From pragmatism to passion: changing partner preferences in Dutch matrimonial and contact advertisements, 1841–1995

ABSTRACT

This article explores the evolution of partner preferences in the Dutch mating market from 1841 to 1995, focusing on the rise of the cultural ideal of love-based partner selection. The study examines two hypotheses related to partner selection in modernization theory: the romantic-love hypothesis and the status-attainment hypothesis. These hypotheses are tested with descriptive and multivariate analysis using a database on matrimonial and contact advertisements. The study shows that the transition from instrumental to romantic partner choice was a gradual process, that in the Netherlands only gained momentum in the latter half of the twentieth century coinciding with processes of individualisation and the rise of the welfare state, a century later than earlier literature suggests. First instrumental preferences declined, then romantic preferences grew in importance. Furthermore, while the status-attainment hypothesis predicts an increase in educational preferences over time, the growth was slower and later than expected. These findings challenge both the romantic-love and status-attainment hypotheses and suggest an adjusted combined hypothesis. The study also highlights the role of social status and gender in partner preferences. Advertisers from higher social classes were more likely to express romantic preferences, indicating that wealthier individuals were earlier on inclined to turn love into the cornerstone of their relationships. Women exhibited a stronger emphasis on social status and educational level than men, highlighting gender role ideas of the time with women’s financial dependence guiding their preferences for a partner that could provide for them and their children. Moreover, the research demonstrates that religion and civil status did not significantly influence the internalization of love-based partner choice. The article concludes by emphasizing the significance of matrimonial and contact advertisements as valuable sources for comprehending the quest for a partner in the past and suggests potential avenues for future research based on these advertisements.

KEYWORDS:

• Partner choice
• human mating
• instrumental partner preferences
• romantic partner preferences
• matrimonial advertisements
• romantic-love hypothesis
• status-attainment hypothesis

View correction statement:

Correction

1. Introduction

On 25 December 1864, the following advertisement was printed in the Dutch newspaper Nieuwe Rotterdamsche Courant: ‘Attention! A person of decent lineage, 29 years old, of the Protestant religion and in the possession of some money, asks a woman, who can bring with her 8 to 10 grand, to marry him’ (see Illustration 1). Approximately 130 years later, on on 19 February 1994, the following advertisement appeared in NRC Handelsblad, the successor of the Nieuwe Rotterdamsche Courant: ‘This is the moment: meeting someone! Man desires a LAT-relationship in the Randstad [the urbanized regions of the Dutch provinces of Noord- and Zuid-Holland and Utrecht]. My key words: unconventional, calm, balanced, happy, lazy, down-to-earth, cosmopolitan, vulnerable, introvert, philosophical, ironic, 1.69 m, 44 years old, academic degree, hard working. Do you, woman, also value: conception and imagination, media and titbits, wine and walking, tenderness and passion, words and quietness, seriousness and laughter, stars and city?’ (see Illustration 2).

Illustration 1. Matrimonial advertisement, Nieuwe Rotterdamsche Courant, 25 December 1864.

Illustration 2. Contact advertisement, NRC Handelsblad, 19 February 1994.

Throughout history, the quest for a partner plays an important role in virtually everyone’s life. However, how individuals go about finding a partner, what they are looking for in a partner, how they formulate what they are looking for, and how they present themselves as partner seeking individuals changes over time, as is illustrated by the two newspaper advertisements above. Moreover, although partner selection is often perceived as a private choice in recent Western European history, it is strongly influenced by societal structures and influential others, such as kin, friends, neighbours, and colleagues. What type of partner individuals prefer and how they try to find and attract a partner, not only changes over time but also varies between regions, social groups and gender, along the three dimensions of partner selection: (1) the needs an individual hopes to meet, (2) individual’s preferences regarding a partner, (3) and the meeting opportunities. These three dimensions of partner selection interact in turn with the historical context, economic circumstances, societal expectations, cultural and gendered beliefs, the availability of potential partners and interests of third parties (Kalmijn, 1991; Van Bavel, 2021).

Although family historians and historical demographers have meticulously explored who married whom using marriage records, much remains unknown about what people were looking for in a partner, how they negotiated their personal preferences in a context of dominant norms and expectations, what strategies they used in the mating market and how this has changed over time. This article explores long-term trends in explicit preferences for a partner in the Dutch mating market between 1841 and 1995, focusing on an important topic in the history of partner selection: the rise of the cultural ideal of the love-based partner choice. The ideal of the love match emerged in eighteenth-century England and subsequently spread across Northwestern Europe and North America following the rise of the male breadwinner model and the growing importance of emotions, domesticity, and family relations (Coontz, 2005, pp. 145–160; Matthijs, 2001, pp. 240–243). In this article, we aim to trace the internalization of this new cultural ideal in the Netherlands employing a unique source that has hardly been used in historical research: matrimonial and contact advertisements.

This study tests two hypotheses related to modernization theory that are specifically relevant to explaining differences in partner preferences over time: (I) the romantic-love hypothesis and (II) the status-attainment hypothesis (Maas & Van Leeuwen, 2019; Smits et al., 1998). The romantic-love hypothesis argues that due to several modernization processes – such as industrialization, urbanization, educational expansion, the rise of wage labour, mass communication and transportation – the need and opportunities for parents and institutions to control young people’s partner choice decreased. Growing personal autonomy stimulated a search for individual happiness, among others, through a love-based relationship. According to the status-attainment hypothesis, the importance of ascribed status characteristics, i.e. the inherited social status and wealth from the parents, decreased in the labour market in favour of achieved characteristics, e.g. education and/or a certain career. Hence, one may expect achieved characteristics to have become also more important features in the mating market, while ascribed characteristics in turn became less important when it comes to partner choice. The two hypotheses are not mutually exclusive, and can potentially coexist, as the underlying processes in principle can occur simultaneously (Smits et al., 1998).

In this article, we test both hypotheses by investigating, on the one hand, whether a shift from instrumental to romantic partner preferences occurred in Dutch matrimonial and contact advertisements, and on the other hand by analysing whether a change from ascribed to achieved status criteria in terms of partner preferences is observed in the data. Moreover, we analyse the determinants of partner preferences by ways of multivariate analysis.

The article has the following structure. First, we will sketch the historical context in which the studied changes in partner preferences occurred. Then we will present a literature review on partner selection and marriage. Subsequently, we will describe and explain the data and methodology. We then present the results of our quantitative analyses, and we will conclude with a summary of the main findings, and a reflection on the possibilities offered by matrimonial and contact advertisements for future historical research on partner selection.

2. Historical context

A major shift in the history of marriage was the emergence of love-based marriage partner choice. People have always fallen in love, but in pre-industrial societies partner choice was mainly driven by instrumental motives (Coontz, 2005, pp. 123–130). Marriage was too important as an economic and political institution to be based primarily on love. However, from the late eighteenth century onwards marrying for love became a new cultural ideal in Western societies driven by two seismic social changes: the spread of wage labour and new political and philosophical ideas (Coontz, 2005, pp. 145–148). Secular and progressive Enlightenment thinkers argued that social relationships, including those between women and men, should be organized on the basis of justice and reason, rather than force. With regard to partnering, they fostered the idea of marriage as a private agreement between two individuals instead of a contractual link in a political and economic system. Influential thinkers like Jean-Jacques Rousseau and Marquis de Condorcet advocated marrying for love and the pursuit of happiness as a legitimate goal (Coontz, 2005, pp. 145–149). Wage labour made young people less dependent on their parents for the start of their own household and nuclear family (Coontz, 2005, p. 146; Puschmann, 2020).

Due to the emergence of modern transportation and communication in the nineteenth century, the marriage market widened as the pool of potential partners increased (Kok & Van Leeuwen, 2005). Consequently, the average distance between birth places of spouses in the Netherlands increased systemically from 15–20 kilometres in the first three quarters of the nineteenth century to 25–30 kilometres in the 1920s (Ekamper et al., 2011). Moreover, during the nineteenth and twentieth centuries increasingly more people – especially women – received an education and they were educated for a longer period of time. As a result, schools and colleges became increasingly important places for meeting a potential partner (Kok & Van Leeuwen, 2005).

The emergence of the cultural ideal of love-based marriages, coupled with expanded avenues for meeting potential partners, indicates a substantial increase in the freedom and opportunities available to individuals in the realm of mating. Partner choice, however, did not all of a sudden become unrestricted in the nineteenth-century Netherlands, nor did the interests of individuals and their families change from one day to the other. The idea of marrying someone on the basis of love broke with centuries of tradition and sparked a lot of criticism from individuals and institutions that feared romantic partner choice would undermine the social and moral order and threaten the stability of marriage (Coontz, 2005, pp. 149–154).

Marriage advice literature, often compiled under the supervision of religious authorities, kept considering alignment of religion and social class as a pre-requisite for a harmonious and love-based marriage and parents continued to have a great deal of influence on their children’s partner choice (Van Poppel, 1992, pp. 40–62; Van Tilburg, 2005, pp. 234–237). Especially in the wealthy classes, there was ongoing parental control and influence. Young women from the elite usually did not leave the house without the company of family members or a chaperon, and consequently elite singles had limited opportunities to meet with potential partners besides the private (dance-)events within their own community under the watchful eye of family, and acquaintances (Stokvis, 2005a, 2005b). Working class youth enjoyed much more freedom in terms of partner choice, as their parents had limited means to control them. They could meet potential partners in pubs, in church, at dance venues, fairs and all kinds of festivities (Stokvis, 2005a).

When it comes to the demographics of marriage, several major changes took place during the research period that are interrelated with broader social-cultural, economic and political developments (Zwaan, 1993b). From the mid-nineteenth century onwards, marriage rates increased, and the average age at first marriage decreased, signalling a gradual disappearance of the Western European marriage pattern (Engelen, 2009, pp. 29–56). Whereas the average age at first marriage in the mid-nineteenth century had been around 29 years for men, and 27 years for women, this decreased to 24.9 years for men and 22.7 years for women between 1965 and 1970 (Zwaan, 1993b).

