Long-term demographic change and local socio-cultural patterns: marriages and household structure in 18–20th century Hungary

In past societies, where the replacement of the population was linked first of all to the institution of marriage, and where households (the units of co-resident and co-operating people either related to each other or not) were the core elements of societies with respect to demographic reproduction, work, and consumption, marriage customs and the rules of household formation had a direct and important impact on population development. On the one hand, the length of time spent in marriage and the proportion of those who never married directly affected the level of (marital) fertility. On the other hand, households were not only the framework of reproduction, work, and consumption, but they were influenced by and related to the types of subsistence, taxation, the rules of inheritance, or the demographic composition of the given societies. Therefore, the study of marriage customs and household structure in the past can reveal some key elements of the societies under investigation, and the literature related to the topic is extremely rich. In addition, since the 1960s research on marriage and household formation has become a crucial part of historical demography and family history. John Hajnal’s and Peter Laslett’s approach, in spite of the strong criticism, has remained an inevitable reference point so far. Hajnal’s attempt to create an explicative model of marriage customs and household formation rules, or Laslett’s classification of family households were an important phase of research because they incited a long series of comparative studies in this respect. As a consequence of the great S T U D I A


Introduction
In past societies, where the replacement of the population was linked first of all to the institution of marriage, and where households (the units of co-resident and co-operating people either related to each other or not) were the core elements of societies with respect to demographic reproduction, work, and consumption, marriage customs and the rules of household formation had a direct and important impact on population development.On the one hand, the length of time spent in marriage and the proportion of those who never married directly affected the level of (marital) fertility.On the other hand, households were not only the framework of reproduction, work, and consumption, but they were influenced by and related to the types of subsistence, taxation, the rules of inheritance, or the demographic composition of the given societies.Therefore, the study of marriage customs and household structure in the past can reveal some key elements of the societies under investigation, and the literature related to the topic is extremely rich.In addition, since the 1960s research on marriage and household formation has become a crucial part of historical demography and family history.John Hajnal's and Peter Laslett's approach, in spite of the strong criticism, has remained an inevitable reference point so far.Hajnal's attempt to create an explicative model of marriage customs and household formation rules, or Laslett's classification of family households were an important phase of research because they incited a long series of comparative studies in this respect.As a consequence of the great S T U D I A "Przeszłość Demograficzna Polski" 37, 2015, nr 3 DOI: 10.18276/pdp.2015.3.37-01amount of results related to Western Europe, marriage customs (first of all, age at first marriage and the final percentage of the never married) and household formation rules have been regarded as the key elements of the 'Western' demographic regime.At the same time, Hajnal's model has been refined several times, the cultural determinants of marriage and household structure have been completed by many other factors (see later in detail).But most of the experiences refer to Western Europe, and, in spite of some really considerable efforts, we know much less Central or Eastern Europe in this respect.This paper is a modest contribution which may help to better understand the characteristics of Central European marriage and household formation.
In this paper, I try to reconstruct the main contours of marriage customs and household formation by using and analysing the data of a region situated around the capital (Buda and Pest or Budapest from 1873 onwards), and consisting of about two hundred settlements (on the territory of the historical county of Pest-Pilis-Solt-Kiskun).The analysis focuses on two periods: the first is the end of the 18 th century when a greater amount of sources suitable for studying household structure and marriage customs emerged, and the second is the end of the 19 th century and the beginning of the 20 th century since when the series of official and professional population censuses have been available.Besides 18 th century household lists (Conscriptiones Animarum) and 19 th century censuses, I shall use the data of parish registers and vital statistics, and those of an 18 th century cadastral survey, and 19 th century statistics of food prices.All the data used in the course of the analysis are aggregated on settlement level, therefore this time the stress has been laid on the study of spatial differences, and, in some cases, changes over time.
This paper aims at testing the relevance of the Hajnal model at settlement level, following the county level analysis of Tamás Faragó (Faragó 2003) at a considerably less aggregated level.First of all, I have presumed that according to the Hajnal model one should find more or less clear ethnocultural patterns of marriages and household structure in a typical Central European multi-ethnic and multilingual region, in the middle of the historical Hungarian Kingdom.Here, migratory movements in the 18-19 th centuries mixed "Western" and "Eastern" populations (Germans, Hungarians, Slovaks, Serbs and Croats).Therefore, if ethnic or in certain cases denominational factors influenced marriage rules or household formation, that would be a strong argument supporting the Hajnal model.On the contrary, if other (economic, social, or regional) determinants emerged, that would clearly falsify the prevalence of the dichotomous model.Similarly, if we found the ethnic patterns of marrying and forming households persisting in the longer run, that would support Hajnal's approach which supposed an "Eastern" type of marriages existing until the second half of the 20 th century (Hajnal 1965: 101;Csernák 1997: 342;Tomka 2013: 50-60).By contrast, if these factors were altering in the shorter or longer term, that would refute the long term existence of an 'Eastern' marriage pattern at least in the studied area.

