Front Cover
 Table of Contents
 Two kinds of preindustrial household...
 Demographic foundations of new...
 Fertility decline in Cuba: A socioeconomic...
 Women in agriculture: Counting...
 Mortality transition in South and...
 Preliminary estimates of fertility...
 Joseph Townsend on poverty, prudence,...
 Book reviews
 Chinese population policy: A People's...
 The Heidelberg manifesto: A German...
 Immigration policy of West...
 Quotations: Versailles Communique,...
 Authors for this issue
 Back Cover

Title: Population and development review
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00088804/00001
 Material Information
Title: Population and development review
Abbreviated Title: Popul. dev. rev.
Physical Description: v. : ; 24 cm.
Language: English
Creator: Population Council
Publisher: Blackwell Pub., etc.
Place of Publication: New York
Frequency: quarterly
Subject: Population research -- Periodicals   ( lcsh )
Bevolking   ( gtt )
Sociaal-economische ontwikkeling   ( gtt )
Population -- Recherche -- Périodiques   ( rvm )
Genre: periodical   ( marcgt )
Dates or Sequential Designation: v. 1- Sept. 1975-
Issuing Body: Issued by: Population Council.
General Note: Has occasional supplements.
 Record Information
Bibliographic ID: UF00088804
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 01613929
lccn - 76640030
issn - 0098-7921

Table of Contents
    Front Cover
        Front Cover 1
        Front Cover 2
    Table of Contents
        Table of Contents 1
        Table of Contents 2
        Table of Contents 3
        Table of Contents 4
    Two kinds of preindustrial household formation system, by John Hajnal
        Page 449
        Page 450
        Page 451
        Page 452
        Page 453
        Page 454
        Page 455
        Page 456
        Page 457
        Page 458
        Page 459
        Page 460
        Page 461
        Page 462
        Page 463
        Page 464
        Page 465
        Page 466
        Page 467
        Page 468
        Page 469
        Page 470
        Page 471
        Page 472
        Page 473
        Page 474
        Page 475
        Page 476
        Page 477
        Page 478
        Page 479
        Page 480
        Page 481
        Page 482
        Page 483
        Page 484
        Page 485
        Page 486
        Page 487
        Page 488
        Page 489
        Page 490
        Page 491
        Page 492
        Page 493
        Page 494
    Demographic foundations of new sex roles, by Kingsley Davis and Pietronella van den Oever
        Page 495
        Page 496
        Page 497
        Page 498
        Page 499
        Page 500
        Page 501
        Page 502
        Page 503
        Page 504
        Page 505
        Page 506
        Page 507
        Page 508
        Page 509
        Page 510
        Page 511
        Page 512
    Fertility decline in Cuba: A socioeconomic interpretation, by Sergio Diaz-Briquets and Lisandro Perez
        Page 513
        Page 514
        Page 515
        Page 516
        Page 517
        Page 518
        Page 519
        Page 520
        Page 521
        Page 522
        Page 523
        Page 524
        Page 525
        Page 526
        Page 527
        Page 528
        Page 529
        Page 530
        Page 531
        Page 532
        Page 533
        Page 534
        Page 535
        Page 536
        Page 537
        Page 538
    Women in agriculture: Counting the labor force in developing countries, by Ruth B. Dixon
        Page 539
        Page 540
        Page 541
        Page 542
        Page 543
        Page 544
        Page 545
        Page 546
        Page 547
        Page 548
        Page 549
        Page 550
        Page 551
        Page 552
        Page 553
        Page 554
        Page 555
        Page 556
        Page 557
        Page 558
        Page 559
        Page 560
        Page 561
        Page 562
        Page 563
        Page 564
        Page 565
        Page 566
    Mortality transition in South and East Asia: Technology confronts poverty, by Lado T. Ruzicka and Harald Hansluwka
        Page 567
        Page 568
        Page 569
        Page 570
        Page 571
        Page 572
        Page 573
        Page 574
        Page 575
        Page 576
        Page 577
        Page 578
        Page 579
        Page 580
        Page 581
        Page 582
        Page 583
        Page 584
        Page 585
        Page 586
        Page 587
        Page 588
    Preliminary estimates of fertility decline in India during the 1970s, by Anrudh K. Jain and Arjun L. Adlakha
        Page 589
        Page 590
        Page 591
        Page 592
        Page 593
        Page 594
        Page 595
        Page 596
        Page 597
        Page 598
        Page 599
        Page 600
        Page 601
        Page 602
        Page 603
        Page 604
        Page 605
        Page 606
    Joseph Townsend on poverty, prudence, and population growth: An eighteenth century tract
        Page 607
        Page 608
        Page 609
        Page 610
        Page 611
        Page 612
    Book reviews
        Page 613
        Page 614
        Page 615
        Page 616
        Page 617
        Page 618
        Page 619
        Page 620
        Page 621
        Page 622
        Page 623
        Page 624
        Page 625
        Page 626
        Page 627
        Page 628
        Page 629
        Page 630
        Page 631
        Page 632
    Chinese population policy: A People's Daily editorial
        Page 633
        Page 634
        Page 635
    The Heidelberg manifesto: A German reaction to immigration
        Page 636
        Page 637
    Immigration policy of West Germany
        Page 638
    Quotations: Versailles Communique, US Supreme Court
        Page 639
        Page 640
        Page 641
        Page 642
        Page 643
        Page 644
        Page 645
        Page 646
        Page 647
        Page 648
        Page 649
    Authors for this issue
        Page 650
        Page 651
    Back Cover
        Page 652
Full Text





Volume 8 Number 3
September 1982

John Hajnal Household
formation patterns in
historical perspective
Kingsley Davis and
Pietronella van den Oever
Demographic foundations
of new sex roles
Sergio Diaz-Briquets and
Lisandro Perez
Socioeconomic factors in
Cuba's fertility decline
Ruth B. Dixon Countin
women in the agriculture
labor force
Notes and Commentary
L. Ruzicka and H.
Hansluwka on persistent
high mortality in South
and East Asia
Data and Perspectives
A. Jain and A. Adlakha on
estimates of fertility
decline in India
Archives Joseph
Townsend on poverty
and population growth
Book Reviews by Garrett
Hardin, Ester Boserup,
Roland Pressat, and others
Documents Chinese
population policy
Heidelberg Manifesto





The Population and Development Review seeks to
advance knowledge of the interrelationships
between population and socioeconomic
development and provides a forum for discussion of
related issues of public policy.

Published quarterly under the editorial auspices of
the Center for Policy Studies of The Population
Council, One Dag Hammarskjold Plaza. New York,
NewYork 10017 USA

Editorial Committee
Paul Demeny. Chairman Geoffrey McNicoll
Mead Cain Michael P. Todaro

Managing Editor
Ethel P. Churchill

Advisory Board
Ester Boserup Akin L. Mabogunje
Gustavo Cabrera Milos Macura
John C. Caldwell Carmen A. Miro
Richard A. Easterlin Asok Mitra
Jose Encarnacion. Jr. Simeon H. Ominde
Simon Kuznets Masri Singarimbun

Signed articles are the responsibility of the authors.
Views expressed in the Review do not necessarily
reflect the views of The Population Council.

Direct manuscripts, comments on articles, and
correspondence to:

Population and Development Review
The Population Council
One Dag Hammarskjold Plaza
New York. NewYork 10017 USA
Subscription information appears on the inside
back cover.

Volumes are available on microfilm from
Xerox University Microfilms. 300 North Zeeb Road.
Ann Arbor. Michigan 48106.

Population and Development Review (ISSN 0098-7921) is
published quarterly in March. June, September. and
December at $14.00 for a one-year and $24.00 for a two-
year subscription by the Population Council. One Dag
Hamma'skjold Plaza. New York. NY 10017 USA. Application
to mal! at second-class postage rates pending at New York.
NY POSTMASTER: Send address changes to Population and
Development Review. One Dag Hamnmarskjold Plaza. New,
York. NY 10017.

g 1982 by The Population Council. Inc. All rights
reserved. Printed in the United States of America.

Library of Congress Catalog Card
Number: 75-27311
ISSN 0098-7921





Volume 8 Number 3
September 1982

Two Kinds of Preindustrial Household
Formation System 449
John Hajnal
Demographic Foundations of New
Sex Roles 495
Kingsley Davis
Pietronella van den Oever
Fertility Decline in Cuba: A
Socioeconomic Interpretation 513
Sergio Diaz-Briquets
Lisandro P6rez
Women in Agriculture: Counting the
Labor Force in Developing Countries 539
Ruth B. Dixon

Notes and Commentary
Mortality Transition in South and East
Asia: Technology Confronts Poverty 567
Lado T. Ruzicka
Harald Hansluwka

Data and Perspectives

Preliminary Estimates of Fertility
Decline in India During the 1970s 589
Anrudh K. Jain
Arjun L. Adlakha


Joseph Townsend on Poverty,
Prudence, and Population Growth: An
Eighteenth Century Tract 607

Book Reviews

Ashok S. Guha, An Evolutionary View of
Economic Growth
Garrett Hardin

Theodore W. Schultz, Investing in People:
The Economics of Population Quality
Ester Boserup

Roderic Beaujot and Kevin McQuillan,
Growth and Dualism: The Demographic
Development of Canadian Society
Roland Pressat

Abdel-Rahim Omran, Population in the
Arab World: Problems and Prospects
Allan G. Hill

K. Srinivasan and S. Mukerji (eds.),
Dynamics of Population and Family
Welfare, 1981
John C. Caldwell

Daniel Callahan and Phillip G. Clark (eds.),
Ethical Issues of Population Aid: Culture,
Economics, and International Assistance
Jonathan Lieberson

Short Reviews


Chinese Population Policy: A People's
Daily Editorial

The Heidelberg Manifesto: A German
Reaction to Immigration










Immigration Policy of West Germany 638

Quotations: Versailles Communique,
US Supreme Court 639

Abstracts 643

Authors for This Issue 650

Two Kinds of
Preindustrial Household
Formation System

John Hajnal

A similar discovery about household size has been made in
the last few decades about widely different societies, namely parts of Europe in
preindustrial times, India, and China. Until recently it was widely believed that
in preindustrial Europe, as well as in India and China, large households used to
be the norm. The discovery that the average household size was on the order of
only five persons, therefore, came as a surprise.
The traditional household formation systems of India and China are sim-
ilar to each other (at least in the aspects dealt with in this paper). But the
household formation systems of Northwest Europe in the seventeenth and
eighteenth centuries were very different, though they yielded households of
roughly the same size. The aim of this paper is to describe and contrast two
kinds of household formation systems in quantitative terms. The term "house-
hold formation system" is used to indicate that the intention is to compare
modes of behavior that result in the formation of households of various kinds,
as well as to compare the results of that behavior. A household formation
system is defined by household formation rules, as described in section 1.
Despite gaps in the evidence, enough data have now been accumulated
by historians to support strongly the conclusion that the household formation
systems of all populations in preindustrial Northwest Europe shared certain
common features that distinguished these populations from those of India,
China, and many other preindustrial populations.
The term Northwest Europe as used here covers the Scandinavian coun-
tries (including Iceland, but excluding Finland), the British Isles, the Low
Countries, the German-speaking area, and northern France. This area showed
the European pattern of late marriage, as much evidence now confirms, back
into the seventeenth century. The data on household composition in this area
show, among other characteristics to be described, a high proportion of "ser-



Two Kinds of Preindustrial Household Formation System

vants" and very small numbers of households comprising more than one mar-
ried couple. The late age at marriage and household composition are clearly
related and reflect the distinguishing features of the preindustrial Northwest
European household formation systems, as explained below.
The term preindustriall Northwest Europe" denotes Northwest Europe
in the seventeenth and eighteenth centuries.' This is the period for which there
is an adequate body of data. No attempt will be made to discuss household
composition in Southern Europe or in Finland and the Baltic countries during
those centuries even though these areas are also mostly areas of late marriage
by the end of the nineteenth century.2 Age at marriage is a crucial variable in
household composition, and one might expect that the whole region that dis-
played the European marriage pattern could be treated as a unit. The reason for
singling out Northwest Europe is twofold, namely lack of data for the other
areas mentioned and evidence that in the seventeenth and eighteenth centuries
household formation systems in at least some parts of these areas did not have
the distinguishing Northwest European features. There is evidence that in the
eighteenth century and earlier the first marriage of women in parts of southern
France and in Italy occurred at a younger age.3 There is also evidence that the
composition of households did not everywhere in southern France and Italy
display the Northwest European characteristics. In Finland and the Baltic coun-
tries also there is evidence from the eighteenth century of both earlier marriage
and household composition patterns departing from the preindustrial Northwest
European norms. There are also large areas (for example, the Iberian penin-
sula) from which no household data for the relevant period appear to have been
published. It may well be that the distinguishing features of Northwest Euro-
pean household formation systems were to be found in the seventeenth and
eighteenth centuries among populations outside Northwest Europe as that re-
gion has been defined here. We must wait for further research to decide this
question. By the end of the nineteenth century the European pattern of late
marriage certainly extended beyond the boundaries of Northwest Europe.
Northwest European populations in the seventeenth and eighteenth cen-
turies were both demographically and economically "traditional" societies.
They had a "young" age structure (and, of course, the age structure has a
strong effect on household size and composition). Northwest Europe was at
that time largely preindustrial in the sense that populations almost everywhere
were largely rural. "Productive" economic activities (farming, fishing, crafts)
were mainly carried on in and by households and not in enterprises or work-
places specialized for the purpose (plantations, factories, offices, and the like).
However, not all households functioned as productive enterprises. For this
paper, households are "housekeeping units," as is explained below.
This paper attempts to contrast the household formation systems of pre-
industrial Northwest Europe with the household formation systems of a number
of other populations, which also showed these premodern demographic and
economic characteristics. In particular, all the populations treated in this paper
had a "young" age composition (something like 43 or more percent of the
population under age 20). They were all populations of high fertility compared


John Hajnal 451
with the levels of fertility found in Europe today. All of them also had much
higher mortality rates than those of modern Europe. All were predominantly
rural, and the "productive" economic activities of these rural populations were
mainly carried out in and by households (sometimes with two or more house-
holds combining for farming and other "productive" purposes). However,
many households in most of the populations discussed were not economically
The household formation systems of all the populations outside North-
west Europe dealt with in sections 3 and 4 below shared certain common char-
acteristics (as described in section 1 below). Household formation systems
sharing these characteristics will be called "joint household systems." The
different kinds of data that are available for different populations have largely
dictated the comparisons with Northwest Europe that it has been possible to
For comparisons of household structure between very different societies
it is desirable to be explicit about what is meant by "household." The intention
has been to use data that treat each "housekeeping unit" as a separate house-
hold. The matter is dealt with in greater detail in Appendix 1.
In some populations there is a substantial difference in household size
and structure between urban and rural areas. The aim of this paper is to de-
scribe household formation in the rural areas. So far as possible, data for rural
areas have been used. Where the urban population constituted only a small
proportion of the total population, data for the total population reflect the be-
havior of the rural component. This is mostly the case with the populations
covered in this paper.
The emphasis has been placed as far as possible on data covering popula-
tions of 5,000 or more people, rather than on data from individual commu-
nities, which are often used in discussions of demographic and other statistical
measures for past centuries. The household composition data for small commu-
nities display substantial variation (both between communities and over time in
the same community) even when vital rates and household formation rules are
identical and unchanged.4
This paper, like all work dealing with historical household data, owes a
great deal to the work of the Cambridge Group for the History of Population
and Social Structure. In particular the volume Household and Family in Past
Time (Laslett and Wall, 1972) put the historical study of households on a new
factual basis. The theme of the present paper bears an obvious affinity to Chap-
ter 1 of Peter Laslett's book Family Life and Illicit Love in Earlier Generations
(1977). I have tried to put emphasis on other kinds of data and other types of
questions than those dealt with in the two works just mentioned.5

1 Household formation rules
By a joint household we mean a household comprising two or more related
married couples; a simple household, correspondingly, is one that contains
only one married couple or none at all. The Northwest European household

Two Kinds of Preindustrial Household Formation System

formation systems operated in such a manner as to produce very few joint
households. The majority of persons, even of those who survived to middle
age, were never members of a joint household.
The joint household systems did not normally produce a situation in
which the majority of households were joint at any one time (though there have
been joint household systems that have operated in that way); however, under a
joint household system, the majority of people were members of a joint house-
hold at some stage of their lives.
The two kinds of household formation system being compared can both
be characterized by three rules of normal household formation behavior, as
1 Formation rules common to Northwest European simple household
A Late marriage for both sexes (mean ages at first marriage6 are, say,
over 26 for men and over 23 for women).
B After marriage a couple are in charge of their household (the husband
is head of household).
C Before marriage young people often circulate between households as
2 Formation rules common to joint household systems
A Earlier marriage for men and rather early marriage for women (mean
ages at first marriage are under about 26 for men and under 21 for
B A young married couple often start life together either in a household
of which an older couple is and remains in charge7 or in a household
of which an unmarried older person (such as a widower or widow)
continues to be head. Usually the young wife joins her husband in the
household of which he is a member.
C Households with several married couples may split to form two or
more households, each containing one or more couples.
These rules have been stated here in general qualitative terms. There was
much variation, within both kinds of formation system, between different
areas, and also probably over time, in the way in which the same rule was
carried out, for example in the age at which changes specified in the rules took
place. (This is illustrated by the discussion below of the splitting rule 2C.)
Moreover, the two sets of rules are not complete in that they would not suffice
to determine the movement of individuals between households. It is not possi-
ble for everyone to remain until death in the household into which he or she
was born unless moved by marriage, entry into service, or the splitting of
households. Additional movements not governed by the rules stated are certain
to occur. But the rules listed suffice to determine the features with which this
paper is concerned.

John Hajnal

The rules require some explanation.
The Northwest European rule 1B, that a married couple are in charge of
their own household, implies that upon marriage, either (a) a new household
was created, or (b) one spouse joined the other in a household in which there
had been no married couple, or (c) if they took over a farm run by the parents
or a parent of one of them, the parent or parents retired when the young people
married. (The very small number of households in preindustrial Northwest
Europe in which married children are recorded as members of households
headed by the parents of one of them seems to reflect temporary circum-
stances.) The practice of retirement by contract was known in most, if not all,
of Northwest Europe. A farmer would make a contract with his heir by which
he transferred the farm to the latter in return for a commitment that the heir
would maintain him (and his wife if she was alive) from then on; that is, would
provide lodging, income, food, firewood, and so on. (A retirement contract
could also be made by a widow surrendering her farm or could be made with
someone who was not the heir.) In some parts of Northwest Europe this kind of
retirement was common.8
It is the third rule for each of the two kinds of system (i.e., 1C and 2C)
that will probably appear most surprising. The argument here is that in each
case the rule was essential to the operation of the kind of household formation
system in question.
The circulation of servants in Northwest European systems is dealt with
further in section 5 below. Servants are found in substantial numbers concen-
trated at young adult ages throughout preindustrial Northwest Europe. It seems
highly probable that the circulation of servants made possible the late age at
marriage, for service provided a function for young unmarried adults. Because
of the institution of service, young men and women were able to move away
from farms and villages where their labor could not be effectively used. On the
other hand, the availability of servants provided an adjustable labor supply to
those farm or craft households where the number of family members available
for work was too small.
The formation rules listed above (1A, 1B, and 1C), which were common
to all of Northwest Europe, had no tendency to create very large households.
Very large households did occur in Northwest Europe because some wealthy
and important householders employed many servants, but such large house-
holds were not the inevitable result of the common Northwest European house-
hold formation rules. Rule 2B for joint household systems, however, will
produce some enormous households unless splitting occurs. Under this rule
households can continue to grow with all male descendants of an original an-
cestor bringing their brides into the household. The result can be households
comprising brothers, cousins, second cousins, and even more distant relatives
with their wives and children. This is not only a theoretical possibility, but
will, except for splitting, occur quite often. Even under conditions of high
mortality, some men will have several sons surviving to adulthood. (It should
be remembered that the variance of the total number of children born to men is

Two Kinds of Preindustrial Household Formation System

much larger than for women. Men who marry a number of times can continue
to father children until an age well past the limits of the female reproductive
period.) If a man has a number of surviving sons who bring their brides into the
household, and if some of the sons have several surviving sons in their turn, a
very large household will result. A joint household system, therefore, needs
the splitting rule, 2C. The descriptive literature concerning joint household
systems contains references to the splitting of households in all societies in
which joint household systems operated."
The practices in regard to the splitting of households have a crucial effect
on the size and composition of households under a joint household system. If
young couples tend to split away even before the husband's father has died, and
if brothers surviving their father's death tend to split not many years later,
households under a joint household system will be comparatively small. If,
however, not only brothers, but the children of brothers tend to stay together,
much larger households will result. This matter will be further dealt with in
section 4 below.
A full description of a household formation system would involve dis-
cussion of what happens when death or some other circumstance leaves indi-
viduals or groups (e.g., old people without children,10 widows with young
children) that cannot subsist as households because they no longer have enough
ablebodied members. Such individuals or groups must be attached to other
households, have new individuals added to them, be combined with other simi-
lar groups to form new households, or be maintained on their own by being
supplied with food and other necessities by relatives or by some other means.
It is obvious that such collapses of households occurred fairly frequently
under the Northwest European rules, where there was at most one married
couple in the household. The retirement contracts already mentioned some-
times met this kind of situation, as did the hiring of a servant. Moreover, in
Northwest Europe such needs were widely met by communal provision" (a
matter discussed further in section 6). Under joint household systems, the ex-
tinction of households is also a common occurrence (though fewer people are
left stranded than under Northwest European rules). The tendency described
above for some men to have large numbers of male descendants within a few
generations is counterbalanced under conditions of slow population growth
(that is, given a rough balance between fertility and mortality) by a high pro-
portion of male lines becoming extinct. The process of extinction leaves behind
individuals or groups incapable of functioning as independent households.12 A
full specification of household formation rules would include what might be
called household dissolution rules.
It is often said that household formation varies with economic condi-
tion-in particular, that the rich can maintain larger households than the poor.
There is some truth in this proposition; in Northwest Europe an important so-
cial differentiation was reflected in the fact that those who had land employed
as servants the children of those who did not. Yet it is remarkable that (with the
possible exception of some very small aristocratic groups) all layers of the rural

societies dealt with in this paper, from the rich to the very poor, followed the
same household formation rules. In particular, even the rich did not form joint
households in Northwest Europe; on the other hand, under joint household
systems even the poor did.13
Needless to say, there is much variation within each of the two basic
kinds of systems. But it is hoped that the data quoted below will show not only
that the basic distinction is valid, but will also give a useful quantitative picture
of the operation of the two kinds of systems and their varieties (a quantitative
picture with a number of surprising features).
Finally, it should be explicitly stated that there are other kinds of house-
hold formation system besides the two considered here.14 Moreover, there are
populations to which it would not be appropriate to apply the household con-
cept used in this paper.
The study of households is still in its infancy in comparison with the
study of demographic topics such as fertility or mortality. What is said here is,
therefore, tentative in many ways. The methods of classification and analysis
to be used in the interpretation and comparison of household data from differ-
ent societies are likely to undergo much modification and refinement in the
future. It remains to be seen, in particular, whether the category of joint house-
hold formation systems as used in this paper, that is, household formation
systems for which rules 2A, 2B, and 2C hold, is a useful one. Joint household
systems in this sense have occurred under a variety of very different conditions
widely separated in time and space.15 It may turn out, when statistical data on
households for many more populations have been analyzed, that it is not fruit-
ful to group together all the populations exhibiting those household formation
rules that for the purposes of this paper are the defining characteristics of joint
household systems.

2 Household composition in Northwest Europe
Unfortunately, there are no data for large populations tracing the movements of
individuals between households as they pass through life. We must infer the
consequences of household formation behaviors from censuses and similar data
sources that show the distribution over households of all individuals in a popu-
lation at a particular time. We can thus verify indirectly that the household
formation rules stated were in fact in operation. A particularly useful form of
analysis is provided by classifications of populations by relationship to the head
of household. Such classifications display characteristic differences between
the Northwest European systems and joint household systems. We may sum-
marize these differences as follows: (1) Populations following joint household
systems have much higher proportions of joint households, as one would ex-
pect; however, they do not, on that account, necessarily have larger households
on average. (2) Households under the two kinds of formation systems are made
up of different sorts of individuals. In joint household systems almost all
household members are relatives of the head. There are substantial numbers of

John Hajnal


Two Kinds of Preindustrial Household Formation System

such relatives in addition to the wife and children of the household head. (3) In
Northwest Europe, on the other hand, the composition of households is differ-
ent. The numbers of relatives other than the wife and children of the head are
small; instead we find substantial numbers of servants and also some other
persons called lodgers (or some equivalent term), who may or may not be
related to the head of the household (e.g., they may be farm laborers). Some of
these persons (e.g., farm laborers' families living on the farm but doing their
own cooking) should probably be counted as separate households.16
We shall proceed by considering in detail one population illustrative of
each of the two kinds of household formation system. Data for other popula-
tions will then be more briefly reviewed. The present section considers data for
rural Denmark as representative of Northwest Europe.

There were official Danish censuses in 1787 and 1801. The data are of excep-
tionally high quality for the time and are suitable for our purposes in a number
of ways. For example, marital status was explicitly recorded at the original
enumeration; it does not need to be indirectly inferred when the data are re-
analyzed-for instance, by presuming that a man and woman are married if
they are entered in succession at the beginning of the household list.17 Not only
is the original Danish census material of high quality, but it has been carefully
retabulated from the original documents by H. C. Johansen'8 of Odense Uni-
versity. His painstaking analysis underlies much of what follows.
Johansen has analyzed a sample of 26 rural parishes, whose population
(a total of some 7,000 persons), as he shows, resembles the whole Danish rural
population in many characteristics. We may confidently take his figures as a
picture of the rural Danish population as a whole. Some 80 percent of the
population of Denmark lived in rural parishes.
Table 1 shows how the rural population in Denmark at the end of the
eighteenth century was distributed by relation to head of the household. The
table shows the numbers in each relationship category per 100 households.19

TABLE 1 Rural Denmark (26 parishes): persons per 100
households by relationship to head, average for 1787 and 1801
Married heads Other Other
and wives heads Children relatives Servants Others Total
Males 88 5 99 8 50 5 255
Females 88 7 96 15 40 9 255
Both sexes 176 12 195 23 90 14 510
SOURCE: Computed from Johansen (1975), p. 148.

