity for spontaneity and empathy in an erotic relationship” (Shorter, 1975, p. 15), implies equality between husband and wife (Giddens & Pierson, 1998). Spontaneity and empathy are not present in a power relation, “otherwise the emotional encounter would be impossible” (Shorter, p. 16; Sieder, 1978). At the level of daily routines, romantic love means conversation, and, to feed this conversation, people must have experiences in common. The experience of belonging to the same age cohort contributes powerfully to this commonality: Similarity of values and opinions about marriage and the family, tastes in leisure time activities, life experiences, and so forth lead to mutual confirmation of each other's behavior and worldviews and enlarge opportunities to participate in joint activities.

In line with this argument, historians have considered high proportions of marriages in which men were much older than their wives as indicative of a low standard of marital sexuality (Knodel, 1988; Mitterauer & Sieder, 1983). The equality and intimacy between husbands necessary, for example, for facilitating discussions of sexual and reproductive matters, can more easily be reached among age-homogamous couples (Safilios-Rothschild, 1972). In contrast to this, in case marriage is primarily viewed as instrumental, as an economic contract, the economic characteristics of the potential spouses are the main criterion in the partner selection process. Romantic love is in that case only of secondary importance, and age differences are considered less important.

For these reasons, we assume that age homogamy will be the result of a partner selection pattern that is based on less instrumental and egalitarian principles. But these kinds of relationships may, of course, also be stimulated among age-homogamous couples.

A methodological reason to study age homogamy is its important advantages relating to the accessibility, coverage, and quality of information. Marriage registration was already institutionalized in the beginning of the 19th century, and this registration covered the whole population—not only a privileged group. Hence, comparisons of the level of age homogamy over time and region and for different social groups are possible using large-scale demographic databases.

Why Would Marital Relationships Become Less Instrumental?

A fundamental factor underlying the EMP was the instrumental, economic attitude toward marriage. For many people, the self-sustainability of the new household was a major concern informing the decision to marry. Of equal importance was the requirement that the marriage should not undermine the economic security of the domestic cells in which the couple formerly lived. This was in particular the case when one's family owned property because ill-considered marital decisions could have negative consequences. Thus, marriage was an affair of prudence rather than of passion, an economic contract rather than a romantic adventure (Ariès, 1975; Borscheid, 1986). Patriarchy was perhaps not intrinsically connected to the EMP, yet it was the way in which this instrumental marriage pattern was applied. Patriarchy was a central element of traditional marriage and hence a key element in the marriage legislation (Napoleonic Civil Code).

One consequence of the instrumental view on marriage was older age at marriage. Instrumental considerations such as assuring the self-sustainability of the new household were objectives that required careful arrangements, and this simply required time. Waiting for the best moment, the best partner, and the availability of economic resources led to postponement of marriage.

This instrumental view on marriage made sense in times of poverty and an insecure standard of living (Van de Putte, 2005). It also made sense in a society in which economic opportunities were limited and in which property determined one's life chances. In particular, in societies in which economic survival largely depended on options provided by the family and the local community, it was recommended to hold a careful, family-oriented approach on any decision. This marriage pattern was also underpinned by a firm cultural basis (Lynch, 1991). For example, in the vision of the Catholic Church love was not the essential fundament of marriage—it was seen as a duty rather than as a condition for marriage (Cloet & Storme, 2000).

In the Low Countries this marriage pattern came under pressure from the second half of the 19th century onward. The clearest indication of the erosion of the EMP was the decrease in the age at marriage and the percentage never marrying. One explanation could be to simply see this as a relaxation of Malthusian principles: Because of the new economic circumstances it was easier to fulfill the requirements of the EMP. Various mechanisms may have played a role in this process. The extension of the wage economy and the declining importance of inheritance and employment within family businesses made people less dependent on a limited number of economic niches, pushed young people to pool their incomes and therefore to marry early, and at the same time decreased the influence of parents on the marriage decisions of their children. Yet the effect of the wage economy should not be overestimated because it was strong in many areas long before the second half of the 19th century—which makes it unlikely that the growth of wage economy as such might explain the shift in the marriage pattern. Industrialization was also a key factor in economic change. Especially some Belgian cities experienced a genuine takeoff, which had a dramatic impact on the size of the wage economy in those cities. Yet, in most of these cities (Ghent, Liège, Verviers), industrialization also predated the change of the marriage pattern. Industrialization is thus unlikely to be the sole explanation. Perhaps more important is that the new economic circumstances after some time went together with an increasing standard of living (Segers, 2003). The rise of the standard of living did occur in the same time period as the change of the marriage pattern. Particularly during the years 1850–1873 and after 1890 the progress was so large that nearly all groups benefited from it, including the industrial, the artisanal, the commercial, and the agricultural sectors (Oris, Alter, & Neven, 2005).