Results of former research
Research on marriage and the rules and patterns of household formation have been a central field of family history and historical demography since the 1960s.Therefore, the literature related to the topic is extremely rich, the survey of which cannot be the task of this paper.Besides some general statements on international literature, I would like to concentrate on Hungarian research findings and the gaps and unresolved problems in this respect. 1ccording to the Hungarian social historian Gyula Benda, the research in this field can be described by two different approaches (Benda 2002: 109-110).The first is a very strong intention to create models and classifications which make comparative studies of European societies or different (European and non-European) cultures possible.Thus, the research on household structure and family history has been dominated by the works of John Hajnal and Peter Laslett since the 1960s (Hajnal 1965(Hajnal , 1982;;Laslett 1972Laslett , 1983Laslett , 1988)).Despite their model having been later modified and refined (Laslett 1983), its culturalist and static approach and simplifying classification have remained, and have been strongly criticised so far. 2 As a consequence of the creation of the 'West-East' model, a great number of case studies and micro-level analyses were carried out in the last some decades, which succeeded in revealing the complexity of European marriage customs and household formation rules. 3 Benda regarded these local studies as the other ruling approach in family history.Naturally, to prove the existence of more differing models means the falsification of the simplifying dichotomous Hajnal-model, and to stress economic and social factors in household formation refutes the culturalist explanation which dominated Hajnal's and Laslett's works.In addition, dynamic approach and the understanding how households worked Péter Őri and their structure changed over historical time or household life-cycle got more and more stress during the last two or three decades.According to this approach, household structure was a continuously changing phenomenon, in which demographic, cultural, and economic determinants (for instance the rules of inheritance, mortality, migration, or the continuously changing labour-force demand of the family households4 ) played the decisive role, instead of exclusively cultural and mental ones.So -instead of revealing the regional patterns of family and household formation -recent research has aimed to understand the dynamics and functioning of households.Since longitudinal series of household lists or series of regularly repeated population censuses have seldom survived, the most frequent method is the analysis of cross-sectional data by age of household members (synthetic cohort approach), focusing on structural changes over the household life-cycle (Berkner 1972;Lundh 1995;Reher 1997;Faragó 2006;Szołtysek 2008;Pakot 2013;Őri-Pakot 2014).
In Hungary5 at the macro level Tamás Faragó proved that the frequency of complex household forms increased in Hungary between 1787 and 1828 (1977). 6lso at macro level, Faragó tested the relevance of the Hajnal line using data of eighteenth -century enumerations (2003). 7His results showed that there were regional patterns of marriage and household formation in eighteenth century Hungary, in which migration and mixing of ethnic groups played a decisive role.Thus, Faragó simultaneously refuted the dichotomous model and the mechanical regional division, and supported what may be termed the culturalist explanation of Hajnal.Important research efforts took place at the micro-level from the 1970s as well.A series of analyses using eighteenth-century household lists (Status Animarum) was published in the 1970s and 1980s.As a summary, Andorka and Faragó considered Hungary as being in a central position between the "West" and the "East" in terms of marriage customs and household formation (1983).Péter Pozsgai tried to clarify the household notion used by 18-19 th century Hungarian census takers.He carefully compared successive censuses in a North Hungarian micro-region and suggested the use of Chayanov's approach to understand household dynamics, and the revision of Laslett's typology in order to classify the households of a different -"Eastern" -society (Pozsgai 2000(Pozsgai , 2001a(Pozsgai , 2001b).8Ildikó Husz analysed a longer series of church household lists in Zsámbék, a large village close to Budapest, inhabited mostly by eighteenth-century German settlers (2002).In accordance with Faragó's results, she managed to prove that household structure changed over time in the first half of the nineteenth century, that is to say, more complex households became more frequent until around 1830, after which time the opposite development could be observed.In her view, the main cause of this process was the changing difficulty of leaving the parental home (particularly for non-heirs in a system of stem inheritance).In this respect, the role of cultural factors (the system of inheritance) and some exogenous ones (economic development, possibilities of having new plots and those of out-migration and mortality) appears to be important.Gyula Benda studied the social changes of a small Transdanubian town (Keszthely) between the middle of the eighteenth and nineteenth centuries.The process of inheritance was particularly stressed in his work on demography and household structure (Benda 2002(Benda , 2008)).He emphasised the role of farming and subsistence in household formation, which was reflected by differences by district within the town.He drew attention to the role of socio-professional status, the complexity of the larger settlements in economic and social terms, and to that of inheritance in forming spatial separation and household structure.In his view, the process of inheritance may cause the split of households as regards co-residence or sometimes common farming, and may partly explain the cyclical changes of household structure in accordance with two different types of succession (stem inheritance and equal partition among sons).Their works -together with Levente Pakot's newest paper using Chayanov's approach (2013) -open up new perspectives to understanding household dynamics, and draw attention to important new aspects in order to go beyond seemingly infertile attempts of regional division and cultural classification.
