The patriarchy index: a comparative study of power relations across historical Europe

This article stands at the confluence of three streams of historical social science analysis: the sociological study of power relations within the family, the regional demography of historical Europe, and the study of spatial patterning of historical family forms in Europe. It is a first exercise in the design and application of a new ‘master variable’ for cross-cultural studies of family organization and relations. This indexed composite measure, which the authors call the Index of Patriarchy, incorporates a range of variables related to familial behaviour, including nuptiality and age at marriage, living arrangements, post-marital residence, power relations within domestic groups, the position of the aged, and the sex of the offspring. The index combines all these items, with each being given equal weight in the calculation of the final score, which represents the varying degrees of sex- and age-related social inequality (‘patriarchal bias’) in different societal and familial settings. In order to explore the comparative advantages of the index, the authors use information from census and census-like microdata for 91 regions of historical Europe covering more than 700,000 individuals living in 143,000 domestic groups, from the Atlantic to the Urals. The index allows the authors to identify regions with different degrees of patriarchy within a single country, across the regions of a single country, or across and within many broader zones of historical Europe. The unprecedented patterning of the many elements of power relations and agency contained in the index generates new ways of accounting for both the geographies and the histories of family organization across the European landmass.

asserted that the crux of family system analysis should be the contours of parental authority, the latter emphasized the effects of inheritance patterns. The discussion was given a new impetus in the studies of Ruggles, who argued that family structure and living arrangements could be most profitably analyzed from the perspective of the elderly. Focusing on the co-residence of the elderly not only minimizes the effects of variation in demographic conditions on indicators of family structure (Ruggles, 2009(Ruggles, , 2010(Ruggles, , 2012; it can also shed light on contrasting systems of social security and family welfare provision (Cain, 1991;Laslett, 1988;Szołtysek, 2012a).
Members of the Eurasia Project on Population and Family History recently went even further by stipulating that one of the most important characteristics which distinguishes various family systems is the sequence of individual life-course transitions (Dribe, Manfredini, & Oris, 2007). Recently, Kok (2009) added yet another building block to these theoretical considerations by looking at family systems through the prism of illegitimacy patterns.
Meanwhile, Wall (1995;earlier Laslett, 1983) pushed the discussion forward by investigating 'domestic co-residence' using a matrix of 'statuses', 'functions', and 'relationships'. According to Wall (1995), this matrix provides a comprehensive account of the various 'attributes' of family systems, including the welfare capability of the family, the household as a work unit, the status of women within the family, the patterns of marriage and household formation, the household as a kin group, and inequalities between households. For each attribute of the family system, Wall (1995, pp. 21-30) proposed a range of measures, along with a description of the target population for each of these measures.
While the approaches mentioned above suggest that there is a wide range of angles from which family systems may be analyzed, none appear to be fully adequate when the goal is to measure differences in familial organization on a large comparative scale. While many scholars have made valuable proposals for measuring family systems across time and space, each of these approaches tends to favour one aspect of the family system, while neglecting the others. For example, although Wall's 'disaggregation' of the family system into its constitutive elements (29 variables in total) is conceptually very useful, this approach cannot be easily scaled to measure the family system characteristics of multiple societies. If, as has been suggested, variation in family organization in different societies implies the coexistence of a number of different elements in many different permutations and combinations, then Wall's approach cannot tell us how to classify various societal family constellations (for similar dilemmas in cross-cultural research, see Whyte, 1978).
Technically speaking, most of the developments in the measurement of historical family systems reviewed so far stemmed from and were designed for studies of a single community or a small group of communities (Ruggles, 2012). However, the ongoing revolution in the availability of census and census-like microdata across time and space (Ruggles, 2012) opens up unprecedented opportunities for a revitalization of the family system debate in historical demography. This, however, requires us to develop new approaches and new tools. 1 We argue that the only solution to such challenges is to design a 'master variable' which can be employed in cross-cultural studies of family systems by applying it to harmonized data sets covering multiple settings. This measure has to be: 1. holistic -it has to capture critical aspects of familial behaviour without being overloaded; 2. feasible -it has to be easily derived from historical census-like microdata with often limited information; 3. quantifiable -it must be possible to calculate it from basic numerical variables derived from individual-level sources; and 4. comparable -it must yield quantities that can be easily compared across time and space, and between societies.
In order to meet these requirements, we suggest an indexed composite measure that incorporates a selection of variables related to familial behaviour. We call this measure the Index of Patriarchy. The index is based on a wide range of variables pertaining to the spheres of nuptiality and age at marriage, living arrangements, post-marital residence, power relations within the domestic group, the position of the aged, and the sex of the offspring. Our measuring device combines all of these items in order to facilitate analyses of the complex reality of family systems. The different items that constitute the index are given equal weight in the calculation of the final scores, which should reflect the varying degrees of patriarchal bias in various societal and familial settings.
In order to explore the comparative advantages of using this index, we drew on census and census-like microdata for 91 regions of historical Europe covering more than 700,000 individuals living in 143,000 domestic groups, from the Atlantic to the Urals. The data used in this study were collected within the Mosaic project, which was started in 2011 at the Max Planck Institute for Demographic Research in Rostock 2 , itself using the experiences of a global community of researchers involved in international data infrastructure projects like Integrated Public Use Microdata Series (IPUMS) and The North Atlantic Population Project (NAPP) (see Szołtysek & Gruber, in press). The project gathers, harmonizes and distributes (openly) surviving census and census-like materials from historical Europe. The microdata samples included in the Mosaic database are very similar in terms of structure, organization and available information. In each case, the listings describe the characteristics of the individuals grouped into households, through which the interrelationships of the individuals within the households can be determined. All of the census samples define a household as a group of people sharing a place of residence. There is a core set of variables common to virtually all of the data sets, including the relationship of each individual to the household head, and each inhabitant's age, sex and marital status. In addition, there are many important variables available for subsets of the censuses, such as occupation, birthplace, year of immigration, religion and ethnicity (Szołtysek & Gruber, in press). As the Mosaic data files have a common data format based on the standards established by IPUMS and NAPP, they can easily be compared with information from these databases in the future.
For all of the available data sets from the Mosaic project, we have computed a list of wellspecified variables. Our Index of Patriarchy is based on these variables. The index allows us to identify regions with different degrees of patriarchy within a single country, across the regions of a single country, and across and within broader zones of historical Europe.
In what follows, we first discuss the elements of patriarchy and present a list of variables for measuring those elements. These variables are then compared to each other in order to determine whether they correlate to each other. Next, we present our Index of Patriarchy alongside the preliminary results of our application of the index across multiple spatial settings in historical Europe. At the end of the article, we discuss the implications for comparative research on historical family systems.
In this first report, the index is applied only to historical European data. Although we hope that we shall be able to deal with non-European and contemporary data in the future, these further applications -as one of the anonymous reviewers of our work remarkedare likely to pose challenges sufficiently specific to warrant their separate discussion. Also, The patriarchy index the geographical coverage of Europe itself remains incomplete, as no data from England, the Netherlands, Scandinavia and Spain were included.
Although there is the temptation to delay the current publication in order to have more comprehensive European coverage (but what would be the data threshold we should stop at?), we have decided to resist this. Given that, in principle, the article represents an exercise in the design and first application of the new index, in order to explore its comparative advantages it is crucial that the data employed contain enough variation to show the differential operation of the measure proposed. We believe that this condition has been fulfilled (see the 'Data' section below). Since the rationale for building the index has been made transparent in this article, scholars are encouraged to test it on other existing and future data sets, provided that they fulfil the necessary requirements.
Our index is built only of variables which can be derived from routine historical census or census-like microdata. This implies, in the first instance, that non-observable determinants of the observable demographic and residential configurations are not accounted for in the index -for example, parental control over marriage, actual inheritance patterns, or the availability of kin for co-residence. This also necessarily confines our attention to actual behaviours and not to behavioral norms, which are not always adhered to. The challenge of comparing the results of the index to patriarchy research based on other sources, such as parental power or inheritance patterns, remains a task to be taken up in the future.

