Chinese Historical Microdemography - Couverture rigide

Joint Committee On Chinese Studies (U. S.)

Couverture rigide

ISBN 10 : 0520083067 ISBN 13 : 9780520083066

Editeur : University of California Press, 1995

Synopsis

Using local studies to answer global questions, this compilation challenges traditional notions concerning historical Chinese population trends. Genealogies, epitaphs, and household registers are some of the local and primary materials used to examine the important issues of fertility, mortality, family structure, and migration patterns.

A propos du livre

À propos de l'auteur

Stevan Harrell is Professor and Chair of the Department of Anthropology at the University of Washington and coeditor of Chinese Families in the Post-Mao Era (California, 1993).

À propos de la quatrième de couverture

Using local studies to answer global questions, this compilation from eight scholars takes on traditional notions concerning historical Chinese population trends. Provoking rather than defining, these studies challenge some of the prevailing theories on demographic rates and family structure in late imperial China; they challenge the ideas that the Chinese were a low-fertility population and that population growth in the late imperial period was interrupted by severe mortality crises. Using local and primary materials - genealogies, epitaphs, and household registers - this collection examines and explores the important issues of fertility, mortality, family structure, and migration patterns. With the family-level data from those unique sources, this book investigates and illuminates the demographic processes behind late imperial China's population growth.

Extrait. © Reproduit sur autorisation. Tous droits réservés.

Chinese Historical Microdemography

By Stevan Harrell, editor

University of California Press

Copyright 1995 Stevan Harrell, editor
All right reserved.
ISBN: 0520083067


One
Introduction: Microdemography and the Modeling of Population Process in Late Imperial China
Stevan Harrell

The study of Chinese historical demography is still in its infancy.1 Since the publication of Ho Ping-ti's impressive summary work, Studies on the Population of China , in 1959, however, a sort of "standard model" of what happened has been accepted in most Western scholarship. This model sees a gradual but bumpy rise in population, from about 50,000,000 in the Han dynasty to over 430,000,000 in the mid-Qing, and the numbers rise and fall with the fortunes of the dynasties (Ho 1959:27778).

This model has recently been challenged explicitly from two directions. Bielenstein (1987) has asserted that the whole idea that mortality crises were serious enough to cause declining population for more than a few years is one that does not make sense either in terms of what we know about demographic rates or in terms of the figures we can derive from censuses and tax-registration records throughout various dynasties. He sees instead a curve that although bumpy is always rising, with changes in slope but never an actual turn downward. At the same time, Skinner (1987) has challenged the standard model from a different direction: he has shown through an internal statistical analysis that some of the records thought to be among the most accurate used by Ho and others to formulate the standard model are in fact based not on any kind of empirical population counts but on a simple algorithm applied annually to the previous year's records. Skinner's demonstration was for Sichuan in the mid-Qing only, but if those records are spurious, Skinner asks, how can we trust other records from remoter times based on even more obscure methodologies?

There is also a third, and more implicit, challenge to the standard model: a challenge that accuses the model less of inaccuracy than of incompleteness and oversimplification. The standard model has been accepted on the rather naive assumption that processes were similar everywhere in the country, that

the effects of the dynastic cycle were similar within and across regions. But we now know from the grand conceptual model put forth by Skinner (1985) and from various locally based empirical studies that the economic conditions faced by different regions, or by core and peripheral areas within the same region, were in fact different in their nature and in their timing. We would expect regions to be affected differentially not only by factors that raise or lower fertility and mortality but also by migration, as indeed Ho pointed out in his original work (Ho 1959:13668). But as in the case of other factors, migration has been studied from data that purport to portray its aggregate results rather than from descriptions of its actual process.

If the standard model stands challenged, then, it is more than anything because it is based on aggregate figures compiled by governments for taxation and land-registration purposes rather than on any kind of population records built upon life histories of individuals. The great advances made by European historians of population in the last two decades have been based on individual records, upon the detailed analyses of small populations taken as representative of a particular section of the larger whole. For certain European populations we now know not just aggregate counts but also the fertility, nuptiality, and mortality processes that have contributed to these counts. (See Stone 1981.) If we are going to understand population process in China, we are going to have to study the process at the individual and family level. A start has been made in this direction for the twentieth century. The articles published in Hanley and Wolf (1985) represent the beginning of historical microdemography in China. In their introduction to that volume, for example, Wolf and Hanley develop the idea that there was a "Chinese regime" of demographic process, with basic features of early and universal marriage, moderate to high marital fertility,2 and joint (or grand) family organization.3 They contrast this regime with the "Japanese regime" of later marriage, low fertility involving conscious fertility limitation, and stem family organization, as well as with the northwestern European regime described by Hajnal (1965), in which marriage is late and not universal, merital fertility high, and family organization primarily nuclear (Wolf and Hanley 1985: 16).