From the late 1960s on, the Second Demographic Transition unfolded: ages at marriage rose, fertility declined below the replacement level, and diverse living arrangements beyond the traditional heterosexual marriage became more common (Coontz, 2005, pp. 263–280; Lesthaeghe, 2010). In 1990, the average age at first marriage had risen to 30.3 years for men and 28 years for women. Furthermore, marriage rates plummeted, and marriages increasingly ended in divorce. Simultaneously, alternative relationship forms such as cohabitation and LAT-relationships gained in popularity (Coontz, 2005, pp. 263–280; Zwaan, 1993b). The main driver of the Second Demographic Transition was a cultural shift from altruistic to postmodern norms and attitudes stressing individuality and self-actualization (Coontz, 2005, pp. 263–280; Van de Kaa, 1987, p. 5; Zaidi & Morgan, 2017).

Another key element of the Second Demographic Transition is the disconnection of marriage and procreation. This disconnection was the result of a seemingly irreversible shift in ideas and behaviour regarding sexuality and relationships that took place in the Netherlands from the late 1960s onwards and the arrival of efficient means of contraception (Knijn & Verheijen, 1988, pp. 37–53; Kok & Leinarte, 2015; Kraaykamp, 2012). This shift first and foremost developed among younger, higher educated people in the cities. Marriage, love and sexual intercourse which had been inextricably connected for centuries in Western Europa, became disconnected during the sexual revolution of the 1960s and 1970s. A radical re-evaluation of the place and meaning of sexuality occurred. Whereas for centuries sexuality had been mainly perceived as a necessary evil, a more positive perception and mindset regarding sexuality emerged. Consequently, sexuality also became more visible in society and was more openly discussed. Moreover, out-of-wedlock fertility increased, as more and more couples became parents without being married (Zwaan, 1993a).

In the late 1960’s gender role ideals also started to change. Although Enlightenment thinkers already argued for a love-based marriage, they did not imply similar roles for men and women. They believed that both sexes were equal, but that men and women had different, complementary, natures (Coontz, 2005, pp. 145–160). From the middle of the nineteenth century on, this perspective stimulated the rise of separate spheres for men and women, a public and a private one, first and foremost among the bourgeoisie. Men participated in public life and acted as breadwinners, whilst women were increasingly confined to their homes as housekeepers and guardians of moral integrity (Coontz, 2005, pp. 161–165; Kloek, 2009, pp. 179–205; Puschmann, 2020). This male breadwinner model was initially unfeasible for the lower classes, as working-class families depended on the income of all family members. However, this changed gradually from the late nineteenth century on, when real wages of men increased and a social security net emerged, resulting in the male-breadwinner model coming within reach for most households. The ideal of the separate spheres reached its heyday in the 1950s (De Regt, 1994).

In the wake of ongoing individualisation – increasing emphasis on an individual’s personal needs and desires apart from their role within the larger entity of the family – and the Second Feminist Wave of the late 1960s, more egalitarian gender roles emerged (Portegijs, 2010; Weusten, 2011, pp. 29–72). From the 1970s onwards, the gender gap in higher education in Europe and North America decreased and eventually even reversed. For the first time in history, more women enjoyed higher education than men and the share of couples in which wives have higher educational attainment than their husbands is now greater than those in which husbands have a higher educational attainment (Grow et al., 2015; Van Bavel et al., 2018). However, despite changing gender role ideals, and increasing female participation in higher education and the labour force, Dutch female labour force participation remained low for a long time compared to other European countries (female Dutch labour market participation only reached the European level of 54% in 1990). Moreover, the trend of greater participation of women in the Dutch labour forces essentially came down to an increase of women working part-time. The proportion of Dutch women working full-time only rose marginally compared to other European countries (Kraaykamp, 2012; Portegijs, 2009, pp. 6–7; Zwaan, 1993a). Moreover, despite increasingly egalitarian gender role ideals, Dutch women still tend to do much more household and childcare chores compared to their male partner (Dekeyser & Puschmann, 2021; Jonkers, 2021).

3. State of the art

3.1. The transition from instrumental to romantic partner choice

According to the American historian Coontz (2005, p. 147), the rate of adoption of the new norms associated with the love match varied across regions and social classes. The celebration of the love match reached a fever pitch in England as early as the 1760s and 1770s and soon the romantic ideal spread in North America. Other western countries gradually followed suit. The turning point in Germany is commonly acknowledged to be the Age of Romanticism, during which emotional and physical love became the primary requirement for marriage (Borscheid, 1986).

Historical demographers and family historians often assumed that the ideal of romantic partner choice became internalized among Western populations in the nineteenth century when domesticity, sentiments and family relationships became increasingly valued, and the male breadwinner model started to take root (Van de Putte & Matthijs, 2001, pp. 240–243; Van de Putte et al., 2009). Certain indirect measures, such as the rise of age homogamy, point indeed in that direction, as they seem to indicate that the functional and unequal relationship between spouses was increasingly replaced by more equal arrangements, which were economically less logical (Van de Putte & Matthijs, 2001; Van Poppel et al., 2001). Moreover, Lystra (1993) showed on the basis of love letters that many married couples in the nineteenth-century United States were indeed soul mates, who shared intimate information in their personal communications which they would not share with parents, siblings and friends, including information on erotic and sexual experiences they had experienced together. This implied that couples truly loved each other, but was it also love that brought them together? There are many indications that love was certainly not yet the main determinant of partner choice in the nineteenth century (Kok & Van Leeuwen, 2005; Metzlar et al., 2023; Puschmann, 2020). Following the romantic-love hypothesis, one would expect that social, religious and ethnic heterogamy would have increased. After all, true love is boundless. However, mixed marriages in terms of social status, religion, ethnicity, and migration status remained the exception to the rule in nineteenth- and early twentieth-century Western Europe (Borscheid, 1986; Maas & Van Leeuwen, 2005; Puschmann et al., 2016). Although, marital homogamy seemed to remain strong, research revealed major differences between social groups (Dribe & Lundh, 2005; Lippényi et al., 2017; Roikonen & Häkkinen, 2019; Seiler, 2019).

The historian Peter Borscheid (1986) analysed the transition from instrumental to romantic partner choice in Germany on the basis of a combination of quantitative and qualitative sources and he concluded that love was not yet a prerequisite for marriage in nineteenth-century Germany. Partner selection was still driven by such factors as money, social status, and prestige. According to Borscheid ‘the majority of the population endorsed the idea of love, but at the decisive moment of partner choice discarded it as a dream’. This suggests that – at least in Germany – the internalization of the romantic-love match occurred only in the twentieth century.

Long-term historical research on partner preferences before paring and how these changed over time is still in its infancy. The scant literature available indicates that most people adhered to instrumental preferences in partner choice until the final decades of the twentieth century (Beauman, 2011, pp. 191–194; Lippmann, 2021). Lippmann’s (2021) underscores that partner preferences in France only started transforming in the late 1960s when economic preferences gradually gave way to personality preferences, following major family and demographic changes in society. The British historian Beauman (2011) argues that three-hundred years of English matrimonial press consistently shows that men seek youth and offer financial security, while for women it is the other way around. She emphasizes the importance of youth in the male search, along with its cue, physical attractiveness, through an evolutionary lens, interpreting it as a sign of fertility; the capacity to provide healthy children. For women, men had to fulfil the one evolutionary criterion that women consistently had: the ability and means to feed and support the family. According to Beauman, mating preferences before pairing only started to change at the end of the twentieth century, driven by an increased female earning potential.

For the Netherlands, it remains to be explored when the transition from instrumental and romantic partner choice took place and what factors drove this transition. Matrimonial and contact advertisements are an ideal source for this purpose as partner seeking individuals could formulate the characteristics of the partner they were looking for.

3.2. The transition from ascribed to achieved status

Based on the above, one might conclude that love is simply not as status blind as some believe it to be. According to the status-attainment hypothesis, economic and rational motives remained important factors in nineteenth- and twentieth-century partner selection. However, in the nineteenth century, due to modernization, characteristics that determine success in the labour market – and therefore also in partner selection – gradually changed from ascribed characteristics (social status and parental wealth) to achieved characteristics (e.g. education and career) (Hendrickx et al., 1995; Knigge et al., 2014; Smits et al., 1998). This hypothesis builds on the pioneering work by Blau and Duncan from 1967, whose status-attainment model reconceptualized social mobility in terms of the influence of individual’s socioeconomic origins and other attributes, such as education, region and migration, on the chances in life and more specifically on occupational status (Blau et al., 1992). They argued that with the advent of industrialization and educational expansion, a man’s own characteristics became more important indicators of present and future success. Subsequent research has built upon and refined the model, giving rise to various hypotheses, including the status-attainment hypothesis.

Evidence in support of the status-attainment hypothesis varies. Although, in general, for most industrialized countries indeed a decline in the importance of social background for marriage choice has been observed (e.g. for the Netherlands: Uunk, 1996; for the United States:, pp. 54–56; Kalmijn, 1991), but trends in educational homogamy do not point in one direction (Grow et al., 2015; Kalmijn, 1998; Ultee & Luijkx, 1990). Maas and Van Leeuwen (2019), p. did find strong indications in support of the status-attainment hypothesis in the Dutch context. They found evidence that men’s occupational status (achieved status) became more important in partner selection and that the occupational status of their fathers (ascribed status) became less important from 1865 onward in the Netherlands when modernization accelerated. In general, for the 1813–1922 period they found a positive relation between modernization and partner selection based on achieved characteristics and a negative relation between modernization and partner selection based on ascribed characteristics. However, for the larger cities they also found some support for the romantic-love hypothesis indicating that modernization caused a decrease of both ascribed and achieved status. This implies that the two hypotheses do not necessarily exclude one another.

Although the results of Maas and Van Leeuwen mainly support the status-attainment hypothesis, the slight indications found for the romantic-love hypothesis can serve as support for the combined hypothesis formulated by Smits et al. (1998). Smits et al. (1998) analysed the degree of educational homogamy in sixty-five countries based on data from surveys carried out in the 1970s and 1980s. They concluded that the status-attainment hypothesis is supported only if the least-developed countries are compared with countries at an intermediate level of development. When comparing countries at an intermediate level of development with the most developed countries, they only found evidence for the romantic-love hypothesis. Smits et al. attempt to reconcile the romantic-love and status-attainment hypothesis by arguing that during the process of modernization first the shift from ascribed to achieved status occurred. They argue that in the first phase of modernization, education (achieved status) becomes the dominant criterion for socioeconomic success, whilst the importance of ascribed status continuously decreases. In later stages of the modernization process – when high standards of living are virtually guaranteed for everyone – the importance of achieved status is replaced by a shift to more individualistic values.