Our notion of patriarchy (and why we wish to measure it)
Our notion of patriarchy departs from the often value-laden, monolithic and ideologically determined discourse of Western feminism (see Walby, 1990;cf. Kandiyoti, 1988, pp. 274 -275). Instead, we treat the concept simply as a useful descriptive tool for discussing social patterns in a comparative perspective. In line with a number of recent theorists, we see patriarchy not as having a single form or site, but as encompassing a much wider realm (cf. Joseph, 1996;Kandiyoti, 1988). 3 According to Therborn (2004), patriarchy has two basic intrinsic dimensions: 'the rule of the father and the rule of the husband, in that order' (p. 13). As such, it refers to generational and conjugal family relations or, more clearly, to generational and gender relations. Thus, the term also encompasses the domination of men over each other based on the seniority principle followed in many patrilineal and patrilocal societies. Similarly, Joseph (1996) defined patriarchy (albeit in the Arab context) as the prioritization of the rights of males and elders, and the justification of those rights within kinship values that are usually supported by religion.
This multifaceted notion of 'patriarchy' echoes Wolf's (2005) argument that the kind of authority parents might have over their children is the additional crucial building block of any patriarchal order, next to the conjugal balance of power. Wolf distinguishes between 'state patriarchy' and 'property patriarchy'. Under the former, parents are at liberty to exploit their children in return for allowing political or military superiors the right to exploit them on their own terms. The general societal outcome of parental practices in these settings (for example, in China) usually implies omnipotent parental interference in the life-course decisions of their offspring -i.e. leaving home, marriage and household formation. Under what Wolf calls 'property patriarchy', parental authority is weak because it rests primarily on the premise of property control, without recourse to the sanction of higher authorities.
The multifaceted nature of patriarchy was best captured by Kaser and others, who argued (in the Balkan context) that 'it is insufficient to understand patriarchy simply as the rule of the father, the eldest, or the husband', but that it is instead necessary to look as well at the formalized rules based on patriarchal concepts -i.e. inheritance rules, child obedience, marriage arrangements, residence at marriage, the presence or absence of institutionalized sexual asymmetries, and the obedience of women (Kaser, 2008, p. 33; see also Therborn, 2004, p. 13). By applying such a holistic perspective, they were able to tackle the patriarchy problem from various angles, looking not only at the complex set of hierarchal values embedded in a social structural system defined by gender and age, but also at the broader social units and social behaviour through which patriarchy in the Balkans attained its historical form -i.e. a form characterized by interlocking relationships of patrilinearity, patrilocality and a patriarchally oriented common law (Halpern, Kaser, & Wagner, 1996, p. 427). 4 However, an approach used to gain a better understanding of local meanings of patriarchy may be less helpful when the task is to measure comparatively the 'intensity' of patriarchy across time and space, especially among historical societies. While many conceptualizations of patriarchy have provided rich, detailed descriptions of gender and generational relations (see, for example, Halpern et al., 1996;Kaser, 2002Kaser, , 2008Miller, 1998;Mitterauer, 1999;for Russia, see Worobec, 1995, pp. 175-216;also Wolf, 2005), they are of little help in comparing and mapping across space historical forms of power relations between the generations and the sexes. 5 As the term 'patriarchy' suggests the existence of a complex social system, approaching patriarchy comparatively and globally (if only at the European level) will inevitably imply a reduction of its internal intricacies to a set of characteristics or aspects which can be studied across space and time. Such a reductionist approach would also appear to be necessary due to another set of constraints. Since it is very unlikely that a large number of comparative holistic reconstructions of gender and age relations across the multitude of historical societies (or even their subpopulations) will ever become available (cf. Miller, 1998), scholars may be forced to use routine records from census microdata to generate spatially sensitive descriptions of historical gender and generational indicators at the local level. This can now be done more effectively, as the ongoing revolution in census microdata (Ruggles, 2012) has made it feasible to assemble for the first time a very large amount of comparable individual-level data for continental Europe in pre-industrial times (see the Mosaic project website). These localized indicators of gender and age relations can not only greatly enhance the historical reconstruction of different family systems, household and regional economies, and power dynamics. The availability of such indicators may also become critical to the analysis of historical crosscountry differentials in fertility, social mobility, human capital formation and parental control.
In order to show how our own approach addresses these challenges, we will first provide an overview of our 'patriarchal' variables. The variables are grouped into four 'clusters', or the four 'domains' we believe capture the four major dimensions of 'patriarchy': domination of men over women; domination of the older generation over the younger generation; the extent of patrilocality; and the numerical balance of the sexes (see Figure 1). Combinations of the many different elements of power relations and agency contained in the composite Index of Patriarchy will substitute for rather vague notions of 'family systems' and 'patriarchy'. By revealing important features of both female and male autonomy in a given context, this approach may also point to new ways to account for the geographies of family organization across historical Europe. 6 For some of our variables, different definitions might be possible. However, to the best of our knowledge, the definitions we have chosen remain either the most meaningful or the least exposed to the biases inherent to cross-sectional census-like microdata. For example, for the most part, we preferred individual-level age-specific measures over householdlevel variables, because the former tend to minimize the effect of variation in demographic conditions on indicators of family structure (Ruggles, 2009, p. 252). Nevertheless, when dealing with prospective non-European and contemporary data, some revised definitions might prove more meaningful.
3. Measures used 3.1. Cluster 1: Domination of men over women 3.1.1. Proportion of female household heads (female household heads) Patriarchal hypothesis: only men can be household heads.
Description: this is the proportion of all female household heads among all adult (aged 20 þ) household heads of family households. We use an age-standardized measure to account for different age structures in different societies at different points in time.
This measure should be negatively correlated with patriarchy, as in truly patriarchal societies women would not be allowed to become household heads under most circumstances. Female headship is widely regarded as a key element of social structure, and as a positive indictor of the extent to which women are able (and willing) to manage an independent livelihood; thus, the existence of female headship tells us a lot about the options available to women more generally (Ogilvie & Edwards, 2000;Szołtysek, 2009;Wall, 1981). In traditional agrarian societies, the position of head of household had a quasi-public character (Wetherell & Plakans, 1998). Meanwhile, the reproduction of classic patriarchy is dependent on the operations of patrilocally extended households headed by men. Whereas in the patriarchal context most men can expect to be 'patriarchs' at some point in their life cycle, most women are denied the opportunity to hold a formal public position of economic power (Moghadam, 2004, p. 141).

Proportion of young brides (young brides)
Patriarchal hypothesis: a lower female age at marriage facilitates male domination.
Description: this is the proportion of ever-married women in the 15 -19 age group. We use this measure instead of the Singulate Mean Age at Marriage because: . it is less affected by missing information about the marital status of women in the 40-50 age range; . it does not reflect the developments over several decades, but instead presents data for the most recent years (Schürer, 1989;Smith, 1978); . it is less affected when there is only a small number of cases; and . it is used by the Social Institutions and Gender Index. 7 This measure should be positively correlated with patriarchy because we assume that in strongly patriarchal areas women would be married as soon as possible. In societies in which property and other rights are transmitted through men, the production of male children is critical. Early arranged marriages of daughters reduced the household economic burdens that came with supporting females who were destined to marry and leave the home in any case, and whose children would contribute neither income nor offspring to their father's natal group. 'Since unmarried women are social liabilities', Dyson and Moore (1983) argued, 'their marriage costs (dowry requirements) increase with age, providing another reason to ensure early marriage' (p. 48). Finally, as early marriage increases the period during which females can produce male heirs for the groups into which they marry, a prospective spouse is most in demand by other families when she is young (Davis & Blake, 1956). 8 3.1.3. Proportion of wives who are older than their husbands (older wives) Patriarchal hypothesis: the husband is always older than his wife.
Description: this is the proportion of all of the wives who are older than their husbands among all of the couples for whom the ages of both partners are known. If a husband is married to more than one wife, only the first wife is considered here. We use an agestandardized measure to account for different age structures in different societies at different points in time.
This measure should be negatively correlated with patriarchy because we assume that in truly patriarchal areas men would not marry women older than themselves. We further assume that, in intensely patriarchal societies, male remarriages involving much younger partners will enjoy broad societal acceptance, and will therefore be prevalent. There is a widespread consensus among sociologists, anthropologists and historians that the magnitude of the age difference between spouses is an indicator of the level of equality in the relationship between a man and a woman (Carmichael, 2011;Skinner, 1993;Van de Putte et al., 2009). Laslett, for example, argued that a relatively small age gap between spouses in historical north-western Europe indicated a societal preference for companionate marriage (Laslett, 1977, pp. 90, 99-101;similarly, Hajnal, 1965, p. 129). Meanwhile, Cain asserted that large age differences in favour of men can be seen as an indication of a patriarchal family system (Cain, 1988; see also Cain, Khanam, & Nahar, 1979).
3.1.4. Proportion of young women living as non-kin (female non-kin) Patriarchal hypothesis: a woman cannot live outside the home of her or her husband's relatives.