Children, it should be noted, are defined in this kind of tabulation by
relationship to the head, not by being under a certain age. No doubt the cate-
gory included not only biological children of the head, but others in an analo-
gous position in the household, such as children of the wife by a former


John Hajnal 457

husband, or even children of a remarried wife's former husband by his first
wife. The category "servants" will be further discussed below.
Table 1 represents the situation at the end of the eighteenth century.
However, we can go back in time by one and one-half centuries and add a
comparable set of figures for rural parishes in the Danish island of Moen in
1645. They result from an enumeration carried out for tax purposes by the local
clergymen, and the surviving listings were analyzed with exemplary care early
in the twentieth century by E. P. Mackeprang. Table 2 compares Mackeprang's
data with the situation at the end of the eighteenth century. ("Other relatives"
and "others" have had to be combined in one category.) In this case, it is not
clear from the original listing just which persons constitute a separate house-
hold, and the figure for average household size depends, in part, on a guess of
the number of households made by Mackeprang.
TABLE 2 Rural Denmark, 1645 (Island of Moen) and 1787/1801
(26 parishes): persons (both sexes) per 100 households by
relationship to head
Married heads Other
and wives heads Children Servants Others Total
Island of Moen 175 12 249 62 25 523
26 parishes 176 12 195 90 37 510
NOTE: Figures for Moen are based on data for five rural parishes covering 4,014 persons.
SOURCE: Data for Island of Moen computed from Mackeprang (1907), p. 258, using Mackeprang's
estimate (p. 260) of the number of independent households among laborers ("Husmaend" in Danish).
Data for 26 parishes: see Table I above.

The only substantial differences between the Moen data and the later
Danish parish sample lie in the larger number of children in Moen in 1645,
compensated, in part, by a smaller number of servants.20 (See section 6.)
The categories distinguished in Table 2 are categories of household
membership that were taken for granted in the seventeenth and eighteenth cen-
turies in Northwest Europe; as we shall see, they differ from those appropriate
in joint household systems. In the very first work ever written on demography,
John Graunt wrote (Observations on the Bills of Mortality, p. 60): "I imagined
S. .there were about eight persons in a family, one with another, viz the man
and his wife, three children, and three servants, or lodgers."21 (His numbers
for children and servants or lodgers are somewhat high, a natural illusion for
various reasons.) It is noteworthy, in light of the comparisons to be made with
joint household systems, that Graunt assumes that relatives other than children
can be ignored and that servants (or lodgers) are present in the household in
substantial numbers, although they are not permanent members by virtue of
In the Danish data in Table 1 "other relatives" amounted to under 5
percent of the population, a clear indication that there could have been few
joint households.
The data enable us to study directly the way in which the married were
distributed between households. The percentage distribution of married men by

Two Kinds of Preindustrial Household Formation System

relationship to the head was as follows in 1787 and 1801 (the percentages are
based on 2,606 married men for both censuses combined):

Heads 93.4
Sons of heads 0.7
Fathers of heads 1.9
Relatives other than fathers or sons 0.1
Servants 2.0
Others 1.9
Total 100.0

Most of the couples who were not in charge of their own household
were, as Johansen points out, old people or very recently married (servants,
laborers). A married servant did not live with his wife in his master's house-
hold. Johansen suggests that the married servant went home periodically to his
wife from his master's house.
It is likely that some of the retired married parents and some of the other
married men (who were often laborers) formed independent housekeeping
units of their own and were not fully integrated in the households to which they
were allocated at the census (this matter is discussed in Appendix 1). It thus
seems that households with two related married couples must have formed well
under 4 percent of all households. It is clear, therefore, that the great majority
of married men became heads of their own household upon marriage.
It hardly needs emphasis that Denmark was a country of late marriage
for both sexes, with mean ages at first marriage at some 30 to 31 years for men
and 26 to 28 years for women in the population to which our tables relate.22
The remaining Northwest European household formation rule (namely
the circulation of young people as servants) shows its effects in the numbers of
servants in Tables 1 and 2. This rule will be treated in detail in section 5 below.
There are data from other parts of Northwest Europe similar to those that
have just been reviewed for Denmark, namely classifications of the population
by relationship to the head of household. It would be tedious to go over this
evidence in detail; a brief review is found in Appendix 2. The features illus-
trated by the Danish rural population, which contrast with what is found under
joint household systems, apply throughout Northwest Europe. Few households
comprise more than one married couple. Households consist largely of heads
and wives, their children, and servants. The numbers in categories other than
these, such as lodgers or retired parents, are in some cases rather larger than in
the Danish data, and the mean number of persons per household is also greater.
In such cases, some of these additional elements probably constituted separate
households23 (according to the definition of a household as a consumption
unit). The size of households and their composition was in these cases proba-
bly, in reality, closer to the Danish situation than the data appear to show at
first sight.
Another kind of data that confirm our conclusions regarding Northwest
European household systems consists of classifications of households by types,


defined in terms of the relationship between the individuals they contain (nu-
clear households, extended family households, and so on). The classification
system for households worked out by the Cambridge Group for the History of
Population and Social Structure has been applied by historians to a large num-
ber of communities. This evidence has been assembled by Peter Laslett in his
book, Family Life and Illicit Love in Earlier Generations.24 A slight question
of terminology arises in relating that evidence to the argument of this paper.
Households that comprise more than one married couple do not form a separate
category in the classification scheme of the Cambridge Group. The category
"multiple family households" in that classification includes all households that
comprise two or more conjugal family units connected by kinship or by mar-
riage. Most conjugal family units will comprise a married couple, but other
groups, such as a widow and her child, also constitute conjugal family units.
The category "multiple family households" thus includes not only all joint
households, but, in addition, some households containing only one married
couple. The fact that the number of multiple family households in all North-
west European communities so far studied is so small is thus strong evidence
that the proportion of joint households must have been small (say less than 6
percent) .2

3 Household composition in joint household
systems: India and China
The Danish household composition data were presented for the purpose of
contrasting these with the situation in a society with a joint household system.
Our example is India. We shall show that while Indian households are no larger
than in preindustrial Northwest Europe, the proportion of joint households in
India is substantial.
The Indian censuses are too well known to require description here. In
1951 and 1961, tabulations regarding households were obtained from samples
of the census schedules. The tabulations on households were different at the
two censuses, the sampling was different, and the quality of the data varied
between parts of that vast country.26
India in 1951 is not quite an ideal comparison for our purposes. The
country has, of course, a joint household system of enormous importance in
that it affects so large a population. In 1951 India was still an overwhelmingly
rural country of small villages (over 80 percent of the population was classified
as rural); yet "modem" influences may, to some slight but unknown extent,
have affected the traditional household formation system. There is, however,
no reason to believe that in the past households were larger or that a higher
proportion of them was joint.27
The composition of Indian households in 1951 is presented in Table 3.
Comparison of Tables 1 and 3 shows that the average number of persons
per household was about the same (about five in each case). The numbers of
heads and their wives, and the numbers of children per household were also

John Hajnal


Two Kinds of Preindustrial Household Formation System

about the same. It is to be noted, however, that whereas in Denmark the num-
bers of children of both sexes were roughly equal, in India sons far exceeded
daughters. The main reason is that the daughters had moved out of their origi-
nal household on marriage, become daughters-in-law, and, thus, been included
under "other relatives" in Table 3.

TABLE 3 Rural India, 1951: persons per 100 households by
relationship to head
Married heads Other Other
and wives heads Children relatives Unrelated Total
Males 71 19 110 48 3 251
Females 71 10 81 74 3 239
Both sexes 142 29 191 122 6 490
SOURCE: Census of India, 1951, vol. 1, India, Part IA, Demographic Tables, Table C.I(ii).

A basic difference between Northwest European and Indian household
composition is that, in Northwest Europe, servants (and to a smaller extent
some other categories of "unrelated" individuals) take the place of most of the
"other relatives" in Indian households. Indian households consist almost en-
tirely of related persons with only very few household members (well under 2
percent) unrelated to the head.
As will be noted below, some of the servants in Northwest European
households were related to the head of household, but unlike "relatives," such
as daughters-in-law in India, they did not have permanent household member-
ship by virtue of the relationship.
It is well known that India satisfies the characteristic of joint household
systems by which both sexes marry early. For our purposes, it is not marriage
as recorded in Indian censuses that is relevant (which in European terms is
more like an irreversible betrothal). We must take as "marriage" the movement
of the bride into the husband's household, which takes place later than the
formal marriage. Even so, the "effective" marriage in this sense occurred for
both men and women at a much earlier age, on average, than marriage in
Northwest Europe.
In Denmark, as we have seen, married people were almost all in charge
of their own household. This was not so in rural India: only some 64 percent28
of married men in 1951 were heads of households. Consequently, there were a
substantial number of households in which two or more married couples lived
together. The number and types of joint households (in terms of the relation-
ship between the married men living in them) enumerated at a census depend
not only on the types of joint household created, but also on the extent to which
couples survive and the circumstances under which couples living together
decide to split.
There is some information on the numbers and types of joint households
in India. In analyses of the 1961 census household composition, the numbers of
married sons and other married relations of the head were obtained. Some 67

percent of married men in 1961 were heads of households; 22 percent were sons
of the head of household; and 11 percent were related to the head in other ways.
Per 100 households there were 24 married sons and 12 married men related to
the head in other ways. The majority of joint households were therefore formed
by married sons living with their fathers. But there were also substantial num-
bers of married brothers and married couples related in other ways living to-
gether. Nevertheless, splitting occurred at an early enough stage so that the
average household size remained on the order of five.29
A more detailed picture of the frequencies of various types of joint
household in India is available from a survey carried out in the state of Ma-
harashtra in 1947-51 (Table 4). The sample comprised some 12,000 households
selected from about 74 villages.30 The great majority of households (some 77
percent) at any one time contained only one married couple, or none. How-
ever, it can be inferred that many couples at some stage form part of a joint
household, and while the father-son type is the most frequent, other combina-
tions occur in nonnegligible numbers.

TABLE 4 India and Nepal: percentages
of households with different numbers of
married couples in varying relationships
(Maharashtra) Nepal
1947-51 1976
No couple 19 17
One couple 58 63
Two couples
Father and son 10 1
Two brothers 5 16 15
Other relations 1
Three couples 5 3
More than three
couples 2 1
Total 100 99
Number of households
in sample 12,030 5,537
SOURCES: India: Dandekar and Pethe (1960); Nepal: World
Fertility Survey (1980).

Table 4 also shows the distribution by numbers of couples in households
of Nepal in 1976. These data are from a sample household survey taken as part
of the World Fertility Survey. The results are remarkably similar to those of the
Maharashtra survey. Over 90 percent of the Nepalese are reported as Hindus,
and Nepal displays a marriage pattern of the Indian type.31 The Nepalese rural
population according to the survey constitutes over 97 percent of the total
Nepalese population. The mean number of persons per household was 5.2.
Finally, we present some household composition data for traditional
China. Because of the enormous demographic significance of China, it seems

John Hajnal


Two Kinds of Preindustrial Household Formation System

worthwhile to present a few figures even though they are subject to a number of
uncertainties both in interpretation and regarding their representativeness. The
data come from the China Land Utlization survey of 1929-31.32
Demographic data were collected in 1929-31 and analyzed for over 100
rural localities spread over 16 provinces with a total population of over
202,000. These data have recently been subjected at the Office of Population
Research at Princeton to modern methods of analysis developed for dealing
with imperfect data.33
This analysis reveals a pattern of early marriage for both sexes (mean
ages at first marriage of 21.3 years for men and 17.5 years for women)34 and of
high birth and death rates (each estimated at 41 per 1,000 population). In the
survey, a household was defined as consisting of "all persons living and eating
together, including non-relatives such as hired laborers." In spite of the clause
italicized, the number of nonrelatives enumerated was very small, as shown in
Table 5, constructed on the same lines as Tables 1 and 3.

TABLE 5 Rural Chinese communities, 1929-31 survey:
persons of both sexes per 100 households by
relationship to head
and Other
wives Children relatives Unrelated Total
Whole survey 190 238 94 8 530
"South" 190 250 142 1 583
"South Eastern
Hills" 191 205 68 6 470
SOURCE: Taeuber (1970).

Apart from the data for the whole survey, Table 5 also gives figures for
two of the seven regions into which the data were divided for analysis. These
regions are the extremes, that is, those that have respectively the largest and the
smallest numbers of persons per household.35
In comparison with Northwest European data, the Chinese figures pre-
sent much the same features as the Indian ones, namely presence of substantial
numbers of relatives and very small numbers of unrelated individuals. It is
especially conspicuous (though not shown in the table) that there were virtually
no women unrelated to the head of the household; the few unrelated household
members were men. In the whole survey there were about 12 times as many
unrelated males as females. The absence of unrelated persons from traditional
Chinese households, as well as some of the other features found in Table 5, can
be documented from Chinese populations that were under Japanese rule and for
which statistical data were collected by the Japanese administration.36

4 The age at becoming household head
Within any one population, joint households are, on average, larger than sim-
ple households. Yet households under a joint household formation system are,
on average, not necessarily larger than were households under the Northwest


European simple household systems. Nor is the distribution by size distinctly
different under the two kinds of household formation system, as Table 6 shows.
Northwest European data have been italicized.37
TABLE 6 Selected populations: distribution of
households by size (percentages of households in each
size group)
Number of persons Average number
of persons
1-3 4-6 7-9 10+ Total per household
100 English
1574-1821 36 42 17 5 100 4.8
Rural India, 1951 34 43 17 6 100 4.9
Denmark, 1787
(26 rural parishes) 30 43 21 6 100 5.2
Taiwan, 1915 30 42 18 10 100 5.3
Norway, 1801
(3 areas) 21 46 24 9 100 5.7
SOURCES: See sources to Tables 1 and 3, and Appendix 2, Tables A and B; for data
for Taiwan see Barclay (1954).

This seems paradoxical. Indeed, a number of papers using models have
been devoted to showing how much larger households would be, on average, if
married children joined the households of their parents (as is the case under the
joint household systems) than if they formed independent households at mar-
riage (as was the case in Northwest Europe). These models assume "other
things equal"-in particular, the same age at marriage.38 As the two kinds of
household formation systems in fact operated, the Northwest European ones
could create households as large, on average, as those under joint household
One way to gain insight into the situation is to compare certain kinds of
movements between households under the two kinds of formation system. Sup-
pose, for example, that a girl aged 17 from household X becomes a servant in
household Y under a Northwest European system; and consider for comparison
under a joint household system a girl aged 17 from household X' who is mar-
ried and joins her husband in household Y', which is headed by his married
father. The number of persons in household X could be the same as in X' and
the number of persons in household Y could be the same as in Y'. Then the
effect of the movement between households on the distribution of households
by size would be exactly the same in the two cases. Yet in one case, but not in
the other, a joint household would have been created.
Another way of shedding light on the apparent paradox is to consider the
age at which men become household heads. The ways in which household
headship is attained constitute an important difference between the two kinds
of household formation system.
The reason why the age at which headship is attained is relevant to the
size of households may be seen most easily if simplifying assumptions are
made. Assume that (1) all household heads are men; (2) every man becomes a

John Hajnal


Two Kinds of Preindustrial Household Formation System

head of household if he survives long enough; (3) once a man is a household
head, he remains a household head. The number of heads of household is, of
course, equal to the number of households. Then we know that3

Mean number of persons per household =
Total population Total population
No. of households No. of heads

If men become heads of household later in life, there will, at any one
time, be fewer household heads and thus fewer households. Hence, the mean
size of household will be greater.40
It is not easy to find data that show how entry into household headship
varies with age under our two kinds of household formation system. Under
Northwest European systems men mostly became heads of household at first
marriage; entry into headship was thus concentrated into a comparatively nar-
row age range. This may be shown by the kind of data given in Table 7 for the
Danish rural population in 1801.

TABLE 7 Danish rural parishes, 1801: relation between
entry into marriage and into headship (males)
Percent of all males in age group who are:
Heads of heads
Age group Ever married household of household
18-22 2 2 2
23-27 23 19 18
28-32 56 52 51
33-37 74 74 72
38-42 90 90 88
43-47 91 90 88
48-52 94 90 90
53-57 95 93 91
58-62 96 88 88
SOURCE: Data received from Professor H.C. Johansen.

As shown in Table 7 (the first and last columns of which are graphically
presented in Figure 1A), both marriage and the attainment of headship were
very largely concentrated in the 24 to 40 age range. All three columns of Table
7 show very similar figures. (Only two columns are represented in Figure 1A; if
all three columns were shown, the resulting three curves would not appear
clearly distinct.)
The only comparable set of data for a joint household system known to
me comes from a population not yet discussed: that of fifteenth century
Tuscany (Italy), whose rich records (compiled in connection with taxation)
have been analyzed in magnificent detail by Herlihy and Klapisch-Zuber.41 The
rural population of fifteenth century Tuscany displays all the marks of a joint
household formation system in our sense.

FIGURE 1 Relation between entry into marriage and into
headship: (A) Denmark, 1801, and (B) Tuscany, 1427-30
100 1

Denmark, 1801


10 20 30 40 50 60

- Ever married

----- All heads

Tuscany, 1427-30

42 -3

|/ /


I I I I I/

10 20 30 40 50 60 7
- Ever-married heads

Table 8, which exactly parallels Table 7 (and is graphically represented
in Figure 1B), relates to a part of this Tuscan rural population, namely the
population of the countryside around Pisa.42 Here, the mean number of persons
per household was 4.7. (The data cover some 3,900 households.)

TABLE 8 Tuscany, 1427-30: relation between entry
into marriage and into headship (males)
Percent of all males in age group who are:
Heads of heads
Age group Ever married household of household

18-22 15 14 3
23-27 53 30 16
28-32 74 45 36
33-37 85 57 48
38-42 93 70 67
43-47 96 74 72
48-52 95 84 81
53-57 93 86 82
58-62 95 90 87

SOURCE: Klapisch and Demonet (1972).

Under joint household systems, marriage is not, in most cases, the point
of time at which headship is attained. Two other routes to headship predomi-
nate, namely (1) succession to headship when the head dies, or (2) becoming
head of one of the households formed by splitting a larger one. These processes
are spread out over a much wider age span than marriage.
Table 9 presents the comparison between the rural Danish and the rural
Tuscan populations in a somewhat different fashion (in part repeating informa-


M 80

. 60

0 40

& 20



John Hajnal


Two Kinds of Preindustrial Household Formation System

tion from Tables 7 and 8). It shows that in rural Tuscany there was no relation-
ship between marriage and headship. Of the young married men, only a small
proportion were heads, with this proportion slowly growing up to men in their
60s. The mean age at entry to headship in the Tuscan rural population was
probably on the same order of magnitude as in the Danish rural population, that
is, about 30 years of age. (It is of interest that in Denmark, but not in Tuscany,
the proportion of heads diminishes again around age 60. This is apparently due
to retirement.)

TABLE 9 Comparison of Tuscany, 1427-30, and Danish rural
parishes, 1801
Percentage of heads among:
All men in age group Ever-married men in age group
Age group Tuscany Denmark Tuscany Denmark
18-22 14 2 20 (83)a
23-27 30 19 30 78
28-32 45 52 49 91
33-37 57 74 57 96
38-42 70 90 71 98
43-47 74 90 75 97
48-52 84 90 85 96
53-57 86 93 88 96
58-62 90 88 91 91
aBased on five heads among six ever-married men.

The distribution of households by the number of married couples in the
household in fifteenth century Tuscany is not so very different from twentieth
century India. A rough comparison between the data given in Table 4 for the
Indian state of Maharashtra and the fifteenth century Tuscan population is
shown in Table 10. (This table refers to the whole Tuscan population, not just
the rural part.)

TABLE 10 India and Tuscany: percent
distribution of households by number of
couples in household
(Maharashtra) Tuscany
1947-51 1427-30
No couple 19 23
One couple 58 58
Two couples
Father and son 10 11} 15
Two brothers 5 16
Other relations 1
Three or more couples 7 4
Total 100 100
SOURCES: Maharashtra: Table 4 above; Tuscany: Herlihy and
Klapisch-Zuber (1978), p. 482.


John Hajnal

In spite of some lack of comparability in the two classifications of house-
holds by type from which Table 10 has been constructed, the broad similarity
between the two distributions seems beyond doubt.43 It is very striking how
similar a result has been produced by two joint household formation systems in
two such widely different cultures.
There seems to be only one population subject to a joint household sys-
tem for which a direct analysis has been made of the frequency of the two
modes of attaining headship that have just been mentioned, namely succession
and splitting. Peter Czap has studied populations of Russian serfs on two es-
tates in the first half of the nineteenth century.44 He was able to utilize succes-
sive enumerations of sufficient detail to distinguish between the two modes of
accession to headship. Before quoting his data on this point, we first give a
brief description of the household formation system of the serf population.
Households were much larger than among the populations so far considered.
The mean number of persons per household was over nine. From Czap's care-
ful description of the operation of these households, it is clear that they were
fully integrated households, in spite of their size.
Some household composition data for the serfs on one estate are pre-
sented in Table 11,45 which follows the pattern of Tables 1 and 3.

TABLE 11 Russian serfs (Mishino Estate in Ryazan Province),
1814: persons per 100 households by relationship to head
Married heads
and wives Other heads Children Other relatives Others Total
Males 62 23 130 207 10 432
Females 62 15 79 313 14 483
Both sexes 124 38 209 520 24 915
NOTE: The data relate to 1,173 persons in 128 households.
SOURCE: Computed from Czap (1982a), Table 11 (same as Table 6 of Czap [1982b], "The perennial
multiple family household . .").

The figures for heads and their wives and children in Table 11 are broadly
of the same magnitudes as in other populations covered in this paper.46 It is the
large numbers of "other relatives" that make the serf households so big. In the
serf households there were very few, if any, persons unrelated to the head.
(The category "others" in Table 11 comprises persons whose relationship to the
head could not be determined.)
Indeed it can be shown, since Czap gives a breakdown of household
members by category of relationship, that in the serf households the great ma-
jority of members were not merely related to the head, but were in fact fairly
closely related to him or her. Some 85 percent of household members (other
than heads) were wives of heads, children of heads, or the spouses of children,
grandchildren, or nephews or nieces. The remaining 15 percent included many
whose relationship to the head was only slightly more remote, such as great-

Two Kinds of Preindustrial Household Formation System

The number of married men per household was high in the serf popula-
tion (on average about two married men per household),47 and much higher
than in the other joint household systems covered by this paper. A related
index, proportion of joint households, was also much higher than in other joint
household populations. Of the serf households, some 75 percent were joint48 at
any one time, compared with 15 to 30 percent in other joint household systems
(as shown, for example, in Tables 4 and 10). Because joint households are
larger than simple households, the great majority of the serf population at any
one time were members of joint households. Most serfs were members of a
joint household either all their lives or for most of their lives.
What, in terms of household formation behavior, were the causes of
large household size, large numbers of married men per household, and a very
high proportion of joint households? The mean age at first marriage was under
20 for both sexes. The fact that both men and women became parents so young
meant that they survived their sons' marriages by a longer period than in popu-
lations where marriage occurs at later ages; this tends to create more house-
holds comprising married sons, living under the headship of their father, than
would occur with later marriage. There was also a tendency to delay the split-
ting of households far beyond the point at which it would have occurred in
other joint household systems. Married brothers, as well as cousins, stayed
together in the same household in much larger numbers than in other joint
household systems.
We now return to the age at attaining headship. Czap found that in the
period 1782-1858, there were 343 cases on the Mishino estate in which head-
ship was attained by succession (at ages ranging from 12 to 92) and 112 cases in
which it was attained by splitting (at ages ranging from 18 to 77). When split-
ting occurred among the Russian serfs, the new daughter households were
usually also joint households. In both cases (succession and splitting), the
mean age of the new head was 46. This mean age contrasts with a mean age
around 30 in the data for the fifteenth century Tuscan population, a mean that
may be presumed to be the right order of magnitude for most joint household
systems. Of course, under the serf system a far higher proportion of men died
without ever becoming heads than under other household formation systems.
Most of Peter Czap's work so far has been devoted to the peasants of one
estate, Mishino, in Ryazan province; his data for this estate are the basis of our
description of serf households. Czap has also analyzed four enumerations for
an estate in the province of Tver. In the 1816 enumeration for that estate, he
found that the mean household size was 9.1 and some 75 percent of all house-
holds were joint.49 It must be presumed that the striking household formation
system revealed by Czap's detailed study of the Mishino estate was shared by
other serf populations. Indeed it seems likely that something like the Mishino
household formation system prevailed among populations numbering in the
millions. Average household size can be calculated from aggregate data on
population and numbers of households that were compiled and published in the
1850s and early 1860s by an organ of the Tsarist administration, the Imperial


John Hajnal

Central Statistical Committee. From these data Czap shows that large mean
household sizes (of eight or more persons per household) prevailed over sub-
stantial parts of Russia in the 1850s, although there were also regions where
average household size was rather lower (between six and seven persons per
In addition, there is evidence suggesting that there may also have been
populations outside Russia with household formation systems similar to those
of the serfs studied by Czap. Censuses taken at the end of the eighteenth cen-
tury, in the territory then under the Hungarian crown, covered Croatia, where
mean household size was as large as among the serfs studied by Czap, and the
number of married men per household was also of the same order of magni-
tude. Data from these censuses have been published by the Hungarian Statisti-
cal Office. Table 12 summarizes the 1787 count.