This relaxation of the EMP could mechanically have led to an increase in age homogamy. If, for the above-mentioned reasons, the age at marriage decreased, this would have changed the age structure of the marriage market in such a way that marrying an age peer became structurally easier. Indeed the frequency of age-homogamous marriages depends heavily on the distribution of the ages at marriage of both spouses. If many men and women marry early, marriages are “squeezed” into the younger age cohorts. Even if they have no special preference to marry someone of a specific age, these numerous youngsters have more chance to marry an age peer (Berardo et al., 1993; Van de Putte, 2005).

To sum up, the central claim underlying this economic perspective is that people no longer needed to wait too long to marry. A complementary view is that cultural developments even stimulated the rise of the idea that marriage should be based on less instrumental relationships. There are various but closely related ideas on this topic. Perhaps the strongest claim is that romantic love became the basis of marriage. Marriage became a matter of affection and personal compatibility. The spouse was viewed as the only, true partner with whom a unique relationship was established (Shorter, 1975). This new marriage model has often been interpreted as a recent “invention” (Giddens & Pierson, 1998). Such a statement is hard to prove. Nevertheless many authors observed indications of a new meaning of marriage and marital life that was based on “growing intimacy, emotionalism and sentimentalism of family life” (Mitterauer & Sieder, 1983, p. 60) or, in one word, romantization (Coontz, 2000), particularly in the second half of the 19th century (Matthijs, 2001).

There are slightly different views on the resulting balance between instrumental and romantic partner selection. First, romantic partner selection is not necessarily completely incompatible with instrumental partner selection. Even if individuals look for partners on the basis of romantic selection criteria, they may be searching primarily within a group of suitable potential partners. Romantic partner selection does not necessarily imply that all instrumental considerations have disappeared. Second, a less idealistic interpretation of the change in the basis of partner selection has been formulated by Gillis (1985) and Borscheid (1986). In their vision, romantic love became accepted as a norm among the lower classes but did not become the true basis of its marriage behavior. Other authors have argued that, at the best, the lower class marriage was a union of two comrades rather than soul mates or lovers (see Tilly & Scott, 1978). Third, the new view on marriage has also been interpreted within the framework of the companionate marriage model. The latter model is, however, somewhat paradoxical. It might appear as a patriarchal variant that stresses the male breadwinner aspect and female dependency and a more egalitarian variant that stresses the companionship between spouses. The “classic” (bourgeois) family, which was built around the strict application of formal criteria of what constituted a decent family (Gillis), indeed combined a sharp division of labor (the male breadwinner model) with a companionship as friends or lovers (Cherlin, 2004). In short, the companionate marriage model also suggests that the view on marriage changed less dramatically than the romantization thesis implies.

In line with the meaning of age homogamy, we expected a rise of age homogamy on the basis of this cultural view. Furthermore, also on the basis of this cultural view, we expected a connection between the decline of the age at marriage and the rise of age homogamy. Less instrumental motives such as romantic love and comradeship seem to have stimulated early marriage (Matthijs, 2003; Perrot, 1989). Love is impatient and urgent. For that reason the Catholic Church, for example, warned against the blind love, the levity, and the exaggerated positive evaluation of physical beauty by the “friskily, playful youth” (Cloet & Storme, 2000, p. 20). Yet we argue that increases in the preference for an age-homogamous marriage, brought about by a cultural change, are not limited to those who marry at an early age.