There has been particular progress in recent decades in understanding one aspect of this demographic regimethe developmental cycle of the Chinese family. Beginning with the work of Maurice Freedman in the postwar period (1958, 1966) and proceeding to the empirical studies of Arthur Wolf (1985a),Wolf and Huang (1980), Myron Cohen (1968, 1970, 1976), Burton Pasternak (1983), and others, we have gained an understanding of the demographic and ecological/economic conditions under which the Chinese joint family system does or does not reach full fruition. We find in general that prosperity itself encourages the development of joint families, both by increasing fertility and by delaying family division, and that

certain economic conditions, particularly those that encourage pooling of either labor or capital, also delay family division (Cohen 1976; Harrell 1982). We have come, in fact, to the point of challenging earlier assertions by anthropologists Francis L. K. Hsu (1943) and Fei Xiaotong (1939), who surmised that the joint family was an exclusive preserve of the elite. Wolf (1985a) in particular has shown that opportunities to form joint families extended to all but the poorest strata of the population in the nineteenth and twentieth centuries, and Lee's recent work (Lee and Eng 1984, Lee 1990) has demonstrated a developmental cycle for nineteenth-century Liaoning farmers in which families commonly grew to the joint stage of married brothers and their wives living together, and often even beyond that.

These initial steps toward understanding the actual processes behind China's population history are very useful, but they also lack something vital to the whole enterprise: historical depth. That their detailed analyses are based entirely on late-nineteenth- and twentieth-century records leaves open the possibility that something very different might have been going on in earlier periods. If we wish to understand the bumps in the long-term population curve, we need to extend to earlier periods the study of the same factorsfertility, nuptiality, mortality, family organization, migrationthat have made at least provisional sense of the twentieth-century data.

The conference that gave rise to this volume was suggested by Professor G. William Skinner on the basis of his hope that the study of genealogies (and secondarily other available historic microsources) would be able to bring the detailed examination of particular cases to bear on the general problems of China's population history. All of the papers represented in this volume approach Chinese population history from this perspective: they consider detailed evidence for small populations in order to observe trends in the regional history of populations. To understand how they do this, we must first look at the nature of the sources.

Genealogies And Household Registers As Sources For Demographic Research

Genealogies (more properly, genealogical records) have been compiled by Chinese lineages since the Song period (Ebrey 1986). By the Ming and Qing, it is quite probable that most large lineages in China (and a great number of smaller ones) compiled genealogical records of some kind. These records range in size from hand-scribbled notes on odd scraps of paper, kept by a few families to keep track of their ancestors' birth and death dates, to multivolume, printed sets published by large and wealthy southern clan associations. However, neither of these extremes is particularly useful for demographic research: the scribbled scrap has too few cases and the

encyclopedic reference probably included people in return for a subscription fee rather than according to any real genealogical principle.

What interests the demographic historian is the midsized genealogy containing a few hundred to a few thousand entries. These genealogies were compiled and usually printed (though sometimes handwritten) by literate members of large local lineages and typically included all members of the lineage on purely genealogical grounds. Such genealogies still survive in great numbers in the home towns and villages of the lineages themselves (Wang, chap. 8), in spite of the great amount of genealogy-burning that went on during the Great Proletarian Cultural Revolution. A not quite so large but still impressive number are found in libraries both inside and outside China, and many of these, in turn, have been collected by the Genealogical Society of Utah, a research arm of the Church of Jesus Christ of Latter-Day Saints.4

A good genealogy is much more than just a table of names and dates. It typically includes a series of prefaces by famous members or friends of members of the lineage, a history of the lineage, biographies of prominent members, detailed diagrams of how to find the graves of certain important ancestors, descriptions of ancestral halls, texts to be used in hall and grave worship, rules of conduct for lineage members, tables of land held by the lineage or by its constituent branches, and other documents relating to history, administration, and ritual. It is a wealthy lode of information about local society and history, one but minimally utilized by historical researchers so far.

For demographic purposes the most important part of the genealogy is the shixi tu (chart of the system of generations) and/or shixi biao (table of the system of generations), which contain detailed entries for each male member of the lineage who reached a certain age, as specified in the rules of compilation. A complete entry for a lineage member typically contains the following information: the man's personal name (ming ), courtesy name (zi ), and style (hao ), if he had one; his generation and birth-order rank within that generation (hang ); the name of his father and his birth order among the surviving sons of that father; if he was adopted, the name of the natural father; any official titles or imperial honors the man may have held; the year, month, day, and time of his birth; the year, month, day, and time of his death; and the place where he was buried. The entry also includes the name of his first wife, occasionally together with the name of her father and her birth order among her sisters; the year, month, day, and time of her birth and death, her place of burial; and the number of her sons, along with their names (a few genealogies list daughters and the men they married). Then the information is repeated for any subsequent wives the man might have had after his first wife's death and for any concubines who bore sons.5

On the surface, this kind of data appears to be a historical demographer's dream; there are thousands of cases, each gives detailed information, and birth and death dates are given directly and do not have to be deduced from self-reported ages. One must remember, however, that genealogies were compiled not for the convenience of historical demographers but in order to record the birth and death dates of ancestors to whom lineage members owed worship obligations. There are thus certain gaps. Although women marrying into the lineage, the wives and mothers who eventually become ancestors, are recorded, in many genealogies the information on these women is much less complete than that on their husbands and sons. However, the daughters of lineage members, who eventually move out and become someone else's ancestors, are rarely recorded and then not in much detail.6 Since children who die young are typically not accorded full ancestral rites, most genealogies have a compiler's rule that excludes from full entries those boys who died before a certain age, usually eighteen or twenty sui .7 The purpose and methods of compilation thus introduce gaps and biases into the available data.