In addition, various scholars have demonstrated the importance of gender differences with respect to social status in partner choice. Men and women value different characteristics in a partner of the opposite sex (Greenlees & McGrew, 1994). Until the late 1960s, in most Western societies, marriage was based on the sex-specific division of paid and domestic labour. Consequently, women were generally more focused on socioeconomic resources and thus social status compared to men who tended to be more concerned with women’s capacities to perform domestic labour and bear children (Beauman, 2011, pp. 191–194; Van Poppel et al., 2001). It has been suggested that as a result of changing gender roles, the reversal of the gender gap in education and women’s increasing participation in the labour market, men’s and women’s partner preferences became increasingly similar as women became less dependent on men for their economic well-being (De Hauw et al., 2017). Potentially these changes have resulted in a convergence in the importance that men and women attach to the socioeconomic resources of their partner (Grow et al., 2015; Kalmijn, 1998). However, empirical historical research on the consequences of the shifting gender balance on partner selection is rare.

3.3. Studies on partner selection using matrimonial and contact advertisements

Quantitative historical research on long-term trends in partner preferences using matrimonial and contact advertisements, especially in the Dutch context, is still in its infancy. Most historical research on matrimonial advertisements has been done in the English context, which is not surprising as these advertisements first appeared in England. They were first published in the seventeenth century and the advertisements swiftly gained popularity, leading to the emergence of a dedicated matrimonial press in the second half of the eighteenth century (Zeegers, 1990, pp. 16–17). The booming world of English matrimonial advertisements created an early public debate about gender roles and potential risks of this new form of dating. The English studies are often mainly descriptive and focus on these debates (e.g. Beauman, 2011; Cocks, 2009). For France, noteworthy exceptions include the aforementioned study by Lippmann (2021) and research conducted by Gaillard (2018). Both scholars compiled their datasets by extracting matrimonial advertisements from a particular newspaper. Gaillard employed a mixed-methods approach, examining 4,000 advertisements published between 1921 and 1940. Lippmann analysed a staggering 340,000 matrimonial advertisements spanning the period from 1928 to 1994, using innovative dictionary-based methods, such as text mining. In the Netherlands, Stokvis (2011) conducted a small-scale analysis of matrimonial advertisements from 1825 until 1925 exploring how their form and content changed during this century and what these changes tell about the self-image of the advertisers and their wishes for a partner and relationship. According to Stokvis, the shift in terminology in his sample set over time reflected the growing importance of achieved status over ascribed status. Psychologist Wil Zeegers (1990) conducted large-scale research on Dutch matrimonial advertisements albeit only for the second half of the twentieth century. In his fascinating and detailed study Zeegers distinguishes seven categories that together form the template of all contact advertisements. Moreover, he provides a clear overview of the biggest shifts observed in the advertisements from 1945 to 1990, such as (1) the shift from marriage as virtually the only and self-evident goal of the advertisements to a variety of relationship goals, (2) a shift from social status to educational level (3) and the shift from emphasis on decency, good manners and occupation to personal character traits, leisure activities and interests. However, his analysis lacks a systematic analysis of differences between sexes, in geographical origin, religious background, civil status and relationship types.

Outside the field of the Humanities, the potential of matrimonial and contact advertisements for the study of partner selection has been recognized more. A wealth of literature from other fields mainly focusses on the second half of the twentieth century, when marriage became less popular, alternative relationships emerged and the advertisements became constructions of self-identities (Coupland, 1996; Frey, 2015; Greenlees & McGrew, 1994; Kana & Mberia, 2018; Mukhopadhyay, 2012; Polzenhagen & Frey, 2017; Privat, 1987; Rinnekangas, 2013; Wiederman, 1993). Greenlees and McGrew (1994), for example, analysed sex differences in human mate preferences and intrasexual competition in contact advertisements published between 1987–1989 from an evolutionary biologic perspective. In line with basic evolutionary principles, they found that women were more likely to seek financial security and an older mate, and requested investment from a potential partner, whereas men were more interested in physical appearance and sought a younger partner. Moreover, they found that regarding self-presentation partner-seeking individuals tend to emphasise those aspects individuals of the opposite sex seek in a partner. When presenting themselves, men tend to offer financial security and willingness to invest more often than women, whereas women emphasise their physical appearance more than men do. This study, however, lacks a historical perspective and an analysis of the consequences of the shifting gender roles on partner selection.

4. Sources and methodology

Matrimonial and contact advertisements have barely been used by family historians and historical demographers. Historical research on partner choice has mainly focused on the end result of the quest for a partner, as reflected in, for instance, marriage records, population registers or censuses. Partner preferences are deduced from data on married couples. Matrimonial and contact advertisements provide, by contrast, a much more direct insight into what people were looking for in a partner and how they formulated their partner preferences. In the advertisements, the actors provided basic information about themselves like sex, age, socioeconomic background, and religion. Subsequently, they specified the characteristics sought in a partner. Lastly, the desired type of relationship was stated.

Matrimonial and contact advertisements have a number of features that make them particularly suited for the study of partner preferences. First of all, due to spatial and financial constraints, advertisers presumably only listed the characteristics believed to be essential to the advertiser and the potential respondents. Secondly, the partner preferences were specified – in the context of dominant discursive norms – by the historical actors themselves and do not have to be deduced or assumed. It is, however, important to be aware that the advertisements are not necessarily a pure expression of individual preferences. The advertisers’ preferences are mediated (consciously or unconsciously) by genre and societal conventions (Zeegers, 1990, pp. 47–66).

Furthermore, the source material has some (potential) selection biases, for example, regarding the representativeness of the advertisers for the whole partner-searching population of society. Matrimonial advertisements were initially often seen as an improper partner selection strategy and there were prejudices that individuals who used them had more difficulty finding a partner than others (Stokvis, 2011; Zeegers, 1990, pp. 11–13). However, in the Netherlands, these advertisements became an increasingly common way to find a partner from the late nineteenth century onwards. In the 1920s and 1930s, meeting a potential spouse by way of an advertisement was not only common, but also popular (Stokvis, 2011). Figure 1 displays the average age at first marriage in the Netherlands of men and women and the average age of single advertisers in our database.¹ This graph shows indeed that the average age of the advertisers is considerably higher than the average age of marriage. This might indicate that the advertisers had a relatively hard time in finding a partner and that these were not the most popular candidates in the marriage market. Other factors, such as living in relatively remote places or having limited financial means, might have been at play too.

Figure 1. Average age single advertisers and average age at first marriage by sex, the Netherlands, 1841–1995.

The question is whether the partner preferences articulated in the matrimonial and contact advertisements evolved in the same way as preferences in the broader Dutch mating market. Given the scarcity of literature and long-term historical data on partner preferences prior to pairing it is hard to quantify this potential bias. The expense associated with placing an advertisement may have served as a barrier for individuals from lower social classes. Publishers typically imposed charges per word or line in advertisements with a stipulated minimum requirement for the number of words or lines. Examples from the Algemeen Dagblad show prices ranging from Dfl. 0.25 per line in 1843 (with a minimum of seven lines) to Dfl. 0,53 per line (with a minimum of two lines) in 1970. To draw a comparison, the prices of bread stood at Dfl. 0.10 in 1840 and Dfl. 0.36 in 1970, which means that the costs of placing an advertisement became relatively less expensive over time.² Another potential bias is related to the relatively low number of female advertisers in the nineteenth century. In the Netherlands, larger numbers of women only really started placing matrimonial advertisements in the twentieth century (Stokvis, 2011). Furthermore, as the advertisement were anonymous messages, the question whether the advertisers were successful in their search for a partner remains unanswered.

Despite (potential) biases, advertisements combining partner preferences and self-descriptions are one of the few sources that exist that can help to disclose fundamental changes in partner preferences. After all, placing an advertisement is a ‘real-life’ act with genuine consequences. The advertisements are therefore likely to reflect true partner preferences (Greenlees & McGrew, 1994). Moreover, as the cultural expectations of what ideal male and female partners constitute is shaped by the historical context, as well as what is appropriate to mention in a matrimonial or contact advertisement and what is not, these advertisements provide a ‘snapshot’ of our (matrimonial) culture at a specific time in history (Eagly et al., 2004; Zeegers, 1990, pp. 9–10).

As a result of the mass digitization of historical newspapers and magazines, large quantities and long-time series of marriage and contact advertisements are easy to access. For this article we used a dataset we constructed ourselves with matrimonial and contact advertisements from the collection of Delpher, the digital database of the Dutch Royal Library (Koninklijke Bibliotheek) for Dutch books and periodicals.³ At the time of construction of the project database, the main collection contained 1.7 million newspapers from the Netherlands, Dutch East Indies, Antilles, North America and Surinam printed between 1618 and 1995, covering 15% of all Dutch newspapers published in that time period.

For our study, Dutch newspapers published in countries outside the Netherlands were ignored. Furthermore, since no single newspaper in the Delpher collection covers the period under study, matrimonial and contact advertisement were extracted from as many national and regional Dutch newspapers as possible. This approach would broaden the geographical coverage at the level of individual advertisers, as the publishing companies of the national newspapers were exclusively situated in two of the eleven provinces: Noord-Holland and Zuid-Holland. Advertisers from outside these provinces probably had easier access to offices of local newspapers. The research period ranges from 1841 to 1995. While a handful of early matrimonial advertisements were discovered before 1841, an increase in their prevalence did not occur until the 1840s (Stokvis, 2011). The end date of 1995 is related to the shift to online dating taking over the printed matchmaking market in the 1990s (Lippmann, 2021).

In order to find and extract the matrimonial advertisements from Delpher we used software developed at Radboud University in Nijmegen.⁴ This software can conduct more complex search queries than the search tool offered by Delpher and saves all information from the advertisements in a database, including the link to the page of the original advertisement in the newspaper. The software selects unique hits and avoids multiple inclusion of the same advertisement in the database as much as possible. For sampling we defined various search queries, which are discussed in Appendix 1, including potential pitfalls of the sampling strategy. The results of our queries were meticulously scrutinized for temporal and geographical coverage, coverage by gender, social status and profile of the main national newspapers, see Appendix 1.