In this paper, I consider marriage customs and household formation as the two coherent parts of the same phenomenon, and concentrate on their determinants instead of classifying the studied communities by the aspects of Hajnal or Laslett.
Regarding that former research referred to a more aggregated level (country or counties, e.g.Faragó 2003), or some villages were studied at individual level (one important exception is the analysis of a national-wide sample taken from the individual data of census 1869 -Őri-Pakot 2014), and more refined statistical methods have hardly been used so far (except for Faragó 2003 andŐri-Pakot 2014), this paper may offer some new aspects to discriminate between the determinants of marriage customs and household formation rules.It uses data aggregated at settlement level in a very heterogeneous area, the number of settlements used in the analysis is higher than that in most of the researches, and the statistical method used here (multiple linear regression analysis) makes possible that at the same time we take more factors into account.The results must not be regarded as decisive ones with respect to the relevance of the Hajnal model, but we have to take into consideration that aggregate level data we have at our disposal cannot result in more robust regression models.But, in this respect, the most important aspect is whether ethnocultural or denominational factors or economic, social, or regional ones played some role.The results gained from the analysis can support Hajnal's approach or can refute it which means that the problem must be studied further at individual level (like in the case of the census data from 1869 -Őri-Pakot 2014).
The area under study: county Pest-Pilis-Solt-Kiskun in the 18 th and 19 th centuries 9Pest County had become one of the largest counties of Hungary, both in terms of area and of population, by the end of the 18 th century.Its territory was 10,711 km 2 before 1876 and 12,010 km 2 after that.In geographical terms, it is a highly varied entity, divided into two parts by the North-South flow of the Danube, while the Northern, hilly district of the county is also markedly different Long-term demographic change and local socio-cultural patterns... from the southern one, which is a part of the Great Plain of Hungary.From an agricultural point of view, the presence of the great river and the difference in topography and vegetation between North and South are highly important, as is the closeness of Buda and Pest, centres of administration and market.The original structure of the population and its towns and villages changed a great deal during 150 years of Turkish rule (in the 16 th and 17 th centuries).The region around Buda and Pest became more scarcely populated, and in the southern, flat part of the county, this condition continued to prevail in the long term, even after the war of liberation.Thus, the original density of villages and towns was never restored.Part of the population was killed, some escaped or moved to safer towns or villages which enjoyed privileges.Thus, a typical set of geographic, social, and economic conditions developed on the Plain, which is characterised by large flat stretches of bare land, poor in wood, with little rain, where the soil is not well suited for cultivation as it is often sandy or sodic.The Turkish rule led to the increase and consolidation of features such as low population density, widely scattered, relatively large towns and villages (Figure 1).This state of affairs was not fundamentally altered by the in-migrations of the 18 th century -the previous, originally scarce network of settlements was not restored (Beluszky 1999: 88-89), and right up to the middle of the 19 th century the most profitable economic activity was extensive stock breeding.These conditions also had their social aspects.Most of the population inhabiting the plain consisted of free peasants and craftsmen of the market towns in a relatively loose dependence on the landlords, and thus in a rather favourable legal position.The special conditions of the Plain can be described in categories of backwardness, peripheral position, or a frontier type existence (Beluszky 1999).What appears significant from our point of view is that those districts of the county which are on the Plain show a marked difference from the northern, hilly sections, even at first sight, and this suggests the existence of a unique demographic structure.
The original Hungarian population of the county, which had mostly converted to Calvinism in the 16 th century, only survived in the large market towns of the Plain and on the protected, marshy islands and floodplains of the Danube.In the southern part of the county this is where we find smaller villages in the traditional sense of that word (under a 1000 inhabitants).In those parts of the county which are close to Buda and Pest, as well as in the Northern, hilly parts, partially or fully depopulated villages were repopulated after the liberating wars (from the late 17 th century onwards).The new settlers significantly altered the ethnic and denominational character of the area: the organised repopulation of the Buda and Pest region resulted in Catholic German villages, while other parts became populated by Catholic and Lutheran Slovakians, Catholic Croatians, and Orthodox Serbs, while several of the estate centres (e.g.Óbuda) saw the emergence of major Jewish communities in the 18 th century.