That genealogies are compiled for ritual rather than demographic reasons also contributes to a further problem: the incompleteness of records. Not all entries in any genealogy contain the full range of data; studies (Harrell and Pullum, chap. 6; Telford, chap. 3; Harrell 1987) show that wives' entries usually contain less complete data than their husbands' and that the entries of men who had many descendants are usually more complete than those of men who had no sons or whose lines of descent died out after a few generations. To perform demographic analyses based on genealogies, then, one must infer what is not there; one must strike a delicate balance between making unwarranted conclusions based on incomplete data and making no conclusions at all because the data are incomplete. One must find ways to infer without imputing too much. The papers in this book that are based on genealogies (Harrell and Pullum; Telford; Liu; and Wang) continue a line of research that began with papers by Liu Ts'ui-jung (1985) and Harrell (1985), both of whom attempted to derive demographic information from genealogies, something that had not been done previously. Each of us made a different assumption: Liu took the data in the genealogies as complete and calculated actual vital rates, whereas I assumed that we could not know the nature of the biases and thus attempted only comparisons of one subpopulation with another and refused to estimate actual rates. It now seems clear that we both had much to learn. Liu's rates were probably off because she did not allow for infant mortality in her computations of such measures as life expectancy and total fertility. My analysis was overly cautious, not even looking for ways to estimate actual rates, and thus not allowing comparison of the Chinese population with other historic populations.

In this volume, however, having learned from those early works, all the authors working with genealogies have calculated demographic rates based on some more realistic assumptions about the relationship between what actually happened and what is recorded in the genealogies. Many of the assumptions used are no doubt still somewhat off the mark; but we are closer than we were, and we can begin to use genealogical data to ask questions about topics of general interest such as long-term fertility and mortality trends, the nature of the developmental cycle, and the demography of overseas migration.

As for genealogies, so for other kinds of historical demographic sources. At least two papers in this volume (Ebrey and Lee) rely partly or wholly on sources other than genealogies: Ebrey on epitaphs and Lee on the household registration records of the Qing imperial banner system. Like genealogies, these sources were not compiled for demographic purposes and thus can be assumed to have certain biases, as will be explained more fully by the authors in their papers. But as does the work based on genealogies, the work based on epitaphs and records attempts to understand and adjust for the biases introduced by the nature of the data and thus helps us come close enough to what must have been the historic reality that we can ask questions that should be of interest to historical demographers and demographic historians everywhere.

A Proposed Demographic Regime For Late Imperial China

The first use to which the papers in this volume put demographic data derived from genealogies and similar sources is in making preliminary guesses about the nature of the demographic regime in the empire as a whole, or at least in its constituent geographic regions. This demographic regime has two salient characteristics. First, the regime is driven primarily by mortality factors rather than fertility factors: age-specific marital fertility in particular seems to remain remarkably constant, whereas mortality changes cause differences in years of exposure and cause net reproduction to rise or fall. Second, the factors that drive this regime rise and fall with the regional manifestations of the dynastic cycle. Using the specific data and calculations from the empirical studies of fertility and mortality found in the papers by Ebrey, Telford, Lee, Liu, and Harrell and Pullum, as well as other published and unpublished work, I want to formulate a preliminary model of this regime based, on a curious and very significant fact: even though genealogies and household registers studied so far encompass a tiny fraction of the Chinese population (a few tens of thousands out of a few hundred million, or perhaps a one-in-ten-thousand sample), they all show more or less the same trends happening at the same time. This fact suggests that

these sources, no one or few of which may have any significance in itself as far as empire-wide trends, when combined force us to sit up and take notice. Is it a coincidence that they all show the same patterns? At least we have data that are reasonably accurate (Harrell 1987; Telford 1990), though not representative, on which to base further research.

The model I wish to propose on the basis of these data is thus a simple one: population growth did indeed go up and down with the dynastic cycle,8 and it did so as a consequence of a particular combination of vital rates. In periods of economic prosperity and political peace, mortality went down for both sexes at all ages, age at marriage for males went down, and the proportion of married males went up. Conversely, in periods of turmoil, age-specific mortality went up at all ages for both sexes, male age at marriage increased, and the proportion of married males declined. Throughout all this, female age at marriage remained fairly constant, as did age-specific rates of marital fertility. These factors in themselves, of course, would be sufficient to cause twists and turns in rates of population growth. Mortality increases at premarital ages would preclude people from being born, and mortality increases at marital ages would eliminate women from the reproductive pool if they or their husbands died, since widow remarriage was frowned upon (though hardly unknown) in the traditional family system. Increased male age at marriage would also increase the likelihood of early widowing, and would thus lower crude birth rates in the absence of changes in marital fertility. It would be nice to know more about the reasons for the lower proportion of males marrying. Was this caused by a higher rate of female infanticide or differential care that discriminated against girls, which would mean that fewer males could marry because there were fewer women to marry (assuming the impossibility of polyandry, which was rare enough to be nothing but a curiosity among the Chinese), or did it somehow derive purely from economic conditions that either skewed the marriage market, leaving unmarried females at the top and males at the bottom, or otherwise made it impractical for young people to arrange marriages? A skew in the marriage market would mean that a lowered proportion of males ever married happened along with a continued virtually 100% marriage rate for women; difficulties in arranging marriages would lead to a lower nuptiality rate for women also. Since genealogies typically don't give us very much information about daughters, we have no way of telling. So let us look at the figures we do have, for age-specific mortality rates, for nuptiality of males, and for age at marriage for both sexes.