Each record in the database refers to one matrimonial or contact advertisement and contains the newspaper name, date, location of the publisher, distribution area (national/regional) and edition (day/morning/evening). The search queries provided us with links to entire newspaper pages, for the twentieth century often containing multiple matrimonial or contact advertisements. From each page we randomly selected one advertisement. ‘False positives’ or type I errors, i.e. texts not being a matrimonial or contact advertisement, and erroneous double hits were manually removed from the database, as were advertisements that did not include any partner preferences. The total database contains 7,469 Dutch matrimonial and contact advertisements published between 1841 and 1995. The numbers of advertisements per decade vary but represent a sizeable sample that covers a wide range of advertisements from a large number of national and regional Dutch newspapers (see Appendix 3).

All personal characteristics of the advertiser were manually coded and inserted in the database: sex, age, religion, civil status, and occupation of the advertiser, as well as the preferred age and religious denomination of potential partners. For more details on this approach, reference is made to Appendix 2. To determine the social status of the advertiser we coded his/her occupation using HISCO and classified each occupation according to HISCLASS_5.⁵

Based on the outcome of two pilot studies and following the example of Lippmann (2021, p. 842), we formulated distinct categories for classifying the partner preferences included in the advertisements. We identified eleven categories, which were divided into two groups: (1) instrumental and (2): romantic preferences, see Table 1. The group ‘instrumental’ contains seven partner preferences labelled as financial status, social status, educational level, health, children, geographical origin and religion. The group ‘romantic’ encompasses all non-instrumental preferences, notably appearance, intimacy, hobbies/interests and character. In this article, references regarding education are considered instrumental and related to the earning capacity of a potential partner. We recognize that these preferences can also be viewed as non-instrumental, where advertisers seek affinity of mind. Moreover, preferences regarding a specific religious denomination are viewed as expressions voiced in a strictly segmented society and therefore considered instrumental. A large part of the research period is characterized by the process of pillarization, which took place in the Netherlands between roughly 1880 and 1960. This era witnessed the Dutch social and political structure being compartmentalised into distinct pillars based on religious or ideological affiliations, whereby each pillar maintained its own separate and distinct institutions and organizations. Furthermore, we decided not to classify preferences regarding age, recognizing the presence of valid arguments supporting their categorization in both the ‘instrumental’ and ‘romantic’ group. For instance, male advertisers looking for potential partners that were young(er) perhaps may have indirectly referred to the capacity to provide children (considered an instrumental preference) or to attractiveness and physical appearance (characterised as a romantic preference). For each advertisement, the preferences were dichotomously coded as ‘yes’ (advertiser set positive or negative condition) or ‘no’ (advertiser did not set condition). For detailed information on the classification and grouping of the advertisers’ preferences and the methodological choices we made, see Appendix 2.

Table 1. Categories of instrumental and romantic partner preferences.

Partner preferences – group ‘Instrumental’	Partner preferences – group ‘Romantic’
Financial status	Appearance
Social status	Intimacy
Educational level	Hobbies/interests
Health	Character
Children
Geographical origin
Religion

Figure 2 shows the distribution of male and female advertisers in the dataset. In the nineteenth century, the advertisements were placed almost exclusively by men. From the late nineteenth century, and especially the early twentieth century onwards, it became increasingly common for women to place an advertisement, although women remained clearly underrepresented. This is reflected in the composition of our dataset. As our search queries were equally targeted to male and female advertisers and the advertisements were randomly selected from one newspaper page, we expect our male to female ratio to be a rather good representation of the entire corpus of nineteenth- and twentieth-century advertisements.

Figure 2. Number of advertisers by sex, 1841–1995 (N = 7,469).

Our methodology is centred around the romantic-love and status-attainment hypotheses. First, we test the romantic-love hypothesis using descriptive statistics to explore the timing and rate of the inclusion of preferences (individually and combined) from the group ‘romantic’ compared to preferences from the group ‘instrumental’.

Since there are no absolute measures for calculating the degree to which advertisements were ‘instrumental’ and/or ‘romantic’, the romantic and instrumental categories covered by an advertisement are expressed in percentages. In order to calculate the percentage of ‘romantic’ content in each advertisement, the following formula was used: the number of romantic categories covered by an advertisement divided by the total number of categories covered by that advertisement (with a maximum of eleven, corresponding to the eleven categories that we defined). As the percentages ‘instrumental’ and ‘romantic’ together form 100%, the percentage ‘instrumental’ was calculated as: 100% minus percentage ‘romantic’.⁶ In this way, an advertiser can theoretically get a score of 100% ‘romantic’ and 0% ‘instrumental’, or vice versa. In reality, it is a sliding scale to do justice to the fact that instrumental and romantic are not absolute concepts.

Subsequently, linear regression models are fitted to measure the association between, on the one hand, time, geography and individual characteristics of the advertiser and, on the other, the dependent variable: the percentage ‘romantic’ of the advertisement. The independent variables in the OLS regression are the year of publication of the advertisement (time), the location (province) of the publisher (geography), the (national/regional) distribution of the newspaper (geography) and the sex, age, social status, religion and civil status of the advertiser.⁷

We expect to find a positive relation between the degree of romance in the advertisements and the period of publication, in the sense that over time advertisements become more romantic. Next, we expect a positive relation between the degree of romance in the advertisements and newspapers produced in the western provinces of Noord- and Zuid-Holland. After all, modernization first hit the big cities, which -in the Netherlands- are located in the western provinces (Van de Putte et al., 2009). Forerunners with regard to the level of romanticness are expected to be found among the young and single male advertisers, among those with a Protestant and bourgeois background compared to other groups in society that had sociocultural and/or economic reasons to be more conservative. Maslow’s theory on the hierarchy of needs (Maslow, 1943) argues that an individuals’ physiological and material needs must be satisfied first before focusing on less material psychological and self-fulfilment needs. In line with this theory, we expect advertisers from higher social classes – whose material needs are most likely secured – to focus more on romantic partner preferences compared to advertisers from lower social classes for whom marriage might serve more as a way to fulfil their material needs, thus emphasising more instrumental partner criteria. Furthermore, for the majority of the research period, a great number of women were dependent on their husbands for their social status and financial security (Coontz, 2005, pp. 145–165; Kloek, 2009, pp. 179–205). Consequently, the financial situation and social status of a partner were of greater importance for women compared to men when scouring the marriage market. For widows and divorced women, finding a male breadwinner often was a dire necessity, while widowers wanted someone to take care of their offspring. Finally, we expect Protestant advertisers to be more romantic than Catholics in the mating market, as according to literature, age homogamy, a sign of a less instrumental partner choice, grew more rapidly in regions in which few Catholics lived (Van de Putte et al., 2009).

Next, we use binary logistic regression to determine whether an association between time, geography and five individual characteristics of the advertisers (sex, age, social status, religion and civil status) and the inclusion of specific romantic and instrumental preferences in the advertisements can be ascertained. We fit separate models for character, appearance, hobbies/interests, intimacy, financial status and social status. In this way, we want to investigate which factors were associated with the incorporation of these specific preferences in the advertisements.

Secondly, the status-attainment hypothesis is tested using descriptive statistics to illustrate a potential shift from the preferences ‘financial status’ – used as a proxy for ascribed status as financial status is mainly based on family capital and intergenerational transmission of wealth – to the criterion ‘educational level’ (achieved status).

Subsequently, binary logistic regression is applied to determine whether an association between time, geography and five individual characteristics of the advertisers (sex, age, social status, religion and civil status) and the inclusion of financial or educational preferences existed. The latter model aims to shed light on what was driving the transition from ascribed to achieved status characteristics in the mating market.

5. Results

First, we present the descriptive results on the romantic-love hypothesis, i.e. the transition from instrumental to romantic partner preferences. Figure 3(a) shows the percentage of the advertisements specifying instrumental preferences: financial status, social status, educational level, health, children and religion. Advertisers almost never included preferences regarding the geographic origin of the preferred partner. Therefore, this preference is not included in Figure 3(a). Figure 3(b) displays the romantic partner preferences: character, hobbies/interests, intimacy and appearance. During the nineteenth century, advertisers especially often specified preferences regarding the financial status (criteria regarding income and/or capital) of their potential partner. Furthermore, towards the end of the nineteenth century, preferences regarding social status increased considerably and remained stable until the mid-twentieth century. Whereas social status preferences increased during the nineteenth and early twentieth century, the importance of the preference ‘financial status’ decreased. Perhaps bluntly asking about one’s financial situation became a social taboo (earlier), while asking about social status was more accepted at times, as it is a more indirect way of getting insight into one’s financial situation. The increase in financial preferences during the 1880s and 1930s can be attributed to the agricultural crisis at the close of the nineteenth century and the impact of the Great Depression in the 1930s. These economic challenges amplified the importance of finding a financially stable match. Although the importance of the criterion ‘social status’ rapidly declined after World War II, it remained the most important preference in the advertisements until the 1970s. The criterion ‘religion’ shows an upsurge between 1920 and 1950, the heydays of pillarization in the Netherlands. Around 1880, coinciding with the emergence of the welfare state and an augmentation in real wage for men, preferences related to children started to decrease. The presence of children probably became less of a financial concern. Preferences in the category health, in our study references to the physical condition of the preferred partner excluding appearance (included in the group ‘romantic), as representation of the working and earning potential of someone, remained a seldom-specified criterion for partner selection throughout the entire research period. Perhaps articulating a preference for potential partners with a specific social and/or financial status was deemed adequate for conveying expectations regarding earning potential. As we did not classify personal characteristics of the advertisers other than sex, age, religion, civil status and social status, our data does not allow us to assess if disability was a topic that was taboo-ed.

Figure 3a. Instrumental preferences in advertisements per decade (% of N), 1841–1995.

Figure 3b. Romantic preferences in advertisements per decade (% of N), 1841–1995.