TABLE 12 Territory of Hungary, census of 1787
Population Persons per Married men
(000s) household per household
Hungary proper 6,085 5.22 1.05
Transylvania 1.372 5.03 1.03
Croatia 617 8.33 1.70
61 "Free Royal Cities" 485 4.45 0.84
Total 8,559 5.28 1.06
SOURCE: Hungary (1960).

The brief summary in Table 12 shows how Croatia stands apart. There
are counties within Croatia that were more extreme than Croatia as a whole.
Pozsega (population 64,000) recorded a mean number of persons per house-
hold of 10.2, with an average of 2.0 married men per household. In Zagreb
(population 150,000) the corresponding figures were 10.6 and 2.3. The Croa-
tian population comprised large numbers of serfs (Czap suggests that house-
holds would have been smaller among his serfs if they had been free to form
more separate households, as indeed they did when serfdom was abolished). A
Slavic tradition shared with the Russians may be relevant to the interpretation
of this phenomenon within Croatia.
In Hungary proper, the figures for different counties on size of house-
hold and married men per household suggest that in some places joint house-
hold formation systems more similar to those of other populations considered
in this paper may have been in operation. Detailed analyses of listings of inhab-
itants for three villages for years between 1792 and 1816, in combination with
family reconstitution materials, confirm this picture. They show early marriage
(the average age at first marriage was under 24 for men and under 19 for
women), an absence of unrelated household members, and substantial numbers
of joint households (though, as in India and other joint household populations,
the joint households constituted a minority of all households)."

Two Kinds of Preindustrial Household Formation System

5 The circulation of servants
Many of the terms describing relationships within a household have much the
same meaning across different societies and across the centuries (with no doubt
as to what are equivalent words in different languages). For example, "chil-
dren of the head" are probably very much the same category in all the societies
considered in this paper.
Servants are a characteristic and, on average, a substantial component51
of rural preindustrial Northwest European households; but unlike the word
"child," the term "servant" and its equivalents in other European languages
are apt to be misunderstood. The term refers to an institution that, so far as is
known, was uniquely European52 and has disappeared.
The servants who are recorded53 as household members in the data for
preindustrial Europe were in the main not servants in the now customary mean-
ing of the term, that is, people ministering to the personal comforts of the more
prosperous section of the population. They participated in the productive
tasks-mainly in farming or craft activities-of the households in which they
lived. It must be remembered that at the time "production" was largely carried
on in households, and in such households there was no sharp division between
activities now classified as "production" and those classified as "consump-
tion." Servants lived as integrated members of the household; in particular they
often participated in meals.
Servants were regarded as members of their master's household (or
"family" in the language of the time). The sentence quoted from John Graunt
in section 2 above may serve as an example. Religious writings emphasized
that in the eyes of God, masters had a responsibility for the moral welfare of
their servants similar to that which they had for their children.
The number of servants does not by itself bring out the significance of
the circulation of servants between households in preindustrial Europe. For
example, Table 1 showed that in Denmark at the end of the eighteenth century
there were on average 90 servants per 100 households, so that 17.6 percent of
the total population were servants. But the servants were concentrated at ado-
lescent and young adult ages.
Table 13 shows that in rural Denmark in 1787/1801 well over 50 percent
of those who survived past adolescence were in service at some point in their
lives. Now the proportion of servants in the Danish population at that period
was higher than in many other parts of Northwest Europe. But it was always
true that a very substantial proportion of young men and women experienced
service at some point in their lives. Table 14 summarizes some data (similar to
those of Table 13) that are available for other populations.54 (An additional set
of rates for Iceland is given in Table 17.)
Servants, as has already been mentioned, were almost always unmar-
ried, especially female servants. "All masters discourage the marrying of their
male servants, and admit not by any means the marrying of the female, who are
then supposed altogether incapacitated for their service." So wrote David


John Hajnal

TABLE 13 Danish rural parishes, 1787
and 1801: servants as percent of total
population in each age-sex group
Age group Males Females
5-9 4 4
10-14 36 26
15-19 52 50
20-24 56 51
25-29 43 28
30-34 23 13
35-39 14 6
40-44 6 5
45-49 6 4
50-54 5 3
55-59 5 2
SOURCE: Johansen (1975), p. 158. Values have been aver-
aged for 1787 and 1801 from data given only as integers.

Hume in an essay entitled "Of the populousness of ancient nations" (published
in 1742),55 and a variety of data bear out his words.56 Service was in general a
stage for young people between leaving home57 and marriage, that is, a stage in
the life cycle. Servants in preindustrial Northwest Europe were, in a phrase
coined by Peter Laslett, "life-cycle servants." Servants often moved repeatedly
between households; they were members of their master's household not by
any permanent right, but by virtue of a contract usually fixed for a limited
It was not only the poor and landless whose children went into service.
Those who operated their own farms and even farmers with large holdings sent
their children into service elsewhere, sometimes replacing them with hired
servants in their own household.58 Under certain conditions a high proportion
of all servants were the children of farmers, although, of course, the children of

TABLE 14 Servants as percent of total population
in each age-sex group
Iceland, 1729 1801 9 Flemish villages, 6 English commu-
(3 counties) (3 areas) 1814 nities, 1599-1796
Age group Males Females Both sexes Males Females Males Females
10-14 21 20 10 14 5 5 4
15-19 33 34 32 38 31 35 27
20-24 39 44 33 48 36 30 40
25-29 34 32 19 35 25 15 15
30-39 12 24 8 23 6 6 7
40-49 9 17 3 8 2 2 2
SOURCE: Iceland: Iceland, Statistical Bureau (1975); Norway: Drake (1969) (for the data on Iceland and
Norway see also Appendix 2); Flemish villages: private communication from Richard Wall. The villages
and the data are discussed in an article by Wall to be published in R. Wall, J. Robin, and P. Laslett
(eds.), Family Forms in Historic Europe (forthcoming); English communities: Laslett (1977), Table 1.7.

Two Kinds of Preindustrial Household Formation System

those with little or no land (here called laborers) were more likely to go into
service than farmers' children since they did not have the option of employ-
ment on their parents' farm. This matter was first documented in the pioneering
study by Mackeprang (1907) of the 1645 data for the Danish island of Moen
(see Table 2).
The basic data for Mackeprang's simple calculation are shown in Table
15. Mackeprang assumed that the total of unmarried young people at each age
was divided between farmers' and laborers' children in the same ratio as at age
0-4; that is to say, some 18 percent of the total in each age group would have
been laborers' children. Thus of the total of 557 aged 10-14 some 103 would
have been laborers' children. Since there were 66 laborers' children in that age
group living at home, there would have been 37 in service. One then reaches
the picture given in Table 16.

TABLE 15 Denmark, Island of Moen, 1645: numbers of
unmarried young people (both sexes) at home and in
Farmers' children Laborers' children
Age group at home at home Servants Total
0-4 478 108 0 586
5-9 458 97 20 575
10-14 368 66 123 557
15-19 175 6 165 346
20-24 93 8 110 211
Total 1,572 285 418 2,275
SOURCE: Mackeprang (1907), p. 263.

Of course this method of calculation involves several questionable as-
sumptions (e.g., in relation to migration), but the general picture can hardly be
in doubt. Because laborers formed only a small minority of the population,
most servants must have been the children of farmers. A similar situation ob-
tained in Iceland and probably elsewhere in Europe.59
There was a great increase in the proportion of the landless population in
Denmark (and indeed elsewhere)60 during the seventeenth and eighteenth cen-

TABLE 16 Denmark, Island of Moen, 1645: estimated distribution
of servants by the social class of their parents
Numbers of servants who were Percent who were servants among
children of unmarried children of
Age group Farmers Laborers Farmers Laborers
5-9 11 9 2 8
10-14 86 37 19 36
15-19 107 58 38 91
20-24 79 31 46 80
Total 283 135
SOURCE: Calculated from Table 15 as described in the text. (The figures differ very slightly from
Mackeprang's own results.)


John Hajnal

turies. In the 1801 sample of Danish rural parishes some 50 percent of the
newborn were laborers' children, as compared with 18 percent on the island of
Moen in 1645. Under the circumstances of Denmark in 1801 a much higher
proportion of servants must have been drawn from among the children of la-
To summarize, the characteristics of the institution of service in the rural
populations of preindustrial Northwest Europe may be delineated by the fol-
lowing statements. (1) Servants were numerous, apparently always constituting
at least 6 percent, and usually over 10 percent, of the total population. (2)
Almost all servants were unmarried and most of them were young (usually
between 10 and 30 years of age). (3) A substantial proportion of young people
of both sexes were servants at some stage in their lives. (4) Most servants were
not primarily engaged in domestic tasks, but were part of the work force of
their master's farm or craft enterprise. (5) Servants lived as members of their
master's household. (6) Most servants were members of their master's house-
hold by contract for a limited period. (7) There was no assumption that a
servant, as a result of being in service, would necessarily be socially inferior to
his or her master. The great majority of servants eventually married and ceased
being servants. Their social class before service (i.e., usually the class of their
parents) and their social class after service could be the same as their master's
(and in some Northwest European populations at some periods this was not
infrequently the case).
Servants (or persons whose designation can be translated by that word)
are of course found outside Europe. But elsewhere, and in particular in so-
cieties following a joint household system, the servants represent different
kinds of phenomena both in function and in scale from the Northwest European
variety. Nowhere in data for agricultural joint household populations is there
any suggestion of service along the Northwest European lines, where a high
proportion of men and women spent some portion of their lives circulating
between households in that condition. Purely domestic servants (often espe-
cially numerous in cities) or household heads whose occupation is "servant"
are clearly different from Northwest European rural servants. So is the "estate
servant caste," a distinct population on the Russian estates whose serf popula-
tions have been studied by Peter Czap. Types of "servants" in another joint
household system, namely traditional China, are described by J. L. Watson
(1980). It is clear that these kinds of "service" (e.g., hereditary domestic ser-
vants constituting rather less than 2 percent of the population and not living in
the master's household) bore no resemblance in scale, function, or method of
recruitment to the Northwest European variety. In these and other instances the
groups designated "servants" do not have several of the features that, as de-
scribed in the last paragraph, characterized service in preindustrial Northwest
Europe. In other preindustrial populations such "service" as existed did not
redistribute a large part of the young adult population between households.
It would be impossible in a society where women married very early for
them to spend time as servants in the Northwest European sense and on such a

Two Kinds of Preindustrial Household Formation System

Another argument in support of the conclusion that the Northwest Euro-
pean type of service could not have existed in the joint household systems is
based on the very small numbers of individuals "unrelated" to the head found
under the joint household systems. The number of such persons in the data for
joint household systems is much less (usually under 2 percent of the total popu-
lation) than the number of servants in preindustrial Northwest Europe.
It may be objected to this argument that some servants were relatives of
the head. It is indeed true that some servants were even close relatives. Ann
Kussmaul has pointed out that kinship affected servant hiring in two ways:
"Servants found places with their own kin; masters found servants through
theirs." A higher proportion of servants were relatives of their master than
would have been the case had servants been recruited by random selection.
Nevertheless the great majority of servants were probably either unrelated to
their master61 or much less closely related than household members in joint
household systems. In any case servants were present in the household on a
temporary basis under contract; they did not have a permanent right to house-
hold membership by virtue of their relationship.
The circulation of servants between households seems to have been an
essential feature of the Northwest European household systems. The connec-
tion between late marriage and the existence of life-cycle service has already
been pointed out. Another reason for regarding life-cycle service as an essen-
tial ingredient of the household system is that the other Northwest European
formation rules (rules 1A and 1B of section 1) would frequently have resulted in
households consisting of a couple and their very young children-a unit with-
out a sufficient number of persons capable of doing the work needed to run the
farm. The possibility of hiring servants overcame this problem.62
The Northwest European systems, like the joint household systems,
could operate under conditions very close to subsistence level, as in Iceland.
Probably they could exist under such conditions only because of the institution
of life-cycle service.
The institution of service has been neglected by scholars. Further study
is likely to reveal links between service and many other aspects of the function-
ing of preindustrial Northwest European societies. The paragraphs that con-
clude this section briefly and tentatively indicate some links of this sort that
would be worth investigating.
Much migration in Northwest Europe was the migration of servants, a
large-scale migration of young unmarried adults that had no parallel in joint
household populations. The seventeenth century migration of servants, and
indentured servants in particular, to the New World had its origins in the in-
stitution of service. From Britain and many other parts of Northwest Europe
great numbers of servants crossed the Atlantic. It is thought that more than half
of all those who came to the North American colonies south of New England
were servants.63
Service was in general about as common among women as among men.
This was an aspect of the apparently greater independence of women in prein-

dustrial Northwest Europe compared with women in joint household popula-
tions. While in service, women were not under the control of any male relative.
They made independent decisions about where to live and work and for which
employer. There was also financial independence even though women ser-
vants' wages were lower than men's. Savings accumulated during service were
probably often a substantial contribution to the economic basis of a woman
servant's subsequent marriage. This was probably the reason why women mar-
rying laborers (for whom such a contribution was important) were on average
older at first marriage than women marrying farmers. The future wives of la-
borers would have needed a substantial period of service to accumulate the
necessary savings.
The fact that marriage in Northwest Europe joined together two mature
adults must have affected considerably the nature of the relationship between the
spouses. Indeed this relationship was generally initiated by way of a period of
courtship, whereas in the joint household populations the practice of marriage
arranged by the parents seems to have been nearly universal, often arranged
marriage in the full sense that the couple had little or no acquaintance with each
other before the wedding. The joint household system must also affect the rela-
tionship between husband and wife after marriage because, in contrast with the
Northwest European situation, the young couple are often not the only couple in
the house. The young husband's parents will often be in charge of the household.
The young wife is under the dominance of her mother-in-law at an age at which,
in Northwest Europe, she would often have been in service under an unrelated
mistress. Her husband may continue to have a closer relationship with his
mother, who is present in the household, than with his wife.
Servants were one of the forms of hired labor employed by Northwest
European farmers. There were also day laborers, who were often married.
(Day laborers were usually former servants who had married.) In joint house-
hold populations, households consist almost entirely of relatives, and almost
all work is done by household labor, that is, family labor. There are probably
substantial differences between an agricultural household economy with a con-
siderable element of hired labor and one relying very largely on family labor.
This difference is the core of a "theory of peasant economy" elaborated in the
1920s by the Russian economist Chayanov, whose views have attracted sub-
stantial attention in recent years. Chayanov did not, of course, refer to joint
household formation systems by that term. But something like that was clearly
what he had in mind when he referred to "peasant farms in Russia, India,
China." He discusses the development of a joint household, illustrating his
argument by a "theoretical scheme" (it would nowadays be called a model) in
which a joint household created by the marriage of the head's son lasts for eight
years before splitting takes place.64
Chayanov argued that the concepts of economic theory (rent, capital,
price, etc.) had been developed in the framework of an economy based on
wage labor. From the point of view of the present article one may add that it is
hardly surprising that Adam Smith, Ricardo, and their successors based their

John Hajnal


Two Kinds of Preindustrial Household Formation System

theories on the social system they knew, namely that of Northwest Europe.
However, the notions of rational behavior developed by economists were not
applicable, Chayanov argued, in a different kind of society, in the circum-
stances of "peasants," that is, members of joint household populations. If
Chayanov was right, Northwest Europeans must have differed fundamentally
in their economic behavior from joint household populations, a difference that
must be of great significance for economic development.

6 Origins of Northwest European systems
and consequences for fertility regulation
In this final section, additional aspects of the Northwest European type of
household formation systems are treated, with emphasis on future research.
There is a great deal of information on households in historical times waiting to
be analyzed. Much progress has been made since the availability of com-
puterized data processing, and it seems certain that very much more informa-
tion will be available in only a few years' time. In particular a great deal more
will probably be known about household composition in Southern Europe in
past centuries.
The Northwest European kind of household formation system was found
over a contiguous area in the seventeenth century and probably derived every-
where in this area from a common origin. This kind of household formation
system presumably arose only once in human history. By contrast it seems
possible that many societies have independently developed household forma-
tion systems conforming to the joint household formation rules.
There were indeed differences in household composition between vari-
ous parts of the region here termed Northwest Europe, but the emphasis in this
paper has been on features common to the whole region. The distinctive char-
acteristics of the common Northwest European household formation systems
will emerge more clearly when populations adjacent to Northwest Europe, and
particularly populations of Southern Europe, have been thoroughly studied
with regard to household composition in past centuries. There were in Southern
Europe household formation systems that did not conform to the Northwest
European rules, but that were probably much more similar to the Northwest
European systems than were the joint household systems (for example, there
probably were some "life-cycle" servants). The way in which the distinctive-
ness of Northwest European household formation systems has been presented
in this paper may have to be modified when Southern European systems have
been thoroughly studied.
How long ago did distinctive features of the Northwest European house-
hold formation systems emerge? It may be that substantial data on household
composition for whole communities from earlier centuries will be discovered,
but even in the absence of direct data some inferences about household forma-
tion systems may be possible. For the Northwest European household forma-
tion systems were associated with features that can, to some extent, be traced
in legal and other records for centuries earlier than the seventeenth. Three

features may be mentioned. (1) The presence of large numbers of servants in
the specific Northwest European sense. This feature has already been exten-
sively discussed. (2) Retirement contracts. This feature has been briefly re-
ferred to. If children stay with their parents in a joint household even after they
have become adults, the new generation can take over the work as the parents
age and become weak. The children provide for their parents' needs within the
undivided household. In a system where the children move out of the parents'
household, the consequences of aging cannot be dealt with in this way. On the
one hand, control over land and other resources must in some way pass to the
younger generation so that the society's productive activities can be effectively
carried out. On the other hand, the needs of the old must be provided for by
transferring to them some of the output of the working population. One way of
dealing with these problems was retirement of the head of the household in
return for the guarantee of upkeep by the incoming head. This kind of retire-
ment by contract with the incoming head was not the only solution, and retire-
ment shows only partially in the kinds of data that have been analyzed in this
paper. (That is why retirement was not included explicitly among the house-
hold formation rules listed.) (3) Public provision for the poor, usually by the
local community. This seems to have been present over at least a large part of
Northwest Europe in the seventeenth and eighteenth centuries. It is tempting to
see a connection between this feature and the fact that under the Northwest
European household formation systems the needs of the old, the widowed, and
of other disadvantaged persons would not be met in the same way as in joint
A comparative survey of public provision for the poor in different so-
cieties that attempts to establish how far such provision existed throughout
Northwest Europe, and whether it failed to exist outside the region, would be
of great interest. For the moment let us suppose that such provision was in fact
found everywhere in Northwest Europe in the seventeenth and eighteenth cen-
turies. To what extent did public provision for the poor exist in earlier cen-
turies? (The post-Reformation enactments, such as the Elizabethan poor law in
England, may in part have continued what had already existed.)
It might be possible to investigate how far evidence for the three features
mentioned (servants, retirement contracts, provision for the poor) can be traced
in earlier centuries. For England at least, it seems likely that these features can
be traced back for perhaps four centuries prior to 1600.65 It may well turn out
that aspects of the Northwest European household formation systems can be
shown to be very old indeed. (It would not follow from this fact that, if house-
hold composition data were discovered for say twelfth century England, they
would in all respects be similar to those found in the seventeenth and eight-
eenth centuries. For example early marriage might have been more common in
the early Middle Ages than in later centuries, with consequences traceable in
household composition data.)
It is already clear that Northwest European household formation systems
operated in economic conditions (such as those of Iceland around 1700) that
were very "primitive" by the standards of the more advanced economies of

John Hajnal


Two Kinds of Preindustrial Household Formation System

eighteenth century Europe. The distinctive features of the Northwest European
household formation systems may have had their origins early in the develop-
ment of European societies with economies based on household farms.
We turn now to a discussion of the consequences of the institution of
service for fertility control. It was suggested in section 1 that the institution of
service in Northwest Europe made late marriage possible. It probably also
made possible variation over time in the age at marriage and the proportion
remaining unmarried. Such variation could occur in response to changing eco-
nomic circumstances. When conditions were unfavorable for setting up new
households, some who might have married directly from their parents' home
probably went into service, and others remained longer in service than they
would have done had earlier marriage been possible. As servants, the unmar-
ried could move to farms and localities where their labor was most useful.
If this suggestion is correct, then the institution of service played an
important role in the demography of preindustrial Northwest Europe. For it
constituted part of the much discussed mechanism that, by varying the age at
and extent of marriage, adjusted fertility and thus population growth in re-
sponse to the economic conditions. It was probably because of service that
Northwest Europe could operate with a balance between birth and death rates
established at a lower level than prevailed in other preindustrial societies.
This suggestion is open to investigation, in principle at any rate. With
suitable data it would be possible to trace relations between variations over
time in the proportion of the population who were in service, the age at mar-
riage, fertility, and household composition. Delayed marriage did not, how-
ever, inevitably involve going into, or remaining in, service, and other factors
besides the age and incidence of marriage powerfully influenced the number of
servants. Thus, too close a correlation between variations in the incidence of
late marriage and of service must not be expected.
The institution of service was common to the whole of Northwest Eu-
rope, an area within which a considerable variety of social conditions was to be
found. The response to adversity by way of increased service and delayed
marriage may have operated in rather different ways in different countries.
We present brief and preliminary examinations of three contrasting situa-
tions. The first example is that of Iceland in the early eighteenth century. The
censuses of 1703 and 1729 encompassed a period that saw change of just the
kind that should display the phenomenon under discussion. The 1703 census
showed large numbers of servants, a very high proportion unmarried, and an
age structure indicative of a very low birth rate in the years before the census.
The low birth rate was presumably a reaction to the appalling economic condi-
tions,66 due in part to climatic disasters, that caused the Danish government to
order the taking of the census. In 1707-8 there was a further disaster, a severe
smallpox epidemic. The returns of the 1729 census have survived for only three
counties covering about one-fifth of the total population. In those counties the
population in 1729 was some 20 percent below the 1703 level. But the 1729
census shows signs of a sharp recovery of the birth rate. H. O. Hansen, who
analyzed the 1729 material for the Icelandic Statistical Office, drew attention to

the close spacing of the births of the children surviving to the census, an appar-
ent increase in the proportion of the population who were married (especially
within the young age groups), and the enlarged proportion of the population
under age 15. These facts, he wrote, indicate "the existence of an extremely
high level of marital fertility, intensive formation of young families, i.e., a
drastic fall in the age at marriage and consequently an immediate and explosive
growth of the population through the birth rate just after the crisis."67
Was the reduction in age at marriage associated with a decline in service?
The proportion of servants in the population was 19 percent in the whole of
Iceland in 1703, but only 17 percent in 1729 in the three counties whose data
have survived. The age-specific picture shown in Table 17 suggests a substan-
tial shift in the age pattern of service between the two censuses (although the
change may be due in part to the change in geographical coverage, apart from
the possibility of defective data). The 1703 proportions in service are lower
than the 1729 ones at ages under 20, but higher at ages over 25. It is as if there
had been a shift to a pattern of leaving home earlier and marrying earlier. An
increase in the proportion of women married68 at ages 25-39 equal to the de-
crease between 1703 and 1729 in the proportion in service in that age range
would have made a substantial contribution to the birth rate.

TABLE 17 Iceland, 1703 and 1729 (3 counties):
servants as percentage of population
in each sex-age group
Males Females
Age group 1703 1729 1703 1729
10-14 2 21 1 20
15-19 24 33 20 34
20-24 44 39 41 44
25-29 45 34 45 32
30-39 25 12 32 24
40-49 12 9 20 17
SOURCES: Computed from Iceland, Statistical Bureau (1960), Table V, and Iceland,
Statistical Bureau (1975), Tables 1 and 9.

Iceland at the beginning of the eighteenth century was very "backward"
both demographically (with very high mortality) and economically, being en-
tirely rural with a near total dependence on sheep farming and fishing. England
in the seventeenth and eighteenth centuries was a very much more "advanced"
society. Wrigley and Schofield, in their recent fundamental and revolutionary
reconstruction of English demographic history from 1541 to 1871, have pre-
sented a picture of substantial variation over time in the rate of population
growth due to variations in birth rates. These fertility variations were in turn
largely due to variations in the marriage rate, which reflected changes in age at
marriage and proportions marrying. Wrigley and Schofield interpret these
changes in marriage rate as a lagged response (the lag is of the order of 25-30
years) to economic changes as shown in an index of real wages.

John Hajnal


Two Kinds of Preindustrial Household Formation System

Demographic changes in seventeenth and eighteenth century England
proceeded in long-term swings. There was apparently a period of fairly rapid
growth in the second half of the sixteenth century. Then there was a prolonged
period of late marriage, with resulting low fertility and low population growth,
in the seventeenth century. This period was followed by a century of increasing
marriage rates and rising population growth. The mean age at first marriage for
both men and women appears to have been some 1.5 years lower in the period
1750-99 compared with the period 1650-99. The proportion of persons of both
sexes who had never married by age 40-44 is estimated to have fallen from
some 25 percent in the middle of the seventeenth century to some 5 percent in
the second half of the eighteenth century.69
Was the change in marriage habits in England between the seventeenth
and eighteenth centuries accompanied by a diminution of the proportion of
young people who entered service and of the length of time spent in service?
Some evidence suggesting that it was is set out in Table 18, which is taken from
the work of Richard Wall. He has compared household composition data from
eight local lists of inhabitants spread over the period 1650-1749 with data from
lists of 19 communities in the period 1750-1821. The lists used for this com-
parison were selected for their high quality from a much larger number of
English lists that have been preserved and analyzed (see Table B in Appendix
2). Unfortunately the sets of lists for the two periods come from two non-
overlapping sets of communities so that, in the comparison between the two
sets, the change over time may be confounded with some local variation. We
do not know how closely the household characteristics of the two sets of lists
approximated those of the whole English population in the two periods being

TABLE 18 English communities in two periods: persons of both
sexes per 100 households by relationship to head
Heads and Other
wives Children relatives Servants Lodgers Total
1650-1749 163 177 16 61 26 444
1750-1821 175 209 22 51 24 481
NOTE: The figures for 1650-1749 are based on eight lists covering 866 households that comprised 3,850
persons. The figures for 1750-1821 are based on 19 lists covering 1,900 households that comprised 9,133
SOURCE: Wall (1982).