The patriarchy index
Description: this is the proportion of women aged 20 -34 who live as non-kin, usually as lodgers or servants. These women are not controlled by their relatives or by their husband's relatives. We use an age-standardized measure to account for different age structures in different societies at different points in time.
This measure should be negatively correlated with patriarchy because we assume that in intensely patriarchal societies young, unmarried women tend to be controlled by relatives, and are prevented from living with or working for non-relatives, especially before marriage. This prohibition may stem from cultural concerns about female virginity, and is thus related to the fear -based on either religious prescriptions or the structural difficulties of dealing with illegitimate children -that a girl could become pregnant before marriage (see McDonald, 1985;Schlegel, 1991). Equally, it could result from larger societal constraints that preclude the emergence of any role in the social structure for single, post-pubescent women other than that of wife or daughter (De Moor & Van Zanden, 2010;Klep, 2005;Szołtysek, 2014).
3.2. Cluster 2: Domination of the older generation over the younger generation 3.2.1. Proportion of elderly men co-residing with a younger household head (younger household head) Patriarchal hypothesis: the oldest man is always the household head.
Description: This is the proportion of elderly men (aged 65 þ) living in a household headed by a male of a younger generation. Only family households are considered here, and the elderly men must be relatives of the household head. We have chosen to analyze generations and not ages because we consider the generational difference to be more important than the age difference between men.
This measure should be negatively correlated with patriarchy because we assume that in intensely patriarchal areas no younger man is permitted to become a household head as long as an older male household member is alive. Research on patriarchal family arrangements in historical Europe has shown the pervasiveness of cultural traditions in which power relations within domestic groups were structured according to strict seniority. Thus, younger couples were not permitted to assume the headship when older couples were present in the household. This tradition has been found in both 'archetypical' joint-family societies and in societies with various 'stem-family' arrangements -i.e. families in which one adult son remains at home while his father continues to act as the head of the household, even after the son's marriage (Engelen & Wolf, 2005;Szołtysek, 2014; see also Halpern et al., 1996;Szołtysek & Gruber, 2014).

Proportion of neolocal residence among young men (neolocal)
Patriarchal hypothesis: sons cannot establish their own household on marriage.
Description: this is the proportion of ever-married household heads among evermarried men in the 20-29 age group. This measure only applies to family households and is an age-standardized measure that accounts for different age structures in different societies at different points in time.
This measure should be negatively correlated with patriarchy because it is assumed that in strictly patriarchal societies sons with living fathers are permitted to establish their own independent households only under exceptional circumstances. As Wolf (2005) has argued, in a very practical sense, 'how young people marry, when they marry, and where they reside after marriage will reflect the extent to which their society empowers parents' (p. 225). In domestic groups in which the 'vigorous authority of the senior patriarch' is enforced (Seccombe, 1992, p. 42), the authority structure prevents offspring (and sons in particular) from early independence because male children (as well as grandchildren) are capital resources and, like all capital resources, they are more rather than less desirable. Given the benefits of keeping sons on the patrimony for longer periods of time to assist with the low technology and extensive forms of agriculture typical of the many areas of pre-industrial Europe, parents were highly unlikely to have encouraged their sons to leave home and live independently. The result is not only early marriage, but also the widespread practice of post-marital subordination of sons to the older generation, in which the older male deploys the younger male for his own benefit (Niraula & Morgan, 1996;Szołtysek, 2014;Verdon, 1998;Wolf, 2005). In patriarchal societies in which property is controlled by the elders, an underlying structure of opportunities is created in which many young people have to wait a long time before they can step into 'dead men's shoes' and take over a parental (or fraternal) holding.

Proportion of elderly people living with lateral relatives (lateral)
Patriarchal hypothesis: some sons tend to stay in the household even after the death of their father.
Description: this is the proportion of elderly people (aged 65 þ) living with at least one lateral relative in the household. Lateral relatives are defined as siblings, uncles/aunts, nephews/nieces, great-nephews/nieces, cousins and other distant relatives (including inlaws). In addition, two married relatives of the same generation form a lateral extension (this applies to lineal relatives: children, parents, grandchildren and grandparents). This measure only applies to family households.
This measure should be positively correlated with patriarchy because we assume that in intensely patriarchal areas some men will not establish their own households at all, or will have to wait until late in life. Given that patriarchy has often been assumed to have strong associations with patrilineality and patrilocality, the centrality of brother -brother relationships to the reproduction of the patriarchal system seems obvious. Although the idea of a simple correlation between any descent system and actual social relations is not universally accepted (Schubert, 2005), it is nevertheless often assumed that agnatic, patrilineal and patriarchal environments helped develop the specific psychosocial dynamics of brother -brother relationships. Such relationships -which Joseph has called a 'patriarchal connective mirroring' -tend to be characterized by the staged sequencing of brotherly competition, giving way to strong brotherly solidarity which can surpass other supposedly intimate relationships in the family or domestic group (such as those between husband and wife) (see Joseph, 2001). Finally, in patrilocal multiple-family households, which provided the best environment for patriarchal complex values to emerge, the prolonged co-residence of married brothers in the domestic group often fostered a complex web of kin relationships, domestic hierarchies and economic dependencies between various types of lateral kin (Collver, 1963;Czap, 1982;Halpern, 1977).

Proportion of elderly people living in joint residence ( joint family)
Patriarchal hypothesis: all sons have to stay in the household of their father.
Description: this is the proportion of elderly people (aged 65 þ) living with at least two married children in the same household. This measure only applies to family households. This measure should be positively correlated with patriarchy because we assume that in truly patriarchal areas no sons will leave their parental household, either because they have internalized the idea of paternal power and joint residence or due to economic or legal restrictions. Joint-family types of living arrangements -i.e. co-residence with at least two married offspring (preferably sons) -have commonly been seen as being the locus of archetypical patriarchal relationships (Caldwell, 1982). Joint families could be found in many different societies of Eurasia, from the nomadic tribes of the Middle East to the Slavic serf agriculturalists and the ancient civilizations of the Far East. The common features of joint families around the globe include the co-residence of two or more nuclear families; the patrilineal succession of family titles and property; a tendency to keep sons on the patrimony and virilocal household formation; a tendency to unify the joint domestic group around some common economic project; a tendency towards fission at some point in the developmental cycle; a marginal position for female siblings; and a tendency to recruit workers from among kin rather than from among wage labourers (Szołtysek & Gruber, 2014). It is in this context that the concept of patriarchy has often been evoked, becoming a convenient shorthand for the presumed distinguishing traits of joint-family relations (Erlich, 1966;Halpern et al., 1996;Kaser, 1992;Mitterauer, 1999).

Proportion of elderly people living with married daughters (married daughter)
Patriarchal hypothesis: all daughters move into their husband's father's house.
Description: this is the proportion of elderly people (aged 65 þ) living with at least one married daughter in the same household among those elderly people who live with at least one married child in the same household. This measure only applies to family households.
This measure should be negatively correlated with patriarchy because in intensely patriarchal areas it is expected that all daughters will leave their parental household on marriage. According to the prevailing principle of patrilocality, on marriage a woman will move into the household of her husband or her husband's father. Although in some patriarchal societies the uxorilocal residence of sons-in-law was used as a substitute for a missing male biological heir, in intensely patriarchal societies there was a strict preference for patrilocal marriages of the male offspring, and for the exclusion of daughters from reproduction within the natal group (Szołtysek & Gruber, 2014). This variable can also be seen as a proxy for inheritance patterns in peasant societies (i.e. who inherits the parental farm).

Cluster 4: Son preference
3.4.1. Proportion of boys among the last child (boy as last child) Patriarchal hypothesis: after the birth of a daughter, parents will try to have another child.
Description: this is the proportion of boys among the last children (if the last child is one of a set of siblings of both sexes, he or she will be excluded from the analysis). So far, this measure has been restricted to the children of household heads because the analysis is much more complicated for other relatives. The analysis is restricted to the 10-14 age group because, in the younger age groups, we cannot know whether the last child really is the last child and, in the older age groups, we cannot know whether one of the children has already left the parental household through marriage or going into service. This measure only applies to family households.
This variable is also used in the Social Institutions and Gender Index, but this index takes advantage of contemporary household surveys, which make it easier to identify the last child. This measure has already been used in an analysis of the fertility decline in a Serbian village, which yielded a male surplus for almost all birth decades between 1850 and 1939. After 1879, the proportion of boys among the last children was always 60% or more (Wagner, 1984, p. 232).
We would expect to find that this measure is positively correlated with patriarchy. Son preference is considered to be an inherent feature of patriarchal family settings, as it generates strong disincentives to raise daughters, while valuing adult women's contributions to the household (see, for example, Das Gupta et al., 2003). In patriarchal environments, sons are critical to families in a variety of ways, including for the continuity of the lineage, performing ancestor-worship rites and providing support in old age. As daughters normally cannot perform these functions, and are therefore of far less consequence to families than sons, in truly patriarchal areas parents should be more inclined to cease reproduction after the birth of a son than after the birth of a daughter (Larsen, Chung, & Das Gupta, 1998).