Mortality

Several studies on mortality suggest a common pattern: mortality was moderate in the mid-Ming, rose with the interdynastic turmoil of the seventeenth

century, went down again quite rapidly with the Qing pacification, and then rose slowly from the first half of the eighteenth century to the end of the Imperial period. Let us look at some of the case studies that support this generalization.

Telford (chap. 3) finds big net changes in reproduction rates during the crisis period of 163545 in Tongcheng, Anhui. Before this period, reproduction rates were fairly high, though they did show a steady decline in the sixteenth century, perhaps owing to gradually deteriorating social conditions in the middle and late Ming. Then, with the interdynastic crisis, reproduction rates dropped suddenly and were even negative for a single cohort whose prime reproductive years came during the transition. After social stability was restored at the beginning of the Qing, the rates went up again and remained high for the remainder of the period studied by Telford, which ends before the decline of the Qing would be expected to become visible. When Telford examines the causes of the decline in a multiple regression analysis, he finds the variable that explains the greatest portion of the variance is the average age of the wife at the birth of the last son, which, given the natural fertility that Telford has found in this population and the general ban on widow remarriage, should be a fairly accurate proxy for the age at which the first parent dies. The next most important factor in his stepwise analysis is the percent of men unmarried, which is of course a nuptiality factor. Last comes the wife's age at marriage (also see Table 3.3).

Similarly, Liu Ts'ui-jung (chap. 4) shows for a single lineage in Shaoyang, Hunan, that both mean age at death and life expectancies at particular ages follow the same kind of curve. High in the early Ming, they go down steadily and then bottom out in the cohorts that show the effects of the Ming-Qing transition. Then they rebound again at the beginning of the Qing before starting another decline that continues throughout the eighteenth and nineteenth centuries. (See Table 4.9.)

In a previously published paper, Liu (1985) used different measures to demonstrate similar findings for two lineages in Xiaoshan, Zhejiang. In Table 1.1, reproduced from her paper, survivorship values can be seen to go down between the mid-eighteenth century and the mid-nineteenth.

Harrell and Pullum (chap. 6) have found similar figures for death rates in three lineages in Xiaoshan County, Zhejiang. Unfortunately, their material does not extend back beyond. 1600, but it shows the expected pattern of fairly high death rates 1600-50, much lower rates 16501700, and then a progressive climb afterwards. Because of some questions about the data in the Wu lineage genealogy, the figures for the Wus stop at 1800 (seeTable 1.2).

Another analysis, by birth cohort, for another lineage in the same area shows a similar pattern, though perhaps not as clearly, given that the Ns are very small for the Lin lineage (see Table 1.3).

TABLE 1.1. Survivorship Values (1x ) for Two Xiaoshan Lineages   Females Males   Shen Lineage Xu Lineage Shen Lineage Xu Lineage Ages 1695 1709 1785 1799 1695 1709 1785 1799 1695 1709 1785 1799 1695 1709 1785 1799 1519 35643 32900 37773 27308 38795 32680 38970 33575 2024 34488 31583 36710 25750 37863 31438 38050 32388 2529 33200 30058 35488 24050 36738 29955 36940 30755 3034 31745 28295 34155 22310 35508 28353 35725 28565 3539 30083 26265 32723 20495 33878 26373 34183 25753 4044 28170 23920 30883 18573 31495 23748 32048 22318 4549 25958 21273 28228 16510 28228 20458 29208 18368 5054 ___ ___ ___ ___ 24063 16608 25603 14143 5559 ___ ___ ___ ___ 19193 12460 21278 10003 SOURCE : Liu Ts'ui-jung 1985:59, Table 2.17.

TABLE 1.2. Total Death Rates for Males, Ages 2099, Shi and Wu Lineages ( 1000 Person-Years) Years Shi Wu 16011650 34 23 16511700 13 17 17011750 24 23 17511800 24 29 18011850 29 __ 18511875 53a __ SOURCE : Harrell and Pullum, chapter 6, Table 6. 1. a Includes the mortality crisis years of 186265, when the Taipings occupied Xiaoshan.

Finally, we have the analysis by Lee, Campbell, and Anthony (chap. 7) of mortality in a Liaoning peasant population in the late eighteenth and early nineteenth centuries. Although the time span of the analysis is short, there is a definite trend downward in life expectancy from the first two decades of the nineteenth century to the second two. We eagerly await the extension of this analysis to the 1870s.