In the sets of parishes covered by Table 18, servants formed 14 percent of
the total population in the lists for 1650-1749, that is, the period of com-
paratively late and little marriage, whereas only 11 percent of the population
were servants in 1750-1821.70
Following the argument given at the beginning of section 4, one would
expect that, in two Northwest European populations with the same age struc-
ture, households would on average be larger in the population where men
married later and less-populations where there were fewer heads of house-
hold. Thus, according to this argument households should on average have


John Hajnal

been larger in the late seventeenth than in the late eighteenth century. How-
ever, the change in age at marriage between the seventeenth and eighteenth
centuries did not leave the age structure unchanged. The birth rate was substan-
tially lower in the earlier period and hence there were fewer children. The data
in Table 18 conform to this expectation; households appear on average to have
been larger in the later, high-birth-rate period, mainly because of the larger
number of children.71
Finally, one may look back to Table 2 and apply the same considera-
tions. Table 2, like Table 18, gives data for two periods in one of which there
were fewer servants, but more children, than in the other. In rural Denmark in
1787/1801, where there were more servants, fertility was much lower than on
Moen in the seventeenth century. (Only 24 percent of the population were
under age 10 in 1787/1801, compared with 30 percent on Moen in 1645.) Thus
in Denmark also the incidence of service seems to have varied inversely with
marriage and fertility.
Joint household populations and Northwest European populations must
have reacted in fundamentally different ways to adverse economic difficulties
and particularly to difficulties resulting from population growth. In joint house-
hold populations an increased population could result in the underemployment
of married adults in circumstances that, in Northwest Europe, would have
caused pressure to delay marriage. The institution of service was probably an
essential part of the mechanism by which marriage could be delayed, with the
result that population growth was under partial control. Populations with joint
household systems lack that mechanism.

Appendix 1
The definition
of a household
What is a household? If one wishes to make
meaningful statistical comparisons of house-
hold size and composition between cultures
and across centuries, it is necessary to use a
household concept that is appropriate to all the
societies being compared, and to be able to
suppose that the statistical data used are the
result of enumeration or recording procedures
in which the chosen concept has been at least
roughly applied.
One important restriction on the concept,
for the purposes of the present paper, arises
from the fact that it is necessary to allocate
every person (or virtually every person)72 in
the population of an area uniquely to one
household. Given this constraint and the na-
ture of the available data, there seems in effect
no choice. A household, for our purposes,
must be defined as a housekeeping or con-
sumption unit. A definition of this type is used

in the majority of modem censuses. The es-
sential characteristic of a household in this
sense has often been taken to be the eating of
meals together by all members of the house-
hold, or the sharing of meals deriving from a
common stock of food. (When shared con-
sumption is taken as the defining characteristic
of a household, it is in general tacitly assumed
that spouses and, more particularly, parents
and their young children are in the great ma-
jority of cases in the same household. This as-
sumption holds true for the populations dealt
with in this paper.)
The notion that a household comprises all
those who share their food with one another
has clearly been current among many peoples
over many centuries and is embodied in such
descriptions of a household (or "family")73 as
comprising those eating from a common pot or
sharing the same wine and bread. Economic

Two Kinds of Preindustrial Household Formation System

dependence on the household head is, under
preindustrial conditions, a closely related no-
tion. A people's own notion of who con-
stituted members of one household must have
had a great influence on the process of compil-
ing censuses and administrative records in
which individuals were to be listed by house-
The definition of a household as consisting
of those depending on one head and sharing
the same stock of food is already explicitly
formulated in the instructions for the censuses
carried out in Austria-Hungary toward the end
of the eighteenth century. These censuses cov-
ered large populations of diverse nationalities,
cultures, and patterns of household formation.
The essence of the definition used in these
Austro-Hungarian censuses was as follows:
"All those should be counted in one house-
hold, and accordingly entered on the same
household schedule, who do not cook for
themselves, but are nourished under one mas-
ter or mistress on the same table and bread,
whether they be married or not." Where more
than one married couple share the same table
and bread, they are thus counted in the same
household, as are living-in servants or others
who do not cook for themselves.75 By this def-
inition it was possible to have more than one
household in the same house or to have one
household occupying several dwellings. In the
territories under the Hungarian crown enumer-
ated in 1784-87, the numbers of households
per house were as follows:

Hungary 1.24
Transylvania 1.12
Croatia 0.96
61 "Free Royal Cities" 1.59

Thus, in Croatia, where households con-
tained an average of 1.7 married couples,
there were more houses than households. At
the other extreme, the sharing of a house by
several households was most common in the
The effects of different definitions and of
inconsistencies in their application in the enu-
merations are probably greater in Northwest
Europe than in other populations with which
this paper is concerned. One reason is that in
joint household systems, households consist
almost entirely of members who have a perma-

nent right to belong by virtue of relationship
(including adoptive relationship). Such a
household is a fully integrated unit, leaving lit-
tle room for doubt who its members are. In
Northwest Europe, on the other hand, there
were frequently to be found living in the same
farm, house, or group of buildings groups of
individuals not sharing fully in one integrated
household. In such cases the subsidiary group
(usually the smaller one) could be, and was by
contemporaries, treated variously either as a
separate household or as part of the main
household. One such sort of group were re-
tired persons-in particular, parents who had
handed over their farm to the heir in return for
a contract guaranteeing an allowance. This
category was fairly frequent in some areas of
Northwest Europe. The extent to which a wid-
owed retired person or a retired couple, possi-
bly with children or servants of their own,
lived apart and led an independent existence in
terms of eating meals and so forth76 must have
Laborers with their families living on the
farm for which they worked constituted an-
other sort of subsidiary group whose relation-
ship to a larger household could be ambigu-
ous. The Danish scholar E. P. Mackeprang, in
his analysis of the 1645 census of the island of
Moen, noted that "husmaend" (laborers) were
sometimes enumerated with the farm house-
hold and sometimes on their own, with incon-
sistent enumeration practices between the
parishes77 of the island.
General terms such as "lodgers," and
equivalent terms in other languages, often
cover subsidiary groups whose integration
with the main household may be questionable.
Interesting light on the possible effect of
different definitions of a household in instruc-
tions to census takers is shed by a comparison
between the Danish censuses of 1787 and
1801. At the 1787 census it was specified that
"farmers with their households should be
counted separately and laborers with their
households separately . ." In 1801 there
was an additional instruction that "if several
households live in a house or farm, then every
household should be counted separately, to-
gether with the persons belonging to it."
Johansen78 analyzed the allocation of indi-
viduals to households at the two censuses and
concluded that the 1801 addition to the in-
structions resulted, in particular, in the sepa-
rate identification of households of laborers

John Hajnal

living on farms. He believes that in the 1787
census the number of households distin-
guished was some 2 to 3 percent too low.
Changes in definition also affected the number
of retired people classified as living independ-
Archives in European countries have pre-
served many documents listing individuals by
households, documents deriving from the tra-
dition in the Roman Catholic church by which
the parish priest was supposed to prepare, and
revise regularly, a list of his parishioners, not-
ing who had been confirmed, attended com-
munion, and so on. This regular check on
religious practice was instituted later than the
registration of baptisms and marriages, but in
1614 the promulgation of the Rituale Ro-
manum under Pope Paul V formalized a long-
standing tradition and laid down procedures
for the priest to follow in maintaining a regis-
ter of souls (Liber Status Animarum), along
with registers of marriages, baptisms, and bur-
The crucial feature, from the point of view
of the present paper, was that the register was
to be kept by households, presumably for the
same practical reasons that censuses and sur-
veys enumerate people by households. How-
ever, no definition was laid down of what
constitutes a household. The Rituale Ro-

manum of 1614 prescribed that the Liber Sta-
tus Animarum be kept in the following
manner: "Each household is to be separately
noted, with a space left between each one and
the following one; the name, surname, and age
is to be listed of each individual who is of the
household or lives in it temporarily.""79 There
follow instructions for the notations recording
conformity to the various religious obliga-
tions. This Catholic tradition continued to
have effect after the Reformation in Protestant
countries, and listings were compiled by cler-
gymen there on similar principles.80
For most of the older European listings of
individuals by households, whether they be
general government censuses, administrative
documents, taxation registers, confession
books, or whatever, there is no explicit defini-
tion of what was to be taken to constitute a
household. Sometimes there is uncertainty
about just which of the persons listed the origi-
nal compiler of the documents intended to be
included in one household. But where the cat-
egories of persons mentioned (such as retired
people or farm workers) who are likely to have
formed separate housekeeping units are pres-
ent in substantial numbers, we may suspect
that some of the "households" are larger units
than they would be had a strict housekeeping
unit definition been adhered to.

Appendix 2 Household
composition data
for Northwest Europe

The following three tables summarize house-
hold composition data, such as were given for
Denmark in section 2, for other Northwest Eu-
ropean countries in the seventeenth and eight-
eenth centuries. Only brief comments have
been added. The categories of relationship to
the head, distinguished in the analyses on
which the tables are based, vary somewhat.
The figures for Iceland81 and Norway
(Table A) are, like the data for Denmark
quoted earlier, modern reanalyses of the origi-
nal schedules of all or parts of complete na-
tional censuses. The data for Iceland in 1703
cover a population of 50,358. Paupers were
enumerated on a separate register. (The major-
ity of them were allocated by the parish
authorities to stay with individual house-
holds.) They are included in Table A under the

heading "others." The 1729 figures relate to
the three counties for which the original sched-
ules have survived. These three counties had a
population of 8,077 in 1729; at the 1703 cen-
sus their population had been 10,107.
The Icelandic household composition data
show considerable differences between 1703
and 1729. As explained in section 6, these dif-
ferences are probably attributable not so much
to the difference in geographical coverage as
to the considerable changes that took place be-
tween 1703 and 1729. The age structure
shown by the 1729 data is very different from
that revealed by the 1703 census.
The Norwegian figures in Table A relate to
three rural areas selected for analysis by M.
Drake. There are notable differences between
the household composition data for the three


Two Kinds of Preindustrial Household Formation System

areas. Nevertheless, for the purposes of this
paper I have combined the data for all three
areas (with a combined population of some
21,000). The figures are not based on the total
population of the areas concerned, but cover
only the households of farmers and crofters.
These, however, account for 98 percent of the
The average size of Norwegian households
in the three areas covered in Table A was 5.7,
rather larger than in Denmark; but in the Nor-
wegian case it seems likely that some of the
retired couples and lodgers formed separate
households, so that the true size according to
the household definition used in this paper
would be smaller (see Appendix 1).
The English data in Table B come from
analyses by the Cambridge Group of listings
of individuals for 100 communities with a total
population of some 68,000. The listings are of
varied character and spread over the years
1574-1821. Where relationship to the head is
not clearly specified in the original documents,
it is sometimes "presumed" on the basis of
indirect evidence; for example, members of a
household with the same surname are pre-
sumed to be relatives. Those placed in the col-
umn headed "lodgers" in Table B are
individual lodgers; groups of two or more
lodgers were omitted from the analysis. (Such
groups comprised less than 2 percent of the
total population.) The treatment of the data in
these and other respects is fully described in
Laslett and Wall (1972).
The figures for North Holland (Nether-
lands) in Table B are based on data for 2,367
households comprising 8,842 persons from
listings for 1622-1795. In this case those indi-
vidual households appear to have been se-
lected for inclusion in the analysis where the
information given was adequate. The house-
holds thus selected could be compared in cer-
tain characteristics (average size, proportion
of the population in service, predominance of
female servants) with data covering some
8,500 households in the same area collected
by the demographer Struyck around 1740. The
two sets of data agree in these characteristics.
In the low mean household size of 3.7, and
the very great predominance of female ser-
vants (not shown in the table), these data for
rural North Holland differ markedly from any
other data of high quality for a large rural pop-
ulation anywhere outside the Netherlands in
preindustrial Northwest Europe. In the Nether-
lands also, other eighteenth century data,

where so far analyzed, except in one area
(Friesland), show mean household sizes over
4.5 and more than 10 percent of the population
in service.83
The rural area in North Holland to which
the data in Table B refer was a very special one
and perhaps should not be regarded as a
fully preindustrial rural area (although it is of
course included in preindustrial Northwest Eu-
rope by the purely chronological definition
used in this paper). The province of Holland
was already heavily urbanized in the seven-
teenth and eighteenth centuries, with more
than half the population living in towns. There
was a great deal of nonagricultural employ-
ment in the countryside; agriculture itself was
dominated by market-oriented livestock farm-
The figures labeled North Brabant in Table
B are for six villages in the West of North Bra-
bant in the Netherlands. They cover popula-
tions of 8,805, 10,745, and 7,688 in the years
1750, 1755, and 1800. The sources are tax
registers compiled on an annual basis and ap-
pear to be of very high quality.
The data for the six Flemish villages (now
in Belgium) are from a census taken in 1796
under the French administration. The total
population covered is not precisely given.
Finally, Table C gives two examples of a
classification using somewhat different cate-
gories employed in the seventeenth and eight-
eenth centuries. One set of figures comes from
records compiled for tax purposes for estates
in lower Austria at the end of the seventeenth
century. They have been analyzed by Mit-
terauer. I have combined the figures for all the
estates, covering 4,942 persons in all. The
second set of figures comes from Stissmilch's
classic treatise and covers 9,690 persons in 50
villages near Berlin in 1738. In each case the
basic categories are "men," "women," "chil-
dren," "servants," and "inmates." Siissmilch
also distinguishes retired people as a separate
category and subdivides the other categories
It is not clear just how many households
are involved and exactly what sorts of persons
the categories "men" and "women" (present
in nearly equal numbers with slightly more
men) included. For rough comparability with
the other tables, all figures have been shown
per 100 "men," on the assumption (also
argued by Mitterauer) that "men" and
"women" represented mainly heads and their

TABLE A Iceland and Norway: persons per 100 households
classified by relationship to head

Married heads Other Other
and wives heads Children relatives Servants Others Total
Iceland, 1703
Males 69 17 100 11 52 30 279
Females 69 14 102 27 66 57 335
Both sexes 138 31 202 38 118 87 614

Iceland, 1729
(3 counties)
Males 78 14 111 5 31 21 266
Females 78 8 112 7 61 42 308
Both sexes 156 22 223 12 98 63 574

Norway, 1801
(3 areas)
Males 93 4 118 11 19 32 277
Females 92 3 107 15 32 44 293
Both sexes 185 7 225 26 51 76 570

NOTE: The column headed "others" covers paupers in Iceland (see text), but lodgers and retired people
in Norway.
SOURCES: Iceland, Statistical Bureau (1960), p. 19; Iceland, Statistical Bureau (1975); Drake (1969),
Table 18.

TABLE B England, the Netherlands, and Belgium: persons per 100
households classified by relationship to head (both sexes)

Heads and Other
wives Children relatives Servants Lodgers Unidentified Total

English communities,
1574-1821 163 203 16 63 7 23 475
North Holland,
1622-1795 156 176 5 24 13 374
North Brabant
(6 villages)
1750 160 200 20 70 10 460
1775 170 230 10 70 10 490
1800 170 220 10 70 10 480
Flanders, 1796
(6 villages) 170 240 30 40 0 480

NOTE: The ratios on which the figures for North Brabant and Flanders are based are available to one significant digit
less than other similar figures in this paper; in effect, they give the numbers of persons per 10 households. A "0" has
been added to each such figure for better comparability.
SOURCES: Laslett and Wall (1972), Table 1.13, p. 83; Laslett and Wall (1972), Tables 12.9 and 12.11, pp. 316-317;
Klep (1973), p. 58; Vandenbroeke (1976), p. 276.

TABLE C Lower Austria and villages near Berlin:
persons per 100 "men"

Men and Retired
women Children Servants Inmates persons Total

Estates in Lower
1695-96 197 224 88 45 554
50 villages near
Berlin, 1738 199 245 141 48 26 659

SOURCES: For Austria: Mitterauer (1973), pp. 214-215; for villages near Berlin: Siissmilch (4th edition,
1775), vol. 2, p. 277.

Two Kinds of Preindustrial Household Formation System

A shorter version of this paper will appear in a
volume entitled Family Forms in Historic Eu-
rope, edited by Richard Wall in collaboration
with Jean Robin and Peter Laslett (Cambridge
University Press, forthcoming 1982).
This paper was written during a stay at the
Office of Population Research, Princeton Uni-
versity (supported by the Andrew W. Mellon
Foundation), and at Rockefeller University
(supported by the Ford Foundation). Without
the help of many people in both institutions
(especially the extraordinary amount of assis-
tance given by the library staff at the Office of
Population Research) the paper could not have
been written. I wish especially to record my
gratitude to Joel E. Cohen, whose initiative
and warm support made possible my period of
work in the United States. A Nuffield Founda-
tion Fellowship in 1974-75 and a Simon Fel-
lowship at the University of Manchester in
1953-54 made possible some earlier work on
this topic.
1 The adjective preindustriall" has,
however, usually been omitted; the terms
"Northwest Europe" and "Northwest Euro-
pean" will refer to this area of Europe in the
seventeenth and eighteenth centuries unless
otherwise stated.
2 See Hajnal (1965). The present paper
may be regarded as a sequel to that paper.
3 R. M. Smith (1979) reviews this evi-
dence. For the contrast in household structure
between northern and southern France in the
seventeenth and eighteenth centuries, see
Flandrin (1976), pp. 241-242.
4 Information on the "random variation"
in household characteristics to be expected in
small populations may be found in Wachter et
al. (1978).
5 My debt to the Cambridge Group goes
far beyond the reading of their publications. I
have benefited greatly from many discussions
with Peter Laslett and other members of the
group (especially Richard Smith and Richard
6 "Mean ages at first marriage" is in-
tended to refer to the mean ages at first mar-
riage experienced by a cohort passing through
life. Usually, only other types of indexes of
mean age at marriage will be available. (The

singulate mean age at marriage will be some-
what higher than the mean age at marriage of a
cohort, since the SMAM, in effect, assumes
that there is no mortality within the age range
where marriage occurs.)
7 There is no accepted term for the cou-
ple who are jointly in charge of a household; it
might be useful to adopt the term "house-
keepers," quoted by Richard Wall from a his-
tory of the parish of Corfe Castle in England
(see Laslett and Wall [1972], p. 166).
8 The custom whereby one son remains
at home and takes over the farm when his fa-
ther retires has often been regarded as an es-
sential characteristic of the "stem family" (at
least on some definitions of that contentious
term). In some parts of Northwest Europe,
such "stem families" were numerous. On the
other hand, the kind of "stem family" arrange-
ment in which one son remains at home and
marries while his father continues as head of
the household after the son's marriage did not
occur in Northwest Europe. It is incompatible
with rule IB. No kind of stem family system
can be classified as a joint household system in
our sense. In a stem family only one heir re-
mains in the parental household after mar-
riage. Under a joint household system all sons
normally bring their brides into the household
by rule 2B.
9 For some references see Wheaton
(1975), p. 619.
10 If there is no son in a household, but
there is a daughter, the daughter can bring her
husband into the household. If there are no
children at all, some equivalent of adoption
can be resorted to. Devices of this sort occur in
all the populations studied in this paper.
11 For an excellent brief description of
communal provision for the needy in prein-
dustrial England, see R. M. Smith (1981), pp.
12 For an extensive discussion of the im-
portance of household extinction and splitting
in their effects on social differentiation in one
joint household system, that of rural Russia
toward the beginning of this century, see
Shanin (1972), especially chapters 5 and 6.
13 Shah (1973), pp. 167-169, discusses
the effect of wealth and poverty on the founda-


John Hajnal

tion of joint households in India. He stresses
the influence of caste. Herlihy and Klapisch-
Zuber (1978) give extensive data on household
composition by economic status in fifteenth
century Tuscany.
14 As mentioned in note 8, a stem family
system in which a single heir remains in the
household with his spouse after marriage
while the old head does not retire does not fall
under either of our two kinds of household for-
mation system.
15 There have undoubtedly been other so-
cieties that have followed the joint household
formation rules besides those discussed in sec-
tions 3 and 4 of this paper. Two further sets of
data may be mentioned here. (1) The figures
on households in Anatolia (Turkey) given in
Stirling (1965), pp. 37-38, are clearly char-
acteristic of a joint household system. (2) The
data in the fascinating study by K. Hopkins
(1980) suggest that the population of Roman
Egypt in the first three centuries of the
Christian era followed a joint household sys-
tem. (The household data are on pp. 328-334
of Hopkins's article. His figures are, in my
view, entirely compatible with mean ages at
first marriage appropriate to a joint household
16 The way such subsidiary units are re-
corded in the original enumerations, and how
they ought to be treated, are discussed in Ap-
pendix 1.
17 Thus, for example, the Danish data
show the number of married servants; but it is
not possible to obtain this information from
listings in which only a servant's relation to
the household head (i.e., the fact that he is a
servant) is recorded, but not his marital status.
18 See Johansen (1975), Chapter 10. I owe
thanks to Professor Johansen for supplying ad-
ditional materials and answering questions
about his data. I am grateful to Ulla Larsen for
helping with the Danish text.
19 The numbers in the table are averages
of the number of persons in each category per
hundred households, calculated separately for
1787 and 1801. The 1787 and 1801 figures are
very similar; differences between them are dis-
cussed in Appendix 1.
20 The difference may be slightly exag-
gerated by the figures as given. There were, in
1787 and 1801, grandchildren, children of

other relatives, lodgers, and others who were
not included among "children." Their coun-
terparts in 1645, may, however, have been
classified as children.
21 Graunt uses the term "family" as was
universal usage at the time for what today
would be called a "household."
22 Johansen (1975), p. 85.
23 This matter is discussed in Appendix 1.
24 See Laslett (1977), Tables 1.1 and 1.2.
Table 1.5, containing data on "other" rela-
tions, is also relevant.
25 Of course the evidence, especially for
the seventeenth century, is patchy for much of
Northwest Europe. It seems to me conceivable
that evidence will come to light showing that
there were areas with a somewhat larger pro-
portion (than 6 percent) of retired parents liv-
ing in fully integrated households with married
children. Even then, the number of joint
households would be much smaller than under
joint household systems. (If one excluded
from the "joint household" category house-
holds in which the son is head and the father
has retired, then the contrast between North-
west Europe and, e.g., India and China would
be even clearer.)
26 The information about households to
be found in the Indian censuses up to, and in-
cluding, 1961 is surveyed in Dandekar and
Unde (n.d.). In 1951 a household was defined
as "all persons who live together in the same
house and have a common mess," that is, who
took their food from a common kitchen.
27 Data from past censuses in India on the
size of households and on the numbers of mar-
ried men and widowers per household suggest
that there has been no decrease in size or in the
extent of "jointness." (See Dandekar and
Unde [n.d.].) Shah (1973) compared house-
hold data from a census of a village in Gujarat
in 1825 with an enumeration of the same vil-
lage in 1955. The average household size was
4.54 in 1825, compared with 4.61 in 1955. He
concluded from a variety of indications that
"complex" households (roughly our joint
households) were no more frequent in 1825
than in 1955.
28 This figure is not given directly in the
census, which gives the number of male heads
and a single total for female heads and wives
of heads combined. Using these figures in con-


Two Kinds of Preindustrial Household Formation System

junction with the total number of households
(i.e., the total number of heads), it is possible
to deduce the number of married male heads.
This number can be compared with the total
number of married men. This sequence of
steps is complicated, when applied to the 1951
Indian census, by the fact that household data,
on the one hand, and marital status data, on
the other, were obtained from different sam-
ples of the census schedules.

29 The mean number of persons per
household was slightly larger according to the
1961 census, namely about 5.2, than in 1951.

30 No details of the sampling methods or
other survey procedures are given in the report
(cited in Table 4). However, in various char-
acteristics (distribution of households by size,
number of couples per household, etc.), the
results conform to the census. They are taken
as representative of conditions elsewhere in
India in Dandekar and Unde (n.d.), p. 58. The
comparison with Nepal given in Table 4 also
suggests that the results of the survey give the
right orders of magnitude. In 1961, 70 percent
of married men in rural Maharashtra were re-
ported as household heads, just a little higher
than for all India.

31 Nepal resembles India in that (a) mar-
riage occurs very early (the mean age at first
marriage for Nepalese women was 16 years ac-
cording to World Fertility Survey data); (b)
marriage is virtually universal (98 percent of
Nepalese women were reported as having been
married among those aged 30-34 and in older
age groups); and (c) there is frequently a sub-
stantial delay between marriage and the onset
of cohabitation; see World Fertility Survey
32 The survey is described in Buck (1937)
(see especially Chapter 13, by F. W. Note-
stein). For comments on the representative-
ness and quality of the data, the papers by
Taeuber (1970) and Barclay et al. (1976)
should also be consulted.

33 See Barclay et al. (1976).

34 These figures are singulate mean ages
at marriage, implying that the average ages at
first marriage would be even lower.

35 The "South" comprised six localities

with a total population of 11,107 in Kwangtung
and Fukien. The "South Eastern Hills" com-
prised four localities with a total population of
7,680 in Kiangsi and Chekiang.

36 See, in particular, Barclay (1954).

37 The population of Taiwan was a popu-
lation of Chinese culture and followed a joint
household system. The other data are dis-
cussed elsewhere in the paper.