Thus, the territory of the county unites vast flatlands which are part of the Great Plain of Hungary, the frequently flooded, marshy meadows and floodplain forests of the Danube, as well as some medium high mountains covered in forest.Farming ranged from extensive stock breeding through traditional crop cultivation to forestry, fishing, vine-growing, and vine-making.The type of farming might have been important with respect to marriage and household structure: according to the results of Hungarian researchers, traditional crop cultivation and extensive stock breeding in some places went together with more complex household structure (and earlier marriage), whereas forestry, fishing, or vine-growing were linked to the dominance of smaller (simple family) households (Faragó 1977;Benda 2002;Őri 2005;Őri-Pakot 2014).Naturally, the link between the two phenomena (ecotypes versus marriage and household structure) was indirect, the labour-force demand of the farms worked together with the special social conditions, inheritance system, size of plots, the number of farm animals, the possibilities of out-migration, etc.In some cases ethnicity and type of farming more or less covered each other: Germans in the northern hilly part of the county dealt first of all (although not exclusively) with vine-growing, the Slovak communities among the mountains lived on forestry, while Hungarian villages or market towns on the Plain could be characterised by stock breeding or crop cultivation.Multiple regression analysis can help to separate these composition effects in the course of the analysis.This paper cannot resolve all problems, but can be a good starting point for further analysis in this respect.
Of the major denominational groups of Hungary the only one we do not find here are the Greek Catholics, of all its ethnic groups only Romanians and Rusins are not represented.Therefore, to some extent, we may regard this area as one representing the country as a whole, while the settlement-level analysis may assure a deeper knowledge of local context, but also a considerable number of variables and communities with very differing economic, social, ethnic, and cultural conditions.

The demography of marriages (county Pest-Pilis-Solt-Kiskun in the 18-19 th centuries)
In this part, first I concentrate on the short term fluctuations of marriages, and the connection between the changes of marriages and grain prices.Changes in marriages are contrasted to those of death and births.The main goal of this kind of comparison is to better understand the role of marriages in this heterogeneous Central European context.The question to be answered is whether marriages played a regulating role in the given demographic regime by markedly reacting to changes in food prices or not.I would like to know not only the mean age at first marriages or the percentage of the never married, but also the stability or flexibility of these phenomena.Secondly, other features of marriage (age at first marriage and the percentage of the never married) are studied which can be described by the summaries of 18 th century enumerations (Conscriptio Animarum), and those of some late 19 th century population censuses (1890 and 1900 in this case).
Regarding short-term changes of marriages, we have to take into account the effects of economic conjunctures and the changing level of living standard, and mortality as well.In the following part, I deal with the correlation between the grain prices and the number of marriages, births, and deaths.Although in this paper the stress is on marriages, it seems to be worth dealing with the changes of all these demographic events in order to better understand the role of marriages in 19 th century Hungary.We have at our disposal the collection of the demographic events for all settlements of the county from 1828 onwards (Klinger 1972(Klinger -1984)), and the collection of food prices on the local market of the town of Pest between 1791 and 1870 (Kőrösi 1873).In the case of prices and demographic events alike, we used the yearly values compared to that of a nine-year period, in the middle of which the given value stands.The correlation between the changes of grain prices and demographic events (first of all marriages) has been studied in the case of Budapest and its rural neighbourhood (the county).In both cases, the direct impact of the changing prices (the changes of demographic events in the same year) and their lagged effect (observable in the next year) have been examined. 10 The relationship between the economic and demographic fluctuations, the short or medium term impacts of economic crises (high food prices) on demographic development have been studied for a long time.The analysis of the impact of economic crises on demographic events was one of the most important starting points of modern historical demographic research (Meuvret 1946;Goubert 1960).In their view, economic crises and mortality crises went hand in hand, but the effects on nuptiality or fertility were also considerable.Later on, a lot of evidences were found that great mortality crises were caused rather by epidemics so the correlation between food prices and rising mortality was not too strong. 11At the same time, the correlation between economic and demographic variables cannot be neglected, but the strongest link existed between prices or real wages and births and marriages.Births were much more sensitive than marriages in this respect (Bengtsson 2004: 38).But the correlations revealed at the macro level are weak and contradictory since economic fluctuations measured by the level of food prices differently affected the members of the given populations according to their age, sex, social or household position.The individual or household level decisions on food distribution also had demographic consequences besides the purely biological and economic factors (Bengtsson 2004: 39). 10 About the methodological aspects of this kind of analysis see Roehner 1990.About the results and shortcomings of the macro analyses of short term relations between prices and demographic events see Bengtsson-Reher 1998 and Bengtsson 2004.  1See for example Galloway (1988), Livi Bacci (1991).A detailed overview of the problem can be found in Bengtsson-Reher (1998) and Bengtsson (2004).In this case, I used a very simple method to reveal the correlation between changing prices and demographic events.The method used here is not suitable to reveal causal links between the studied phenomena, only common parallel changes can be pointed out.At this time, I could distinguish only between the capital and its rural background.In the studied period some economic crises with high grain prices occurred (in 1847, later in 1854, 1861 as a consequence of wars) and serious cholera epidemics also devastated partly as a consequence of wars and economic crises (in 1849, 1855, 1866).