At the same time, however, there are now grounds for believing that the downturns in Chinese reproduction rates at times of dynastic or regional crises were probably not sufficient to cause the kind of massive population declines posited by the standard model for the Ming-Qing transition, for example, or for the Taiping rebellion of the mid-nineteenth century. Pop-

TABLE 1.3. Death Rates of Adult Males, by Birth Cohort, Lin Lineage Birth Cohort Death Rate 16001649 31.5 16501699 13.5 17001749 26.2 17501799 27.7 18001849 24.5 18501899 58.0 SOURCE : Harrell and Pullum, chapter 6, Table 6.1.

ulation growth rates declined during these crises, to be sure, but it is now doubtful whether they declined to below replacement level for more than a few years. Keith Hazelton's analysis in an unpublished paper (n.d.) provides the first hard evidence (at least for some of the genealogical populations), and it finds that these mortality crises were not as severe as had previously been thought. Even in the Ming-Qing transition, the actual population decline was short-lived, and the precrisis level was reached again within a decade. The genealogical populations examined here might well have weathered the crises better than the populace as a whole, but even with this caveat it still seems improbable that there were ever any large-scale or long-lasting population declines in late imperial Chinese history. We may have been misled previously by the data, which consisted of impressionistic descriptions, surely exaggerated, and population registers, surely rendered less complete by the conditions of disruption.

Nuptiality

Mortality is one component of the temporal variation in reproduction rates; nuptiality is another. We have already mentioned the figures cited by Telford, in which, after mortality, the most important factor in explaining differential rates of population growth from one period to another is the percentage of a cohort of males who married. Other scholars have found the same thing.

Liu Ts'ui-jung (chap.4) provides a suggestive analysis for several lineages scattered throughout central and southern China. Although she does not give figures for percentages of men who died unmarried, she does record figures for percentages of heirless males for the five lineages (see Table 1.4).

A previous paper (Harrell 1985) shows the same trends through a different kind of procedure. In that paper, I demonstrate that differential growth rates of different lineages and lineage branches are correlated with a series of factors; including the number of consorts (wives or concubines)

TABLE 1.4. Percentages of Heirless Males, by Cohort, Five Lineages Cohort Zhu Zhao Xu Li Mai 15481597 7.8 25.8 0 4.7 13.5 15981647 21.9 6.7 0 13.7 21.2 16481697 16.1 11.6 13.5 9.6 16.2 16981747 15.9 13.5 18.5 18.2 21.0 17481797 16.9 19.4 22.8 20.9 20.1 17981847 21.0 21.9 26.2 24.2 31.0 SOURCE : Liu, chapter 4, Table 4.3.

brought in per man. Although this analysis shows lots of variation according to the differential economic successes of different branches at different times, the overall trends are quite clear, as can be seen from Tables 1.5 and 1.6.

That the trends at the beginning of the period studied are not as predicted is almost certainly because these lineages were getting started at that time, and men who found successful lineages or are the sons and grandsons of founders of successful lineages are by definition more successful reproductively than the average. What is more interesting is the steady decline in almost all branches after 1700; the decline becomes particularly noticeable in the nineteenth century.

Telford (chap. 3) provides similar evidence for the Tongcheng lineages in the late Ming and early Qing in a tabulation by birth cohort of percentages of sons unmarried (see Table 3.3). Notice the clear pattern in Table 3.3. The variable being measured is the percentage of the sons of the members of that particular birth cohort who remain unmarried; the changes over time thus reflect trends forty to sixty years after the date indicated in the left column. Thus percentages of unmarried men are very high during the turbulent final decades of the Ming, begin to fall around 1650, the date of the Qing consolidation of power, and continue to drop to a plateau that lasts until about 1700; it is not clear if the slight rise in the unmarriage rate of the sons of men born in the latter half of the century (men who would be marrying in the early and middle 1700s) is the beginning of a longer Qing downswing, but it is a hint in that direction.9

Age at Marriage

Age at marriage is our final factor. Because the birth dates of the wives and mothers of lineage members are usually listed in genealogies, this is one area where we can look at both male and female figures. And we find that the age at marriage also plays a large part in determining reproduction rates.

TABLE 1.5. Number of Consorts per Man, Lin and Wu Lineages   Lineage or Fang Cohort Lin 1 Lin 2 Lin 3 Wu 16011650 .93 1.00 2.00 1.83 16511700 .87 .90 1.4 1.36 17011750 .97 .85 1.2 1.02 17511800 1.00 .73 1.2 .84 18011850 .77 .62 1.0 .74 SOURCE : Harrell 1985:105, Table 4.18.

TABLE 1.6. Ages at First Son's Birth, Xiaoshan Wu Lineage Cohort Father's Age Mother's Age 16011650 27.5 24.6 16511700 26.8 23.8 17011750 28.9 25.3 17511800 31.3 26.1 18011850 33.1 24.6 SOURCE : Harrell 1985:102, 104, Tables 4.14 and 4.16.