38 See, for example, Burch (1970).
39 Here, as throughout this paper, it is as-
sumed that every person is allocated to one
household and that the institutional population
can be neglected.
40 It would be tempting to conclude that
under our simplifying assumptions we should
Number of heads = Total male population
above the mean age of accession to headship.
This relationship will, however, not hold ex-
actly in general, not even in a stationary popu-
lation, although it will be correct in a
stationary population where there is no mor-
tality over the age range in which accession to
headship occurs.

41 See Herlihy and Klapisch-Zuber
42 The figures for this part of the Tuscan
rural population were given in an earlier article
by Klapisch and Demonet (1972). The full
publication of the material in Herlihy and
Klapisch-Zuber (1978) does not contain the
corresponding data for the whole Tuscan rural
population, but the graph on p. 490 of the
book suggests that figures for the whole
Tuscan rural population would be very similar
to those in our Table 8.

43 The classification used for the Tuscan
data by Herlihy and Klapisch-Zuber (1978) is
fortunately detailed. It largely follows the
Cambridge Group classification, which was
mentioned at the end of section 2 (i.e., the
classification is in terms of the number and
type of conjugal family units in the household,
rather than in terms of the number of couples).
The figures in Table 10 have been produced by
combining suitable categories from the Tuscan
data, as follows:


John Hajnal

categories, Herlihy
and Klapisch-Zuber
Group in Table 10 (1978), p. 482
No couple All of 1 and 2 plus 3c
and 3d
1 couple 3a and 3b plus all of 4
2 couples: father 5a ("verticaux A deux
and son noyaux")
2 couples: 2 brothers 5c ("horizontaux a deux
3 or more couples 5b ("verticaux a trois
noyaux ou plus") plus
5d ("horizontaux A trois
noyaux ou plus")

44 Professor Czap very kindly made two
of his papers available to me before publica-
tion, namely Czap (1982a, b). He also helped
me with a number of queries.
45 As can be seen from the note under
Table 11, that table is based on rather smaller
numbers of households and of persons than the
other tables in this paper. However, Czap has
analyzed eight enumerations between 1814 and
1858 for the serf population of Mishino estate.
This cumulative body of evidence makes it
clear that the picture given in Table 11, as well
as the other facts quoted in the text (age at
marriage, number of married men per house-
hold, proportion of joint households), is not
the result of random fluctuations associated
with small sample size. Some changes over
time can be deduced from the sequence of
eight enumerations, but these changes are not
discussed in this paper.
46 The figures for heads and children in
Table II, while of the same broad orders of
magnitude as elsewhere, present some special
features. The number of married heads is
smaller and the number of "other heads" (in
this case widowed heads) is larger than
elsewhere, presumably because headship was
on average entered into much later in life than
elsewhere. The great predominance of sons
over daughters among the children of the
heads reflects the very young ages at marriage
in the serf population-daughters became
daughters-in-law, and hence "other relatives"
for the purposes of Table 11, early in life.
47 Czap utilized the classification scheme

of the Cambridge Group (see end of section 2)
and hence gives data not on the numbers of
married men per household, but on the number
of conjugal family units per household, that is,
a slightly higher figure.
48 Czap, utilizing the classification
scheme of the Cambridge Group, gives not the
proportion of joint households, but the propor-
tion of multiple households (i.e., the house-
holds containing more than one conjugal
family unit), which is slightly larger.
49 The Tver province estate, whose enu-
merations Czap has analyzed, contained far
fewer households and people than the Mishino
estate, to which most of his work has been de-
voted. He has also analyzed three more enu-
merations of this estate (for 1834, 1850, and
1856), in addition to the 1816 enumeration.
The 1816 enumeration covered 46 households.
Given the small numbers of households cov-
ered, it is difficult to reach firm conclusions.
50 Andorka (1975). I have also used a
mimeographed paper kindly given to me by
Dr. Andorka, "Micro-demographic researches
in Hungary (Family reconstitution and types of
household structure)." Numerous surviving
documents from eighteenth and early nine-
teenth century Hungary are relevant to the
study of household composition. Much work
on these materials has been done by Hungarian
scholars. I have not had the opportunity to
study most of this work, but I felt justified in
including the points mentioned in the text.
51 See Tables 1 and 2 and Appendix 2.
There is much additional evidence. Laslett
(1977), Table 1.6, contains a summary of his-
torical materials relating to the proportion of
servants in the populations of various commu-
nities. Sogner (1979) and Berkner (1972) also
contain additional data. The presence of farm
servants as a substantial part of the agricultural
labor force can still be traced in nineteenth
century censuses in Northwest European coun-
tries; see, for example, Knodel and Maynes
52 The institution of service in the North-
west European sense was, of course, shared by
Northwest European populations overseas.
See the chapter on "Masters and servants" in
Morgan (1966).
53 This paper is concerned only with the


Two Kinds of Preindustrial Household Formation System

meaning of the word "servant" as applied to
persons recorded as servants in documents that
list individuals by households. There has been
some controversy about the meaning of "ser-
vant" in seventeenth century English usage.
This matter is discussed in Appendix I in
Kussmaul (1981). I am grateful to Dr.
Kussmaul for sending me a typescript of her
book in advance of publication.
54 Mitterauer (1973), p. 205, notes that in
two seventeenth century Austrian villages
more than 50 percent of young people between
18 and 20 were in service. It should be remem-
bered that the proportion of young people who
experienced service at some point in their lives
must have exceeded (perhaps sometimes con-
siderably) the highest proportion of servants in
any one age group. For some who were ser-
vants at a younger age would have left service
before reaching the age at which the propor-
tion was at its maximum. Others may have en-
tered service for the first time after the age of
the maximum proportion.
55 The quotation occurs in vol. 1 of Hume
(1875), p. 387; see also Population and Devel-
opment Review 3, no. 3 (September 1977):
56 For example, in the sample of 26
eighteenth century rural Danish parishes ana-
lyzed by Johansen, the percentage distribu-
tions of male and female servants by marital
status were as follows (1787 and 1801 figures

Male Female
Single 95.2 98.3
Married 3.9 0.9
Widowed 0.9 0.8
Total 100.0 100.0

57 Concerning the ages at which young
people left home, see Wall (1978).
58 These facts are documented in
Kussmaul (1981), Chapter 5. Children of
craftsmen and tradesmen in England also be-
came farm servants. How parents sent their
own children into service and hired other ser-
vants, as well as other aspects of seventeenth
century family life, is vividly illustrated in
Macfarlane (1970).
59 Kramer (1957) says of servants in a
part of Germany: "Mostly [grosstenteils] they

were the sons and daughters of farmers, later-
born children who could not inherit the paren-
tal farm and who for that reason entered serv-
ice till the opportunity of a favorable marriage
arose" (p. 155).
60 See, for example, Sogner (1979) and
Berkner (1972), p. 409.
61 The identity of surname between mas-
ter and servant has been used as an indicator of
relationship. The proportion of cases of mas-
ter-servant surname identity ought, however,
to be compared with the proportion of identi-
cal surnames among randomly selected pairs
of individuals in the same population. Only
then can valid inferences be made about the
role that relationship played in placing ser-
vants with particular masters.
62 The hiring of day laborers is another
possibility. But the system worked in areas
where there were very few laborers, for exam-
ple, one of the Norwegian areas, Heroy, stud-
ied by Drake (1969).
63 See A. E. Smith (1947), p. 4. For the
contrasting migration patterns in different
household formation systems, see Todd
64 See Chayanov (1966). The "theoretical
scheme" of peasant household development is
discussed on pp. 57, 245, and elsewhere.
65 See Homans (1942), pp. 144-149 (re-
tirement contracts), and p. 210 (servants). For
retirement contracts in England at a later pe-
riod see Howell (1976), pp. 126-130, and
Spufford (1976), pp. 174-175. The paper by
R. M. Smith (1979) contains further material
and references on servants in medieval Eng-
land. This paper also discusses data directly
bearing on age at marriage in the Middle Ages
and treats a number of related features of me-
dieval society.
66 Early eighteenth century Iceland's low
proportion married, apparently following upon
catastrophic conditions, is reminiscent of Ire-
land in the mid-nineteenth century. In Ire-
land's case a high proportion of the unmarried
were "drained away" by overseas migration,
but this migration was a continuation of the
tradition of transatlantic migration by Irish ser-
vants since the seventeenth century. So here,
also, there was a link with the institution of
67 Iceland, Statistical Bureau (1975), p.


John Hajnal

68 The marital status data of the 1703 and
1729 Icelandic censuses are problematic since
there was no specific instruction to record mar-
ital status.
69 Wrigley and Schofield (1981), pp. 255-
260. Figures quoted for mean ages at marriage
are based on 12 reconstruction studies.
70 The reduction in the proportion of ser-
vants may have been larger than Table 18
shows. In an earlier study based on more lists
(i.e., with a less strict elimination of lists of
inferior quality), Wall (1979) found that 18
percent of the population were servants in
1650-1749, compared with 11 percent in
1750-1821; but this fall in the proportion of
servants is "to some extent simply an artifact
of the inclusion of some London parishes in
the first period, but not in the second." It was
this earlier paper by Wall that was used by R.
M. Smith in his recent (1981) paper. Smith's
paper was in part based on an earlier version of
the present paper, a version that did not yet
include the present treatment of the variation
in the number of servants over time.
71 Chapter 6 of Kussmaul (1981) dis-
cusses other evidence on variations in the inci-
dence of service in England between the
sixteenth and nineteenth centuries.
72 In the populations dealt with here, in-
stitutional households are not sufficiently im-
portant to affect in any appreciable way any of
the quantities with which we are concerned.
73 In past centuries, the term "family"
and its equivalents in other European lan-
guages commonly denoted what is now called
a "household." In referring to historical texts
below, I have felt free to use "household"
where "family" occurs in the original.
74 The phrase "a uno vino e uno pane"
became the title of a celebrated article on
households as recorded in the taxation records
of medieval Tuscany; see Klapisch and De-
monet (1972).
75 The German version of this definition,
as used in Austria in 1777, is quoted in Mit-
terauer (1973), p. 177. The Hungarian version,
as used in the censuses of the territories under
the Hungarian crown in 1784-87, is given in
Hungary (1960), p. 8. The two versions are
virtually identical. Mitterauer argues that in
Austria there was a development from a com-
monly accepted larger concept of the house-
hold in the seventeenth century (the notion of

"das ganze Haus") to the narrower concept of
the housekeeping unit in the eighteenth.
76 Mitterauer (1973), p. 302, quotes a
study of Austrian contracts of retirement. The
contracts exemplify a wide range of pos-
sibilities, from complete integration of the re-
tired with the household to which they handed
over the farm, to substantial separation of the
retired couple in a house of their own.
Johansen (1975), p. 145, for the Danish rural
parishes (with 1787 and 1801 data combined),
gives a figure of 46 retired people living as
part of a larger household, compared with 32
living independently. See also Berkner(1976),
p. 93.
77 Mackeprang (1907), p. 258. Mack-
eprang's estimate of the number of "hus-
maend" with independent households was
used in Table 2 above.
78 Johansen (1975), pp. 144-145. For a
discussion of the treatment of subsidiary
groups in households employed in the Danish
census of 1769, see Elklit (1978).
79 I have broken up the sentence in trans-
lation for clarity. The original is as follows:
"Familia quaequae distinct in libro notetur,
intervallo relicto ab unaquaque ad alteram sub-
sequentem, in quo singillatim scribantur
nomen, cognomen, aetas singulorum, qui ex
familiar sunt, vel tanquam advenae in ea vi-
vunt" (quoted in Mols [1954-56], vol. 3, p.
80 See, for example, Laslett (1977), pp.
54ff.; Utterstrom (1965), pp. 533-534; Mols
(1954-56), vol. 1, pp. 75-102.
81 I owe thanks to Dora Bjarnason and
Helle Degnbol for help with Icelandic mate-
82 The Norwegian Central Bureau of Sta-
tistics has now published a retabulation of the
entire 1801 census (the English title is Popula-
tion Census 1801-Reprocessed). This mate-
rial is much more extensive than any other
data so far available on households in prein-
dustrial Northwest Europe. Unfortunately it
came to my knowledge too late to be taken
account of in this paper.
83 The article by van der Woude in Laslett
and Wall (1972) compares the data for North
Holland with figures for some other areas in
the Netherlands. Much more information
(published in Dutch) on North Holland data is
given in van der Woude (1972).


Two Kinds of Preindustrial Household Formation System

Andorka, R. 1975. "Peasant family structure in the 18th and 19th centuries" (in Hun-
garian). Ethnographia, nos. 2-3: 340-365.
Barclay, G. W. 1954. Colonial Development and Population in Taiwan. Princeton.
-- A. J. Coale, M. A. Stoto, and T. J. Trussel. 1976. "A reassessment of the
demography of traditional China." Population Index 42, no. 4 (October):
Berkner, L. K. 1972. "The stem family and the developmental cycle of the peasant
household." American Historical Review 77, no. 2 (April).
-- 1976. "Inheritance, land tenure and peasant family structure: A German re-
gional comparison," in Family and Inheritance, ed. J. Goody, J. Thirsk, and E.
P. Thompson. Cambridge.
Buck, J. L. (ed.). 1937. Land Utilization in China. Nanking.
Burch, T. K. 1970. "Some demographic determinants of average household size: An
analytic approach." American Sociological Review 13, no. 3: 347-363.
Chayanov, A. V. 1966. The Theory of Peasant Economy, ed. D. Thorner, B. Kerblay,
and R. E. F. Smith. Homewood, Ill.
Czap, P. 1982a. "A large family: The peasant's greatest wealth," in Family Forms in
Historic Europe, ed. R. Wall, J. Robin, and P. Laslett. Cambridge.
---. 1982b. "The perennial multiple family household." Journal of Family History
no. 1 (Spring): 5-26.
Dandekar, K., and D. B. Unde. (n.d.). "Size and composition of households." Census
of India, 1961, vol. 1, Monograph No. 9. New Delhi.
Dandekar, V. M., and V. Pethe. 1960. "Size and composition of rural families."Artha
Vijinana 2, no. 3 (September).
Drake, M. 1969. Population and Society in Norway 1735-1865. Cambridge.
Elklit, J. 1978. "Household structure in Denmark 1769-ca. 1890," in Chance and
Change: Social and Economic Studies in Historical Demography in the Baltic
Area, ed. S. Akerman, H. C. Johansen, and D. Gaunt. Odense.
Flandrin, J. L. 1976. Families. Paris.
Hajnal, J. 1965. "European marriage patterns in perspective," in Population in His-
tory, ed. D. V. Glass and D. E. C. Eversley. London.
Herlihy, D., and C. Klapisch-Zuber. 1978. Les Toscans et Leur Families. Paris.
Homans, G. C. 1942. English Villagers of the Thirteenth Century. Norton.
Hopkins, K. 1980. "Brother-sister marriage in Roman Egypt." Comparative Studies in
Society and History 22, no. 3 (July): 303-354.
Howell, C. 1976. "Peasant inheritance customs in the Midlands, 1280-1700," in Fam-
ily and Inheritance, ed. J. Goody, J. Thirsk, and E. P. Thompson. Cambridge.
Hume, D. 1875. Essays. Moral, Political and Literary, ed. T. H. Green and T. H.
Grose. 2 vols. London.
Hungary, Library of Central Statistical Office and Archives Section of Education Min-
istry. 1960. Az els6 magyarorszagi nepsztmllrds (1784-1787) (The first Hun-
garian census 1784-1787). Budapest.
Iceland, Statistical Bureau. 1960. Manntalid 1703 (Population census 1703) (Statistics
of Iceland II, 21). Reykjavik.
1975. Hans Oluf Hansen, Population Census 1729 in Three Counties (Statis-
tics of Iceland II, 59). Reykjavik.
Johansen, H. C. 1975. Befolkningsudvikling og Familiestruktur i det 18 Arhundrede.


John Hajnal

Klapisch, C., and M. Demonet. 1972. "'A uno vino e uno pane.' La famille rurale
toscane au d6but du XVe si&cle." Annales E. S. C. 27, nos. 4-5 (July-October):
Klep, P. M. M. 1973. "Het huishouden in westelijk Noord-Brabant: Structuur en ont-
wikkeling, 1750-1849." A.A.G. Bijdragen, vol.18. Wageningen (Netherlands).
Knodel, J., and M. J. Maynes. 1976. "Urban and rural marriage patterns in imperial
Germany." Journal of Family History 1, no. 2 (Winter): 129-168.
Kramer, K. S. 1957. Bauern und Burger im nachmittelalterlichen Unterfranken.
Kussmaul, A. 1981. Servants in Husbandry in Early-Modern England. Cambridge.
Laslett, P. 1977. Family Life and Illicit Love in Earlier Generations. Cambridge.
and R. Wall (eds.). 1972. Household and Family in Past Time. Cambridge.
Macfarlane, A. 1970. The Family Life of Ralph Josselin. Cambridge.
Mackeprang, E. P. 1907. "Et Brudstykke af en Folketelling Fra 1645" (A fragment of a
census of 1645). Nationalokonomisk Tidskrift, Series III, vol. 15, no. 3:
Mitterauer, M. 1973. "Zur Familienstruktur in landlichen Gebieten Osterreichs im 17
Jahrhundert," in Beitrige zur Bevolkerungs- und Sozialgeschichte Osterreichs,
ed. H. Helczmanowszki. Vienna.
Mols, R. 1954-56. Introduction a la demographic historique des villes d'Europe du
XIVe siecle au XVIIIe siecle. 3 vols. Louvain.
Morgan, E. S. 1966. The Puritan Family (new edition, revised and enlarged). New
Shah, A. M. 1973. The Household Dimension of the Family in India. New Delhi.
Shanin, T. 1972. The Awkward Class. Oxford.
Smith, A. E. 1947. Colonists in Bondage. Chapel Hill.
Smith, R. M. 1979. "Some reflections on the evidence for the origins of the 'European
marriage pattern' in England," in The Sociology of the Family, ed. C. Harris.
-- 1981. "Fertility, economy, and household formation in England over three
centuries." Population and Development Review 7, no. 4 (December): 595-622.
Sogner, S. 1979. Folkevekst og Flytting. Oslo.
Spufford, M. 1976. "Peasant inheritance customs and land distribution in
Cambridgeshire from the sixteenth to the eighteenth centuries," in Family Inher-
itance, ed. J. Goody, J. Thirsk, and E. P. Thompson. Cambridge.
Stirling, P. 1965. Turkish Village. London.
Siissmilch, J. P. 1775. Die Gottliche Ordnung in den Veriinderungen des Menschlichen
Geschlechts. 3 vols., 4th edition. Berlin.
Taeuber, I. B. 1970. "The families of Chinese farmers," in Family and Kinship in
Chinese Society, ed. M. Freedman. Stanford.
Todd, E. 1975. "Mobilitd et cycle de vie en Artois et en Toscane au XVIIIe siecle."
Annales E. S. C. 30, no. 4 (July-August): 726-744.
Utterstrom, G. 1965. "Two essays on population in eighteenth century Scandinavia,"
in Population in History, ed. D. V. Glass and D. E. C. Eversley. London.
Vandenbroeke, C. 1976. "Bevolkingsstruktuur en gezinstypologie in Zuid-Oost-Vlaan-
deren op het einde van het Ancien Regiem." Handelingen van de Geschied- en
Oudheidkundige Kring van Oudenaarde 18, no. 2 (December): 243-303.
van der Woude, A. M. 1972. "Het Noorderquartier." A.A.G. Bijdragen, vol.16 (in 3
parts). Wageningen (Netherlands).


Two Kinds of Preindustrial Household Formation System

Wachter, K. W., with E. A. Hammel and P. Laslett. 1978. Statistical Studies of Histor-
ical Social Structure. New York.
Wall, R. 1978. "The age at leaving home." Journal of Family History 3, no. 2:
1979. "Regional and temporal variations in English household structure," in
Regional Demographic Development, ed. J. Hobcraft and P. Rees. London.
1982. "The household: Demographic and economic change in England," in
Family Forms in Historic Europe, ed. R. Wall, J. Robin, and P. Laslett.
Cambridge (forthcoming).
Watson, J. L. 1980. "Transactions in people: The Chinese market in slaves, servants
and heirs," in Asian and African Systems of Slavery, ed. J. L. Watson. Oxford.
Wheaton, R. 1975. "Family and kinship in Western Europe." Journal of Interdisciplin-
ary History 5, no. 4 (Spring): 601-628.
World Fertility Survey. 1979. N. Goldman, A. J. Coale, and M. Weinstein, "The
quality of data in the Nepal fertility survey," Scientific Reports, no. 6 (Decem-
-- 1980. M. Kabir, "The demographic characteristics of household popula-
tions," Comparative Studies, no. 6 (May).
Wrigley, E. A., and R. S. Schofield. 1981. The Population History of England
1541-1871. Cambridge.

Foundations of
New Sex Roles

Kingley Davis
Pietronella van den Oever

A fundamental dilemma in advanced societies today is
whether all ascription of status on the basis of sex will disappear or whether
some division of labor between men and women will be retained. Hardly any-
one believes that the traditional system will, or can, be brought back com-
pletely, but many people doubt whether all assignment of rights and
obligations on the basis of gender can be abolished. They feel either that se-
lected traditional elements must persist or that a new utopian structure of sex
roles based on modern technology must be worked out; and in either case they
rest their argument on biological differences between the sexes that cannot be
totally ignored in the division of labor. If they are right, then the debate shifts
from the question of whether men and women shall be treated exactly alike in
all assignment of tasks to the question of what would be an equitable division
of labor between the sexes.
The present paper does not address these issues in their entirety; rather, it
considers an important but somewhat neglected aspect of them: the effect on
sex roles of some important demographic changes in advanced industrial so-
cieties-changes such as increased longevity, widening sex differences in mor-
tality, aging populations, and low fertility. The argument is that such changes
give rise to new circumstances between men and women that force alterations
in sex roles. This in turn brings forth ideological developments such as the
feminist movement. It is not that the demographic trends are perceived by each
individual and thus consciously adapted to, but rather that they arise spon-
taneously in advanced societies and alter the conditions of life. Ironically, they
strengthen the public demand for equal treatment of men and women while
widening the differences between them.


Demographic Foundations of New Sex Roles

The demography of age and sex

The basic demographic changes in modem times are familiar. In addition to
those already mentioned, they include decreases in household size, unbalanced
sex ratios, shortened reproductive activity, and near universal survival to old
age. What is less familiar is the tendency of these developments to affect the
sexes differentially and thus to alter the traditional organization of sex roles. In
analyzing the effects, we can start with any demographic change we wish,
since the changes are all interrelated. Let us begin with mortality.

New patterns of mortality

Everyone knows about the dramatic fall in human mortality during the last two
centuries. Many also know about the greater fall in female than in male mor-
tality, but few realize the pace and scope of that change. In Japan, for instance,
the male-female difference in expectation of life at birth increased from six
months early in this century to over five years recently.' Elsewhere the shift has
been comparable. As a result, female longevity now substantially exceeds male
longevity virtually throughout the industrialized world. In the white population
of the United States in 1980, the difference in life expectancy was 7.6 years; in
France in 1976 it was 8.0 years; and in the USSR in 1971-72 it was 10.0. As
Table 1 shows, the female advantage characterizes developing as well as devel-
oped countries, but to a lesser degree.

TABLE 1 Difference between male and female life
expectancy in developing and developed countries in
the 1970s
Life expectancy Difference as
at birth (years)a Difference percent of male
Male Female (years) life expectancy
16 developing
countries 60.5 63.9 3.4 5.7
20 developed
countries 68.4 74.8 6.4 9.4
aUnweighted averages of individual country figures.
SOURCE: Calculated from United Nations Demographic Yearbook 1978, Table 22.

Naturally, the difference in mortality shows up not only in life expect-
ancy but also in crude death rates. However, since the female population is
older than the male population, the difference in crude rates grossly understates
the real difference in mortality. Thus in the Netherlands in 1977 the male crude
death rate was 27 percent higher than the female rate, but if the female popula-
tion had had the same age structure as the male population, the male rate would
have been 72 percent higher.