Sex ratio of youngest age group (sex ratio)
Patriarchal hypothesis: girls are treated worse or are considered to be of lesser importance than boys.
Description: this is the sex ratio (number of boys to 100 girls) in the youngest age group (0 -4). We are investigating the youngest age group because the effects should be most marked in this age group. This measure only applies to family households.
This variable is also used in the Social Institutions and Gender Index, but this index uses all of the age groups. As the score of the index is particularly influenced by the sex ratio at young ages, we use only the youngest age group.
This measure should be positively correlated with patriarchy because we assume that increasing patriarchy will lead to higher female mortality or the under-registration of females. Since the path-breaking research of Amartya Sen on the 'missing women', the explicit or implicit notion of patriarchy has been fundamental to research on the distorted sex ratio in Asia and elsewhere (Das Gupta, 2005;Guilmoto, 2009;Lynch, 2011). As women tend to be treated worse than men in patriarchal societies, we can assume that female mortality will be higher than male mortality -i.e. that females are neglected, given less or inferior food or even killed as infants. In societies in which less importance is placed on women, there may also be higher rates of under-reporting of females in censuses. We cannot distinguish between these two possible reasons for the underrepresentation of women in censuses. However, both reasons are associated with a patriarchal regime as understood here.

Age standardization
Theoretically, the index we are proposing should be applicable to any kind of human society, as long as some basic requirements are met (sufficient population size and the availability of microdata which cover the whole population and report each person's sex, age, marital status and relationship to the household head). 9 Among the challenges we face in creating such an index is that the age structures of societies may differ, and these differences could heavily affect the results of the index for the given society under investigation. There are several ways we can control for the age distribution: by restricting the analysis to one age group, age standardization, and regression (see Ruggles, 2012, p. 431).
Some of our measures are restricted to a single age group and some use age standardization. The standard population should not be based on only one historical population, but should cover the whole of Europe, because our data now cover the whole of Europe. As real populations are always affected by fertility or mortality crises and the migration flows of the preceding decades, a constructed population is better suited for our purposes. We have therefore chosen the age structure of a stable population, the Model West, with a mortality level 6 and the rate of population growth of 5 per 1000 (Coale & Demeny, 1983, pp. 60, 110).
We have chosen the Model West because this model is based on the largest number of life tables, including tables that cover populations in western, northern, and northeastern Europe, as well as some populations outside of Europe. It can be seen as being the most general of the four models (United Nations, 1983, pp. 12-13). 10 The mortality level 6 translates into a life expectancy at birth of 32.5 years for females and 30.1 years for males. This can be seen as representative of populations in eighteenth-and nineteenth-century Europe. The rate of population growth of 5 per 1000 is closest to the growth rate of the population in Europe for the period 1750-1950: 0.68% (Livi-Bacci, 2012, p. 25). Table 1 presents the detailed list of regions included in the data pool used here as a basis for calculating the index. These regions are conceptualized either as spatial units for data sampling (for example, or Hungary, 1869, based on administrative units used at the time of the census (for example, Braclav Governorate, 1795; Schleswig and Holstein, 1803), or on geographic considerations (for example, . Table 2 shows the distribution of these regions across time and urban-rural contexts, and Figure 2 reveals the spatial patterning in the distribution of the data across Europe.

Data
The data cover 91 regions/locations in Europe, from the Atlantic Ocean to the Ural Mountains. A slight majority of the locations included come from the nineteenth century (56%), with more or less equal proportions covering earlier and later periods. The collection contains both rural and urban sites, although rural societies clearly predominate (82%).
Certain limitations of the analysis merit attention. Despite the significant expansion and scope of the Mosaic database, from which these data were derived, the present collection is still a rather miscellaneous amalgamation of locations (for a detailed description of the data sets, see Appendix 1). The data are largely concentrated in the central continental belt, providing quite good coverage of the French, German, Austro-Hungarian, Polish and Balkan areas. However, some areas which are important for the investigation of the European geography of family systems are not yet included, or the coverage of these areas in the database is very limited. These areas include the Low Countries, which are often assumed to have encompassed the essential features of 'northwest European' family systems (De Moor & Van Zanden, 2010); the Italian territories, which, according to some scholars, exemplify the 'Mediterranean' zone ; and, of course, Russia. As the Mosaic database expands, some of these deficits will be addressed in the future. However, in addition to covering both urban and rural communities, the current database runs across many important fault lines in the European geography of family systems, including places located: 1. eastward of the Hajnal-Mitterauer line (parts of Poland, Russia, Ukraine, Belarus, Hungary and Lithuania); 2. in the Balkans (or even Asia Minor) zone (Albania, Serbia, Turkey, Romania [Wallachia] and Bulgaria); 3. in the 'intermediary central European zone' of Laslett (1983), or Austro-Hungary and German areas, as well as parts of historical Poland; and 4. in western Europe (France).
The collection encompasses societies which varied significantly in terms of the basic principles of family and household organization -i.e. strictly nuclear and neolocal populations (like urban Rostock, but also southern Ukraine, the Braclav area and Podolia); stem-family societies (like those in the area of Münster in Germany, in south-western France and in parts of western Poland); complex societies exhibiting a 'classic' eastern European joint-family pattern (like those in Mishino near Moscow, studied by Czap (1982), or in Polesia in eastern Poland-Lithuania) or Balkan versions of this pattern (Albania and Serbia); and a range of intermediate patterns with varying degrees of intermingling of nuclear-and stem-family organization (Poland proper, Germany in 1846 and Austria around 1910), or stem-and joint-family patterns (Red Ruthenia in Poland and fifteenth-century Italy). Furthermore, even in its present scope, the database covers much of European variability in terms of geographical features, populations, cultures and socio-economic geography  (Jordan-Bykchov & Bykchova-Jordan, 2002) -i.e. plains, mountains, and coastal areas; free and unfree peasantries; a range of ethnicities and religions; and a range of regional patterns of economic growth in the early modern and modern eras. Thus, the internal diversity of the present sample should allow us to investigate the varying degrees to which family power relationships in different contexts were influenced by patriarchal features. Table 3 presents a summary of the descriptive statistics for all of the variables considered for the computation of the index, with basic measures of dispersion at the bottom. Overall, in 91 regions of historical Europe, only 10% of all adult household heads were female. Very low percentages (under 5%) seem to be clustered in the eastern and south-eastern part of Europe: in rural Albanian prefectures and two Albanian cities (Kruja and Kavaja); in the Serbian district; in several regions of historical Poland-Lithuania (both western and eastern); and in two Wallachian regions in Romania. However, low percentages of female headship were also found in one German region (the district of Rheine-Bevergern in the Münster area for both time periods) and in the Italian region of Legnago during the medieval period. Higher proportions of female heads (over 10%) were found in scattered locations in Germany, Hungary and Poland-Lithuania, but also in some areas in Russia. However, most of the locations with the highest proportions (15% and more) of female headship were found in Germany, with only a few exceptions (for example, central Slovakia in 1869).