What stands out regarding the romantic preferences is that they were found throughout the entire research period, at least when it comes to appearance and character. The trend of appearance displays a peak between approximately 1860 and 1900. This pattern could be associated with the advent of photography around 1863, as requests for the inclusion of a photo in response to advertisements were consistently classified as a preference regarding appearance. Requests for a photo suddenly became a hype, that levelled off during World War I and the interwar period, including the Great Depression, only to resurge after World War II. We do not have a clear-cut explanation for the lower percentages of character between 1880 and 1940. While a surge in preferences regarding religion becomes apparent only during the heydays of religious pillarization between 1920 and 1950, the decrease in considerations regarding character from 1880 onwards may somehow be related to the onset of the pillarization process around the same period. It was during this period that clergy highlighted agreement in faith, rather than character traits, as an absolute necessity for a successful marriage (Van Tilburg, 1998, p. 24). Overall, preferences regarding character and appearance seem to have been of lesser importance from the late nineteenth-century till the mid-twentieth century. A systematic incline of all romantic considerations can be observed in the second half of the twentieth century. All in all, the results suggest that the shift from instrumental to romantic partner choice was a gradual process, that started – at least in the Netherlands – only really in middle of the twentieth century.

Figure 4 presents the total percentages ‘romantic’ and ‘instrumental’ partner preferences per decade. The graph shows that the share of romantic and instrumental preferences remained fairly stable until 1940, i.e. between roughly 20–30% (‘romantic’) and 70–80% (‘instrumental’). From then on, the percentage of instrumental preferences declined systematically whereas the percentage of romantic preferences increased steadily. Asides from a small setback in the 1950s, this trend continued during the second half of the twentieth century. The trend accelerated in the 1960s and seems to have stabilized slightly at the end of the twentieth century. In the 1970s, the percentage of ‘romantic’ is for the first time higher than that of ‘instrumental’, thus marking an important turning point. In contrast to what is often assumed by historians like Coontz (2005, pp. 161–176), these results suggest that the transition from instrumental to romantic partner choice was not yet internalized during the nineteenth century. These findings are in line with those of Borscheid (1986), who concluded for nineteenth-century Germany that love did not yet guide partner choice and that instrumental factors, such as money, social status and prestige determined who married whom. Our findings more or less align with those of Lippmann (2021), who argued that partner preferences before pairing in France remained relatively stable for a significant portion of the twentieth century, with noticeable shifts emerging only around the late 1960s.

Figure 4. Total percentages ‘romantic’ and ‘instrumental’ per decade (% of N), 1841–1955.

Table 2 shows the results of the linear regression analysis. Model 1 investigates the association between time and the prevalence of romantic content in the advertisements. As anticipated, a significant positive relation between time and the level of romance is observed. Consistent with both the existing literature and the descriptive statistics outlined in the previous section, the advertisements became progressively more romantic over time. This positive and significant relation persists when the two geographic variables, i.e. ‘distribution’ (national or regional) and ‘location of the publisher’ (province), are added in Model 2. While the effect of ‘distribution’ is not significant, noteworthy effects are observed for the variable ‘location of the publisher’. Compared to the reference category of advertisements from newspaper publishers located in Noord-Holland, those from Limburg, Noord-Brabant, Groningen, Friesland, Overijssel, Utrecht and Zuid-Holland were notably less romantic. However, upon inclusion of the individual characteristics of the advertisers in Model 3, both the variable ‘period of publication’ and ‘location of the publisher’ (except for Utrecht and Zuid-Holland) lose their significance. Among the newly added variables, three produced significant results: (1) advertisements from self-employed farmers and fishermen and unskilled workers and farm workers exhibited a significantly lower prevalence of romantic preferences compared to advertisements authored by the elite, (2) compared to advertisements from widowed people advertisements written by divorced and single people were more romantic, and (3) advertisers adhering to a religion other than Protestantism, Judaism or Catholicism demonstrated a lower prevalence of being romantic than the Protestant advertisers.

Table 2. Multivariate linear regression (dependent variable: percentage ‘romantic’).

		Model 1			Model 2			Model 3			Model 4
		Adjusted R square .007			Adjusted R square .023			Adjusted R square .097			Adjusted R square .099
		Stand. coefficients			Stand. coefficients			Stand. coefficients			Stand. coefficients
Multivariate linear regression		B	Std. error	Sign.	B	Std. error	Sign.	B	Std. error	Sign.	B	Std. error	Sign.
Time	Period of publication	.087	.000	<.001**	.108	.000	<.001**	.098	.001	.231	.173	.000	.000***
Geography	Distribution (reference category: Regional)
	National				−.001	.013	.972	.051	.049	.453	−.007	.029	.826
	Location of publisher (reference category: Noord-Holland)
	Limburg				−.061	.031	<.001***	−.053	.080	.492	−.039	.057	.134
	Noord-Brabant				−.041	.022	<.001***	−.094	.066	.204	−.029	.043	.229
	Zeeland				−.008	.144	.499	.008	.088	.909	−.007	.304	.783
	Groningen				−.034	.027	.008**	.008	.088	.909	−.074	.049	.007**
	Friesland				−.065	.027	<.001***	.076	.102	.215	−.065	.054	.012*
	Drenthe				−.017	.227	.134	.025	.306	.668	−.007	.429	.779
	Overijssel				−.024	.052	.038*	−.114	.140	.057+	−.075	.104	.002**
	Gelderland				−.019	.045	.115	−.089	.179	.134	−.035	.099	.144
	Utrecht				−.046	.038	<.001***	−.071	.128	.233	−.056	.080	.019*
	Zuid-Holland				−.112	.014	<.001***	−.172	.056	.007**	−.117	.030	.000***
Individual characteristics	Sex (reference category: Male)
	Female							−.027	.066	.676	−.075	.040	.005**
	Age							.004	.001	.952	−.054	.001	.075+
	Social status (reference category: Elite)
	Lower middle class							.004	.043	.947	−.025	.024	.316
	Skilled workers							.041	.107	.503	.002	.055	.918
	Self-employed farmers and fishermen							−.164	.066	.009**	−.132	.046	<.000***
	Unskilled workers and farm workers							−.151	.076	.015*	−.230	.039	<.000***
	Civil status (reference category: Widowed)
	Divorced							.189	.123	.002**
	Single							.133	.058	.033*
	Married/other							.033	.129	.587
	Religion (reference category: Protestant)
	Jewish							−.053	.133	.391
	Catholic							.088	.047	.246
	Other							−.126	.078	.083+

Significance: + < 0.1 * < 0.05 ** < 0.01 *** < 0.001.

Upon examining the explanatory power of the models, it is evident that the adjusted R square in Model 1 was 0.007, indicating that 0.7% of the variation in the level of romanticness was accounted for by the period of publication. This percentage increased to 2.3% in Model 2 and further expanded to 9.7% in Model 3. Subsequently, we display the model that exhibited the most robust fit considering all independent variables at our availability. This model, Model 4, shows an adjusted R square of 0.099. It includes the variables ‘period of publication’, ‘distribution’, ‘location of publisher’, ‘sex’, ‘age’ and ‘social status’. The key findings of the linear regression analysis become evident in Model 4, indicating that ‘period of publication’, ‘location of publisher’, ‘sex’, ‘age’ and ‘social status’ are significant predictors of the level of romance in Dutch marriage and contact advertisements published between 1841 and 1995. For predicting the level of romanticism, the other variables were deemed insignificant. Compared to the reference category of advertisements from newspaper publishers located in Noord-Holland, those from Groningen, Friesland, Overijssel, Utrecht and Zuid-Holland tended to be less romantic. As anticipated, advertisements authored by older or female advertisers were notably less romantic than those written by younger or male advertisers. Additionally, advertisements from farmers and fishermen and the unskilled workers and farm workers were significantly less romantic than those from the reference category of the elite.

Next, we discuss the results of the binary logistic regression models displayed in Table 3(a) (Model 1–4) and 3b (Model 5–6). Given that the variables ‘civil status’ and ‘religion’ did not yield any results in any of the models, we decided to exclude these variables from the analyses. Likewise, ‘age’ was excluded from Model 1 and ‘social status’ from Model 4 and 6.

Table 3a. Binary logistic regression (dependent variables: explicated preferences).

		Model 1: Preference Character			Model 2: Preference: Appearance			Model 3: Preference Hobbies			Model 4: Preference Intimacy
		Cox&Snell R square .076			Cox&Snell R square .069			Cox&Snell R square .075			Cox&Snell R square .048
Binary logistic regression		B	Std. error	Sign.	B	Std. error	Sign.	B	Std. error	Sign.	B	Std. error	Sign.
Time	Period of publication	.010	.002	<.001***	.006	.002	.002**	.037	.006	<.001***	.120	.013	<.001***
Geography	Distribution (reference category: National)
	Regional	-.260	.170	.126	.112	.142	.430	.093	.502	.853	-11.361	600.483	.985
	Location of publisher (reference category: Noord-Holland)
	Limburg	-.790	.473	.070+	.798	.285	.005**	-17.771	4618.562	.997	-5.592	2073.541	.998
	Noord-Brabant	-.254	.297	.392	.630	.229	.006**	-.814	1.078	.450	-12.362	1489.706	.993
	Zeeland	-19.881	28382.941	.999	.158	1.427	.912	-17.081	28337.514	1.000	.238	12023.010	1.000
	Groningen	-.269	.320	.400	.204	.247	.410	-.453	.786	.564	10.354	600.483	.986
	Friesland	-.297	.372	.425	.447	.274	.102	-.238	.890	.789	9.518	600.484	.987
	Drenthe	-19.852	40192.970	1.000	21.576	40192.969	1.000	-17.231	40192.970	1.000	-2.751	19350.354	1.000
	Overijssel	-19.613	9268.331	.998	-.172	.533	.747	-17.168	9337.983	.999	-2.280	4306.782	1.000
	Gelderland	-.367	.652	.574	.173	.482	.720	-17.605	8879.429	.998	-2.119	3736.734	1.000
	Utrecht	-2.024	1.031	.050+	-.089	.408	.827	-17.220	7187.513	.998	-14.832	2430.175	.995
	Zuid-Holland	.355	.177	.044*	.449	.150	.003**	.821	.412	.046*	-.567	.303	.062+
Individualcharacteristics	Sex (reference category: Female)
	Male	.594	.129	<.001***	.887	.204	<.001***	-.357	.372	.338	.767	.258	.003**
	Age				-.007	.003	.016	.003	.008	.719	-.008	.007	.276
	Social status (reference category: Unskilled workers and farm workers)
	Elite	1.557	.349	<.001***	1.335	.223	<.001***	.750	.640	.241
	Lower middle class	1.694	.347	<.001***	1.445	.223	<.001***	.896	.631	.155
	Skilled workers	1.581	.441	<.001***	1.119	.322	<.001***	1.109	.871	.175
	Self-employed farmers and fishermen	.827	.473	<.001***	.291	.309	.346	1.143	.828	.167

Significance: + < 0.1 * < 0.05 ** < 0.01 *** < 0.001

Model 1 shows the results related to the variable ‘character’. We found a positive effect of time: the later an advertisement was published the higher the odds were that it contained a preference regarding character. Advertisements from newspaper publishers located in Limburg and Utrecht demonstrated lower odds of containing a preference related to character compared to those published in Noord-Holland, while the opposite was the case in Zuid-Holland. Advertisements authored by males exhibited significantly greater odds of containing a preference related to character, compared to those written by females. Additionally, all social classes demonstrated notably higher odds of specifying preferences regarding the character of a potential partner compared to the reference category of the unskilled workers and farm workers.