Kingsley Davis / Pietronella van den Oever 497
A demographic illusion
Vague awareness of such facts has led to a popular illusion-namely, that the
male-female difference in mortality is causing the sex ratio of whole popula-
tions to become unbalanced. Intuitively, one thinks that if more men than
women die, there will be fewer men than women; extrapolating the trend, one
easily envisages a disconcerting feminization of the population and conjures up
radical changes in sex roles. To be sure, if it came to be that there were four
women to every man, sex roles would be different from those of today.
Although basically false, this idea of progressive feminization contains a
grain of truth. Worldwide data show that there is, on average, a surplus of
females, especially in the industrial nations. For dates in the 1970s, the mean
number of males per 100 females in 21 developing countries was 99.3; in 19
more advanced countries it was 96.1
In view of the pronounced sex differences in mortality, however, the
striking thing about the overall sex ratios is how balanced they are. A sex ratio
in the 90s is, after all, fairly close to parity. Evidently, in the popular notion of
overall feminization, something is being overlooked. What is being overlooked
is, first, the fact that mortality is not the only factor that affects the sex ratio,
and second, that the influence of mortality varies strikingly according to the
age group being considered. As a result, the significant distortions in the sex
ratio occur in particular age groups, and in different directions. It follows that
to discuss the sex ratio without reference to the age structure has little meaning.
For instance, the fact that the entire population of the Netherlands in 1977 had
98.9 males per 100 females meant little to those aged 70 and over, among
whom the proportion was 67.7 per 100.
Historically, as mortality declined, the sex ratio became more feminine
at old ages, but it became more masculine at young ages. This is illustrated by
the white population of the United States. If we exclude the effects of migra-
tion and population growth, the sex ratio changed as follows:2

Males per 100 females
1910 1980
Under age 50 101.9 103.7
Age 50 and over 94.2 84.9

At ages under 50, the sex ratio became more masculine despite the fact that
female survival in those ages improved as fast as male survival. Over age 50,
with survival of women rising twice as fast as that of men, the sex ratio became
drastically more feminine.
To understand this movement of the sex ratio in opposite directions ac-
cording to age, and thus to explain why overall feminization is so slight, we
must examine the nature and interaction of the determinants of the sex ratio by

Demographic Foundations of New Sex Roles

Determinants of the sex ratio by age

In a population long subject to constant vital rates and no migration-that is, a
stable population-the sex ratio at any given age is determined by two factors:
the sex ratio at birth and the survival of males and females up to the given age.
Significantly, the second factor can be subdivided into two parts that affect the
sex ratio in contrary ways. The first part is the difference in survivorship be-
tween men and women, and the second part is the mean level of survivorship
for both. The lower the mean level, the more important a given difference will
At exact age 0, there is only one influence-the sex ratio at birth. If no
other factor entered at later ages, the sex ratio at every age would be identical
with the ratio at birth, which means it would be masculine. However, from the
moment of birth the difference in mortality between males and females exer-
cises an influence, and this influence is cumulative. Since the difference nor-
mally favors females, it eventually overcomes the masculine advantage at birth
and, at older ages, makes the sex ratio feminine.
The operation of these factors is shown in Table 2, which uses data from
Norway.3 The first column gives the actual sex ratios by age. These result from
all factors combined, including migration and the absolute level of mortality.
The second column shows the constant sex ratio that would result if the natal

TABLE 2 Determinants of the sex ratio by age,
Norway, 1979
Sex ratio in population if influenced by
Sex ratio Sex ratio Sex difference Sex ratio at
in actual at birth in mortality birth and
Age populationa onlyb only mortality
0-19 105.1 105.9 99.6 105.5
20-29 105.1 105.9 99.0 104.8
30-39 106.4 105.9 98.3 104.1
40-49 102.0 105.9 97.2 102.9
50-59 98.3 105.9 94.0 99.6
60-69 86.1 105.9 83.5 88.4
75+ 62.4 105.9 59.0 62.5
All ages 98.3 105.9 90.8 96.2
aNorway, Central Bureau of Statistics, Population by Age and Marital Status, Decem-
ber 31, 1979, Table 5, pp. 32-35.
bThis is the sex ratio at each age if the sex ratio at birth were the sole determining
'These are the sex ratios that would result from the influence of mortality alone,
obtained by dividing Lx values in the male life table by the corresponding Lx values in
the female table. Since both life tables have a radix of 100,000, the sex ratio at birth
exercises no influence. The life table for 1977-78 is used. Norway, Central Bureau of
Statistics, Vital Statistics and Migration Statistics, 1979, Table 39, pp. 58-59.
'These are the sex ratios that would result from the combined influence of the sex ratio
at birth and, at each age, the sex difference in mortality. Obtained by calculating a
sex-adjusted life table-that is, one in which the ratio between the male and female
radixes equals the sex ratio at birth. (Arithmetically this equals the product of the
corresponding figures in columns 2 and 3, divided by 100.)


Kingsley Davis / Pietronella van den Oever

ratio were the only factor involved. The third column gives the results of mor-
tality alone: since, in the absence of infanticide, from the moment of birth boys
die more frequently than girls, the sex ratio from mortality alone cannot be
masculine at any age. The last column shows the combined influence of the sex
ratio at birth and subsequent mortality. It resembles the first column precisely
because in Norway migration and population growth are minor factors, leaving
the natal sex ratio and mortality as the main determinants.
How important the sex ratio at birth is in keeping the population mas-
culine through at least the reproductive ages can be seen by contrasting the
third and fourth columns. These demonstrate that in the stationary population
the surplus of male babies causes all age groups to be masculine up to about 50
years of age, after which the cumulative effect of greater male mortality be-
comes so large that it makes all older age groups feminine.
We can now understand why, historically, the population under age 50
has become more masculine rather than more feminine. It used to be that a high
proportion of deaths occurred in the young ages. This meant that a pronounced
difference in mortality between males and females could distort the sex ratio
even at young ages. By now, however, death rates have fallen so far that nearly
everybody survives to age 50 (89.8 percent of white males and 94.7 percent of
white females, according to the 1980 US life tables). This being true, there is
no longer room at young ages for differential mortality to make an impact, and
such improvements as are made must be mainly among males. If we imagine
improvement until nobody dies under age 50, then (as comparison of the sec-
ond and fourth columns in Table 2 shows) the sex ratio of people under 50
would be entirely determined by the sex ratio at birth and therefore be even
more masculine than it is now.
It follows that the mortality gap between men and women now comes
overwhelmingly from the older ages. In fact, of the 7.6 year difference in life
expectancy between white American males and females in the United States in
1980, only 2.2 years, or 29 percent, comes from lower female mortality under
age 50, leaving 71 percent to be accounted for by differential mortality over
that age. In Norway, where improvement has gone a bit further than in the
United States, the proportions are even more dramatic-of the 6.3 year differ-
ence in life expectancy at birth, 22 percent comes from lower female mortality
under age 50, and 78 percent from differences at older ages.

Instead of using the sex ratio to describe the numerical relation between the
sexes, one can speak of "surplus" males or females-meaning the number or
proportion above a 50-50 ratio. One can easily convert Table 2 to these terms.
For instance, a sex ratio of 105.9 (in the second column of Table 2) means that
the surplus of males is 5.9 percent, and this is what it would be at all ages if the
sex ratio at birth were the sole determinant. Similarly, in the third column, a
sex ratio of 59.0 at ages 75 and older means that, from the force of mortality
alone, only 59 percent of the women would have a male in the same age group,


Demographic Foundations of New Sex Roles

leaving 41 percent of the women without such a man and therefore "surplus."
If mortality alone governed the matter, there would be an excess of females at
all ages, but the excess would be only 1 percent at ages 20-29, reaching high
percentages only at age 60 and over. It is the sex ratio at birth that makes the
surpluses masculine up to age 45 or 50, after which the cumulative force of
mortality makes the surplus feminine. This means that it is the older ages-the
ones in which mortality is high-that show the greatest imbalances of the
sexes, and these imbalances have increased as the male-female gap in mortality
has widened.
Clearly, then, the remarkable extension of life in the advanced societies
has not been an unmixed blessing. Not only has it achieved a reduction in
fertility and therefore an aging of the population, but by favoring females more
than males, it has produced a revolutionary imbalance of the sexes at older
ages. As the population has aged, this imbalance has affected an ever greater
proportion of the population. Although at younger ages a rise in the male sur-
plus has occurred, at ages beyond 55 both the proportion and the number of
women in the "surplus" category have become very large.

Marriage and the imbalance of the sexes
So far we have considered only imbalances in the sex ratio by age. Now we
must analyze them by marital status as well. One naturally tends to assume that
in any adult age group the sex ratio influences marital chances-that is, that a
"surplus" of one sex or the other reduces the probability of marriage for that
sex. This observation is roughly correct, but it understates the case. Since each
person already has a marital status-either married or unmarried-his or her
chance of getting married depends less on the sex ratio of his or her age group
as a whole than on the sex ratio of those in the group who are unmarried.

Age selection in marriage
It turns out that sex ratio imbalances among the unmarried are greater than can
be explained by the age-sex distribution alone. For instance, among people
aged 60-64 in Norway in 1979, there was a moderate surplus of women-8.2
percent. However, among the unmarried in this age group, the surplus of
women was 45.7 percent. Obviously, something else besides the sex ratio at
birth and differential mortality is at work.
This "something else" is well known. Let us call it "age selection," or
"age choice," in marriage. It is the fact that people do not necessarily take
mates in their own age group. In general, not only are grooms somewhat older
than their brides at first marriage, but after divorce or widowhood a larger
proportion of men than women remarry and, when they do, they tend to marry
women younger than themselves by a margin wider than in their first mar-
riages. With this source of distortion added to the sources already described,
imbalances in the sex ratio of the married and unmarried in some age groups
become spectacular.


Kingsley Davis / Pietronella van den Oever

Table 3 shows the data for two groups of countries, industrial and devel-
oping. Clearly, the married and unmarried populations complement each other.
In the married population the sex ratios are highly feminine at young ages and
masculine at older ages, whereas in the unmarried population the reverse is
true. Among the married, the reason for the female surplus at young ages is
that many young women are married to older men, thus excluding a substantial
portion of men their own age from the married column. At advanced ages the
sex ratio of the married population is masculine-for the same reason: many
men are married to younger women. In the unmarried population, by contrast,
males predominate at young ages. To a minor extent this is because there are
more males than females at those ages. The main cause, however, is that many
males either choose to remain unmarried during their youth or lose out in com-
petition with older men and thus are forced to do so. At older ages, on the other
hand, females predominate both because they are numerous in those ages com-
pared with men and because a sizable portion of men in their age group have
married younger women. In the industrial countries of Table 3, for example, at
ages 20-24 there are, on average, about 60 unmarried women for every 100
unmarried men, while at ages 50-59 the ratio is more than 250 for every 100.

TABLE 3 Sex ratios by age and marital status,
developed and developing countries, ca. 1970
Males per 100 females
9 industrial countries 13 developing countriesb
Age Married Unmarried Married Unmarried
15-19 17.6 114.3 17.0 119.4
20-24 53.5 168.1 46.3 194.3
25-29 86.5 177.6 74.7 215.0
30-34 96.5 133.2 92.4 134.6
35-39 100.7 103.2 102.2 89.6
40-49 100.4 55.7 115.4 53.3
50-59 107.2 39.5 142.2 37.6
60-74 133.4 32.4 194.4 34.2
75+ 200.1 30.2 259.8 41.3
"Australia, Austria, Belgium, Czechoslovakia, Finland, Hungary, Japan, Norway,
"Brazil, Fiji, Malaysia, Mexico, Morocco, Panama, Puerto Rico, Singapore, South
Korea, Syria, Thailand, Tunisia, and Turkey.
SOURCE: United Nations Demographic Yearbook 1978, Table 22. The figures are
unweighted averages, each country counting as one regardless of size.

A glance across Table 3 shows that nonindustrial countries have age- and
marital-status-specific sex ratios roughly similar to those of industrial nations.
Indeed, by comparison it appears that in the developing countries older men are
somewhat more successful in competing for younger women, and this compen-
sates for the lesser male-female differential in mortality. In any case, we are
dealing with a phenomenon so fundamental that it is independent of economic


Demographic Foundations of New Sex Roles

There is, however, one difference between the economically advanced
and the less advanced countries. Because of low birth rates, the advanced na-
tions have much older populations. This means that their excess of unmarried
women in older ages looms much larger as a proportion of their total popula-
tion. Table 4 brings out the difference. Unmarried females aged 50 and over, in
ratio to all females ages 15 and over, are two and a half times more frequent in
the industrialized than in the nonindustrialized countries. On the other hand, as
Table 4 also shows, the surplus of married men at older ages, in ratio to the
total adult male population, is less in the developed countries. In other words,
in developed countries the imposing number of elderly unmarried women is not
due to some aberrant behavior of men but rather to the much greater proportion
of the elderly in these countries. In fact, in industrial countries the propensity
of men to marry younger women is less than it is in traditional societies.

TABLE 4 Surplus married men and unmarried women
aged 50 and over as a percent of the adult population,
ca. 1970
Weighted averages Unweighted averages
9 industrial 13 developing 9 industrial 13 developing
countries countries countries countries
Surplus married
men age 50+
as percent of all
men age 15+ 4.9 5.6 4.5 5.8
Surplus unmarried
women age 50+
as percent of all
women age 15+ 17.3 6.5 14.5 6.0
NOTE: Identity of the countries in each group is the same as in Table 3.
SOURCE: See Table 3.

The contribution of age choice to sex imbalances

In thinking about sex imbalances, people tend to view the involuntary causes-
mainly the sex ratio at birth and the male-female mortality differential-in a
different light from the voluntary cause, age choice. The latter seems somehow
more socially determined, more subject to control. From a policy point of
view, then, we should measure its separate influence.
To do this, we must first get an expected, or theoretical, age-sex marital
distribution derived solely from the sex ratio at birth and mortality. This can
then be compared with a distribution depending on age selection as well. The
difference between the two models will measure the influence of age selection
We begin with the stationary population of a sex-adjusted life table, as
shown in Table 2. To get the first model, we incorporate marital status by
assuming that in each age group the married and unmarried populations have
the same sex ratio as found in the stationary population.4 To get the second
model, we give the married and unmarried populations the sex ratios that they

Kingsley Davis / Pietrorella van den Oever

have in the actual population. Since the sex ratios in the first model are derived
solely from the sex ratio at birth and mortality, while those in the second are
due to age choice as well, the difference between the two models comes from
the latter factor.
The results of the calculations are given in Tables 5 and 6, which show
that the age-choice factor exercises a great influence, especially at either end of

TABLE 5 Sex ratios by age and marital status, with
and without age selection, Norway, 1979
Males per 100 females
Married population Unmarried population
Without age With age Without age With age
Age selection' selection" selection" selectionb
15-19 105.2 9.3 105.2 108.0
20-24 104.8 46.1 104.8 141.5
25-29 104.6 82.6 104.6 165.8
30-34 104.1 97.6 104.1 151.6
35-39 103.8 101.0 103.8 130.7
40-44 103.2 98.2 103.2 123.8
45-49 102.3 99.0 102.3 117.5
50-54 100.7 100.9 100.7 97.4
55-59 98.2 103.7 98.2 74.9
60-64 94.3 110.1 94.3 54.3
65-69 88.7 122.4 88.7 43.3
70-74 80.8 133.6 80.8 35.6
75+ 62.4 171.2 62.4 33.8
aSex-adjusted life table population. From Central Bureau of Statistics (see note c,
Table 2 above), Table 39, pp. 58-59. The ratios are applied to the married and
unmarried populations alike; hence the first and third columns are identical.
hSex ratios of the actual population in 1979. From Central Bureau of Statistics (see
note a, Table 2 above), Table 5, pp. 32-35.

the age scale. As Table 5 demonstrates, if there were no age choice, the sex
ratio of the married and unmarried alike would be predominantly masculine at
young ages and feminine at older ages. This is actually the case among the
unmarried, as can be seen from the last column, but the imbalances are greatly
exaggerated. Among the married the distortion runs the other way: the sex ratio
is very feminine at young ages and masculine at older ages. The table suggests
that older males compete with young ones for young women, contributing to an
excess of unmarried men in the young ages and of unmarried women in older
ages. Between ages 35 and 55, a rough balance is struck. The competition for
younger women by men of all ages, together with the masculinity of the sex
ratio below age 45, maximizes the possibility of mating for women during their
reproductive span.
The same information can be put in terms of surpluses.5 In the married
population the surpluses in any age group run counter to the natural sex ratio.
There is a surplus of married women at young ages, where the sex ratio is
masculine, while married men are in surplus at older ages, where the natural

Demographic Foundations of New Sex Roles

sex ratio is feminine. The surpluses are almost entirely attributable to age se-
lection. With the unmarried, however, the natural sex ratio and age choice both
work in the same direction, and so the surpluses are explained by both age
choice and involuntary factors. Even so, age choice is the larger factor.
For this reason, and because the sex imbalances among the unmarried
are greater than among the married, one can understand why discussions of
marital chances focus on the unmarried. A more important reason, however,
lies in the logic of monogamous marriage. If the entire married population is
taken independently of age, its sex ratio in a monogamous society has to be
100; and, at any particular age, a surplus of married males or females means
only that at least the numbers indicated by the size of the surpluses are married
to persons outside their own age group. With the unmarried, however, sur-
pluses affect the probability of getting married. Insofar as people deem mar-
riage to be desirable, the unmarried who are in surplus are viewed as being
handicapped in reaching this state.
In view of the association of marriage with young adulthood and re-
production, it is odd that people apparently regard being unmarried at an ad-
vanced age as worse than being so at a young age. For instance, the surplus of
unmarried males in ages 20-39 is large, but it is not generally thought of as a
social problem. The reason, doubtless, is that most of the surplus men even-
tually wind up getting married; their problem is temporary. The surplus of
elderly unmarried women attracts more attention for several reasons. First, the
surplus is larger than that of unmarried young males. (In Norway in 1979 there
was a surplus of 162,000 unmarried females age 55 and older, compared with a
surplus of 88,000 unmarried males at ages 20-44.) Second, the likelihood of
these women getting married or remarried seems remote. And third, their un-
married status often imposes on them economic as well as other personal disad-
vantages. Their situation therefore seems more likely than that of young men to
generate proposals for social reform.

Social responses to sex imbalances
Upon learning of unbalanced sex ratios such as those just described, for exam-
ple, many people intuitively jump to certain policy recommendations. They
say that men should, or should be compelled to marry women older than them-
selves, or that a legal limit should be placed on the extent to which the hus-
band's age can exceed the wife's. Such proposals are worthy of study, but in
addition to the evasions and unforeseen consequences they seem likely, if en-
acted, to produce, they address themselves to only one cause of the problem-
age choice. They leave untouched the other two factors-mortality and the
natal sex ratio. Also, deliberate intervention is not the only way social
adaptation comes about. Our view, to be expanded below, is that major
changes in marital patterns and sex roles are resulting less from deliberate
policy than from unconscious adaptation to demographic changes. Finally, de-
liberate measures to control age choice in marriage would have to be quite
dramatic to succeed.


Kingsley Davis / Pietronella van den Oever

Marriage "rules" and their consequences
Consider three hypothetical age-choice rules: (I) the groom cannot exceed his
bride's age by more than five years; (II) men must marry women in their own
five-year age group; (III) men must marry women in the next older five-year
age group. How would each of these rules affect marital chances? Rule I would
have little effect, because most marriages in industrial societies already involve
an age gap of less than five years. In the United States Marriage Registration
Area in 1977, for instance, the difference in mean age of bride and groom was
only 2.8 years (median, 2.3 years).6 In Norway in 1979 the mean difference
was 2.4 years for first marriages and 2.7 years for all marriages.7 A five-year
limit would therefore do little to raise the marital chances of either young men
or older women.
Under Rule II, although the mates could be as much as five years apart,
the average difference would approach zero. As the middle columns in Table 6
show, this rule would reduce but would not eliminate the surpluses of both
unmarried young men and unmarried older women. It would do more for
young men, because their unmarried surplus is due overwhelmingly to age
choice whereas for women over 65 it is substantially due to the cumulative
effect of mortality.
Rule III would virtually eliminate all surpluses. If random choice within
the older five-year age group is assumed, the bride would on average be five
years older than the groom, although she might be as much as ten years or as
little as one day older. As the last column of Table 6 shows, this rule would
especially affect the marital chances of women over 60. Their surpluses would
all but disappear. However, it would not help males 35-44 because these were
mainly born before World War II, whereas the women five years older were
born during the war and were consequently few in number. If it were not for
this special circumstance, Rule III would produce very small unmarried sur-
pluses throughout the age scale.
Still other rules can be imagined, some of them quite complex; but in
general, given a normal age distribution, any rule of marital selection under
which men marry older women will give women a higher number of potential
If age choice were purely habitual or customary, there might be little
difficulty in moving toward Rule III or some similar measure; and if it were
purely economic, there would be a tendency for women, as they achieve finan-
cial success, to marry men younger than themselves. However, three observa-
tions suggest that age choice is not determined solely by either culture or
economics. First, the pattern of age hypergamy is too universal to be attribut-
able to arbitrary custom. Second, there is little evidence that women of high
economic status tend to marry younger men. Third, marriage normally in-
volves a biological component-sexual attraction-that is linked with age. To
the extent that marriage is artificial, as it has been in aberrant groups such as
royalty or in tribes with the levirate, or with ghost marriages or woman-woman
marriages, it need not involve sexual attraction. But since the most fundamen-
tal obligation of marriage is sexual access, sexual attraction is normally an


Demographic Foundations of New Sex Roles

TABLE 6 Unmarried surpluses in Norway, 1979, under
actual patterns of age choice and under two
hypothetical restrictive patterns
Surplus if men
women in next
Surplus if married in older age
Actual unmarried surplus" same 5-year age group" group
% of % of
% of pop. unmarried % of pop. unmarried % of pop.
of that pop. of of that pop. of of that
Age Sex sex that sex Sex sex that sex Sex sex

15-19 M 7.4 7.4 M 5.0 5.1 d
20-24 M 24.4 29.3 M 4.4 5.5 M 5.2
25-29 M 17.2 39.7 M 5.4 17.0 M 7.1
30-34 M 8.3 34.1 M 6.5 28.7 M 1.7
35-39 M 4.5 23.5 M 5.3 38.2 M 27.1
40-44 M 3.6 19.2 M 2.0 12.0 M 21.4
45-49 M 2.8 14.9 M 1.9 10.9 M 2.6
50-54 F 0.5 2.6 M 0.2 1.1 F 7.4
55-59 F 6.2 25.1 F 3.4 15.6 F 13.5
60-64 F 15.0 45.7 F 8.2 31.6 M 4.8
65-69 F 25.0 56.7 F 14.4 39.7 F 0.7
70-74 F 37.1 64.4 F 22.8 52.7 M 2.2
75-79 F 38.6 72.4 F 33.0 59.0 F 1.7
80-84 F 42.3 67.0 F 39.4 58.4 M 4.5
85+ F 44.2 60.8 F 46.2 57.3 F 3.2

"The population in Norway in December 1979. From Central Bureau of Statistics (see
note a, Table 2 above), pp. 32-35. Surplus refers to the excess in the unmarried
population of one sex or another above an equal number, and this surplus is taken first
as a percentage of the total population of that sex in the age group and second as a
percentage of the total unmarried population of that sex in the age group.
'The hypothetical surpluses are those that would arise if all people marry within their
own age group. For instance, at ages 75-79, there were 11,169 more males than
females married. If these had married in the age group, they would have reduced the
unmarried females by that amount. This gives us a reduced surplus of unmarried
females aged 75-79, and a new total unmarried female population in terms of which
the percentage unmarried can be calculated.
cThe hypothetical surpluses are those that would arise if all women married a man in
the next younger age group as long as there were a man there to marry. For instance,
at ages 30-34 there were 152,838 women in the population, and in the next younger
five-year group, 25-29, there were 155,496 males. Under the rule, all the women
could marry, but there would be 2,658 males aged 30-34 who could not. This surplus
divided by the total population of males in the 25-29 group yields a ratio of 1.7
percent. Since we assume universal marriage whenever a mate of suitable age is
available, there is no unmarried population against which to calculate the second
percentage. In other words, the surplus unmarried and the unmarried population are
dlt is assumed that women 15-19 do not marry.

element in mate selection. Thus it is likely that the pattern of age selectivity in
marriage is, at least in part, an evolutionary product. It concentrates the atten-
tion of men, regardless of age, on women in ages in which reproduction is


Kingsley Davis / Pietronella van den Oever

This latter interpretation fits our data on marital choice. For instance, the
main sources of distortion due to age selection are second marriages and male
competition for young women. At young ages, the two sexes strongly attract
each other, and this is when most marriages are consummated. In 1977 in the
United States, 97 percent of the brides and 93 percent of the grooms in first
marriages were under age 30. In such marriages the difference in age is usually
small. The large surpluses among the unmarried are produced by widowhood,
divorce, and remarriage, which allow older men to compete for younger
women. In marriages with one or both partners previously married, according
to US data, the proportion of women under 25 marrying men aged 35 and older
was 7.4 percent, whereas for first marriages for both partners the proportion
was only 0.5 percent.8
Our conclusion, then, is that age hypergamy is too deeply rooted in the
species to be extirpated by legislation. No law will prevent the sexual competi-
tion of men of all ages for younger women, nor is any law likely to cancel the
sex difference in mortality. Control of the sex ratio at birth may be achieved,
but if it is, changes in fashion may cause greater distortions than those found at
present. If preference were given to male offspring, it would make the sex
imbalance worse for young men while alleviating it for older women; if a
switch were made favoring female offspring, there would ultimately be a gen-
eration of women suffering a mate famine. It seems more probable that new
social patterns will gradually emerge-in fact, are already emerging-as unin-
tended and largely unconscious adaptations to underlying demographic
changes. What are these patterns?

New social institutions
Perhaps the main structural adaptation is that marriage-in the sense of a per-
manent sexual and reproductive union-is falling out of fashion. The manner
in which demographic forces are bringing this about is plain. In the past, when
a woman married she could normally expect most of her remaining life to be
spent in wedlock or at least with the children of the union. That, however, was
because her life was shorter and her childbearing more prolonged. Assuming
that she married at age 22, even as late at 1900 she could expect to live only
until age 64. True, she would be widowed by age 60, but the chances were that
she would die before her last child left home. Now, however, according to the
1980 life table for the United States, a white woman marrying at age 22 will
live to age 79.4. Yet she will stop having children at about 30 years of age,
much earlier than her grandmother did. Assuming that the last child will leave
home at age 18, the mother (at 48) will still have 32.5 more years to live free
from caring for children.
What is she to do during this period, which comprises more than half of
her lifetime after marriage? Some of it she can spend with her husband. He will
survive until she is 69. That is long enough, but at 69 she still has an expected 16
years of life remaining, which is 43 percent of what is now (since she has sur-
vived to age 69) her postchild lifetime. And the chances are that the marriage will


Demographic Foundations of New Sex Roles

have broken anyway, long before she reaches age 69. The 1975 risk tables for
divorce in the United States indicate that 47 percent of first marriages ultimately
end in divorce, and that a fourth of them survive only seven years or less.9 On
average the wife will have become a widow or divorce by the time she is 44
years old. If she does not remarry, she will retain that status for 36 years. Her
motivation to remarry may therefore be great, and the chances of it are fair
because, in the middle 40s, the sex ratio of the unmarried is still reasonably
favorable. If she does remarry at, say, age 45, she will still have another 35 years
to live. A part of these would be spent with her second husband, but his age
would be older than hers by a wider margin than in the first marriage, and the
marriage would be more susceptible to divorce or widowhood. She would stand
an even chance of being divorced or widowed a second time within 11 years,
when she would still have more than 26 years of life remaining.
Underlying demographic changes thus reduce the share of marriage and
children in women's lives, and by doing so, lead women to attach less impor-
tance to their domestic role. Women's prospects are that around two-thirds of
their adult years will be spent without children in the household, and possibly
half to two-thirds without a husband. For long periods they will probably be
thrown on their own resources and will need employment. This means that, in
planning their lives, they must look to their own careers as separate individu-
als. For best results, they must choose an occupation early in order to get the
necessary training, and they must enter employment while young and remain
employed consistently in order to build up experience, seniority, reputation,
and whatever other cumulative benefit comes from occupational commitment.
Once under way, the system of change exhibits a dynamic of its own.
Insofar as demographic trends lead women to downgrade marriage and stress
employment, they also lead them to reduce not only their dependence on their
husbands but also their service to them. Men, in turn, are induced to reconsider
the costs and benefits of marriage. They sense that, at older ages, men are
increasingly scarce compared with women, that they do not have to marry to
enjoy female company, and that if they do marry, their role as father and
family head has somehow been eroded. Not surprisingly, the divorce rate rises
to unprecedented levels, making marriage less secure and therefore less valu-
able for both sexes. Marriage undergoes attrition in two ways: it is postponed
or not undertaken at all, and when it is undertaken, it is increasingly brittle.
In such terms one can understand why, in recent years, the proportions
unmarried have risen steeply, especially at ages under 45 (Table 7). Doubtless,
as household surveys indicate, many couples now live together without marry-
ing, but this fact underscores the decline of marriage. Consensual unions are
less stable and less conducive to childbearing than regular unions, a circum-
stance that means low fertility and hence still more freedom for women to
pursue outside interests and contacts.