Descriptive results of the variables used
Overall, about 14% of all of the women in all 91 regions were married between the ages of 15 and 19. The strong spatial differences in female marriage patterns generally reflect the well-known differences between western and eastern patterns of nuptiality, with women in eastern Poland-Lithuania (Belarus and Ukraine), Albania, Bulgaria, Slovakia, Hungary, Serbia and Italy marrying earlier than women in the German areas and France. About one in seven young women lived as a lodger or a servant in the household of people to whom they were not related by blood or by marriage. An east -west divide can be observed in the overall pattern, but with some important ambiguities. All of the 14 regions in which no such women were reported in the census were located the eastern or south-eastern parts of Europe: four Wallachian regions, five Albanian rural prefectures, the Albanian city of Kruja, the Russian villages of 1814, the Serbian villages in both census years, and the Bulgarian Rhodope region. However, the regions with generally low proportions of women living outside of the parental home or the family of procreation (under 15%) were found in a wide variety of sociocultural and demographic contexts, including in several eastern European regions, but also in locations in Germany and France. Finally, regions in which 25% or more of the young women were living in nonfamily and non-kin arrangements were found in Germany and Austria (Styria and Waidhofen/Ybbs in 1910), but also in several locations in western Poland-Lithuania.
About one in seven elderly men lived in a household headed by a man of a younger generation. In five regions no such cases were reported in the census: in the Albanian city of Kruja in 1918, the Russian villages of 1814, central Slovakia in 1869, and German Höhscheid and Brunswick in 1846. In a further 14 locations (primarily in Albania, Romania and eastern parts of Poland-Lithuania), the shares of such cases were negligible. By contrast, in eight German regions (especially in the Münster area and in Schleswig in1803), as well as in several regions of Poland proper, one-quarter or more of elderly men lived in households headed by a younger man.
Slightly more than half of young married men lived neolocally. This is an effect of the variable with the largest range and standard deviation. The lowest proportion by far was reported for the Russian villages in 1814 (3%). Other regions with low shares (under 20%) included the German government district of Arnsberg in 1846, and Polesia and Zhytomyr county in eastern Poland-Lithuania. All of the young married men lived neolocally in the German regions of Brunswick and Koblenz in 1846. At least 90% of these men lived neolocally in another 12 regions -most were in Germany or Austria, but this group also included three Wallachian regions in Romania and Warmia in historical Poland. About one-fifth of elderly people lived in households with lateral relatives. This variable also shows quite a large range and standard deviation, and its spatial distribution does not fully comply with expectations. Lateral relatives were entirely absent in two sites located on opposite sides of the seminal Hajnal-Mitterauer line (in western Slovakia in 1869 and in German Höhscheid in 1846). Of the nine other regions in which less than 5% of elderly people were living with lateral relatives, five were in eastern Europe (two Wallachian regions in 1838 and Transylvania in 1869, all of which are now part of Romania; two western Polish-Lithuanian regions; and four German regions in 1846). By contrast, the real hot spots of co-residence with lateral kin (35% or more) were found almost exclusively in eastern and south-eastern Europe, especially in Albania, Serbia and eastern Poland-Lithuania.
Overall, 6% of elderly people shared a household with at least two married children. In 40 out of the 91 European regions, no such cases were reported, including in all of the Austrian regions, almost all of the German regions, and also the majority of the Romanian regions. By contrast, more than 20% of elderly people lived in such joint residences in 10 regions: in the Ukrainian and Belarusian parts of Poland-Lithuania around 1800, in the Russian villages in 1814, in the Serbian villages in both censuses, in four Albanian rural prefectures, and in the city of Kruja in 1918.
Across all of the data sets, about one-quarter of elderly people who were living with a married child were also co-residing with a married daughter. There is also a considerable degree of variation in this variable: it has the second-largest range and standard deviation. The locations in which no such cases were reported come from three disparate regions in Albania, Serbia and the German region of Brunswick. Six further regions in which less than 3% of elderly people lived with married daughters were already more skewed to the east of Europe (five rural prefectures in Albania in 1918 and the district of Meppen in the Münster area in 1749). In contrast, the regions with the highest proportions of the variable (35% and more) were again more varied. They were found in the majority of the German regions, in Austria, in northern France, and also in Hungary and western Poland proper.
Half of the last children in the 10 -14 age group were boys. In five regions, less than 40% of these children were boys: the German region of Rheine-Bevergern in 1749, the eastern cities and the region of Düsseldorf in 1846, central Slovakia in 1869, and Podolia in Poland-Lithuania. The highest proportion by far (81%) was reported for the Albanian rural prefecture of Puka in 1918. Shares of more than 57% were also reported for another eight regions: the Albanian city of Kruja in 1918, the German regions of Saxe-Gotha and Arnsberg in 1846, the Bulgarian Rhodope region around 1900, the city of Istanbul in 1885, the villages of western Slovakia and the Hungarian Great Plain in 1869, and the Russian villages in 1814.
The sex ratio of the youngest age group was, on average, 103, and was therefore quite

Correlation of measures
In the next step, all of these measures were checked to determine whether they were correlated with each other and, if so, whether they were correlated positively or negatively, as we assumed. Generally, most of the correlations were in line with the assumptions above; only the correlation between the variables 'proportion of female household heads' and 'proportion of elderly men co-residing with a younger household head' was contrary to our assumptions (Table 4).
The four patriarchal variables pertaining to the domain of 'domination of men over women' were all correlated (especially the variables 'proportion of young brides' and 'proportion of young women living as non-kin'). The areas with higher levels of female autonomy in headship also generally had a higher female age at marriage, a higher proportion of wives older than their husbands, and a higher share of young women who were living independently of their immediate kin (i.e. outside of their usual functions/positions as wives and daughters). Similarly, the four variables associated with 'domination of the older generation over the younger generation' were also correlated. In particular, the variables 'proportion of elderly people living with lateral relatives' and 'proportion of elderly people living in joint residence' were highly correlated. We therefore used only one of them for the patriarchy index (the variable 'proportion of elderly people living with lateral relatives'). The two variables of the patriarchal feature 'son preference' were not correlated.
Some of the variables of the different patriarchal domains were also significantly correlated with each other (Table 5). For example, the variable 'proportion of elderly people living in joint residence' was strongly correlated with both variables with the highest correlation coefficient in the domain 'domination of men over women'. This suggests thatin line with previous research and theory -the more a given environment is prone towards joint-family residence, the more likely it is that women in this area will marry early and marry men older than themselves (Berkner & Mendels, 1978). Similarly, the cross-cutting of domains 2 and 4 -i.e. the positive relationship between the 'jointness' of living arrangements and the skewed sex ratios in favour of boys -also complies with previous research findings.
All in all, while the correlations unravelled above are usually not very strong (with some notable exceptions), most of them are statistically significant (out of the 55 correlations in Table 4, 42 -i.e. 76% -were significant). The empirical interrelationships among the different aspects of what we have argued is the general concept of 'patriarchal bias' are generally reassuring in that they validate the usage of the index as one measurement of patriarchy, even if various aspects of patriarchy as defined above may not be equally predictive for the latter's all domains. 11

Calculation of the Index of Patriarchy
In our final step, we created our Index of Patriarchy. It was made up of four components representing different domains of patriarchy: . domination of men over women; . domination of the older generation over the younger generation; . patrilocality; and . son preference.
These components -or sub-indices -consist of the variables described above within these domains of patriarchy. Each variable can have 0 to 10 patriarchy points, and all of the respective variables are added up to obtain the patriarchy points of one feature of patriarchy. All of the variables, except for the last two are turned into patriarchy points in Therefore, at least one region will have 10 patriarchy points in each variable. The patriarchy points for these variables are calculated according to the following formulae: † patriarchy points ¼ RND 10 * proportion maximum proportion for variables positively correlated with patriarchy; and † patriarchy points ¼ 10 2 RND 10 * proportion maximum proportion for variables negatively correlated with patriarchy: For example, the city of Rostock in 1900 had a share of 1.9% of married females in the 15 -19 age group, while in the northern Albanian region of Puka in 1918 56% of these girls were already married. So far, the highest value of this variable has amounted to 66%; in this light, Rostock has 0 patriarchy points for this variable (10* 0.019/0.66 ¼ 0.29 points, rounded to 0 points), while the region of Puka has 8 patriarchy points (10* 0.56/0.66 ¼ 8.48 points, rounded to 8 points).
The last two variables (4.1 and 4.2) are calculated differently because they have a different range. As the minimum value we assume the proportion which is seen as neutral. This is 0.51 for the proportion of boys among the group of last children and 105 for the sex ratio of the youngest age group. All of the proportions below these values are set to these defined minimum values. The maximum value is, again, the maximum achieved for the respective variable. The patriarchy points are calculated for these two variables according to the following formula: † patriarchy points ¼ RND ðproportion 2 defined minimum valueÞ ðmaximum proportion 2 defined minimum valueÞ : The patriarchy points are rounded, which makes the results easier to grasp for each variable. In this way, we obtain 11 categories for each variable ranging from 0 to 10 patriarchy points. A score of 0 indicates the lowest degree of patriarchy, while a score of 10 indicates the highest degree of patriarchy. Using a different number of variables for the four features of patriarchy leads to a different maximum number of patriarchy points for the various features. As Table 5 shows, these four sub-indices are positively correlated, and only the sub-index of 'son preference' has lower correlation coefficients.
Finally, the Index of Patriarchy is calculated by adding up the four sub-indices, but each sub-index is reduced to a maximum of 10 patriarchy points. The Index of Patriarchy can therefore have a minimum of 0 and a maximum of 40 patriarchy points. The index is calculated according to the following formula: † patriarchy index ¼ male domination index 4 þ generational domination index 3 þ ðpatrilocality indexÞ þ son preference index 2 : For example, the city of Rostock in 1900 has a male domination index of 12 points, a generational domination index of 10 points, a patrilocality index of 2 points, and a son preference index of 0 points. These result in a patriarchy index of 8 points according to the formula: 12/4 þ 10/3 þ 2 þ 0/2 ¼ 8.33 points (rounded to 8 points).