Age at marriage cannot be calculated directly from data given in the genealogies, since they only give birth and death dates. But a reasonable proxy can be found in the age of the parent at the birth of the first surviving son. Various methods have been proposed (Liu 1985; Harrell and Pullum, chap. 6; Ebrey, chap. 2) for reckoning back from this figure to age at marriage for fairly large populations, which damp out the effects of individual variation. For our analysis here, what is important is change over time, and for this we can use age at birth of first son. In fact, the direct effect on population is greater for this variable than for age at marriage; age at marriage is only indirectly related to reproduction rates, whereas age at birth of first son is actually a contributing factor to reproduction rates. Let us examine some of the changes over time in age at birth of first son. We can begin with my analysis of the He, Lin, and Wu lineages of Xiaoshan. I have shown in that paper thai the father's age at birth of first son is an important contributing factor to differential growth rates among lineage branches, and other s age at birth of first son makes a weaker or perhaps no contribution (Harrellia 1985:1013). For one lineage, the Wus, there is a definite change over time in the father's age at birth of first son, but there is a much less obvious change for the mother's age (see Table 1.6).

Telford (chap. 3) reports information that sheds doubt on this simple hypothesis; in his stepwise regression analysis he finds the mother's age at birth of first son to be more important than the corresponding father's age. Since his unit of analysis is the individual, however, this variable does not really say anything about changes over time in age at marriage; it just says that women who marry earlier will have more children, or more accurately, that women who have a son earlier will be more likely to end up with a larger number of sons. Telford, unfortunately, does not provide a measure of changes over time in age at birth of first son for either mothers or fathers.

We do have, however, one more set of data that bears on this hypothesis: Ebrey's (chap. 2) analysis of age at marriage in the Song and Ming dynasties, in which she comments that although male age at marriage varied widely, female age varied little. If the late Ming and Qing followed the same pattern, it was likely to be the male age that varied. Incidentally, male age at marriage ought to correlate with nuptial rates: the greater the age difference between spouses, the greater the percentage of males that remain unmarried, which is what we saw in some of the tables on nuptiality.

Possible Sources of Spurious Findings

Given the nature of genealogies as sources, we ought to look for sources of bias before we herald the findings herein as marking a new era in our understanding of China's population history. There are three that come to mind: inaccuracies in genealogies, a skewed sample of genealogies, and the possibility that genealogies are unrepresentative of the total population.

I have treated the issue of inaccuracies in genealogies in a separate paper (Harrell 1987), so I will not elaborate on the issue here. Suffice it to say that the biases inherent in genealogy compilationunderrecording of women and childrenhave been allowed for by the authors of the works from which the figures in this analysis are taken.

The second possible source of spurious findings is that ours is a skewed sample of genealogies; that is, the genealogies sampled here are not representative of the population recorded in genealogiesmembers of medium to large lineages with at least some prosperous and successful members. Since all the genealogies we have examined show the same trends in the same directions at the same times, it is unlikely (though not impossible) that an error in sampling genealogies is the source of the remarkable uniformity of our results. More likely, our results are only representative of the particular macroregions represented by the genealogies used by the authors of this volume. If we had a few from Lingnan, say, or the Southeast Coast, macroregions whose regional economic cycles were quite different from those of the Yangzi and North China regions represented here, then I think we would have made a strong argument for the whole empire. (Liu's analysis

in chapter 4 does include the genealogy of the Mai of Xiangshan in Guangdong, in the core area of Lingnan, but that is the only Lingnan genealogy analyzed in this volume.) As it is, if we do find some sorts of regional variations, they can still be accommodated within the model of the three primary factors affecting rates of growth. Skinner has taught us not to be disappointed to find regional cycles superimposed on empire-wide ones.

If we are fairly certain that we have sampled genealogies fairly, we still need to consider whether or not we are studying an unrepresentative sample of the Chinese population by using genealogies at all. And the answer is yes, of course. Lineages tended to command various kinds of resources in their areas, resources that gave them mortality advantages in the face of bad years as well as fertility advantages in playing the marriage market for more and earlier brides. The question we need to ask is whether the direction of change in the rates we are examining is different in lineages than it is in the population at large.

For mortality, I think we are pretty safe in saying that the direction of change is the same. Telford (1990a) has recently demonstrated that there are considerable differences in mortality rates between elite and non-elite members within Tongcheng lineages, so we are studying a population that is growing faster than the general populace. It is also quite probable that the population decline in the brief periods of serious crisis was less severe in the genealogically recorded proportion of the population than in the population as a whole. But I cannot think of any reason why the temporal trends of mortality would be different for lineage populations than for the rest of the population; there is no reason why their mortality rates would go down in hard times (when everyone else's were going up), and it is even more unlikely that there would be a situation where their rates would increase as the mortality of their less fortunate neighbors went down.

For nuptiality rates and age at marriage, however, the representativeness of lineages for the population as a whole is questionable. If lineages were able to exercise effective dominance, they might have monopolized brides and driven down the age at marriage for their male members and driven up the age at marriage for people in nonlineage communities. Conversely, a decline in lineage dominance would mean an evening out of percentages of unmarried males and of male ages at marriage between lineages and their neighboors. So we need to determine the kinds of circumstances in which lineages were able to dominate a marriage market and to try to plot these against the historical cycles of rises and declines in the population growth rate. Only then Will we know whether the nuptiality variations over time noted here were very important in explaining overall trends in population growth rates in China or whether they were confined to large lineages and other dominant groups within local populations. The distribution of nuptiality variations is an important subject for future research.