The feminist movement
Seen in this light, the rise of the women's movement in recent decades in
industrial societies becomes comprehensible. It seems to be an inevitable re-

Kingsley Davis / Pietronella van den Oever

TABLE 7 Proportion of women not
married, by age, United States and Norway,
recent dates
United States
white women Norwegian women
Age 1967 1980 1972 1980
15-19 89.1 91.2 94.5 97.7
20-24 31.2 54.0 48.4 63.8
25-29 9.4 30.2 19.7 29.8
30-34 6.9 23.1 13.5 18.2
35-44 7.9 18.6 12.1 16.2
45-54 13.8 18.9 17.8 17.8
55-64 29.2 30.0 30.8 28.4
SOURCES: United States: Bureau of the Census, "Marital status
and family status: March 1967," Current Population Reports, Series
P-20, no. 170 (23 February 1968): 12; "Marital status and living
arrangements: March 1980," Current Population Reports, Series
P-20, no. 365 (October 1981): 8. (In 1967 the youngest age group in
the US panel was 14-19 instead of 15-19.) Norway: Central Bureau
of Statistics, Social Survey 1974, p. 8; and Population by Age and
Marital Status, December 31, 1980, pp. 36-37.

sponse to the basic demographic changes we have been describing. Like other
movements, it is an ideological reaction to alterations in the underlying condi-
tions of life. Such alterations make old patterns and ideals unworkable and thus
push individuals into new modes of behavior that, because they are new, have
no ethical justification or support in the received social order. Yet people need
to feel "right" about what they do. When they are forced into new paths, they
welcome a movement that justifies their new behavior and thus morally reas-
sures them. The feminist movement normally supports new female behavior
patterns by showing that these patterns are widespread and hence not aberrant,
that they are justified by the highest moral principles, and that they are opposed
only by people who are misguided. Thus, under the banners of very broad
principles such as "equality," "privacy," and women's "rights," the feminist
movement legitimates a variety of innovations that are in women's interest
under the changed circumstances of contemporary society. These include
bread-and-butter programs like day-care centers and paid maternity leave, as
well as radical measures such as the approval of free abortion, sexual freedom,
childlessness, and affirmative action.
If we try to distill the most extreme, or radical, goal of the feminist
movement, it appears to be the rejection of all division of labor based on sex.
Underlying the specific programs, then, is the general idea that women should
have the same opportunities as men, the same occupations, the same priv-
ileges. Any difference in rights and obligations among people should, accord-
ing to this doctrine, be based on merit, not on sex. Indeed, gender should be
expunged from the language.
In the light of the present analysis, this goal is understandable, because it
extrapolates to its end the actual diminution of marriage and childbearing in
women's lives that we have documented. However, for that very reason it is a

Demographic Foundations of New Sex Roles

questionable goal, for it encounters the stubborn obstacle that it is still women
who become pregnant and bear children. Until advanced technology frees them
from this specialization, a modicum of sex-based division of labor is unavoid-
able if there is to be a birth rate.
Recognizing this difficulty, the feminist movement generally pursues a
less extreme goal. It acknowledges women's special reproductive role but
seeks to maximize the rewards and minimize the costs of that role. In practice it
has done less to maximize the rewards, evidently because it wishes to de-
emphasize women's reproductive specialization and because in the past women
were rewarded for reproduction but in ways that reinforced their ascribed status
and their dependence on the family. Instead, feminist action has concentrated
on lessening the costs of reproduction for women, especially the career costs.
Thus women are not to be penalized in the workplace for pregnancy or parent-
hood, and men, through domestic and child-care duties, are to make an invest-
ment in reproduction that equals that made by women.
While this strategy is in keeping with underlying trends, it nevertheless
encounters an ancient problem. Whenever there is a division of labor, a
troublesome issue arises-the issue of fairness. Is what each side gives and
gets a square deal? In the occupational sphere the issue is resolved by the forces
of the economic and political marketplace, but between men and women re-
ciprocal obligations and rewards have traditionally not been determined that
way. Instead, they have been governed by an institutional structure the very
genius of which was to create solidarity and identity in the family and thus to
protect its members from raw economic and political competition among them-
selves. Now, however, the mutual rights and obligations of men and women
are the subject of dispute, and the question of what is fair or unfair seems
virtually insoluble. As demographic and social change has upset the conditions
supporting the old system, women have turned more and more to earning their
rewards in the marketplace, not in return for reproduction but in return for
activities like those of men. The result, in most industrial societies, is a large
increase in female labor force participation and a drastic decline in fertility.
How to reward men and women for reproduction, at least sufficiently to pro-
vide replacement fertility, is a moot question.
It would be odd if the feminist movement did not overstate its case,
because that is what ideological movements generally do. We must ask, there-
fore, whether from a demographic point of view the idea that except for re-
production men and women are to be treated exactly alike goes too far,
possibly against the self-interest of women-like saying that a fox and a rabbit
are just alike except that one is carnivorous. Fairness is not necessarily
achieved by treating people who are different as if they were just alike. One
complication is that it is hard to set a close boundary on women's reproductive
role. After a child is born, it needs a great amount of attention, including
(according to current medical opinion) breastfeeding. Some economic activi-
ties are more easily combined with this attention than others. Thus reproduc-
tive specialization tends to ramify. Another complication is that there may be


Kingsley Davis / Pietronella van den Oever

biological differences between men and women other than sheer pregnancy and
childbirth. We cannot argue that question here, but we can say that the very
demographic forces that underlie the feminist movement and give women more
opportunity to be like men have, in other ways, widened the differences. The
two sexes have become increasingly distinct in such important matters as lon-
gevity, age structure, and marital chances. To ignore these differences is in
some ways unfair to men and in some ways unfair to women.
A viable division of labor cannot simply be fair; it must also be efficient.
Insofar as women in industrial societies today are not motivated to achieve
replacement fertility but instead are rewarded for nonfamilial activities, they
are curtailing the one activity that most distinguishes them from males and are
acting out the ideology of feminism. In the long run, however, such an ar-
rangement is self-defeating. If it persists, the social order that gave rise to it
will be replaced by another-either one that supports traditional sex roles or
some new order that rewards women adequately for reproduction. The feminist
movement may undertake to construct such a new order. If it does, or if such a
new order evolves by trial and error, it will have to be in keeping with the new
demography found in advanced societies.

1 Differences in life expectancy in Japan
were as follows, at the start of the century and
in modem times:

Male Female
e*O e0, Difference
1902-13 44.2 44.7 0.48
1976 72.2 77.4 5.20
SOURCE: Institute of Population Problems, The
22nd Abridged Life Tables, Research Series No.
194, 15 July 1970, Appendix Tables; United Nations
Demographic Yearbook 1978, p. 408.

2 These are the ratios in a sex-adjusted
life table population.
3 Norway has had minimal influence from
migration in recent decades, and its data on
age, sex, and marital status are excellent.
4 It must be remembered that the question
at issue is how age choice affects the sex ratio
of the married and unmarried population in
each age group. In the absence of such choice,

the sex ratio in these marital categories would
be the same as in the life table population.
5 "Surplus" here is defined as the per-
centage by which a given sex exceeds the
other sex in a given age and marital status
6 National Center for Health Statistics,
Vital Statistics of the United States, 1977, vol.
III, "Marriage and divorce," pp. 1-10, 1-21.
7 Norway, Central Bureau of Statistics,
Population by Age and Marital Status, Decem-
ber 31, 1979, pp. 32-35. See also The Devel-
opment in Marriage and Deaths 1911-1976
(Oslo, Norway: 1978), p. 36.
8 Calculated from National Center for
Health Statistics, Vital Statistics of the United
States 1977, vol. III, "Marriage and divorce,"
p. 21.
9 James A. Weed, National Estimates of
Marriage Dissolution and Survivorship:
United States (Washington, D.C.: National
Center for Health Statistics, November 1980),
pp. 5, 16, 19.

Fertility Decline in
Cuba: A Socioeconomic

Sergio Diaz-Briquets

Lisandro Perez

The recent decline in Cuba's fertility is one of the most
rapid declines on record. From a level of 35 births per thousand population in
1963, the crude birth rate dropped by 60 percent to a low of 14 per thousand in
1980 (Table 1). Measuring the decline over this particular time span, however,
is somewhat misleading. The year 1963 was the peak year of the Cuban baby
boom that began four years earlier and that lifted birth rates throughout the
1960s above the level that prevailed prior to 1959.' Measured from pre-baby-
boom levels, the decline in the birth rate appears less spectacular, but still

TABLE 1 Cuba: crude birth rates, 1953-80
(births per 1,000 population)
Crude Crude
Year birth rate Year birth rate
1953 28.3 1967 31.7
1954 27.6 1968 30.4
1955 27.1 1969 29.2
1956 26.2 1970 27.7
1957 25.7 1971 29.5
1958 26.1 1972 28.0
1959 27.7 1973 25.0
1960 30.1 1974 21.9
1961 32.5 1975 20.7
1962 34.3 1976 19.8
1963 35.1 1977 17.6
1964 35.0 1978 15.3
1965 34.3 1979 14.7
1966 33.1 1980 14.1
SOURCES: 1953-79: Direccidn de Demograffa, Comit6 Estatal de
Estadisticas (Repdblica de Cuba), Anuario Demogr4fico de Cuba,
1979, Havana, 1981, Table 19, p. 67; 1980: Ministerio de Salud
Piblica (Repdiblica de Cuba), Informe Anual, 1980, Havana, 1981,
Table 1, p. 9.


Fertility Decline in Cuba: A Socioeconomic Interpretation

exceptionally rapid. Over the two decades extending from 1958 to 1980 the
birth rate declined by 46 percent, from 26 to 14 births per thousand population.
Age-specific fertility rates in Table 2 confirm that substantial fertility
decline occurred in all age groups, albeit not uniformly. The total fertility
rate-a measure indicating the numbers of children the average woman would
bear at the prevailing level of age-specific fertility rates-declined from 3.7 in
1970 to 1.92 in 1978 and to 1.81 in 1979.

TABLE 2 Cuba: age-specific fertility rates and total
fertility rates, 1970-79 (births per 1,000 women, by
Year 15-19 20-24 25-29 30-34 35-39 40-44 45-59 rate
1970 127.5 227.4 164.6 116.3 73.5 25.8 3.9 3.70
1971 146.5 245.7 174.7 120.3 71.1 25.6 2.8 3.93
1972 143.8 237.6 167.9 113.0 66.1 23.7 2.7 3.77
1973 134.7 208.3 146.1 97.6 56.2 21.2 3.1 3.34
1974 129.7 182.3 125.4 80.2 46.2 16.3 2.5 2.91
1975 127.1 178.3 116.1 67.3 36.6 13.4 2.2 2.70
1976 126.5 170.7 111.2 59.9 32.3 11.1 1.9 2.57
1977 103.3 155.4 96.9 53.3 28.0 9.7 1.7 2.24
1978 83.0 137.2 83.9 45.6 23.9 8.2 1.5 1.92
1979 74.5 131.8 80.2 43.7 22.4 7.6 1.4 1.81
SOURCE: Comite Estatal de Estadistica, in Octavio Avalos Triana Monet, "Evoluci6n
de la fecundidad en el iltimo decenio," Revista Cubana de Administraci6n de Salud 7,
no. 4 (October-December 1981): Tables II and IV, pp. 413 and 415.

Thus, since 1978, fertility in Cuba has been below the level necessary in
the long run to assure the replacement of the population. No other country in
the developing world except Singapore-a country that by most standards
should be classified as developed-has attained such a distinction.2 The esti-
mated total fertility rate for 1979 is the lowest recorded for any developing
country. It is a rate comparable to or lower than that characterizing fertility in
most developed nations.
A number of features of the recent Cuban fertility experience are note-
worthy. First, the rapid fertility decline got under way from relatively low
levels, since the birth rate increase of the 1960s represented only a temporary
disturbance of a long-established downward secular trend.3 Second, the baby
boom itself, if not a unique occurrence in the developing world, was at least a
rare one. Third, contrary to the situation in China, Costa Rica, Taiwan, Sin-
gapore, and other countries undergoing rapid and significant fertility decline,
the fertility decline in Cuba took place in the absence of explicit policies to
produce that effect.4 Fourth, the fertility change occurred in a national context
of considerable social, economic, and political change. Finally, as shown in
Tables 1 and 2, the decline of fertility since 1972 has been extremely rapid.
Our objective in this paper is to offer a socioeconomic interpretation of
that fertility decline, taking into account the peculiar nature of Cuban history
since 1959, the date that marks the beginning of a revolutionary period that

Sergio Diaz-Briquets / Lisandro Perez

entailed a complete revamping of the nation's social and economic structures.
Our explanation focuses largely on the analysis of the profound and dynamic
process of change experienced by Cuban society during this period; only pass-
ing reference is made to the more immediate determinants (notably nuptiality,
contraception, and abortion) through which those processes work their effect
on fertility. These "proximate determinants" have been widely discussed in the

Pre-1959 demographic trends
To understand social and demographic developments in Cuba during the last
two decades, these developments should be seen in a broader historical per-
spective. By any conventional measure, Cuba was well along in its demo-
graphic transition by 1959. Fertility levels in the late 1950s were closer to those
characterizing developed countries than to those generally found in the devel-
oping world.6 Today, in fact, most developing countries have yet to attain the
relatively low fertility levels prevailing in Cuba over 20 years ago.7 Much the
same could be said about mortality. The most reliable mortality estimate avail-
able for prerevolutionary Cuba indicates that in 1953-the last census year
before the revolution-expectation of life at birth for both sexes combined was
58.8 years; in 1960, immediately after the revolutionary takeover, it was esti-
mated to have reached 64 years.8 Even today, comparable life expectancy lev-
els have been reached in relatively few developing countries, mostly in Latin
America and East and Southeast Asia.
Why had prerevolutionary Cuba attained such an advanced demographic
regime? This question, while critical to the understanding of developments
over the last two decades-demographic or otherwise-has rarely been raised,
let alone answered. It is clear that some factors often cited in explaining con-
temporary fertility declines in other developing countries are not part of the
answer. The moderately low fertility levels prevailing in Cuba by the late 1950s
were reached before modern contraceptive methods, such as the pill and the
IUD, were available. Also, Cuban fertility had declined gradually over a pe-
riod of several decades in the absence of any organized programs to provide
family planning information and contraceptive supplies.
The most obvious explanation for the fact that many Cuban women or
couples had acquired the motivation and ability to limit their fertility is that
Cuba, despite frequent allegations to the contrary, had attained a substantial
level of modernization, or that, at the least, a very considerable proportion of
its population had done so. As we have stated elsewhere,

by comparison to other developing countries, Cuba had relatively high levels of
income and consumption, fairly advanced medical and sanitary standards, a
comparatively well developed system of education and other social facilities, a
high proportion of persons of European origin, a fairly irreligious and urbanized
population .. and a significant middle class. The rather large middle class had,
or aspired to, high levels of material consumption, thanks to the country's rela-
tive development and the profound influence that US lifestyles exerted over

Fertility Decline in Cuba: A Socioeconomic Interpretation

Cuban society, and equally important, over the aspirational levels of the lower

This assessment, while accepted by many students of Cuba, differs from the
often-painted picture of prerevolutionary Cuba as a country mired in economic
backwardness where a vast majority of the population existed at the very mar-
gin of survival.
How preferences for low fertility were translated into actual behavior is
not entirely clear and certainly cannot be investigated statistically. Abortion
and, perhaps to a lesser degree, contraception, were probably the most impor-
tant mechanisms involved. Although in theory illegal, abortions were widely
and openly available."' Contraceptives such as the diaphragm and the condom
were used, and many women were sterilized for contraceptive purposes, usu-
ally following childbirth. Because of statistical shortcomings, compounded by
the high prevalence of consensual unions, the role, if any, of changes in age at
marriage and in the proportion of women in reproductive unions in lowering
fertility cannot be ascertained.
One notable characteristic of fertility levels just before the revolution
was that fertility was low not only in the province of Havana, comprising the
country's capital city, but also in the relatively modernized central provinces of
Matanzas and Las Villas, and to a lesser extent in Camagiiey. Only in the more
backward and rural provinces of Pinar del Rio and Oriente, accounting for
approximately 38 percent of the population in 1953, were birth rates still rela-
tively high in 1958." From the provincial birth rate differentials it is obvious
that low to moderately low fertility was not exclusively an urban phenomenon,
although it is safe to assume that within each province fertility was lower in
urban than in rural areas. Indeed, in provinces with high birth rates, low fertil-
ity was in all likelihood limited to cities.
Persons who were medical doctors in prerevolutionary Cuba report that
in at least some cities of these provinces abortion was commonly performed in
health establishments and by private practitioners.12 That some sort of family
limitation was practiced in the high birth rate provinces is also clear from the
levels of the rates: although high by Cuban standards of the time, they were
well below what would be expected in the absence of deliberate birth control
practice by couples.
The historical background sketched in the preceding paragraphs contra-
dicts those who (even when they take note of the secular trend established
earlier) tend to view recent fertility changes in Cuba as resulting exclusively
from the particular political and socioeconomic transformation that took place
in that country following 1959. Indeed, it can be asserted with confidence that
even in the absence of the revolution and the structural changes it produced,
fertility would have continued to decline. Fertility, after all, continued to de-
cline or began its secular trend of decline in other Latin American countries,
such as Chile and Costa Rica, that in the late 1950s resembled Cuba in terms of
a variety of social, economic, and demographic indicators. At that time, fertil-
ity in Chile, although still higher than that in Cuba, had been declining for

Sergio Diaz-Briquets / Lisandro Perez

some years. Fertility in Costa Rica had yet to begin its transition. Significant
fertility declines have since taken place in both countries.13
Fertility declines in Latin America and in other regions of the developing
world have been largely the product of a variety of social, economic, techno-
logical, and programmatic developments that have affected, if unequally, all
parts of the globe. They include dramatic improvements in communication that
have facilitated the flow of ideas and information from country to country, the
increasing legitimacy of fertility regulation as an individual right, significant
increases in levels of educational attainment, the spread of urbanization and
industrialization, the development of highly effective, convenient, and inex-
pensive contraceptive methods, and the emergence of national and interna-
tional population programs. Cuba, even in the absence of the revolution, would
have been unlikely to remain untouched by these developments. Indeed,
viewed from both historical and cross-national perspectives, it is the high fertil-
ity levels of the 1960s-the Cuban baby boom-that constitute the anomaly.
Here the revolution comes into play, for it created social and economic
conditions that not only raised the birth rate in the 1960s but appear to have
contributed to its precipitous decline in the 1970s, especially after 1972. In the
remainder of this paper we attempt to explain how postrevolutionary social and
economic conditions affected Cuban fertility. Our discussion focuses primarily
on the recent fertility decline, but that phenomenon cannot be understood sepa-
rately from the baby boom of the 1960s, since the decline would not have been
as precipitous without the baby boom. As we shall argue, an understanding of
the determinants of the boom helps to explain the extent and speed of the
decline. We start, therefore, with an examination of the Cuban baby boom.

The baby boom: fertility during the 1960s
Broad descriptions of some of the factors determining the sharp, if temporary,
fertility increase in Cuba during the 1960s have been provided by various in-
vestigators.14 These descriptions suggest that during the early years of the revo-
lution the enactment of a series of social and economic measures led to a
radical shift in income distribution. A greater proportion of the national income
accrued to the lower income groups, who as a consequence saw their living
standards rise and their prospects for the future brighten. Under these condi-
tions marriage rates rose, births formerly postponed may have been made up,
and it is likely that at least in some segments of the population, desired family
size rose as couples came to feel that they could afford to have more children.15
Also contributing to the fertility increase were the rigorous enforcement of
previously ignored statutes against abortion and a shortage of contraceptive
supplies caused by the US economic blockade.16
According to the common interpretation of these changes, fertility rose
in Cuba largely because of a general feeling of "euphoria" that engulfed the
country after the revolution," and this effect was reinforced by the decreased
availability of abortion and contraception. Most analysts have focused on the
objective changes that induced those euphoric feelings, putting little emphasis


Fertility Decline in Cuba: A Socioeconomic Interpretation

on the subjective perceptions that fomented such optimism. It is our contention
that the interplay of objective and subjective elements goes a long way toward
explaining the time pattern of fertility change in postrevolutionary Cuba over
the last two decades. We seek to take into account both of these elements in
discussing what we see as the seven most significant factors responsible for the
Cuban baby boom. After describing these factors, we offer an integrated expla-
nation of the increase in fertility during the early years of the revolution.

Income redistribution measures

Income redistribution was effected through some of the best known and most
consequential economic measures implemented by the new Cuban regime.
Most important were the agrarian reforms that transformed large privately
owned landholdings into cooperatives and state farms, or distributed them to
peasants as smaller but viable landholdings; urban reforms that drastically cut
rents and promised eventual home ownership; reductions in the rates charged
for utilities, notably for electricity, water, and telephone service; measures that
caused large-scale emigration (to be treated separately under that heading); and
other measures leading to increased employment.
The redistribution of income resulting from these measures was substan-
tial, involving transfers largely from high- to low-income groups, but also
benefiting a broad segment of middle-income groups. A recent attempt to
measure these distributional changes (Table 3) indicates that the portion of total
income accruing to the poorest 40 percent of the population rose from 6.2
percent in 1953 (the year of the last prerevolutionary census) to about 20 per-
cent in the early 1960s. Most of this change can be presumed to have occurred
since 1959.18

TABLE 3 Cuba: estimated distribution of income, 1953,
1960, 1962, and 1973 (percent of total income ranked
by per capital income)
quintiles" 1953 1960 1962 1973
0-20 2.1 8.0 6.2 7.8
20-40 4.1 12.5 11.0 12.5
40-60 11.0 14.5 16.3 19.2
60-80 22.8 17.0 25.1 25.5
80-100 60.0 48.0 41.4 35.0
Total 100.0 100.0 100.0 100.0
alncome quintiles are ranked from lowest to highest.
SOURCE: Brundenius, cited in note 18, Table 12, p. 43. Original figures for 1962 are
revised in Brundenius, "Growth with equity: The Cuban experience (1959-1980),"
World Development 9, no. 11/12 (1981): Table 6.

Three important points should be noted concerning Table 3. First, the
figures on income distribution tend to understate the true magnitude of re-
distribution because these data do not include the provision of health, educa-


Sergio Diaz-Briquets / Lisandro Perez

tional, and other services that after 1959 were greatly expanded. Since access
to these services is not dependent on the level of per capital (or family) income,
the formerly less privileged social groups have benefited disproportionately
from this change. Second, the income distribution figures do not convey the
change in the volume of income that has been redistributed. There is reason to
believe that in the first few years following the revolution gross domestic prod-
uct grew appreciably, in contrast to the mid-1960s when it may have declined
in absolute size. (The performance of the Cuban economy over the past two
decades will be discussed later in this paper.) Third, the share of total income
accruing to the lowest income groups, after increasing during the first few
years of the revolution, did not grow further in later years. From 1962 to 1973
the pattern of income distribution remained stable. What is critical to note at
this point is that there was indeed a massive shift of income away from the
higher and toward the lower and middle income groups during the first few
years following the revolution.

Social reform
Social reform-in health, education, and other welfare areas-got under way
within the first months following the revolutionary takeover and gradually
gained momentum. The principal thrust was aimed at eliminating past social
and regional differentials in access to basic social services, in particular be-
tween the cities, especially Havana, and the rural regions of the country. Some
of the most notable achievements of the Cuban revolution have been realized in
the areas of health and education.19 The initiation of reforms and the commit-
ment on the part of the revolutionary leadership during the early 1960s to up-
grade basic social services and to place them within the reach of all the
population must have contributed significantly to the feelings of optimism of
the Cuban population at the time.

Early economic performance
Partly induced by the redistributive policies but also facilitated by many other
factors, during the immediate postrevolutionary period and especially in the
first 18 months, Cuban economic performance was remarkably strong.20 Ritter
has provided an excellent review of the main reasons for this strong showing.
Aggregate demand rose in response to the income redistribution measures, and
this led to the fuller use of formerly idle resources in land, labor, and industrial
plant. Other contributing factors were an expansion in government spending,
the coming on line of new productive facilities built in the years just before the
revolution, and the early favorable performance of the agricultural sector. Cer-
tain disinvestments-necessarily temporary-that took place as domestic and
imported inventories were rapidly depleted contributed to economic buoy-
ancy.21 The combined effect of these factors contributed to the early optimism
shared by the people and the revolutionary leadership concerning the state of
the economy and its prospects.