The demographic effects of increasing grain prices were similar in the two studied areas if we consider the whole period.The increase in grain prices went together with the increasing number of deaths and the decreasing number of marriages, but the latter change was much weaker and not significant.A higher number of deaths and lower number of marriages were also observable in the next year, but the correlation weakened especially in the case of marriages.The decrease in birth number started in parallel (increasing number of spontaneous abortions, still births, registration problems), especially in the city, but this decline became significant during the next year (decreasing number of conceptions as a parallel phenomenon with growing prices).Therefore, the impacts that is to say the parallel changes of grain prices and demographic events can be pointed out although the correlations were weak.Economic crises and growing food prices went together with the postponement of marriages and increasing mortality from the beginning of the crises, and this effect lasted for a longer time.High prices affected the number of births, but much stronger that of conceptions.To sum up, the effect of economic changes on demographic development seems to be very probable even in the middle of the 19 th century both in urban and rural populations.But, in all probability, marriages did not function as some preventive checks in the 19 th century, probable mortality and fertility responses were stronger.We may conclude that marriages did not play a stronger regulating role in the studied region (through the changing age at first marriages or the intensity of marrying) in the short run by reacting to the changing living standards.In this respect, this region certainly differed from Hajnal's and others' basically malthusian view on North Western Europe.But there are some other aspects to be taken into account.Economic fluctuation might have affected different social strata differently, but in this case we cannot resolve this question.In the case of marriages, the growing number of widowed persons increased the number of marriages during the time of crises, and this weakened the strength of marriage response, particularly in the next few years.The other aspect is that the strength of the impacts might change over time.In the time of high grain prices in 1846-1847 and 1854-1855 we can observe the decline in marriage numbers (a two times and a half increase in grain prices in 1847 as compared to the average of the pre-crisis ten years went together with an 8 per cent decrease of marriages, in 1854 grain prices doubled compared to the level of the former decade, and the number of marriages decreased by nearly 35 per cent respectively).But this correlation had disappeared by the 1860s, and this weakened the strength of impact as regards the whole period.
Age at first marriage is a basic question in the demography of marriages.On the basis of its differences John Hajnal divided Europe into two cultural patterns (Hajnal 1965).The significance of the phenomenon is also shown by its impact on the level of the completed fertility of married women (Livi Bacci 1998: 100).By using Conscriptio Animarum as a source, we may estimate the age at first marriage at the settlement level for the 1770s, whereas the published census data enable us to calculate singulate mean age at first marriage from 1880 onwards.Table 3 clearly shows that we cannot consider the area situated east of the Hajnal line homogeneous in space and time.First, a clearly different urban pattern with relatively high age at first marriage becomes visible here.In Budapest, not only the males married at a surprisingly high age, but it was the case concerning females during the last decades of the 19 th century.We may consider Budapest an exception but other studies have also detected the existence of a characteristic urban pattern, particularly in the case of males.12Concerning the rural sample (the county), we may observe a real 'eastern' marriage pattern especially in the case of females.Age at first marriages of both sexes which was really low in the 18 th century increased considerably in the last decades of the 19 th century and in the first decades of the 20 th century.That increase finally resulted in a relatively high male age at first marriage after 1900.Therefore, we can conclude that there were important differences in the age at first marriage regarding sex and settlement type, and the changes over time were also considerable, particularly from the second part of the 19 th century onwards.The percentage of the never married around and after age of 50 confirms the results on urban-rural and gender differences observed above as well as the fact of changes over time.In the case of Budapest, obviously the specific labour-force demand of the great and growing city and the strong in-migration as its consequence formed the distribution of the population by marital status.In spite of the minimal increase in the rural percentages, we may conclude that marriage was a more or less general experience in this Central European rural region although the age at first marriage was increasing, and, especially in the case of males, it was not so low around and after 1900 as it was expected on the basis of the former research.