If Telford's analysis is any hint, however, nuptiality should make less difference in the overall growth rate of populations than mortality, which seems to be the prime determiner of total fertility and of reproduction rates in this population.

Implications for the Demographic History of China

We thus have a hypothetical model of population growth in Ming and Qing China, one which is in disagreement with the received wisdom in several ways.

First, there is a real question of whether the population actually dipped from the late Ming to the early Qing. The growth rate certainly went down, but if it ever became negative it probably remained negative for only a few years in any one place and probably never amounted to a large drop in population.

Second, there is a real question of whether the Chinese populations were characterized by low marital fertility or not. I think not. Telford agrees that Chinese were a natural-fertility population: he has come up with estimates of total fertility of somewhere between 8 and 9.5. Estimates to the contrary are of two kinds: (1) the Buck survey, as reassessed by Barclay et al. (1976), which was a retrospective survey in which total fertility is expected to be reported as low because previous infant deaths are not reported (Wolf 1985)10 and (2) Liu Ts'ui-jung's observations (1985) that are based on genealogies that did not record infant mortality. We need more data on the latter to fit rates to model life tables and the like, but my guess is moderate or high and uniform age-specific fertility. Liu's estimates in chapter 5 of this volume of both female and male total marital fertility of around 4.5 to 6.o seem to be at odds with Telford's figures, but Liu has not accounted for infant mortality in her calculations. If a 250 per 1,000 infant mortality rate is figured in, these figures jump to a range of 6.0 to 8.0, which while not high is still much closer to Telford's high range.

Third, as Telford points out (p. 56), population in late imperial China thus appears to be regulated more by mortality factors than by fertility factors. Insofar as reproduction rates change with the dynastic cycle, they reflect not a change in age-specific fertility but rather a change in years of exposure, itself a consequence primarily of mortality levels in the childbearing years and secondarily of age at marriage. Insofar as marriage rates also vary, as shown by Liu in chapter 4, they seem to vary with the degree of imbalance in the sex ratio; higher sex ratios in hard times perhaps reflect increased female infanticide, which itself is a mortality factor.

The only real factor in differential population growth, then, is the change in years of exposure per person. If a person dies early, has a spouse die early, marries late, or doesn't marry at all, total fertility will change; if mortality

is higher for the children, then net reproduction rates will change even if total fertility remains constant. The three factors examined here all contribute to expansion and contraction of years of exposure, and in the preliminary state of our analysis, we can hypothesize that years of exposure vary with the dynastic cycle.

Family Organization And The Developmental Cycle

In light of the various advances made in understanding the developmental cycle of recent Chinese families, as mentioned above, the detailed study of genealogical materials ought now to be addressed to the question of whether the modern-day models apply to Chinese families in past eras. Now that anthropologists have shown convincingly that residence in stem and joint families was part of the life experience of most Chinese in the late nineteenth and twentieth centuries, we need to ask whether residence in extended families was the continuation of a centuries-long pattern or whether the prevalence of complex families in recent years has resulted from lowered mortality due to public-health improvements in places such as Taiwan, a Japanese colony, and Ding Xian (Gamble [1954] 1968), a model of rural reconstruction on the mainland.

Two of the contributions to this volume address this question explicitly: the papers on family and the developmental cycle by Liu Ts'ui-jung (chap. 5) and by Stevan Harrell and Thomas Pullum (chap. 6). Using somewhat different assumptions and methods of calculation, both come up with the same answer: a resounding affirmation of the proposal that the stem and joint family forms were part of the life cycle of most of the families in the lineages studied from late Ming through late Qing times and that the ideal was achievable for centuries before modern health and sanitation measures lowered mortality appreciably. Consider the following figures derived from those two papers. Liu shows that 47.7 to 74.5% of the men in her sample of five lineages experienced life in a three-generation family; in Harrell and Pullum's lineages the values varied from 66 to 80%. Although Liu does not actually compute the chances of living in a joint, or grand, family, she estimates that about 50% of the men had a chance to become grandfathers in such families, since that many men had sons who lived to adulthood. Mortality would of course reduce the number of men who actually attained such particular status, some men were members of grand families at earlier stages of life. Harrell and Pullum's paper (chap. 6) provides more explicit for the possibility of forming joint families: in the three lineages they studied, a little over half the men probably experienced life in such complex households. Although the methods used to calculate these percentage and the detail with which the questions are pursued vary somewhat from author, it is important to note that

all the methods used are quite conservative: they only count a man as having lived in a three-generation family if the man was alive at the same time as his father's father, as his father and his sons, or as his sons' sons. They miss all the stem and joint families headed by widowed matriarchs, which are, at least in most census studies, a plurality if not a majority of the total number of extended families; they also miss those joint families in which one of the brothers has only daughters for several years and the sons, later in the birth order, are born only after the grandparents die. And since these methods assume division at the time of the grandfather's death, they entirely miss joint families of the frrches form, a form that appears to have been common during the Qing period in Liaoning, at least (Lee 1990).