Model 2 displays the outcomes related to the variable ‘appearance’. Over time, there was a slight but noteworthy increase in the odds of stipulating preferences regarding the appearance of the desired partner. Compared to the reference category of advertisements from newspaper publishers located in Noord-Holland, those from Limburg, Noord-Brabant and Zuid-Holland exhibited greater odds of including a preference related to appearance. Advertisements written by males demonstrated significantly higher odds of including preferences regarding a partner’s appearance compared to those written by females. Next, advertisements authored by the elite, the lower middle classes and skilled workers exhibited higher odds of containing specifications related to appearance when contrasted with the reference category of unskilled workers and farm workers.

Model 3 (‘hobbies/interests’) yielded significant results exclusively for the variables time and location of the publisher. The later an advertisement was published, the greater the odds that it included a specified preference regarding hobbies and interests of a potential partner. Advertisements from newspaper publishers located in Zuid-Holland demonstrated higher odds of containing this preference compared to those published in Noord-Holland,

Model 4 pertains to the variable ‘intimacy’. Notably, specifications concerning intimacy increased over time, and advertisements written by males exhibited higher odds of containing a specification regarding intimacy compared to those authored by females. Compared to the reference category of advertisements from newspaper publishers located in Noord-Holland, those from Zuid-Holland exhibited lower odds of including this preference.

Table 3b. Binary logistic regression (dependent variables: explicated preferences).

		Model 5: PreferenceFinancial status			Model 6: Preference Social status
		Cox&Snell R square .247			Cox&Snell R square .080
Binary logistic regression		B	Std. error	Sign.	B	Std. error	Sign.
Time	Period of publication	−.042	.003	<.001***	−.010	.001	<.001***
Geography	Distribution (reference category: National)
	Regional	−.694	.170	<.001***	.481	.063	<.001***
	Location of publisher (reference category: Noord-Holland)
	Limburg	.871	.297	.003**	−.370	.142	.009**
	Noord-Brabant	.304	.244	.214	.433	.121	<.001***
	Zeeland	−20.450	27692.506	.999	.560	.797	.483
	Groningen	.965	.268	<.001***	−.277	.126	.071+
	Friesland	.424	.301	.159	−.417	.123	<.001***
	Drenthe	21.778	40192.969	1.000	.609	1.157	.599
	Overijssel	1.119	.533	.024*	.102	.252	.684
	Gelderland	.684	.512	.182	.662	.253	.009**
	Utrecht	.212	.432	.624	.247	.193	.199
	Zuid-Holland	.499	.166	.003**	−.244	.068	<.001***
Individual characteristics	Sex (reference category: Female)
	Male	.168	.222	.449	−1.002	.065	<.001***
	Age	.011	.003	<.001***	−.001	.001	.454
	Social status (reference category: Unskilled workers and farm workers)
	Elite	.989	.230	<.001***
	Lower middle class	.694	.233	.003**
	Skilled workers	.621	.351	.077+
	Self-employed farmers and fishermen	.459	.314	.144

Significance: + < 0.1 * < 0.05 ** < 0.01 *** < 0.001

Model 5 in Table 3(b) focusing on ‘financial status’ exhibits a relatively high explanatory value, as indicated by the Cox & Snell R Square. Significant results were found for most variables. Over time, there was a significant decline in the odds of including specifications on the financial status of a partner. Advertisements in regional newspapers exhibited lower odds of featuring preferences related to the financial status of a partner compared to those published in national newspapers. Furthermore, advertisements in newspapers published in Limburg, Groningen, Overijssel and Zuid-Holland demonstrated higher odds of containing preferences regarding the financial status of a potential partner compared to those placed in newspapers published in Noord-Holland. Moreover, the older an advertiser the greater the odds that preferences about the financial status of the desired partner were specified. Finally, advertisers from the elite and lower middle class, and skilled workers exhibited higher odds of inquiring about the financial status of a partner compared to advertisers from the reference category of unskilled workers and farm workers.

The final model, Model 6 in Table 3(b), focuses on preferences regarding the social status of a partner. Such specifications were less likely to be found in advertisements from later years compared to advertisements from earlier periods. Contrary to financial status, these specifications were more likely to be present in regional newspapers than in national newspapers. Furthermore, a criterion regarding social status was more likely to be found in newspapers based in Noord-Brabant and Gelderland than in newspapers from Noord-Holland. Conversely, the opposite effect was observed for advertisements from newspapers published in Limburg, Groningen, Friesland, Gelderland and Zuid-Holland: here the odds were lower for a social status specification in an advertisement compared to those from newspapers from Noord-Holland. Finally, male advertisers were less likely to inquire about the social status of a partner compared to their female peers.

Subsequently, the results of the testing of the status-attainment hypothesis are discussed. As illustrated in Figure 3(a), the pattern of ‘educational level’ displays an opposite trend compared to the other instrumental preferences. This might suggest a shift from ascribed status (for instance, financial status) characteristics to achieved characteristics (education). However, although there was a declining trend of ‘financial status’ during the second half of the twentieth century, the upward trend in ‘educational level’ only became prominent in the last quarter of the twentieth century. Consequently, these results do not provide strong support for the status-attainment hypothesis, only a mere indication. The increase of the preferences regarding educational level is much lower and later than one would expect based on the status-attainment hypothesis, considering the fact that the Netherlands started to modernize from the late nineteenth century onward. This finding is, however, interesting as it suggests that the changes in educational assortative mating observed by scholars are only to a limited extent the result of changes in partner preferences regarding educational level. Grow et al. (2015), for instance, observed a shift from homogamy combined with female hypergamy and male hypogamy, which was dominant for most of the twentieth century, to a decrease of homogamy and the share of hypergamic couples (couples in which the woman has lower educational attainment) and an increase in the shares of hypogamic couples (couples in which women have higher educational attainment). Our findings are in line with those of Grow et al. (2015) who demonstrated that changes in educational assortative mating can be explained without any change in male or female partner preferences. A more important factor might be related to meeting opportunities. Thanks to the increase of women in higher education and the labour market, partners started to meet each other increasingly in college or on the work floor. Only later on, individuals’ educational status turned into a partner preference.

Finally, binary logistic regression models regarding educational preferences of a partner were fitted. We display in Table 4 the model with the highest explanatory value. Again, the variables ‘civil status’ and ‘religion’ did not add strength to the model and were therefore dropped. The model reveals that the variables ‘period of publication’ and ‘sex’ are important for explaining differences in the inclusion of educational preferences in the advertisements from the period under study. Over time, there was a significant increase in the odds of specifying preferences about the educational status of a partner. Next, the relation of ‘sex’ with ‘educational level’ is significant and negative, meaning that men were less inclined to include preferences regarding education in their advertisements compared to women. This outcome supports the idea that despite the fact that women became less dependent on men for their economic well-being in the period under study, instrumental preferences, which provided an insight -directly or indirectly- in the socioeconomic position of the preferred partner were still valued more by women.

Table 4. Binary logistic regression (dependent variable: preference ‘education’).

		Model 1: Preference Education
		Cox&Snell R square .049
Binary logistic regression		B	Std. error	Sign.
Time	Period of publication	.032	.009	<.001**
Geography	Distribution (reference category: National)
	Regional	.961	.721	.182
	Location of publisher (reference category: Noord-Holland)
	Limburg	−17.191	4381.760	.997
	Noord-Brabant	.360	.879	.682
	Zeeland	−16.559	28346.650	1.000
	Groningen	−17.405	3398.077	.996
	Friesland	−16.717	4032.140	.997
	Drenthe	−16.673	40192.970	1,000
	Overijssel	−16.956	8590.384	.998
	Gelderland	−16.776	8623.088	.998
	Utrecht	−16.242	6951.314	.998
	Zuid-Holland	.300	.290	.590
Individual characteristics	Sex (reference category: Female)
	Male	−2.414	.438	<.001**
	Age	.004	.013	.778
	Social status (reference category: Unskilled workers and farm workers)
	Elite	16.751	2997.854	.996
	Lower middle class	16.968	2997.854	.995
	Skilled workers	17.446	2997.854	.995
	Self-employed farmers and fishermen	1.298	4537.365	1.000

Significance: + < 0.1 * < 0.05 ** < 0.01 *** < 0.001

6. Conclusion and discussion

The shift from instrumental to romantic partner choice was a gradual process in the Netherlands that started only really in the twentieth century, more than half a century later than the modernization processes had set in. Our results show that first financial status preferences declined, followed by social status criteria. Only in the second half of the twentieth century, the percentage of romantic preferences increased steadily and systematically while instrumental characteristics continued to decline. Contrary to what historians like Coontz (2005) argued, the ideal of romantic partner choice was not yet internalized during the nineteenth century, at least not in the Netherlands. Until the middle of the twentieth century the quest for a partner was mostly guided by such instrumental partner preferences as the financial and social status of a future spouse. These findings suggest that the effects of modernization on partner choice were not immediate but gradual. Only with the ongoing process of individualisation and the expansion of the welfare state in the second half of the twentieth century (more specifically the introduction of the Social Security Act in 1965, which ensured a basic level of welfare for everyone in the Netherlands), romantic partner preferences gained the upper hand. Now, the advertisers, especially women and people from the lower social classes, could also prioritize romantic preferences, as the economic risks associated with an unfortunate economic match were mitigated by the introduction of a social safety net. Given the timing of the shift from instrumental to romantic partner choice, the transition also seems to be related to cultural changes regarding marriage and partner choice and intergenerational changes in the wake of the Second Demographic Transition.