Spatial distribution of the index: discussion
The results of the exploration of the distribution of the index across space are presented in three ways. Figure 3 shows the continuing scale of the extent of patriarchy, as defined here, for 91 European regions (for a complete list of index points for all locations and regions, see Appendix 2). In Table 6, we group all of the regions included in the analysis into four clusters of different levels of intensity of patriarchy: . very low patriarchy, . low patriarchy, . high patriarchy and . very high patriarchy.
These groupings are based on the following considerations: . the mean of the patriarchy index divides high and low levels of patriarchy, and . high and low patriarchy have a range of 1 standard deviation, while all of the other values represent either very low or very high degrees of patriarchy.
Finally, Figure 4 organizes the results included in Table 6 spatially. In our discussion of the search for a pattern of regional variation in patriarchy across historical Europe, we will begin by looking at Figure 3. The observed values of the Index of Patriarchy range from 6 to 33 points. Within the present collection, there were no societies or locations with absolutely no patriarchal features as defined above, and there were no societies or locations to which absolute patriarchal characteristics could be assigned.
In Figure 3, the Index of Patriarchy displays a rather smooth continuum from very low to very high levels of patriarchy. Although it would be an exaggeration to speak of clearcut groupings of regions with high or low patriarchy intensities across historical Europe, certain patterns do emerge, even at this early stage. At the most general level, the ranking of the different regions is broadly consistent with perceptions and insights from the historical demographic science literature on family forms, and seems to confirm the longterm consistency of the dichotomous regional pattern of demographic performance posited in the works of Hajnal (1965 andMitterauer (1999). Indeed, western Europe tended to be much less patriarchal than eastern Europe. Patriarchal features were much more pronounced than elsewhere on the continent as we move east and south of the Danube after it passes through Vienna, and especially east of the Bug River, where Polish and Ukrainian ethnicities converge, and then further into the territories of eastern Europe. This observation is, however, subject to certain qualifications, which should lead us to avoid relying on an overly simplistic understanding of the east -west divide in European patriarchy (see also Szołtysek, 2012b). To date, the lowest patriarchy intensities have been found not in the westernmost country of France, but in Germany and Austria. This is contrary to theories which posit that patterns of family organization in the Germanspeaking areas lie between the western and the eastern patterns. As long as we do not have Dutch, Belgian, Scandinavian or British data, the German-speaking areas will have the lowest levels of patriarchy in the index. 12 These results also seem to run counter to the stereotypical image that the German family has been based on strictly authoritarian principles and strong parental authority (see, for example, Todd, 1985). Especially in the cities, the levels of patriarchy in northern Germany, which is adjacent to Scandinavia,  appear to have been very low. But the degree of patriarchy in Germany was not as low at the Dutch border as it was at the borders with Scandinavia. Generally, levels of patriarchy were higher to the east of Germany and Austria, but there were some regions with low degrees of patriarchy in Romania and Hungary. We can say little about levels of patriarchy in the south because we currently have only one medieval city from northern Italy in the index. In south-eastern Europe, however, we found the largest concentration of regions with very high patriarchy intensities. The Albanian, Bulgarian and Serbian data showed the highest levels of patriarchy. In eastern Europe, only Polesia in southern Belarus and the Russian villages of 1814 scored equally high on the Index of Patriarchy. Most of the other regions in eastern and south-eastern Europe were found to have high, but not very high, levels of patriarchy.
Furthermore, we observed considerable variation within countries and across the macro-regions of Europe. The territory between the Baltic, the Adriatic and the Black Seas (east-central Europe) seems to have been particularly diverse. Depending on how broadly this territory is defined, it might include places with very low levels of patriarchy, like the western and northern parts of historical Poland under Prussian rule in the nineteenth century, or places with moderate to high levels of patriarchy, such as several locations across Hungary, Slovakia and Romania. In fact, historical Poland-Lithuania is the only historical region for which we found a combination of very high and very low patriarchy intensities: we observed very low to low levels of patriarchy on the western and northern outskirts of this region, but much higher levels in the eastern areas of the Polish Republic, inhabited by Belarusian and Ukrainian ethnic groups. These features of the country's 'patriarchal profile' strengthen assertions made elsewhere that, relative to the family organization patterns across Europe, the patterns found in Poland-Lithuania were of a transitory, intermediate nature (Szołtysek, 2014; see also Gruber & Szołtysek, 2012).
Germany was also found to have been very diverse. However, the combination of very low, low and high degrees of patriarchy observed in Germany can be attributed, in part, to two factors: first, Germany had the most regions in the analysis by far and, second, as some of the regions had a rather small population, they may have been exposed to stochastic variation, which could have resulted in artificially high patriarchy scores in the index. More importantly, however, the locations which were found to have had high levels of patriarchy were in Westphalia, a region in which a considerable minority of the population followed principles of stem-family organization (Fertig, 1999;Szołtysek & Gruber, 2013).
Moreover, the eastern 'zone' of high levels of patriarchy displays some interesting similarities with what appears to have been a quite extensive belt of intensely patriarchal features across northern Italy and southern France. Southern France in 1846 had levels of patriarchy that resembled those of some Westphalian locations, as well as those of parts of Poland proper, Slovakia and Hungary. Thus far, the clarity of the east -west division between low patriarchy and high patriarchy levels in Europe remains blurred.
Additionally, the application of the Index of Patriarchy to historical census microdata has revealed spatial constellations which are not entirely congruent with older models of family patterns. For example, some late-marrying, neolocal populations of northern France have scores that are comparable to those of societies of western Poland, in which nuclear households also prevailed, but in which marriage occurred at younger ages and was universal. While Ukrainians from Podolia and Romanians had twice as many nuclear households as Belarusians or Russians, all of these populations were found to have had roughly the same levels of patriarchy, as defined here. In the same vein, the Ukrainians in Podolia and the societies of southern France scored equally high on the index, despite following strikingly different household formation rules (extreme neolocalism in the former case and stem-family rules in the latter). In short, due to the composite character of our measure, many complex intricacies that were latent in the earlier modelling of historical family systems across space are now fully brought to the fore.
The results generally confirm much of our existing knowledge about patriarchy trends: for example, that patriarchy intensities tended to be lower in cities than in surrounding rural areas, and that northern Albania had a very high level of patriarchy. It is nevertheless interesting that central Albania also scored higher on the index than the most patriarchal regions in eastern Europe and other south-eastern European countries.
This provisional picture of patriarchal scores across European geography is, by necessity, subject to change. The Dutch, Belgian, Scandinavian and British data, which will soon be added to the index, will provide extensions or modifications of the western pattern of generally low levels of patriarchy. The same is true of the inclusion of data from Lithuania proper and Kurland (Latvia), which are currently being prepared. These data sets are expected to add further complexity to the mosaic of patriarchal levels within eastcentral and eastern Europe. The same can be inferred about the potential historicity of the patriarchal patterns. Although, at first glance, the appended data (see Appendix 2) do not give the impression that, overall, time has been a very significant variable, 13 in order to validate this claim, it might be necessary to carry out standardized time series of the index for the same locations or regions.