A point of discussion concerns the precise nature of the cultural change. As Marini (1984) made clear, it is extremely difficult to document the existence of norms regarding ages at which people are undergoing role changes. This applies also to the age ranges within which partners have to be selected. To prove the existence of these norms one needs evidence of a collectively shared evaluation that a marriage has to be contracted with a partner whose age falls within a given age range and that sanctions are brought to bear when a marriage is contracted with a person outside that age range. There is indeed evidence that norms regarding age differences existed. The reading of case records, writings of moralists, marriage advice books, and proverbs dating from preindustrial eras provides insight that large age differences between spouses were, without exception, considered detrimental to the purposes of marriage and that the aversion to these relations was strong (de Wildt, 1995; Haks, 1982). Popular proverbs and sayings reflected the low degree of tolerance of these marriages among the population at large (De Cock, 1911). Proverbs depicted marriages between young men and old women as ill-assorted marriages “made by the devil.” Holding the reins of the household—in case the young man married an old woman for her farm, cows, and money—the woman could lord over her husband and reverse the “natural order of things.” From reading these sources one gets the impression that it was only marriages characterized by large age differences that provoked animosity, more so when women were the older partner of the couple. A cultural trend toward a stronger preference for age homogamy could therefore be interpreted as a consequence of a stronger emphasis on existing norms rejecting marriages with extreme age differences. Rather than the development of a new norm, this would imply the realization of preexisting norms. More innovative is the idea that marital relationships should be based on less instrumental and more egalitarian grounds.

Where Did the New Marriage Model Originate?

The susceptibility for a new orientation toward marriage varied over class, place, and religion. We expected that those who belonged to the lower classes (and the middle class), lived in big cities, and lived in non-Catholic regions were most likely to have a marriage pattern, to a lesser degree affected by instrumental considerations already since the beginning of our study period, and they were also most susceptible to initiating the cultural change toward a marriage pattern characterized by more romantic motives.

The degree in which the marriage choices were and remained affected by instrumental considerations differed by class. The elite were associated with patriarchal values and instrumental partner selection strategies (Gillis, 1985; Shorter, 1975) to the extent that it seems rather unlikely that romantic love pushed them rapidly into relationships characterized by age homogamy (Stone, 1977). The farmers' attitude was largely similar (Schlumbohm, 1991). For both of these groups it also applies that they required much more stringent minimum standards for independent household formation.

The middle class was typically associated with respect for the formal family scripts and the idea of a male breadwinner supporting wife and children (Perrot, 1989; Van den Eeckhout, 1993). This was stimulated by the need for distinction with the lower classes and the middle class strategy of imitation of the elite. There were, however, also indications that the middle class took some interest in romantization (see Bulcroft & Bulcroft, 1997) or even was the driving force behind it (Frykman & Löfgren, 1987). The middle class had the cultural capital and economic freedom to put romantic partner selection into practice (Stone, 1977). Their position was therefore not unequivocal. Although they sometimes pictured the aristocracy as an old elite with a degenerated lifestyle, loose morals, and pretentious rituals, Frykman and Löfgren stressed the fact that marriages among the bourgeoisie were “asymmetric alliances.” They observed that “the older man … saw himself as provider and protector of a young and innocent wife” and that “this pattern made marriage into a parent-child relationship rather than a partnership of equals” (Frykman & Löfgren, pp. 101–102).

Probably most susceptible to the new marriage model was the lower class—the group that was least bounded by the protection of property. It is important to stress that also in the lower classes there was some receptivity for the bourgeois ideal of the classical family. The lower classes were keen on gaining respectability, which perhaps can be illustrated by their celebration of family events and their wish to acquire at least some property (Lis, 1977), although it was not easy to put norms about the ideal family life, such as the male breadwinner model, into practice (Van den Eeckhout, 1993). Yet other aspects of the new ideal could be lived up to, such as the “compulsion to marry.” It is important, however, that this urge for respectability was not in contradiction with the previously discussed new marriage pattern, quite the contrary. Both suggest, for example, a very positive evaluation of being married.

Cities were the core sites of economic change. Viewed from an instrumental perspective, age homogamy could more easily and earlier be realized in (industrial) cities because the economic situation here changed more dramatically because of the industrialization and the intensification of the wage economy. But big cities in general were also the core area of cultural change. In the first half of the 19th century, in most cities social life was still traditional and was organized by guilds (Van Isacker, 1978), but this changed dramatically in the late 19th century because of economic changes, population growth, new technology, and the rise of large working-class neighborhoods. These developments gave rise to new forms of recreation and social interaction. Activities organized by individuals and small groups replaced large-scale community events typical for agricultural and artisans' communities. Neighborhood pubs that attracted a mixed audience replaced the artisans' pubs. Young people spent time with friends made at work or school. Leisure possibilities such as dance halls and cinemas became the new meeting opportunities (Tilly & Scott, 1978). Although cities had always been different from the countryside, they became even more the site of innovation. The concentration of people, ideas, and information broadened one's horizon and expectations. People were confronted with different lifestyles. The town's anonymity made social control more difficult (Gillis, 1985; Tilly & Scott). In this way the late 19th century big city offered good conditions for detraditionalization of all kinds of relations.