Searching for the determinants of the age at first marriage for the 18 th century, we used several variables calculated on the basis of Conscriptio Animarum, census 1784-1787 and the cadastral survey in 1789 (Appendix).The statistical method used here was multiple regression analysis in order to take into consideration as many factors influencing marriage customs as it was possible.The goal of the analysis was rather limited in every case: on the basis of these data and variables, the aim of the analysis could not be to fully understand the determinants of marrying, but to see which kind of variables had some explanatory force with respect to marriages.In other words, the results should be considered with regard to the Hajnal model: whether the cultural variables which could be used in the analysis have some explanatory force or other factors had some importance.Thus, the R 2 values do not appear to be of great importance compared to the character of the significant variables.Naturally, this means that the results can be only the starting points of the further analysis, but at this aggregate level and by using these data we cannot reach more.Regarding the age at first marriage, first of all, the role of ethnicity (spoken language or mother tongue) was to be understood.Secondly, the age at marriage of the other sex or the variables on household structure were studied to see if the several elements of the Hajnal model really formed a coherent system in the studied region.
By the help of multiple regression analysis, we can see that the mean age at first marriage of the other sex has proven a significant factor in this respect.It was possible because age difference was not too large and the customs of relatively early or late marriage were characteristic of both sexes.In the case of males, the connection with household composition can be also seen: the more complex the households were, the earlier the males married (and necessarily vice versa, we cannot see the direction of the link).All in all, the elements of a complex marriage and household patterns are visible here.At the same time, regional and economic factors (shore of the Danube, percentage of meadows within the cultivated land and percentage of peasants or copy holders or feudal tenants) and ethno-cultural characteristics also mattered.The high percentage of copy holders (that meant land using or land owning) went together with early marriage, while meadows and the Danube shore which offered poorer opportunities for living on arable lands and where the social composition of the population was more varied were the typical locations of later marriage among males.Regarding the ethnic character of the villages and market towns under study, Hungarians married earlier both in the case of males and females, while especially German women married later as compared to other ethnic groups (Table 5).But we have to take into account, that a greater share of the variance of the age at first marriage cannot be explained by using the available variables (Tables 5, 6).Regarding age at first marriage at the end of the 19 th century (Table 6), we may conclude that the ethno-cultural effect seems to have decreased, only Roman Catholic females showed a significantly greater propensity for early marriage.The effects of age at first marriage related to the other sex (among males) and of household complexity (among females) remained.There was a regional factor which proved significant: the North-western hilly region with many German villages can be characterised by late marriage among females, while in the North-eastern hilly district of the county inhabited mainly by Hungarians and Slovaks, early marriage was much more common.The high proportion of the population living on agriculture resulted in earlier marriage of women, while high marital fertility had a complex and mutual connection with the age at female marriages.The positive correlation between these two factors in all probability means that both high fertility and later marriage were the elements of a traditional demographic pattern which prevailed in the North-western hilly district of the county with small villages inhabited by Roman Catholic (originally German) population, where traditionally late marriage and high marital fertility still remained at the beginning of the 20 th century.All in all, the age at marriage did not depend on exclusively cultural factors, type of settlement (urban or rural), sex, region with its characteristic social and economic conditions, farming, and to some extent ethno-cultural feature, all mattered in this respect.The results support to some extent Hajnal's model concerning the ethno-cultural character of marriage, but, at the same time, draw attention to the simplifying feature of that kind of classification: the differences in marriages by ethnic group can be explained by including much more aspects besides cultural norms.

The characteristics of household structure
Household formation was regarded as a key issue by John Hajnal when formulating his dichotomous model.At the macro level, we can calculate two variables concerning the size and structure of households: mean household size and the number of conjugal units13 per household (the number of married males per household).Although both variables were used many times,14 we have to take the problems relating to them into consideration.
Mean household size or household size itself has by no means a direct correlation with household structure, large households might have consisted of one single nuclear family (Faragó 1977: 108).At the same time, in the age of Joseph II, mean household size had a very strong positive correlation with the variable of household structure (number of married males per household) at the national level and in county Pest-Pilis-Solt as well.In the latter case, the Pearson correlation coefficient was 0.791 between the two variables. 15We may conclude that not every large household was necessarily of complex structure, but most of complex households were large at the same time.Thus, we shall not leave the problem of household size out of consideration although when examining the problem of household structure by using multiple regression analysis, mean household size will be omitted because of the strong correlation between the two variables.