Conclusions
What difference will our efforts to create a patriarchy index make? First, we have shown that it is possible to construct variables for measuring patriarchy. This exercise in quantitative measurement obviously reduces thick descriptions of gender and generational relations to quantifiable, comparable quantities. At the same time, it provides a handy tool that can be used for comparative studies of power relations in historical families and for studies of historical family systems in general.
Applying the composite indexed measure of power relations to historical census microdata has revealed the existence of spatial constellations in family patterns which are not entirely congruent with older models. A further expansion of the database (especially by including materials from NAPP) may lead to the discovery of a much more nuanced geography of European family forms, and may thus help to clarify some pending issues related to the typologization of family systems across Europe. 14 Due to its composite nature, the Index of Patriarchy invites researchers to delve into a domain which is much broader than the usual concerns addressed in historical demography or family history.
The present article deals with the major variations in power relations within domestic groups, and provides the first comprehensive account of their regional prevalence across historical Europe. So far, we have explored these differences through a finely graded analysis of the spatially diverse data set of 91 regions of Europe. The identification of systematic forms of gender or generational bias can help us better understand different societies, past and present. Based on numerical variables that are easily derived from census microdata with only limited information, our Index of Patriarchy generated spatially sensitive descriptions of historical localized gender and generational indicators. Thus, the index allowed us to identify regions with different degrees of patriarchy (as defined here) within a single country, across the regions of a single country, or across and within many broader zones of historical Europe.
The spatial contours of that variation across the landmass of Europe touch on several key aspects of the continent's social, demographic and economic histories. This is because the major variations in the 'intensity' of gender and generational biases ('patriarchal bias') can be seen as critical for explaining a wide range of trends, including cross-country differentials in fertility, social mobility and human capital formation; regional variation in labour relations, agricultural development and gendered well-being; historical patterns of childcare and sex-and age-specific mortality; the persistence of specific cultural norms and ethical frameworks; and the development of corporative institutions (De Moor & Van Zanden, 2010;Duranton, Rodríguez-Pose, & Sandall, 2009;Dyson & Moore, 1983;Greif, 2006). Overall, a comparative study of power relations across historical Europe along the lines we have suggested here may have direct relevance for our understanding of regional disparities in well-being, wealth and inequality.
Explaining the causes of the observed variation in patriarchal intensities is a task for future research. When seeking to identify the factors (economic or cultural) that may have contributed to these patterns, it is important to bear in mind that these regional differences in gender and generational biases lie at the intersection of many complex factors, processes and historical path dependencies, and that disentangling these variables will only be possible through the use of sensitive, multilayered, cross-disciplinary approaches.

Disclosure statement
No potential conflict of interest was reported by the authors. Notes 1. Recently, Ruggles applied two measures of co-residence among the aged to a huge assemblage of census microdata in order to assess the spatio-temporal distribution of stem-and joint-family arrangements. Although these measures represent a huge step forward in accounting for family forms worldwide, they also have some drawbacks (see Gruber & Szołtysek, 2012). 2. See the Mosaic project website at: http://www.censusmosaic.org 3. Thus, our notion of patriarchy corresponds to 'systems of sex-and age-related social inequality', in which individuals have differing levels of access to power, capabilities, prestige and autonomy (Niraula & Morgan, 1996 -but these authors focus exclusively on female autonomy). 4. Compared to this definition, the conceptualization of patriarchy in most demographic studies has been more simplistic. M. Cain, for example, used the median age difference between oncemarried spouses as an indicator of patriarchal structure in a cross-national analysis of fertility in the developing world (Cain, 1988; see also Cain et al., 1979). Cain (1988, pp. 25 -27) has rightly assumed that the age difference between spouses has several attractive features as an indicator of patriarchal structure, which can be used in a comparative demographic analysis. However, he seemed to fail to take into account some other demographic and domestic group characteristics, which are no less inherent in the demographic and familial contexts of peasant societies governed by patriarchal rules as those defined above.
5. To our knowledge, Niraula and Morgan's (1996) article is pioneering in that it explicitly deploys the concept of 'different intensities of patriarchy' in the study of female autonomy in two Nepalese settings; Dyson and Moore, (1983) also do this less explicitly. However, in neither case were these differences in 'intensity' formalized and quantified. 6. Most of the genuine research on the effects of sex-and age-related systems on individual capabilities focused on women, while neglecting the effects of these systems on different generations and on the autonomy of men (see Dyson & Moore, 1983;Niraula & Morgan, 1996). 7. See http://genderindex.org/data. 8. Thus, in a broader perspective, early age at marriage (for both females and males) may result from the domination of the older generation over the younger (see Cluster 2). In those institutional settings where parents were at liberty and able to exploit their children (for example, by being backed up by state officials), the general societal outcome of such practices would usually translate into the parental right to interfere in the marriages of sons and daughters, leading to an early age at marriage, as well as to a lack of celibacy. Unfortunately, neither information about the state's control over marriage nor its delegation to parents can be quantified from our data. 9. We mean here primarily modern state societies or the premodern Eurasian agrarian societies with tributary modes of production. The usefulness of the index when applied to mobile pastoralist groups or foraging societies, for which we know that sufficient data exist (for example, in nineteenth-and early twentieth-century Siberia), is still to be explored. 10. The Model West is derived from the largest number and broadest variety of life tables.
Therefore it is believed to represent the most general mortality pattern, which does not deviate systematically from the standard patterns obtained when all 192 life tables included by Coale and Demeny (1983) are put together, and hence is closer to the standard than those on which other regional sets are based (North, East and South). 11. Otherwise, good bivariate associations do not apply to the two variables purportedly measuring son preference (the correlations were significant in only about half of the casesnote, however, a better match for the cross-cutting of domains). These two variables are theoretically very telling, but are affected by random variation in regions with fewer cases. It is clear that, in fine-tuning the index for global use, these two variables require special attention. This may be compared with Whyte's (1978) research. He compared 52 variables measuring the relative status of women in a cross-cultural survey of 93 pre-industrial societies (the variables included, for example, arranged marriages, inheritance, marriage payments, female control over property, etc.) and found only very few correlations between them (they were hardly ever above 0.3 and most of them were insignificant) (see Whyte, 1978, pp. 96 -106). 12. Our exploratory exercise revealed that this observation might indeed be challenged. In a paper delivered by the authors at the European Society of Historical Demography conference in Alghero, Sardinia, in 2014, the patriarchy index was calculated for the combined Mosaic and NAPP data, showing the lowest values of the index to be in Great Britain and Norway. Although these calculations are still subject to change, it is plausible to expect that areas around the North Sea Basin will generally have lower levels of 'patriarchy', as defined here, than other areas. 13. Although the lowest German patriarchy scores all come from the second half of the nineteenth/ early twentieth century, and the highest from the eighteenth century, the district-level data are not identical. Also, the most contemporary data  are the most 'patriarchal' in the entire collection. 14. This task of linking Mosaic and NAPP data will be the subject of a separate article. 15. See http://www-gewi.uni-graz.at/suedost/seiner/index.html More than two-thirds of Albania is mountainous, especially the northern parts. Most of the western border is formed by the Adriatic Sea, and there are plains along the coast. Durrës is a port city, Shkodra is situated next to a large lake, and Kavaja is not far from the coast. The other three cities included in this study are located in the interior of the country. Shkodra is the only city in northern Albania, while the other five cities are located in central Albania. The cities of southern Albania were not included in this census. The majority of the population surveyed was Muslim (78.2%); only the prefecture of Puka was predominantly Catholic. The only city with a considerable Catholic population was Shkodra (about one-third). The Orthodox population captured in the census was living mainly in urban areas, because the main areas where Orthodox Christians lived at that time were either outside the area covered by this census or areas for which the census originals have not been preserved. The analysis will be done by comparing the different regions in the area covered by the Albanian census of 1918. This area was divided into seven prefectures at that time, and the six cities of this area are separated out based on the assumption that the behavioural patterns of the urban and the rural populations differed. The subprefecture of Gora has been separated from the prefecture of Zhuri because this region was known for having a large number of male migrant workers, which makes it distinct from the neighbouring regions. The analysis is therefore based on eight rural regions and six cities.  This is a sample of the 1910 census of Austria, which was prepared by Peter Teibenbacher for the Max Planck Institute for Demographic Research. As the surviving materials from the 1910 census are very unevenly distributed within the borders of present-day Austria, no representative sample could be made. The data cover villages and market towns with different ecotypes in three Austrian provinces (Upper Austria, Styria and Tyrol), and the city of Waidhofen an der Ybbs in Lower Austria. As the data refer to the administrative boundaries in 1910, some places have fewer inhabitants than the present-day configurations of these locations (for example, Pregarten or Grins).

Bulgaria, 1877-1947
This is a sample of household registers of villages of the Rhodope region in Bulgaria. The sources are civil and church registers, and cover Orthodox Bulgarians and Pomaks (Bulgarian-speaking Muslims) in a mountainous region bordering Greece. The data were transcribed by Ulf Brunnbauer.