We also assume that religion played a role in the adoption of a new marriage pattern. One argument is that the degree to which the couple really had a say in their marriage varied greatly by religion. Consequently, the resistance toward the new marriage pattern may also be expected to differ by religion. This can be illustrated by the differing visions regarding the “need for consensus” between partners. If it is believed that consensus between partners is required for marriage, this means that the parents' (and the groom's) influence and their instrumental preferences are somewhat restricted. From the 16th century on, there was a growing tendency to stress the parental authority in marriage decisions. Lutheran Protestants and Catholics following the Counter-Reformation attacked the former idea that consent of the two parties was the only necessary element in a marriage. In Catholic family ethics, parental authority and solidarity between generations was emphasized. Parents and the local clergy had to keep a tight rein on adolescent children. In religions where parental authority over the children's marital decisions was stressed, more weight was also attached to the power and dominance of the husband over that of his wife. Hence, marriages were favored in which men could more easily exercise this dominance by means of their age advantage. The issue of marital power in relationship to religion has a special significance in the light of the ongoing discussion on the Reformation's impact on the status of women and the nature of the family (McQuillan, 1999). Several historians have argued that the Reformation brought new restrictions on women, confining them to the role of wife and mother under the domination of their husband (see Roper, 1989). Luther in particular has been accused of laying the ideological groundwork for an increasingly authoritarian model of family life. Catholics were not too different it seems. McQuillan (1999) did not, for example, find any marked differences between Catholics and Lutherans in the age gap between spouses. The authoritarian model could be found in Belgium and in the south and the east of the Netherlands, where Catholicism was dominant. In the other parts of the Netherlands, the less stern variants of Calvinism were dominant (De Moor & Van Zanden, 2006). We therefore expect to find a lower level of age homogamy in these Protestant municipalities.

Hypotheses

Our main hypothesis is that age homogamy increased in the second half of the 19th century (Hypothesis 1). We discussed two complementary perspectives regarding the underlying causes of this change. The first one is that, because of changing economic organization and the increase in the standard of living in the second half of the 19th century, a relaxation of the Malthusian marriage principles became feasible. The second view is that a new cultural orientation toward marital choice arose in which less instrumental and more egalitarian principles were important. On the basis of both views the following more specific hypotheses can be derived. We start by specifying hypotheses regarding class differences, which can be expected from the perspective of both views:

If only the first, “economic” view is correct, the following hypotheses have to be confirmed:

If also the second, “cultural” view is correct, the following hypotheses have to be confirmed:

Data

Until recently constraints of time and money necessitated researchers who wanted to study patterns of age homogamy to focus on small communities and limited time periods. During the past decade marriage records were entered into a database within the framework of the so-called GENLIAS project (in the Netherlands) and the Vital Registration project of the State Archive of Bruges, the Demoflandria project of K.U.Leuven University and the Historical Database of the Liège Region (in Belgium). These projects aim to build up a database with genealogical information from the introduction of the vital registration on until the date these data are no longer in the public domain.

The vital registration system was introduced nationwide in the Netherlands in 1811–1812. In Belgium and in those parts of the Netherlands that were an integral part of the French Empire during the Napoleonic era, the Civil Register was even introduced in 1796. As the registration system was functioning flawlessly only after several years had elapsed, we decided to include only information for the years 1812 and later. From the marriage certificates the following information was used: the date and place of marriage and the places of birth, ages, and occupations of brides and grooms. For the Netherlands complete data sets could be used for 5 (of the 11) Dutch provinces: Zeeland, Limburg, Gelderland, Groningen, and Overijssel. The total number of marriages in the database that we use for this analysis is 642,371: 116,694 in Overijssel, 99,325 in Limburg, 203,129 in Gelderland, 126,907 in Groningen, and 96,316 in Zeeland. The Belgian data cover both Walloon and Flemish cities, the Flemish province of West-Flanders, and some Walloon villages. Different sampling strategies were used: For Ghent, 1 in 12 marriage certificates was included; for Leuven and Aalst, 1 in 3; for Liège, 1 in 10; for Verviers, an alphabetical sample was extracted. For the other Belgian municipalities, all marriages are included. The total number of Belgian marriage certificates is 124,041: 114,706 in Flanders and 9,335 in Wallonia.