Besides the period of publication, the sex and social status of the advertiser were important for explaining variations in the degree of romance of the advertisements (Tables 2). With regard to the latter, we found – in line with Maslow’s theory on the hierarchy of needs (1943) and literature (e.g. Borscheid, 1986) – that advertisers from higher social classes tended to be more romantic. These advertisers did not have to worry about the bare essentials of life and could therefore specify partner preferences in the realm of ‘love and intimacy’. Moreover, more well-to-do people were in a position to specify certain standards for a partner because they were more desired candidates in the mating market themselves. The results from the binary logistic regression (Table 3(a)/Table 3(b)) indeed suggest that advertisers from higher social classes overall tended to set higher standards for a partner: their advertisements had higher odds of containing romantic partner preferences regarding character and appearance, as well as preferences regarding the instrumental partner criterion ‘financial status’ compared to the reference category of unskilled workers and farm workers. This implies that romantic and instrumental partner preferences can certainly coexist. Advertisers from the higher social classes were looking for real love, but within their own social class.

Furthermore, our results indeed demonstrate that men and women valued different characteristics in a partner of the opposite sex (Greenlees & McGrew, 1994). The linear regression (Table 2) revealed that the advertisements by female advertisers were less romantic than those by their male peers. The internalization of the love-match ideal by female advertisers was probably hampered – at least in the Netherlands – by country-specific circumstances, such as the high percentages of Dutch women in part-time jobs. In addition, the binary logistic regression (Tables 3(a–b) and 4) revealed that advertisements by female advertisers had greater odds containing preferences regarding social status and educational level compared to those written by male advertisers. Advertisements by men, on the other hand, had higher odds of having specifications about a partner’s appearance, character and intimacy compared to those written by women. These results are in line with the findings of Van Poppel et al. (2001). We argue that these sex differences were mainly the result of gender role ideas. As women were dependent on their husbands for their social status and financial security for most of the research period, instrumental partner preferences naturally also played a bigger role for women when looking for a partner. Men were on the other hand – following Maslow’s theory (1943) – as breadwinners more likely to fulfil their own basic physiological and material needs and were therefore able to focus more on higher order needs when looking for a partner. For future research it would be interesting to further delve into these sex differences and their transition over time. Based on our current conclusion, we expect a convergence of male and female partner preferences in the second half of the twentieth century with the spread of more egalitarian gender role ideas and the increase of female labour market participation. However, we anticipate this convergence to be smaller for the Netherlands compared to neighbouring European countries as a result of the relatively high share of part-time working women in the Netherlands in the second half of the twentieth century.

Although modernization processes, such as industrialization, urbanization, educational expansion, the rise of wage labour, mass communication and transportation, first hit the mainly Protestant western provinces, stimulating individual autonomy and a search for personal happiness in partner choice, religion, as well as civil status did not play an important role in the internalization of the love-based partner choice in the Dutch advertisements. Furthermore, although some significant results were found in the binary logistic regression for newspaper distribution and location of the publisher, for example with regard to financial status preferences and social status preferences, and in the linear regression for location of the publisher, no clear geographical pattern could be discerned, certainly not a pattern that is in line with the romantic love hypothesis.

The results with regard to the status-attainment hypothesis confirm that references to education (achieved status) indeed increased over time, whereas references to financial status (ascribed status) decreased over time. Also, the binary logistic regression (Table 4) revealed that over time there was a significant increase in the odds of specifying preferences about the educational level of a partner. However, the growth of partner preferences regarding educational level was much lower and started much later – in the last quarter of the twentieth century – than expected based on the status-attainment hypothesis. Therefore, we conclude, in line with Grow et al. (2015), that the trend towards educational homogamy is only to a limited extent the result of changes in partner preferences, but rather an effect of changed meeting opportunities resulting from an increase of women in higher education and the labour market. Asides from the period of publication, the advertiser’s sex is also important in explaining differences in the inclusion of educational preferences in advertisements. Men were less inclined to include preferences regarding education in their advertisements compared to women. This again supports the notion that despite changing gender roles and increasing female participation in higher education, women still valued more highly the socioeconomic position – for which educational level is an indirect measure – of the preferred partner compared to men. Once again, this is likely related to the high share of part-time working women in the Netherlands during the second half of the twentieth century: many Dutch women were still largely financially dependent on their partner.

All in all, our results deviate from both the romantic-love hypothesis and the status-attainment hypothesis. The increase of romantic partner preferences and the increase in achieved status preferences occurred late. At best, our results point to an adjusted form of the combined hypothesis (Smits et al., 1998). First, in the early stages of modernization, instrumental preferences and ascribed status gradually lost their importance in partner choice. Only in the second half of the twentieth century, in the wake of ongoing individualisation and the blooming of the welfare state and the Second Demographic Transition, romantic partner preferences surged. Lastly, this trend is followed by an – albeit limited – increase of the importance of achieved status (i.e. education).

Strengthened by the analyses of this article and the results it yielded, we are convinced that matrimonial and contact advertisements are a treasure for (historical) demographers and family historians, that allow us to test theories related to partner selection and human mating in a much more direct way compared to traditional demographic sources. The analyses presented in this article are just the tip of the iceberg of the possibilities. In this article, we only focused on what the advertisers were looking for in a partner in a binary way. For future research it would also be interesting to focus on the self-presentation of the advertisers and the meeting strategies they suggested in the advertisements. Furthermore, in the analyses presented here, we have neglected the textual aspects of the advertisements that make them so interesting. Last but not least, for future research a more qualitative in-depth approach or the use of natural language processing could be applied in order to analyse trends in language use.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction http://dx.doi.org/10.1080/1081602X.2024.2347052

Additional information

Funding

The authors are thankful for the support from HiDO - Scientific Network for Historical Demography (Research Foundation Flanders, Application number [W001821N]).

Notes

1. In Figure 1, all advertisers that did not specify their civil status were considered to be single.

2. Website Centraal Bureau voor de Statistiek (CBS), Statline, ‘Consumentenprijzen; gemiddelde prijzen voedingsmiddelen, 1800–2018’ (Consumer prices; average food prices, 1800–2018), StatLine – Consumentenprijzen; gemiddelde prijzen voedingsmiddelen, 1800–2018 (cbs.nl) (consulted 7 February 2024).

3. Delpher, Delpher » Kranten, Boeken & Tijdschriften

4. The software was developed by Ton van Raaij (Radboud University Nijmegen).

5. For this study, we used HISCO release 2020. See also: Van Leeuwen et al. (2002). HISCO. Historical International Standard Classification of Occupations. Leuven; Van Leeuwen and Maas (2011). HISCLASS. A Historical International Social Class Scheme. Leuven.

6. Binary logistic regression: 1 = specific romantic preference was included in the advertisement, 0 = specific romantic preference was not included in the advertisement.

7. Abbreviation for ‘young man’.

8. Abbreviation for ‘woman’.

9. Abbreviation for ‘young woman’.

10. Dutch abbreviation for ‘seeking to meet with’.

11. Dutch abbreviation for ‘seeking to urgently meet with’.

12. Kranten » Ruim 17.000.000 Nederlandstalige pagina’s | Delpher (website consulted January 2024).

13. Nieuw in Delpher | Delpher (website consulted January 2024).

Appendix 1

Search queries and database

For sampling the online collection of newspapers from Delpher, curated by the Dutch Royal Library (Koninklijke Bibliotheek), searches were conducted using the queries presented in Table A1.

Table A1. Overview of search queries used to construct the dataset.

Queries	English translation of queries
Huwel*k AND kennis AND dame	Marr*ge AND meet AND lady
Huwel*k AND kennis AND jongedame	Marr*ge AND meet AND young lady
Huwel*k AND kennis AND vrouw	Marr*ge AND meet AND woman
Huwel*k AND kennis AND heer	Marr*ge AND meet AND gentleman
Huwel*k AND kennis AND jongeman	Marr*ge AND meet AND young man
((man OR vrouw OR jong OR meisje OR weduwe OR weduwnaar OR j.m. OR vr. OR j.vr.) AND (z.k.m. OR z.s.k.m. OR contact OR kontact OR kennism*) AND (huwelijk OR Huwelijksadvertentie OR contactadvertentie))	((man OR woman OR young OR girl OR widow OR widower OR. y.m.^7 OR w.^8 OR y.w.^9) AND (z.k.m^10 OR z.s.k.m.^11 OR contact OR kontact OR meet*) AND (marriage OR marriage advertisement OR contact advertisement))
((weduwen*) AND (z.k.m. OR z.s.k.m. OR contact OR kontact OR kennism*) AND (huwelijk OR relatie OR partner OR samenwonen OR huwelijken OR huwelijksadvertentie* OR contactadvertenties*))	((widow*) AND (z.k.m. OR z.s.k.m. OR contact OR kontact OR meet*) AND (marriage OR relation OR partner OR cohabitation OR marriages OR marriage advertisement* OR contact advertisement*))

The selection and testing of the search queries were carried out using a smaller dataset built for two unpublished pilot studies, following the approach described below. The typical structure of matrimonial and contact advertisements, as outlined by Lippmann (2021, p. 841), resolves around the supply and demand sides of these advertisements, along with the delimiter. Delimiters are words or combinations of words linking the advertiser’s self-description (supply side) with partner preferences and advertisement goals (demand side). For the pilot studies, we identified the most frequently used words (nouns) representing the actors from the supply side (advertisers) and the potential partners from the demand side, synonyms for ‘man’ and ‘woman’, as well as the most common delimiters and advertisement goals. Employing ‘AND’ or ‘OR’ allowed us to translate this information into various search queries.

It is important to acknowledge that our tool may not have selected advertisements with less commonly applied delimiters or lesser-used words for actors, preferred partners and advertisement goals. Additionally, the software used by Delpher may have introduced errors in the OCR text due to inconsistencies in lay-out and typography in older newspapers, potentially affecting the nineteenth- and early twentieth-century results generated by our tool. Given that our database encompasses advertisements from a substantial number of newspapers (see Appendix 3) spanning approximately one-and-a-half centuries, we cannot define the exact count of marriage and contact advertisements that were not identified by our tool (false negatives, type II errors). Therefore, we do not assert that our dataset comprehensively covers complete series of digitized newspapers accessible in the Delpher collection during the database construction in 2021–2022.