As for the number of conjugal units per household, we face several considerable problems.First of all, it is difficult to separate the regions of complex household structure from those of nuclear family households.Even in Croatia, the number of married males per household was no more than 1.7 in 1787 (Faragó 1977: 109), which fact clearly indicates that only a certain -although considerable -percentage of the households was of complex structure. 16By all means, the border line between the two zones must have been very uncertain and they have become hardly visible by using such crude variables.Similarly, we cannot distinguish between the different sorts of multiple family households: between stem families and joint families.Moreover, the used ratio probably underestimates the complexity of households: one part of the household heads must have been widowed or unmarried persons, and among household members there must have been widowed and unmarried persons around whom nuclear families might have been organised.
Another serious problem refers to the compatibility of 18 th century and 19 th century data.18 th century and early 19 th century sources registered population by families ( familiae) which meant the units of co-resident and commonly working and consuming people either related to each other or not.From 1850 onwards, censuses used -following the Austrian example -the concept of Wohnpartei, which originally might have been the equivalent of nuclear family, but in practice it was not far from the former household category.Anyway, the notion of household became more obscure and the practice of census takers differed to some extent.Sometimes they registered families of more generations, real multiple family households, sometimes persons evidently related to each other and co-operating according to the notes of the enumerators were divided.After 1869, the confusion might decrease, the flat (apartment) was used as a unit of enumeration which was called sometimes Wohnpartei, sometimes household.In 1869 and later, Wohnpartei meant by no means nuclear family although the separation of relatives can be observed many times.Households (the groups of co-resident people) very often contained more nuclear families and co-resident relatives and non-relatives as well.But it is clear that this more obscure notion of household left room for different interpretation of census takers.Therefore, the diminishing household size and the decreasing number of conjugal units per household between 1785 and 1869 in the county (Table 7) may result to some extent from this changing theory and practice.1900, Census, 1910, Census 1920.Otherwise mean household size did not change considerably between 1869 and 1920.The same is true for the variable of household composition: the values are waving, only in the case of the county one can conclude that in the longer run it had diminished by 1920.It refers to the number of married men per household and the number of ever married men per household as well.About the Budapest-county differences it can be told that household structure was stably more complex in the rural background of the capital, whereas mean household size was sometimes higher in Budapest (1880Budapest ( , 1900Budapest ( , and 1910).Here we have to take into account the effects of in-migration, the role of domestic servitude which was much more important in the case of the city.
Focusing on settlement level data and examining the stability of spatial patterns in household size and structure, we consider the significant correlation between two series of data as the sign of remaining relative spatial differences (Table 8).Thus, we can conclude that the spatial structure of mean household size gradually changed in the 19 th century.Between 1869 and 1890 the change was naturally small while between 1785 and 1869 much greater, when comparing 1785 and 1890 the original spatial pattern almost disappeared.At the same time, the spatial pattern of household structure changed less, the correlation remained stronger between 1785 and 1890.But the very strong correlation between household size and structure in the 18 th century had become much weaker by the end of the 19 th century, in other words, households became somewhat smaller, their size was influenced much more by differences in child number or in the number of servants and employees than by the number of co-resident nuclear families.Multiple regression analysis17 carried out for the variable of household structure revealed correlations with ethno-cultural, economic variables and those with marriage customs (age at first marriage) (Tables 9, 10).At the end of the 18 th century household structure had a pronounced ethnic character, households were more complex in the villages where Southern Slav population lived.This ethnic determinant had disappeared by 1890.Similarly, strong correlation existed between mean age at first marriage and household structure: in the 18 th century early male marriage went together with more complex household structure (differences in female marriage were much smaller), whereas the same correlation existed between female marriage and household composition in 1890.But in this case the direction of the effect is not clear: we cannot decide at this level of analysis whether a special cultural pattern of complex household structure made early marriage possible or the custom of early marriage resulted in the higher probability of multiple family households.It is more probable that in some villages the higher share of multiple family households and the custom of early marriage were the elements of the same cultural pattern.In the 18 th century, settlement type (village) had a positive effect on the complexity of household structure.Villages -in contrast to market towns and outskirts -were inhabited first of all by feudal tenants who lived on arable lands, these social and economic elements (landownership and farming) might have influenced household structure as it was also pointed out by other research (Őri-Pakot 2014).On the contrary, the high proportion of vineyards and forest around the villages had significant negative impact on household structure: the social composition of the population, subsistence, and the type of farming did not admit or demand the maintenance of large and complex households.By the end of the 19 th century population size emerged as a significant determinant, that is to say, the larger settlements, market towns situated typically on the Great Plain could be characterized by smaller households of simpler structure.The negative effect of the proportion of Calvinist population seems to be much more a regional phenomenon than a cultural one, but the content of this correlation can be hardly explained here.To sum up, household structure at the end of the 18 th century had a strong ethnic determinant created by 18 th century migration: villages inhabited by Southern Slavs were the typical location of complex household structure.This strong correlation had disappeared by the end of the 19 th century.Otherwise household structure was significantly linked to age at first marriage and type of farming, landownership, and customs of inheritance, which remained hidden in this study, appear to be decisive factors in this respect.In contrast, at the end of the 19 th century the role of population size seems to be more important as a result of the developing social and economic changes in the neighbourhood of Budapest.In other words -similarly to marriage customs -at the end of the 18 th century there were signs of ethnic differences in household structure, but that was resulted from 18 th century migration which mixed culturally very different populations (Germans and Southern Slavs).But beside that, economic factors (the type of farming) also mattered, which is in line with the results of other researches (e.g.Őri-Pakot 2014).Therefore, the background of these 'cultural' differences must be studied at individual level, where a much more complete set of variables are available in some cases (e.g. the censuses in 1850, 1857, or 1869) and the composition effect can be controlled to a larger extent in the course of the analysis.