Context

The Dutch regions selected for this study have a diverse socioeconomic structure. In the coastal provinces of Groningen and Zeeland, the economy was almost entirely dependent on (highly productive) agriculture, with large-scale farms producing specialized primary products. The economies of these regions started to change in the second half of the 19th century, when industrial activities based mainly on agricultural products developed. Both provinces had intensive contacts with the outside world through the well-developed transportation network, the seaports, and their highly market-oriented agricultural activities. In the eastern and southern Dutch provinces (Limburg, Gelderland, and Overijssel), agriculture was the most important activity as well. But in contrast to the coastal provinces, farms there were much smaller, the infrastructure was less well developed, and the productivity of land and labor was much lower. In the Limburg capital city of Maastricht, large-scale industries developed for the first time in the Netherlands.

In the 19th century, Belgium pioneered the industrial revolution on the European continent. The most important industrial cities in Belgium are included in our database: Ghent, Verviers, Liège, and Aalst. Ghent experienced its industrial takeoff from 1800 onward, mainly on the basis of the cotton industry. Verviers was a small center that underwent an early takeoff on the basis of its woolen industry. Aalst was a quiet city in the first half of the 19th century. From 1880–1890 on, industry expanded and factories became larger. Liège was a large city with many professionals and shopkeepers working alongside both artisans such as weapon makers and newer types of workers such as coal miners. During the 1820s mechanization in the textile industry prompted the modernization of the iron industry with its coke-fired blast furnaces as well as of coal mining. The provincial town of Leuven was a medium-sized city that gradually lost the traditional craft and agricultural roots of its economy. Leuven played an important role in administration and education.

Our database also covers the Belgian countryside. The Walloon villages are located in the Pays de Herve (characterized by cattle breeding and proto-industrial textile production) and the Ardennes (characterized by semilandless peasants who lived in miserable conditions). The economy of the coastal province of West-Flanders was mainly based on agriculture. Many people combined agricultural activities with spinning and weaving within a cottage industry framework. In the west and the north of the province, agriculture was much more large scale and market oriented.

Variables

As the hypothesis concerns the straightforward claim that age homogamy increased, we constructed age homogamy as a dichotomous, dependent variable in the analysis, distinguishing between age-homogamous (same age marriages) and age-heterogamous marriages. In the descriptive part, however, we further distinguished between older husband marriages and older wife marriages to provide more empirical background. We opted for a strict definition—and, hence, a strong test—of age similarity. We defined a same age marriage as a marriage in which the age difference between the partners is less than 2 years. If grooms or brides were at least 2 years older than their partner, we defined a marriage as age heterogamous. This strict definition prevents a misinterpretation of a possible trend in age homogamy. If for example we were to use a 5-year difference in age as the criterion, a (very) large part of these same age marriages might, in fact, be marriages where the husband is 3 to 5 years older. Even though this may be, in many societies, the modal value of the age difference between spouses, we think that such a broad definition would leave unanswered the question of whether a rise in age homogamy truly reflects a rise in egalitarian relationships. For the multivariate analysis, we performed some sensitivity tests using alternative age difference criteria (refer to the Results section).

At the individual level we used information on period, class, the age at marriage, and the migration status. Because a general increase in the number of age-homogamous marriages is expected trough the observation period, the period in which the marriage took place was introduced in the model with the continuous variable marriage year, centered around the year 1860 to make the intercept interpretable.

The groom's class was determined on the basis of his occupation. We classified all occupations in a class system applicable for the whole period, namely the Social Power scheme (Van de Putte & Miles, 2005). This scheme uses skill, possession, position within a hierarchical organizational structure, and prestige characteristics as criteria to distinguish classes. Lower class subgroups are unskilled workers (Social Power [SP] Level 1), semiskilled workers (SP Level 2), and skilled workers (SP Level 3). The middle class (SP Level 4) is mainly composed of master artisans, retailers, farmers, and clerks. For this study we excluded farmers from the middle class and put them in a separate category. The elite (SP Level 5) comprises white collar and professional specialists (e.g., lawyers), wholesale dealers, factory owners, and the aristocracy. The latter group, however, is too small to be classified in a separate category.