After testing, we concluded that the chances of false positives (type I errors), advertisements not being a matrimonial or contact advertisement, or double hits, and the insertion of identical advertisements, were minimal. We thoroughly examined the dataset for temporal and geographical coverage, coverage by gender and social status, and the inclusion of national newspapers representing the main ideological pillars in the period under study. Although the breakdown in percentages of newspapers with regional and national circulation (55 vs. 45%) is rather balanced, we identified gaps at the provincial level. Besides this, the total number of female advertisers is relatively small compared to the number of male advertisers (24 vs. 76%), especially in the nineteenth century. This probably reflects the actual situation; specific provinces and women were simply underrepresented in marriage and contact advertisements in that period, the latter consistent with observations by other researchers (e.g. Stokvis, 2011). It is important to emphasize that the regression models are controlled for sex and distribution (provinces).

Furthermore, 22% of the advertisers, not unexpectedly predominantly male, disclosed their occupation, enabling us to determine their social status. The fact that only a small number of women specified their occupation aligns with literature highlighting that for the greater part of the research period, women were increasingly relegated to domestic roles, assuming responsibilities as housekeepers and guardians of moral integrity (Coontz, 2005; Puschmann, 2020. Furthermore, we observed relatively high percentages of advertisers from Hisclass 1 (Elite) and 2 (Lower Middle Class) in the database compared to the other classes. Again, this may reflect the actual trend. The regression models are controlled for social class.

No major issues were found about the temporal coverage and the inclusion of national newspapers representing the main ideological pillars in the research period.

Delpher provides online information about the newspapers in its collection and the corresponding timespans.¹² Individual newspapers cover quite different periods, and none spans the full period under study. Moreover, as newspaper digitization is an ongoing project for Delpher, their online collection is subject to changes, with missing editions and newspapers regularly added.¹³ Replication of the findings using our tool and the Delpher collection is possible, taking into account the above.

Appendix 2.

Classification of personal characteristics and partner preferences

A. Classification of personal characteristics from the advertisers

Table

Characteristic	Options in database	Remarks
Sex	Male Female
Age	Actual age (number) Not mentioned	All references to age besides actual numbers were disregarded. When an age range was given, for instance 20–30, we took the average age, in this case 25.
Religion	Protestant Catholic Jewish Other Not mentioned	When advertisers expressed a preference for a partner belonging to a particular religious denomination, we assigned that denomination to the advertiser, assuming that in those cases individuals only searched for a partner of the same religion. This is particularly during the period of religious pillarization in the Netherlands (from the late nineteenth century till the end of the 1960s) a safe assumption.
Civil status	Single Divorced Widowed Married Other Not mentioned	According to advertisers’ specifications.
Occupation	Actual occupation Not mentioned	In order to determine the social status of the advertiser, we employed the HISCO coding system to categorize their occupation. Each occupation was then classified according to HISCLASS_5: 1 = Elite 2 = Lower middle class 3 = Skilled workers 4 = Self-employed farmers and fishermen 5 = Unskilled workers and farm workers −1 = No code in HISCLASS-5 (primarily because occupation was not included in the advertisement)

B. Classification of partner preferencesTwo pilot studies showed that the partner preferences included in the matrimonial and contact advertisements published between 1841 and 1995 could be classified by creating the following eleven categories.

Table

Category	Remarks
Instrumental preferences
Financial status	References to wealth and income.
Social status	Direct references to specific social class, as well as more subtle, indirect references, e.g. a preference for a ‘Lady’.
Educational level	References to specific education or educational level.
Health	References to the physical condition of a potential partner (except for appearance), as representation of their working and earning potential.
Children	References to the desirability of a partner with(out) children.
Geographic origin	Preference for potential partners from a particular regional area or references to the willingness to relocate to a specific region in- and outside the Netherlands.
Religion	Preferences for a specific religious denomination.
‘Romantic’ (non-instrumental) preferences
Appearance	References to appearance/looks. Asking for a photo from the potential partner is considered a reference to appearance.
Intimacy	References to romance and sex, and other forms of intimacy.
Hobbies/interests	References to hobbies and other interests associated with the spending of free time.
Character	References to personal traits.

A codebook was developed to regulate consistent classification of preferences on the level of individual words.

C. Coding of partner preferences

Table

Category	Options in database
Instrumental preferences
Financial status	Yes No	‘yes’ (advertiser set positive or negative condition) or ‘no’ (advertiser did not set condition).
Social status	Yes No	Ditto
Educational level	Yes No	Ditto
Health	Yes No	Ditto
Children	Yes No	Ditto
Geographic origin	Yes No	Ditto
Religion	Yes No	Ditto
‘Romantic’ (non-instrumental) preferences
Appearance	Yes No	Ditto
Intimacy	Yes No	Ditto
Hobbies/interests	Yes No	Ditto
Character	Yes No	Ditto

Appendix 3

Table

Newspaper	Number of advertisements in dataset
Het nieuws van den dag: kleine courant	1249
Algemeen Handelsblad	914
Rotterdamsch nieuwsblad	783
Het Parool	457
De Telegraaf	370
Nieuwsblad van het Noorden	354
Het vrije volk: democratisch-socialistisch dagblad	284
De Volkskrant	280
Nieuw Israëlitisch weekblad	217
Trouw	210
Opregte Haarlemsche Courant	206
De Tijd: godsdienstig-staatkundig dagblad	199
Haagsche courant	196
Provinciale Noordbrabantsche en ‘s Hertogenbossche courant	154
Nieuwsblad van Friesland: Hepkema’s courant	143
Nieuwe Rotterdamsche Courant	124
Limburger koerier: provinciaal dagblad	120
Leeuwarder courant: hoofdblad van Friesland	111
De courant Het nieuws van den dag	91
De Gooi- en Eemlander: nieuws- en advertentieblad	82
Leeuwarder courant	82
Limburgsch dagblad	68
Arnhemsche courant	65
NRC Handelsblad	52
Nieuwe Venlosche courant	45
Het Vaderland: staat- en letterkundig nieuwsblad	43
Het Centrum	41
Nieuwe Tilburgsche Courant	38
De waarheid	32
Provinciale Overijsselsche en Zwolsche courant	28
Twentsch dagblad Tubantia en Enschedesche courant	26
Eindhovensch dagblad	19
Twentsch nieuwsblad	19
De Graafschap-bode: nieuws- en advertentieblad voor stad- en ambt-Doetinchem, Hummelo en Keppel, Wehl, Zeddam, ‘s Heerenberg, Ulft, Gendringen, Sillevolde, Terborg, Varsseveld, Dinxperlo, Aalten, Breedevoorde, Lichtenvoorde, Groenlo, Neede, Eibergen, Bor	15
Nieuwe Venlosche courant: voorheen Venloosch nieuwsblad – Venloosche courant	15
De Maasbode	14
Leeuwarder nieuwsblad: goedkoop advertentieblad	13
Friese koerier: onafhankelijk dagblad voor Friesland en aangrenzende gebieden	12
Leidsch dagblad	12
Limburger koerier	12
Nieuwe Apeldoornsche courant	12
Dagblad van het Zuiden voor Eindhoven, Meierij, Peel en Kempenland	11
Dagblad van Rotterdam	11
De tijd: dagblad voor Nederland	11
Delftsche courant	10
Het volk: dagblad voor de arbeiderspart?	10
Middelburgsche courant	10
Utrechtsche courant	10
De Tijd De Maasbode	9
Nieuwsblad van het Zuiden: dagblad met ochtend- en avond-editie	9
Provinciale Geldersche courant: Nijmeegsche courant	9
Dordrechtsche courant	8
Tilburgsche courant	8
Dagblad van Noord-Brabant	7
De Noord-Ooster	7
Nieuw Amsterdamsch handels- en effectenblad	7
Oprechte Haarlemsche courant	7
De Zuid-Willemsvaart	6
Delftsche courant: nieuwsblad voor Delft en Delfland	6
Noordbrabantsch dagblad het huisgezin	6
Dagblad van Zuidholland en ’s Gravenhage	5
De Amersfoortsche courant: dagblad voor Amersfoort en omstreken	5
Het joodsche weekblad: uitgave van den Joodschen Raad voor Amsterdam	5
Utrechtsche courant: nieuws- en advertentieblad	5
Advertentieblad van het departement van de Wester-Eems	4
Dagblad voor Noord-Holland	4
De Heerenveensche koerier: onafhankelijk dagblad voor Midden-Zuid-Oost-Friesland en Noord-Overijssel	4
Haarlem’s dagblad	4
Provinciale Geldersche courant	4
Voorwaarts: sociaal-democratisch dagblad	4
Bredasche courant	3
De nieuwe koerier: Maas- en Roerbode: provinciaal Limburgsch dagblad	3
Het dagblad voor IJmuiden – Velsen – Beverwijk – Driehuis – Santpoort	3
Nieuwe Brabantsche courant	3
Provinciale Overijsselsche en Zwolsche courant: staats-, handels-, nieuws- en advertentieblad	3
Dagblad van het Oosten : Almelo’s dagblad	2
De Gooi- en Eemlander	2
De Tijd	2
Drentsch dagblad: officieel orgaan voor de provincie Drenthe	2
Friesche courant	2
Nederlands dagblad: gereformeerd gezinsblad/hoofdred. P. Jongeling … [et al.]	2
Nieuwe Apeldoornsche courant: dagblad voor Apeldoorn en de Veluwe	2
Provinciale Drentsche en Asser courant	2
Recht voor allen	2
Utrechtsche provinciale en stads-courant: algemeen advertentieblad	2
Alkmaarsche courant	1
Apeldoornsche courant	1
Dagblad voor Leiden en omstreken	1
De patriot	1
De residentiebode	1
De Zuid-Willemsvaart: nieuws- en advertentieblad voor handel en landbouw	1
Eindhovensche en Meierijsche courant	1
Friesch dagblad	1
Helmondsch dagblad	1
Leydse courant	1
Nieuwsblad van het Zuiden	1
Opregte Steenwijker courant	1
Roomsch leven	1
Rotterdamsche courant	1
Tubantia	1
Utrechts volksblad: sociaal-democratisch dagblad	1
Utrechtsch katholiek dagblad	1
Utrechtsch provinciaal en stedelijk dagblad: algemeen advertentieblad	1
Venloosche courant	1
Veritas: katholiek 14-daagsch blad voor Maastricht	1
Voorwaarts	1
Total	7469