France, 1846
This sample of the French census of 1846 was created by Rolf Gehrmann for the Max Planck Institute for Demographic Research, and covers 14 villages in 14 départements. The data for 10 of the villages were drawn from the collection of villages selected by Louis Henry for the reconstruction of the population of France from 1670 to 1829 based on family reconstitution. Four of the villages were an extension of regions not covered by this collection or were a replacement for villages for which no census data were preserved for 1846, or for which the census data were of poor quality or were not available. In the end, we chose to use 13 villages in the northern half of France for this article.

France, 1846 and 1856
The files for the small city of St. Emilion were created at the University of Bordeaux. The data were drawn from the French censuses of 1846 and 1856.

France, 1831-1901
The files for these villages from south-western France (the départements of the Dordogne, Gironde and Pyrénées-Atlantiques) were created at the University of Bordeaux. They were part of the French census of the respective years.

Germany, 1690-1713
In 1690, a status animarum was compiled for the parishes of Oldenburger Münsterland. The data for the parishes of Goldenstedt and Lutten were published (Sieve, 2010). In 1703, in the Prince-Bishopric of Münster, a deputy of the bishop visited the parishes of the deaneries of Vechta and Cloppenburg. A status animarum was prepared for this visit, and it has been preserved and published (Kock & Sieve, 2006). In 1713, a deputy of the bishop visited the parishes of the deanery of Cloppenburg. Another status animarum was compiled for this visit, and it has been preserved and published (Sieve, 2010). In addition, a number of status animarum books for several parishes of the Prince-Bishopric of Münster for the years 1709 and 1715 have survived and been published (Tandecki & Cloppenburg, 1995a, 1995b. The data used in this study were drawn from all of these published lists.

Germany, 1749-1751
The status animarum of 1749-50 for the Prince-Bishopric of Münster was the first enumeration of the population of this territory with clear instructions. The data were collected by the Roman Catholic priests at the end of 1749 or the beginning of 1750. They have been almost completely preserved. The data will be published according to the administrative units of that time. The data for the region of Stromberg, in the south-east of this prince-bishopric (Henkelmann & Wunschhofer, 2006), and for some other parishes (Tandecki & Cloppenburg, 1995a, 1995b, have already been published and have been used in this article.

Germany, 1749-1811
The data were drawn from soul listings (status animarum) for three Catholic villages in Baden-Württemberg (which, at that time, were part of the Bishopric of Constance). Oggelshausen belongs to the district of Tübingen, Landkreis Biberach; Dischingen belongs to the district of Stuttgart in the county of Heidenheim; Gögglingen is now part of the city of Ulm. The archival material can be found in the diocesan archive of Rottenburg, call numbers MF Nr. 9 382-9 384 (Oggelshausen), MF Nr. 19 662-19 663 (Dischingen) and MF 16344 -16345 (Gögglingen).

Germany, 1803
The 1803 census of Schleswig and Holstein was conducted two years after the census of Denmark and Norway, and according to the same rules. The data file was prepared by genealogists from Schleswig-Holstein for the Danish Data Archive. We have used only a portion of the data for this article -i.e. data for villages in Holstein and for the city of Altona. Altona was, at that time, the largest city in Holstein and the second-largest city in the countries ruled by the Danish king. It is now part of Hamburg.

Germany, 1846
The census of the German Customs Union (GCU) in 1846 was taken over the course of three days, but each member of the GCU was free to determine the procedure used. The data file was created by Rolf Gehrmann for the Max Planck Institute for Demographic Research, and covers 59 villages in 11 regions and samples of 10 cities. This is not a representative sample of Germany in 1846 in the strict sense: it does not include data for the north because these regions were not part of the GCU at that time, and some of the data for the south and the east are missing because the information was not preserved or the census did not report all of the members of the household individually. For the east, data from the 1858 census can be used as a supplement. The data file for Höhscheid was created by Ralf Rogge for the Max Planck Institute for Demographic Research, and covers Höhscheid, which is now part of the city of Solingen. This city has long been known for manufacturing fine swords, knives, scissors and razors. The census material can be found in the city archive of Solingen, call number StAS Bürgerrolle H 168.

Germany, 1858
The census of the GCU in 1858 was conducted in a manner similar to the census of 1846. The data file was created by Rolf Gehrmann for the Max Planck Institute for Demographic Research, and covers six villages in two regions (the government districts of Danzig and Posen). It is intended as an extension of the sample of 1846 to eastern regions, for which no data from 1846 were preserved.

Germany, 1861
This is the census of the GCU in 1861 for several villages in the government district of Sigmaringen in south-western Germany. The data file was produced at the Max Planck Institute for Demographic Research. The population was mainly Catholic.

Germany, 1867
This census of 3 December 1867 was processed as part of a collaboration between the Max Planck Institute for Demographic Research, the University of Rostock's Department for Multimedia and Data Processing, and the Landeshauptarchiv in Schwerin (Bestand 5.12-3/20 Statistisches Landesamt [1851-1945), with funding from the Ministry for Education, Science and Culture of Mecklenburg-Western Pomerania. The data used in this article cover a sample of rural regions.

Germany, 1900
This census of 1 December 1900 was processed as part of a collaboration between the Max Planck Institute for Demographic Research, the University of Rostock's Department for Multimedia and Data Processing, and the Landeshauptarchiv in Schwerin (Bestand 5.12-3/20 Statistisches Landesamt [1851-1945), with funding from the Ministry for Education, Science and Culture of Mecklenburg-Western Pomerania. The data file covers the whole city of Rostock.

Hungary, 1869
This is a sample of the 1869 census of Austria-Hungary (conducted on 31 December 1869), which was compiled by Péter Ő ri and Levente Pakot for the Max Planck Institute for Demographic Research. The surviving materials of the 1869 census are very unevenly distributed within the borders of the Kingdom of Hungary in 1869, while a large portion of the materials for present-day Slovakia has been preserved. The data cover the territories of present-day Hungary, Slovakia and north-western Romania. The sampling is based on nine regions: four in Hungary, three in Slovakia and two in Romania. The data cover the villages of all religious confessions within the Kingdom of Hungary: Roman and Greek Catholics, Lutheran, Reformed and Unitarian Protestants, Orthodox Christians and Jews. Information on mother tongue or ethnicity was not recorded in the census and could therefore not be used for sampling.

Italy, 1430
The 1430 -1432 catasto of Legnago (which, at that time, belonged to the Republic of Venice) contains information about the population and their property. The data file was created by Gianpiero Dalla-Zuanna. Further information about these data can be found in Dalla-Zuanna, Tullio, Leverotti, and Rossi (2012).

Ottoman Empire/Istanbul, 1885 and 1907
The 1885 (1300 h.) and 1907 (1322 h.) censuses were the first Empire-wide censuses undertaken for purposes other than taxation or military conscription. They were the first censuses to include information about women. The 1907 census is generally the more reliable of the two. The samples cover only 5% of the permanent Muslim population of five central districts of Istanbul. As occupations were recorded for only a small percentage of the respondents, those with non-manual occupations were over-represented. The data have already been used to analyze household structures in Istanbul (Duben & Behar, 1991). This publication provides additional information about these sources.
Poland-Lithuania, 1666-1804 (the CEURFAMFORM Database) The database includes data primarily from late eighteenth-century Poland-Lithuania on 26,655 peasant households belonging to 236 parishes and 900 settlements, with an overall population of nearly 156,000 persons. The data were derived from various types of population enumerations that listed individuals by residential units and where the kinship relationships were made transparent within each domestic group. These data include census microdata that were collected between 1790 and 1792 by the Civil-Military Order Commissions on the territories of the Crown of the Kingdom of Poland (including Ukraine) (49%). The data for the Lithuanian regions came from the materials of the fifth Russian revision list of 1795 (37%). The remaining 14% came from listings drawn from a variety of sources, though they are mainly status animarum or Seelenregister. The census-like microdata applied in this article currently represent the largest collection of population listings for households in this part of the continent. A further description can be found in Szołtysek (2014).

Russia, 1795
This is part of the fifth revision of souls in the Russian Empire, which was conducted in this region shortly after it became part of the Russian Empire following the second partition of Poland-Lithuania in 1793. The data cover 11 villages and have been published (Legun & Petrenko, 2003).

Russia, 1814
Appendix 2. Index of Patriarchy for 91 regions in historical Europe