In the descriptive analysis of the age at first marriage, age was measured using a parametrical variable. In the logistic regression analysis, age at marriage was classified in eight groups: under 20, 20–24, …, 45–49, and above 50. We included age at marriage in this analysis to control for the effect of the age structure of the marriage market because there were no other ways of doing this with the present data. The number of (unmarried) men and women at every age was, for example, not available for all municipalities over the studied period. Migration status was classified in two categories: natives (the place of birth is also the place of marriage) versus migrants (the place of birth differs from the place of marriage). This variable was included because migrants typically have substantially higher ages at marriage.

We used three variables at the municipal level. The distinction between urban and rural communities was based on information on the number of inhabitants at the end of the period (1913): cities (communities with 20,000 or more inhabitants), provincial towns (with 5,000 to 20,000 inhabitants), and the countryside (the rest). To measure the religious composition of municipalities we used the percentage of Catholics in 1910 (classified in four groups: 0%–25%, 25%–50%, 50%–75%, and 75%–100%). The third variable is region (for Belgium: Flanders and Wallonia; for the Netherlands: the provinces).

Unfortunately, we do not possess high-quality indicators of industrialization at the municipal level. Assessing the impact of industrialization was only possible by comparing industrial cities (such as Ghent, Liège, and Verviers from the early 19th century onward and Aalst and Maastricht at later stages) to nonindustrial cities. Given that large-scale industrialization was not geographically widespread, as such this was not too problematic.

Multilevel Logistic Regression

We analyzed the probability that the marriage partners had the same age with a two-level logistic regression model, marriages clustered within municipalities. Observations were spread over 623 municipalities with a minimum of 20 marriages in each municipality. Four different models were tested. In all models we included random intercepts to allow us to examine whether there was a significant difference in the occurrence of age homogamy between the municipalities. In the models, the main intercept then represents the odds across all municipalities for the reference categories in 1860. Model 1 is a basic model to test whether age homogamy increased (Hypothesis 1) and whether there were differences in age homogamy by class of the groom, degree of urbanization, and religious composition of the municipality (Hypotheses 2a, 4.1a, 4.2a). The independent variables were period, migration status of bride, migration status of groom and class at the individual level, the degree of urbanization, the religious composition of the municipalities, and region at the municipal level.

In Model 2 interaction terms were introduced to test whether the period effect differs according to the class of the groom, the degree of urbanization, and the religious composition of the municipality (Hypotheses 2b, 4.1b, 4.2b). In Model 3 we added a random effect of period (a so-called random slope) to investigate whether the period effect on age homogamy differed between municipalities. In this model, the fixed effect of period then represented the general period effect across all municipalities. The results for the period effect per municipality allowed us to evaluate the differences between nonindustrial and industrial cities (Hypothesis 3.1b). The differences between the random intercepts (for the year 1860) permitted us to evaluate whether the level of age homogamy was already higher in the early industrial cities before the second half of the 19th century (Hypothesis 3.1a).

In Model 4 the age at marriage of bride and groom was included to control for the effect of the changing age structure (Hypothesis 3.2). Younger male age groups have better chances of an age-homogamous marriage, and the decline of the age at marriage may therefore “automatically” result in an increase of age homogamy. For women the mechanism is similar as there is not a (perfect) linear effect of age on the chance to have an age-homogamous marriage. Because of the general difference in the age at marriage by sex, chances for women to have an age-homogamous marriage are smaller at very young ages compared to women marrying in their mid-20s. Tolerance tests made clear that there is no multicollinearity problem for this model. The model provides a conservative test because, by included age at marriage, assumed that the decline of the age at marriage is not determined by the cultural change under examination.

First Marriages

Age differences between spouses are usually smaller in first marriages than in remarriages. This is related to the special character of remarriages. These happen on different marriage markets, are evaluated differently by parents and the wider community, and are experienced differently by the spouses. For this reason we restricted the analysis to first marriages only, that is, to marriages in which both husband and wife have not been married before. Including remarriages would also complicate the analysis (e.g., by the necessary inclusion of many interaction parameters).