The findings of Liu and of Harrell and Pullum are of potentially great significance for Chinese family history, since they seem consistent across two rather disparate study populations. At the same time, both sets of findings raise questions regarding the representativeness of lineage populations for studying the total population of China in the Ming and Qing periods. Lineages that had enough wealthy members to be able to fund the compilation and printing of genealogies represented a biased sample of the general Chinese population, as has been demonstrated by Rubie Watson (1985). Even poor families belonging to large and influential lineages had certain economic advantages over poor families who were not members of any lineage at all, and it is quite possible that these economic advantages translated into lower age at marriage or lower age-specific mortality, two of the demographic phenomena that, as we have seen above, contributed to comparative demographic success in traditional China. So although we do not know for sure that lineage members were more likely to form stem and joint families than were their nonlineage neighbors, we can postulate that it was so, and we must be cautious about the overall significance of these findings on complex families.

At the same time, however, Wolf's (1985a) statement of the revisionist position indicated that his attribution of common stem and grand (joint) family structure did not apply to the poorest people, those who had no stable connection with the land or with the community. If we take Wolf's caution into account, then our figures from genealogy-bearing lineages, which after all include both wealthy and poor families, may be reasonably representative of all but the poorest in Ming and Qing China generally, and our case studies certainly contribute to the project of "revitalizing the myth" of Chinese family size.

Migration And Regional Cycles

As have long-term fertility and mortality trends, population migration has been of interest to scholars for a long time (Ho 1959; Averill 1983). But

most previous accounts have been regional in nature, describing the effects on a region of large-scale in- or out-migration but not detailing the family-level processes through which regional population changes were accomplished.

The study of genealogies and similar local sources can contribute to the filling of this gap in the same way that it contributes to our understanding of fertility and mortality. There are, of course, enormous local differences in rates of migration: one village may send most of its inhabitants away while a neighboring village sends nobody. Nevertheless, migration from a region is not randomly distributed; there are patterns across a region as a whole and within certain subregions and ecological niches in that region. And through the study of local-level migration, we can begin to discern these intraregional patterns and then relate them to family-level processes.

The paper by Wang Lianmao in chapter 8 is a first attempt to study the processes of migration from a traditional emigrant area and to relate these processes at the same time to family strategies and to the broader intraregional and regional patterns of variation; the paper is based on a comparison of two genealogies from the Zhangquan subregion of the Southeast Coast macroregion (Skinner 1985:277). On the basis of detailed study of two genealogies in two very different ecological situations within this subregion (the Yans of Anping were a coastal, periurban, commercial lineage, and the Lins of Pushan were an inland, agricultural lineage), we can discern local patterns of participation in the great demographic and economic swings that characterized the Southeast Coast region in the late Ming and early to mid-Qing.

Specifically, the great high tide of migration for the commercial Yans came in the late Ming, when the Southeast Coast macroregion as a whole and the Zhangquan subregion in particular were experiencing an upswing caused by increased coastal trade in southern China and Southeast Asia (Skinner 1985:278). Most of the migrants from this lineage, as Wang demonstrates, went to coastal cities in China or to overseas ports; from anecdotal evidence it is clear that theE went to trade and that many of them made their fortunes at it. The Lins, however, mostly migrated later, during and after the economic disasters of the long and painful Ming-Qing transition, when the Southeast Coast was in in serious economic decline (Skinner 1985:279). Although there was some migration to coastal cities, most was either overseas, where most of the Lin migrants seem to have gone as laborers rather than merchants, or, significantly, to the new frontier of Taiwan, where the purpose of migration was ordinary not to get rich in trade but simply to make a living in the newly opened agricultural areas.

The nature of the family process also differed somewhat in the two lineages; most of the overseas and coastal migration, which was predominant among the Yans, was of single men who intended to return; it was rare

for families to emigrate. In the wave of migration to Taiwan in the Qing in which the Lins participated, however, there were many more families migrating as units, indicating once again the influence of ecology on migration strategies at the family level.

The study of Chinese historical microdemography is, as stated above, still in its infancy; much remains to be done before the educated guesses made in this volume are either rejected or become the historical commonplaces of tomorrow. We hope, however, that we have made a start toward an understanding of the connection between local phenomena and broad national and regional trends in Chinese population history.

Continues...

Excerpted from Chinese Historical Microdemography by Stevan Harrell, editor Copyright 1995 by Stevan Harrell, editor. Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

Détails de l'édition

• Éditeur: University of California Press
• Date d'édition: 1995
• Langue: anglais
• ISBN 10: 0520083067
• ISBN 13: 9780520083066
• Reliure: Relié
• Nombre de pages: 250
• Éditeur: Harrell Stevan
• Coordonnées du fabricant: non disponible
• Personne responsable: Wiley-VCH GmbH; product_safety@wiley.com; ; Boschstrasse 12; Weinheim; 69469; Allemagne