Conclusions
This paper examined the demography of marriages and the basic features of household size and structure in a region around the Hungarian capital, Budapest at the end of the 18 th century and in late 19 th and early 20 th centuries.Changes over shorter and longer time were considered as well as the determinants of the age at first marriage and the composition of households.The area under study was a typical multi-ethnic Central European region where the geographic, economic, and social conditions were also varied.

Péter Őri
The medium-term fluctuation of marriages was not independent of the changes in grain prices, but the effect of rising prices was small, conceptions and deaths appear to have been more sensitive in this respect.This result fits into the view of general marriage which was characteristic of the 'eastern' marriage pattern according to Hajnal's model.Age at first marriage cannot be regarded according to the dichotomous Hajnal model.There were considerable differences between males and females, urban and rural settlements, and, in addition, the changes over time are not insignificant either.By using settlement level data, the role of ethnicity, regional differences, the age at marriage of the other sex, household composition appear to have been important.In the composition and size of households, urban-rural differences were also important, whereas the changes over time seem to have been smaller as compared to the age at first marriage.In the 18 th century Southern Slav ethnicity was an important determinant of multiple family households, this factor had disappeared by the end of the 19 th century.Besides mean age at marriage, in the 18 th century the type of farming, at the end of the 19 th century population size (the level of urbanisation) mattered in this respect.Therefore, the fade signs of cultural patterns appear especially in the 18 th century, the higher frequency of multiple family households were connected, first of all, to Southern Slav communities, whereas later marriage to German settlers, but this ethnic pattern began to disappear in the second part of the 19 th century.Furthermore, other influencing factors also played an important role: type of farming, location, type of settlement, social conditions also affected age at marriage and household formation as well.At the same time, we may come to the conclusion that household formation and age at marriage were closely related to each other, it is worth considering them as the elements of the norms regulating partnership and co-residence.But at the same time, we have to take into account that these norms were created by many factors and cannot be explained exclusively by ethnic origin or place of origin, and this fact does not perfectly fit into the dichotomous model formulated by Hajnal fifty years ago.

Figure 1 .
Figure 1.County Pest-Pilis-Solt-Kiskun (before 1950) within historical and contemporary Hungary, and the settlements of the county

Table 1 .
The correlation between grain prices (on the market of Pest) and demographic events on the territory of county Pest-Pilis-Solt-Kiskun, 1832-1866 * Correlation is significant at the 0.05 level.

Table 2 .
The correlation between grain prices (on the market of Pest) and demographic events in Buda, Pest and Óbuda (later Budapest), 1832-1866 * Correlation is significant at the 0.01 level.

Table 6 .
Multiple linear regression analysis for the variables of the singulate mean age at first marriage, county Pest-Pilis-Solt-Kiskun, 1900 Sources: Census 1890, 1900 in county Pest and Censand 1910.d 1910.

Table 7 .
Mean household size, number of conjugal units per household in 18-19 th centuries, county Pest-Pilis-Solt-Kiskun and Budapest

Table 9 .
Multiple linear regression analysis for the variable of household structure, county Pest-Pilis-Solt, 1774-1789

Table 10 .
Multiple linear regression analysis for the variable of household structure, county Pest-Pilis-Solt-Kiskun, 1890