Descriptive Results

Figure 1 presents the development over time in the pattern of age homogamy. There was a gradual but consistent growth in the proportion of marriages contracted by men and women of about the same age. The preference for same age partners was most visible in the Flemish cities. In the other regions, age homogamy was the highest in Zeeland, Wallonia, West-Flanders, and Groningen; it was the lowest in the provinces of Gelderland, Overijssel, and Limburg. In the Flemish cities, the percentage of age-homogamous marriages increased in one century by about 20 percentage points; in the other provinces it increased by 7–10 percentage points. In this process regional differences hardly changed.

The increase of age homogamy went together with a decrease of older husband marriages and older wife marriages. Figure 2 shows that marriages with grooms being 2 or more years older also showed very strong regional differences: They were less frequent in the Flemish cities, in Zeeland, Groningen, and Wallonia and more common in the eastern and southern Dutch provinces and particularly in West-Flanders. This pattern is illuminating. Older husband marriages were less frequent in areas where cities dominated (Flemish cities, Wallonia) and in the least isolated areas of the Netherlands (Zeeland, Groningen), whereas they were highest in the Catholic provinces (West-Flanders, Limburg) and the most isolated provinces in the east of the Netherlands (Gelderland, Overijssel). For older wife marriages, a strong decrease was visible everywhere (Figure 3). These marriages were less common in West-Flanders and in the Flemish cities than in the Dutch regions.

It was not only age homogamy that changed during the second half of the 19th century. In almost all regions ages at first marriage increased between the beginning of the 19th century and 1860, whereas after that period ages at first marriage strongly decreased (Van de Putte, 2005; Van Poppel, Liefbroer, Vermunt, & Smeenk, 2001). The pattern was similar for men and women, although ages at marriage were generally 2 to 4 years higher for the former. This means that any trend in age homogamy was not the product of a divergent trend of the ages at marriage of brides and grooms. Furthermore, there were large and very consistent regional differences. Men and women in Zeeland and Groningen married on average 2 to 3 years earlier than grooms and brides in Gelderland and Limburg. Bear in mind that precisely the former provinces also showed the highest proportion of same-age marriages, whereas the latter showed the lowest number. This indicates the connection between both characteristics. In the Flemish cities the age at marriage decreased most rapidly—which is again consistent with the strong rise of age homogamy in the same period.

Although it is clear that the decrease in the age at marriage in the second half of the 19th century matched the rise in age homogamy, this does not mean that both characteristics were perfectly related. The age at marriage in the Flemish cities, for example, was certainly not very different from that in Zeeland and Groningen. It is also worthwhile to point out that the increase of the age at marriage in the first half of the 19th century in all areas was not reflected in a decrease of age homogamy in the same period (Figure 1). To assess the relation between both characteristics in depth we have to measure age homogamy controlling for the age at marriage.

Multivariate Analysis

The results for Model 1 are shown in Table 1. As expected there was a period effect. For each year, the odds of marrying an age peer were 1.005 times higher than the year before. If we were to use less strict definitions of age homogamy, the odds of marrying an age peer would have been even higher (b = .0067 and b = .007 for a 3-year and 5-year criterion, respectively; data not shown). This confirms Hypothesis 1, which claims that there was an increase of age homogamy in the second half of the 19th century.

Next we look at the variables at the individual level. Class was associated with age homogamy as expected (Hypothesis 2a). The elite's and the farmers' chances to marry age homogamously were smaller than they were for the skilled workers (the reference group). For the middle class the difference was less strong. The other subgroups of the lower class hardly differed from the skilled workers. The control variable migration status also mattered. Migrant grooms' chances to marry age homogamously were smaller, and, to a lesser extent, this was also the case for migrant brides.

Next we turn to the variables at the municipal level. People living in big cities had better chances of marrying age homogamously, which confirms Hypothesis 4.1a. The difference between medium-sized towns and villages was not significant. Mind that this difference between cities and other contexts was controlled for class. It is not only because of the number of farmers that age homogamy was less frequent in the countryside. The religious composition of municipalities was also important. Compared to the homogeneous Catholic municipalities, people marrying in municipalities with hardly any Catholics were more likely to have an age-homogamous marriage. This is in line with their stronger emphasis on the consent of both partners and confirms Hypothesis 4.2a. Finally region also mattered. Controlled for other variables in the model, the chance to marry age homogamously was the highest in the Flemish municipalities, followed by Zeeland, Limburg, Groningen, and Wallonia. For Overijssel and Gelderland, age homogamy was the least likely to occur. These results are very similar to the ones observed in the descriptive analysis. Only Limburg had a different position because of the control for Catholicism in this model.

In short, the results obtained by applying Model 1 are the ones expected in case the cultural view is also valid. In Model 2 (Table 1) we added interaction terms to test whether the rise of age homogamy differed by category. The exponent (B) estimates for the interaction terms show the factor by which the period effect of the reference category (here 1.004; in the Table recorded as the main effect for period) of the second variable in the interaction (respectively class, size, and religion) needs to be multiplied to find the period effect for the other categories of that variable. All estimates for the interaction between period and class are smaller than one. This means that the period effect was strongest for the skilled workers (the reference category). The differences with the middle class and the elite were, however, not large and not significant. The results for the elite are not in line with Hypothesis 2b. As in our class scheme we included not only the aristocracy but also many people from what other authors have described as the new (professional) middle class, this finding may be related to the composition of this group. Yet the difference with farmers was larger (the odds ratio being only 1.002 compared to 1.004 for the reference group) confirming that farmers were least eager to adopt the new marriage pattern. The other effects were completely in line with the hypotheses. There was a different pace according to the type of municipality. The period effect was strongest in big cities (the odds ratio being 1.007). Also, in middle-sized towns the increase was somewhat larger compared to villages (Hypothesis 4.1b). The new marriage pattern was clearly an urban phenomenon (although to some extent also present in rural areas). Finally, there was also an interaction effect between period and religion (Hypothesis 4.2b). Not surprisingly, the increase was strongest in municipalities with the lowest percentage of Catholics (the odds ratio being 1.006).

Altogether the picture that emerges is very plausible. Age homogamy increased everywhere and in every social group. It was a general change, but the rise was strongest in Protestant areas and in big cities and was slowest among farmers. These interactions are generally in line with what could be expected in case the cultural view on the shift in the marriage pattern is also valid.

We use Model 3 (Table 1) to evaluate the effect of industrialization. The estimates for the variables did not differ substantially from those in Model 2. Table 2 gives an overview of the random effects for the big cities. The intercepts show the differences in the level of age homogamy in 1860 (the reference category for period). After half a century of industrialization, the level of age homogamy was high in Ghent and the nonindustrial cities of Oostende, Aalst, Enschede, and Leuven. The level of age homogamy was lower for the industrial cities of Liège and Verviers. The period effect was strongest for Ghent, followed by Aalst, Leuven, Maastricht, Enschede, and Apeldoorn. The period effect averaged (or was below average) for Verviers, Liège, and the other cities. Thus of the three cities with the strongest experience of industrialization (namely, Ghent, Liège, and Verviers), Ghent was the only town to show a particularly high level and strong rise in age homogamy. Consider, furthermore, that the rise of age homogamy was also present in medium-sized towns and rural villages. In short, we found no support for Hypotheses 3.1a and 3.1b and, hence, no evidence for the view that the change of age homogamy was the direct product of industrialization.

In Model 4 we included age at marriage as a conservative test of the effect of the age structure of the marriage market (Table 3). By including these variables we controlled for the fact that the age of the spouses was related to the chance of attracting an age peer, but we also controlled for the fact that the youngest had a stronger preference for an age peer. Nevertheless, the period effect remained present. Yet we have to point to the lower values of the parameter estimates for migrant grooms, farmers, the elite, the middle class, and cities in Model 4 compared to Model 3. The estimates for the least Catholic municipalities were even no longer significant. This shows that differences in age homogamy were to some extent related to differences in ages at marriage. The Catholic religion, for example, exerted its influence via a (particularly) high average age at marriage of grooms, which was about 2 years higher in the most Catholic municipalities compared to the least Catholic municipalities, resulting in a large age difference.

This analysis confirms that marrying at a relatively old age typically implied marrying a partner who did not have the same age. But the fact that the period effect was still present in Model 4 shows that the rise in age homogamy was not only the mechanical product of a decline of the age at marriage. In other words, there is no support for the hypothesis (3.2) derived from the position that only the economic view on the increase of age homogamy is valid.