Fertility Decline in Cuba: A Socioeconomic Interpretation

Job creation
The income redistribution policies, the favorable performance of the economy,
and the expansion of social services gave rise to an increase in the demand for
labor. As a result of the land reforms, the agricultural sector absorbed more
workers, particularly those who at times of slack agricultural activities had
constituted a pool of surplus labor. In addition, the revolutionary government,
in its attempts to improve the living conditions of the poorest sectors of society
and in fulfillment of social commitments, initiated a program of public works
that had as its principal goal the reduction of unemployment. Job security and
year-round employment were essential components of these programs, espe-
cially since the Cuban economy had traditionally been subject to marked sea-
sonal fluctuations in labor demand, not only in agriculture but also in certain
segments of the urban labor market. Two other important factors leading to
employment growth were the increasing militarization of the country and the
growth of the state bureaucracy. Within a brief period the armed forces, includ-
ing the national militia, grew rapidly. Meanwhile, the size of the state bureauc-
racy increased by leaps and bounds, as the private sector began to be
nationalized and the former market economy was gradually transformed into a
centrally planned economy. In brief, within a very short period of time there
was a remarkable increase in labor utilization, although, as time would prove,
at substantial long-term economic costs.

The income redistribution policy and the increasingly nonpluralistic, radical,
and dogmatic course of the revolution precipitated the emigration of over
200,000 Cubans between 1959 and 1963.22 The detrimental consequences of
this emigration, which deprived the country of many of its former intellectual
and technical elite, have been repeatedly noted; so have some of its favorable
effects for the course of the revolution, notably the fact that the departure of
ideologically hostile elements facilitated the establishment of the new social
system. Largely ignored have been other consequences of this emigration.
Over the years it has helped ease pressures in the labor market. Although pro-
ducing shortages of skilled labor in the short run, emigration created employ-
ment opportunities, which in turn may have contributed to the optimism
concerning economic prospects and, indirectly, to the rise in fertility.

Social mobility
The rapid process of social change sweeping Cuba during the early 1960s was a
classic example of Pareto's "circulation of the elites." An entrenched social
class was dislodged from power and replaced within a matter of months by a
new group of people, of predominantly middle-class origin. The changes at the
top of the social hierarchy were sweeping since they touched practically every
national institution whether political or economic. The old armed forces ceased
to be, as new revolutionary armed forces were created. Political power passed
not just into new hands, but also to a much younger generation. The national-

Sergio Diaz-Briquets / Lisandro Perez

ization of the private business sector, both national and foreign, swept an entire
class out of power. Massive emigration occurred. The end result of these de-
velopments was a broad-based process of social mobility as the slots vacated
by the former elite were filled by other members of society who previously
ranked lower in the social, economic, and political hierarchy. The expansion of
the bureaucracy and the military gave added momentum to this process.
The repercussions of the change at the top filtered down to all levels of
society: those experiencing upward mobility were vacating slots that others
could now occupy. Equally important, the ongoing process of social mobility
had a powerful demonstration effect. It suggested even to those not directly
benefiting from it that social mobility was possible-indeed had been experi-
enced by many of their relatives, friends, and acquaintances who now occupied
posts, had responsibilities, and enjoyed privileges (e.g., access to housing va-
cated by emigrants) undreamed of a few years earlier.

Perceptions of long-term economic prospects
An almost unbelievable naivete in judging the true difficulty of sustained eco-
nomic development-characteristic of the more optimistic world of 20 years
ago, but carried to excess by Cuba's revolutionary leadership-imbued the
Cuban population with wholly unrealistic expectations as to the country's
longer term economic prospects. These perceptions, fueled by the early eco-
nomic successes and articulated in many official pronouncements of the new
leaders, reflected certain misconceptions that were an integral part of the young
revolution's popular political lore. In brief, it was believed that Cuba was a
country abundantly endowed with natural resources, a country that prior to the
revolution was held back in its development by corrupt government officials,
by an economic system that kept the country's rich patrimony in the hands of
the privileged few, and by the exploitation of national resources by foreign
powers, mainly the United States. Once these conditions were changed, it was
assumed Cuba would bloom economically.23
Time and economic realities proved the shallowness of these beliefs. But
the effects of the many promises of material prosperity on the unsophisticated
populace were real. The objective improvements in living conditions that oc-
curred in the early stages of revolutionary change lent credibility to these prom-
ises. The future painted by the leadership and echoed by the media during the
first years of the revolution seemed bountiful indeed.
The flavor of what was promised can best be conveyed by citing some
examples of public statements made by responsible and influential government
officials at that time. Ernesto Guevara, one of the most charismatic leaders of
the revolution in its earliest years and one of the architects of Cuba's initial
strategy of economic development, predicted as chairman of the Cuban delega-
tion at the Fifth Plenary Session of the Organization of American States in
Punta del Este, Uruguay, on 8 August 1961, that by 1980 Cuba would have
achieved "a per capital net income of $3,000 . a diversification of the econ-
omy and elimination of 40 percent of the present housing deficit." He also
predicted that "over these two decades the general growth rate [of the econ-


Fertility Decline in Cuba: A Socioeconomic Interpretation

omy] will be 12 percent, [and] in the industrial field Cuba will be transformed
into the most highly industrialized country of Latin America in relation to its
population. "24
That same month, during the First National Production meeting in
Havana, Regino Boti, then Economic Minister, was promising that by 1965 the
Cuban per capital consumption of clothing and shoes would be higher than in
Sweden, Norway, and Belgium, and that Cuba would be the leader among
Latin American countries in the per capital consumption of electric energy,
steel, cement, tractors and refined petroleum. By about 1970, he continued,
Cuba "would have the highest level of living in Latin America by a wide
margin, as high as that of most countries of Western Europe." Guevara's prom-
ises in Uruguay were also repeated by Boti in Havana.25
Fidel Castro himself, although generally more cautious, made state-
ments expressing similar hopes for the future. For instance, on various occa-
sions in 1960, he promised that the new neighborhoods being planned for
workers would be as fine, and the dwellings as good, as those in Miramar, one
of the more expensive upper middle class areas in Havana.26 The workers were
assured that they would have access to appliances and other goods commensu-
rate with such residences.
These promises, by all evidence actually believed by the leadership mak-
ing them, were widely discussed in the mass media and accepted as a matter of
faith by most Cubans. They reinforced the perceptions of economic improve-
ments, portending a prosperous economic future.

Causes of the Cuban baby boom
The developments discussed in the preceding sections provide a framework
that, we contend, helps explain the Cuban baby boom. By the time of the
revolution, Cuba had already attained fertility levels comparable to those ob-
served in the then developed nations. Therefore, there was considerable scope
for a rise in fertility. Our review suggests that the birth rate increase was the
product of conditions, real and perceived, that led young adults to foresee a
much more promising future, free from many of the uncertainties and con-
straints of a backward and dependent economy. In the new social and eco-
nomic order many of these young adults, particularly in urban areas, were
confident that they would be able to satisfy their material and social aspira-
tions. These aspirations were developed in line with prerevolutionary con-
sumption patterns of the Cuban middle and upper classes to which the Cuban
masses had ample exposure; consumption patterns that, in turn, were shaped
by the example of the rich, consumption-oriented neighbor to the north, with
whom Cuba had very close economic and cultural ties until the late 1950s.
Marriage rates, noted universally for their responsiveness to short-term eco-
nomic fluctuations, responded in the expected way, more than doubling be-
tween 1959 and 1961.27 Childbearing among newly established couples soon
followed, and it is likely that births formerly postponed by earlier-established
unions were made up. Also contributing to the increase in the birth rate was an


Sergio Diaz-Briquets / Lisandro Perez

involuntary reduction in the incidence of abortion and the practice of con-

The baby bust: fertility during the 1970s
Consistent with our explanation of the causes of the Cuban baby boom, we
view the post-baby-boom Cuban fertility trend as resulting from a complex set
of factors that comprise both objective and subjective elements. In particular,
the following discussion of that trend goes beyond the traditional "moderniza-
tion" explanation (the first factor we examine) to arrive at a more comprehen-
sive interpretation of the decline of fertility in Cuba during the 1970s.
We regard the following socioeconomic conditions as the basic factors
underlying the rapid fertility decline of the 1970s: (1) socioeconomic reforms
and modernization efforts; (2) a severe housing shortage; (3) a return of eco-
nomic differentiation; (4) uneven economic performance; (5) low levels of con-
sumption; (6) a low or negative economic value of children; and (7) the
realistic and pessimistic assessments of economic prospects made by the lead-
ers. We discuss each of these factors below, in addition to another factor that,
although it cannot be considered an underlying socioeconomic condition, must
nevertheless be included as an important contributor to the fertility decline: the
increasing availability of abortion and contraception during the 1970s.

Socioeconomic reforms and modernization efforts
By the 1970s the social reforms discussed earlier-especially in education and
health-had produced a generation that was entering the reproductive ages
with relatively high levels of "modernity" as measured, for example, by the
level of educational attainment.
To appreciate the changes brought about by the socioeconomic reforms,
the prerevolutionary socioeconomic differentials should be kept in mind.
While members of the upper social strata in prerevolutionary Cuba undoubt-
edly held modern values and enjoyed modern lifestyles, the social and eco-
nomic inequalities of the prerevolutionary system, especially in terms of
education, meant that other segments of the population-especially the rural
poor-were little exposed to modern values and behavioral norms. The re-
forms instituted in the early 1960s that established a universal system of secular
education diffused these values and norms among the members of the formerly
underprivileged social strata, creating a generation that, in a more uniform
fashion than its predecessors, possesses characteristics of modernity usually
associated with low fertility.
Probably as significant in this area as educational and health reforms was
the spread of the official ideology with respect to the family and sex roles and
measures that reflected that ideology. While the accomplishments in this area
seldom measured up to the official pronouncements, there can be no question
that the government has succeeded in changing public attitudes concerning the
proper relations between the sexes, and that the relative social status of women


Fertility Decline in Cuba: A Socioeconomic Interpretation

has improved.28 This change in the status of women, and especially the in-
crease in the rate of female labor force participation, is likely to have had a
significant influence on lowering fertility.29
The "modernization" explanation of the Cuban fertility decline-with
emphasis on health and educational reform and changes in women's roles and
status-predominates in the literature. Without questioning its validity, we
argue that it is incomplete. That the modernization argument can only partially
explain the fertility decline is suggested by cross-national comparisons of rele-
vant socioeconomic and fertility indicators: Table 4 presents data for Cuba and
three Caribbean and Central American countries with socioeconomic indicators
fairly similar to those of Cuba.

TABLE 4 Cuba and selected countries of Central America and the
Caribbean: socioeconomic and demographic indicators, latest
available data
Expectation Percent of ever-married women
ot bith by highest level of education attained Total
at birth Total
(both Secondary Currently fertility
sexes)a Illiterate' None Primary and higher working' rate'
Cuba 72.8 4.2 53.3 46.7 39.0 1.9
Costa Rica 71.2 8.9 8.2 66.4 25.5 26.5 3.7
Jamaica 70 n.a. 2.2 75.2 22.6 33.9 3.7
Panama 70 9.5 6.7 53.3 39.9 33.1 4.1
NOTE: Educational data for Cuba refer to all women aged 14-49 in 1979. For Costa Rica and Panama
data refer to ever-married women 20-49 years of age. Attained levels, for each of the four countries,
refer to having completed at least a year at the given level.
"Expectation of life-For Cuba projected to 1980: Dirreccion de Demografia, Comit6 Estatal de
Estadisticas (Reptiblica de Cuba) and Centro Latinoamericano de Demografia (CELADE), Proyeccion de
la poblacion cubana 1950-2000. nivel national: metodologia y resultados (actualizacion en junio de
1980), mimeo, Havana, July 1980, Cuadro 18, p. 36; for Costa Rica, estimate for 1974-76: Hugo
Villegas and Carlos A. Valverde, "Life expectancy trends in Costa Rica," Bulletin of the Pan American
Health Organization 13, no. 3 (1979): Table 3, p. 256; for Jamaica and Panama: estimates in Population
Reference Bureau, 1982 World Population Data Sheet, Washington, D.C.
bDireccion de Demografia, Comit6 Estatal de Estadisticas (Repiblica de Cuba), Metodologia y tablas
seleccionadas parte I), Encuesta Demografica Nacional de 1979, Havana, April 1981, Table 20, p. 252.
'Educational attainment-For Cuba: Cuba: el descenso de la fecundidad, 1964-1978, cited in note 54,
Cuadro 11, p. 14; for the other countries: Susheela Singh, "Background characteristics used in WFS
surveys," Comparative Studies: Cross National Summaries, World Fertility Survey, London, no. 4,
March 1980, Table 5, p. 10.
'Work data-For Cuba: Direcci6n de Demografia, Comit6 Estatal de Estadisticas (Reptiblica de Cuba),
La fecundidad retrospective y el deseo de tener hijos de las mujeres en union matrimonial, Encuesta
Demogrdfica de Nacional 1979, Havana, June 1981, p. 33; for the other countries: Singh, cited in note c
above, Table 7.
'Population Reference Bureau, 1981 World Population Data Sheet.

The life expectancy estimates can be regarded as a crude proxy for the
availability of health services. Improvement of the health status of the popula-
tion earned the Cuban revolution much credit, although, as we noted earlier,
with respect to health prerevolutionary Cuba was better off than most other
developing nations. On this score, if allowance is made for data inconsistencies


Sergio Diaz-Briquets / Lisandro Perez

and reference period differences, the four countries appear to be doing more or
less equally well. Infant mortality levels, of course, should not differ signifi-
cantly between these countries if the estimated life expectancies are valid. In
terms of educational attainment Cuba seems clearly more advanced than the
other three countries, but in each of the four, basic educational needs appear to
have been met. Illiteracy levels (not available for Jamaica), for example, are
low (particularly if it is noted that the information includes older women) and
just a few points apart. Labor force participation rates for currently married
women are higher for Cuban women, but not much higher than in Jamaica and
The Cuban total fertility rate, however, is barely half as high as the total
fertility rate in the other three countries. This difference is inexplicable from
the indicators chosen. Thus, although a process of modernization is undoubt-
edly part of the currently very low levels of Cuban fertility, an adequate expla-
nation must also involve other factors.
The key to Cuba's very low fertility, in our opinion, lies in the sharply
differing prospects of Cuban couples regarding the fulfillment of their material
and status expectations during the early 1960s, and in the years since that time.
The former period was one of optimism. The latter, reflecting the revolution's
growing inability to meet material expectations, was a period of pessimism.
Only in the light of this radical change in outlook can the rapid fertility decline
be understood. Had the generation entering reproductive ages in the 1970s been
able to fulfill their aspirations, it is unlikely fertility would have declined as fast
and to as low a level as it in fact has.
In the following sections we briefly review some of the negative eco-
nomic developments that have accompanied the process of revolutionary
change in Cuba. Our description is by no means novel. A number of econo-
mists have already analyzed these factors in detail. What we propose here is to
integrate them into an explanation of Cuba's precipitous fertility decline in the

The housing shortage
The existence of a housing problem in Cuba predates the revolution. In fact,
housing reforms were among the earliest reforms instituted by the revolution-
ary government. As part of the effort to redistribute income and eliminate class
differentials, rental properties were nationalized and the tenants, upon pay-
ments to the state, were granted ownership. A 1960 law also took note of the
country's housing deficit (then estimated to be 655,000 units) and projected a
massive construction program.31
As with other optimistic plans formulated during this period, the en-
visaged surge in housing construction failed to materialize. By 1970 the deficit
of adequate housing exceeded one million units.31 With this housing deficit in
mind, the revolutionary authorities set goals for the construction of 100,000
housing units per year by 1980. In 1980, only 15,000 housing units were fin-
ished. As a result of the low rate of housing construction during the 1970s the
deficit of adequate housing has continued to increase. It can be estimated with

Fertility Decline in Cuba: A Socioeconomic Interpretation

official housing construction figures and housing data from the 1970 census that
between 1971 and 1980 the housing deficit grew by some 50,000 units per year.
By 1980 the deficit stood at approximately 1.5 million units.32 This, despite the
emigration of 220,000 persons during the 1970s that considerably eased the
pressures on the available housing stock.
The problem is most severe in urban areas, particularly in Havana,
where the deterioration of existing housing is a major reason for the critical
need for new construction. According to Roca, "In 1978, the administrator of
the city of Havana estimated that 250,000 dwelling units (50 percent of the
total) were in need of repair and that, of these, 30,000 were propped-up by
beams and 40,000 were declared uninhabitable."33 Numerous accounts by re-
cent visitors to Cuba have noted the hardships many families have to endure
because of the housing shortage.34
The housing situation is likely to have a negative effect on fertility. The
unavailability of new housing for young couples is surely one of the factors
lowering the marriage rate, which by 1978 was down to 6.2 per thousand popu-
lation from levels higher than 10.0 in the late 1960s.35 There is also evidence
that the housing shortage is partly responsible for the high divorce rate, which
stood at 2.6 per thousand population in 1978. This rate represents a fivefold
increase from the prerevolutionary level.36 In a 1973 study of divorce, a sample
of Havana residents cited the unavailability of housing and the need to share
accommodations with relatives as reasons for the increase in the divorce rate.37
It can be pointed out, of course, that in developing countries many chil-
dren and very poor housing commonly go together. Our argument about the
influence of the housing shortage in Cuba on fertility, once again, rests on the
psychological effect of frustrated expectations. With little hope for improve-
ment, many Cubans seem to prefer having fewer children to living in substan-
dard housing or in very cramped quarters.

Return to economic differentiation
By the late 1960s and early 1970s the income redistribution measures had
largely achieved their purpose. Agrarian reforms, free utilities, the rationing
and subsidizing of consumer goods, programs of job creation, and the expan-
sion of educational and health services had redistributed income and reduced
longstanding differentials by class and residence. As MacEwan put it, equality
had become "entrenched."38
After a decade of revolutionary change it had become increasingly clear,
however, that the income redistribution measures were beginning to have ad-
verse effects as well, and indeed, were impairing the ability of the Cuban
economy to grow. Expenditures for social programs, for example, increased
Cuba's foreign debt. Another example was the policy of full employment,
which was maintained at the expense of labor productivity. Concern for these
problems was translated into a change in government policy when, in 1971, the
Cuban leadership began to implement what Mesa-Lago has labeled the Soviet
Economic Reform Model.39 Seeking to remedy problems resulting from labor

Sergio Diaz-Briquets / Lisandro Perez

absenteeism, economic dislocations, and the huge foreign debt, this model
places emphasis on growth based on rational economic planning. This repre-
sented a break with the policy of redistribution. In the words of Mesa-Lago:

The Cubans painfully learned in the 1960s and acknowledged in the 1970s that
when a goal such as more equal distribution is idealistically pushed beyond rea-
sonable limits, disregarding [the] economic costs [of] decline in productivity and
growth, the survival of the whole system is in jeopardy . .. In the current
stage, begun in the early 1970s, goals such as growth have received first priority
while emphasis on other goals-for example, equality-has been considerably

Wage subsidies and programs of labor mobilization were severely cur-
tailed, and a Soviet-type system of performance quotas and more sharply dif-
ferentiated wage scales was introduced. Utilities started charging customers for
services that previously were free. The new policies resulted in higher prices
for consumer goods, tighter domestic budgets, and even some reemergence of
unemployment. One of the principal aims of the policy shift was to establish a
system of material incentives in order to combat absenteeism and raise labor
productivity.41 Indeed, the income redistribution measures that had played a
role in increasing fertility in the 1960s had, by the 1970s, not only run their
course, but were deliberately curtailed. We will address later the implications
of this for fertility.

Uneven economic performance
The favorable economic performance of the Cuban economy during the first
three years of the revolution proved to be temporary. This was not unexpected:
in part, the favorable performance had been spurious since it was achieved
through disinvestments in the capital stock and through depletion of invento-
ries. But, as this became evident by the second half of the 1960s, there were
other and more important factors at work that led to a deterioration of economic
performance. Mesa-Lago provides a fairly comprehensive listing of these fac-

.. .the rapid and wide collectivization of the means of production, which largely
destroyed the automatic mechanisms of the market without an adequate sub-
stitute; the numerous changes . in the model of economic organization and
development strategy, which made it impossible for any policy to consolidate;
the idealistic errors and inefficiencies of the Mao-Guevarist stage . ; the ex-
odus of experienced managerial, technical and professional personnel . ; the
decline of economic and managerial studies . ; the low sugar output in vari-
ous years combined with low sugar prices in the international market; the con-
sumptionist policy of the early years of the Revolution . .; the cost of the
economic embargo imposed by the United States and OAS upon Cuba and of
shifting the entire set of international economic relations away from the United
States towards the USSR and the socialist camp; and the heavy burden of mili-
tary expenditures for internal defense and subversion abroad.42

Fertility Decline in Cuba: A Socioeconomic Interpretation

By the mid to late 1960s the government reversed its earlier policy of
seeking rapid industrialization and placed the hopes for economic recovery and
growth on agriculture and particularly on Cuba's traditional major source of
foreign exchange earnings, sugar. The goal for the 1970 harvest was to produce
10 million tons of sugar, a national record. Achieving that goal in 1970 and in
the years beyond, the leadership argued, would go far toward solving Cuba's
economic problems. It would help to reduce the foreign debt and bring in
much-needed foreign exchange to help develop the country and improve con-
sumption levels. Virtually all sectors of Cuban society were mobilized toward
the realization of the goal: "sucrophilia" reigned.43
The 1970 sugar harvest, although setting a record, fell far short of the
goal. The overcommitment of labor and other economic resources to the sugar
harvest severely disrupted the rest of the national economy.44
The first half of the 1970s did witness an economic recovery, attributable
mainly to the high prices of sugar that prevailed in the world market until about
1975. Also contributing to the improved economic situation were the greater
efficiency of the new model of economic planning; the maturing returns to
previous investments in human and physical capital; the increased flow of
credit from market economies, resulting from the easing of the trade restric-
tions previously placed on Cuba by the United States and the Organization of
American States; considerable Soviet economic assistance;45 and the postpone-
ments of the repayment of the foreign debt to the Soviet Union.46 Levels of
living improved somewhat during this period, as the greater availability of
foreign exchange permitted Cuba to import consumer and capital goods from
Western economies, and as the Soviet Union increased its exports of some of
these goods to Cuba. Household appliances and passenger cars, for example,
while still in short supply, reappeared in relatively large numbers after years in
which their import had been strictly limited.
If the improved economic picture provided the basis for a certain re-
newal of optimism, the renewal was to be short-lived. Starting in about 1976,
the Cuban economy went into a tailspin and Cuban economic aspirations were
once again frustrated. Mesa-Lago has attributed this most recent period of eco-
nomic difficulties to

S. .the decline in sugar prices in the international market; agricultural plagues
S. and problems in the fishing and nickel industries; a sharp reduction of the
flow of credit from market economies since 1979; some complications in the
implementation of the new System of Economic Management and Planning; and
the heavy burden of Cuban involvement in Africa.47

Low levels of consumption
For ordinary Cubans the most important manifestation of these economic prob-
lems was a severe shortage of virtually all consumer goods, including food and
clothing. The onset of that shortage during the second half of the 1960s, and its
persistence thereafter-except for a short respite in the early 1970s-could not

Sergio Diaz-Briquets / Lisandro Perez

fail to dampen the spirit of optimism that had prevailed during the early years
of the revolution.
Those shortages, combined with the government's policy of motivating
people to work through the use of material incentives and a reduction in the
money supply, have priced many "nonessential" goods beyond the reach of
most families. Many of those goods are presumably available through the black
market, and often they are available through the officially established "parallel
market," in which goods are sold at prices substantially higher than those of
rationed goods (in fact, higher than comparable prices in Western economies),
but below the levels of the black market. Average Cuban wages and salaries are
very low in relation to the cost of most of such goods. Mesa-Lago, for exam-
ple, provides data (relating to 1977-78) showing that the price of a 17-inch
black and white television set in the parallel market ranged between 650 and
900 pesos, and the price of a medium-sized refrigerator between 650 and 850
pesos. By contrast, the average monthly wage in 1978 was 140 pesos. Accord-
ing to Mesa-Lago, "the cheapest refrigerator or television [set] cost the equiva-
lent of five months' average wages" even when purchased through the
government allocation plan that controls the distribution of goods among work
places.48 Prior to the revolution a large percentage of Cuba's population owned
such appliances.49 Arguably, part of the original overwhelming popular sup-
port for the revolution was contingent on the belief that aspirations for such
items would be better fulfilled under the new social and economic order.
The shortages and high prices are not limited to "nonessential" goods.
While a number of consumer staples are rationed and marketed at subsidized
prices, many observers believe that the rations are insufficient to satisfy basic
needs, especially if cultural preferences are taken into account. In prerevolu-
tionary Cuba, per capital average yearly rice consumption hovered around 50
kilograms.50 In the official subsidized markets the price of rationed rice, the
basic Cuban foodstuff, is 21 cents per pound, and each person, according to the
rationing guidelines, is allocated five pounds per month (or 27 kilograms per
year). Consumers must supplement their rations by purchasing rice on the
black market, where reportedly rice sells for two pesos per pound.1 Many
other such examples could be cited.
In his recent study, Mesa-Lago observes:

S. .most essential clothing and footwear remain rationed-for example, one
each annually of a pair of pants, a skirt, a shirt, a blouse, a dress, a pair of leather
shoes, and just four meters of fabric . one's quota or option to buy does not
guarantee that either the item is in stock or available in the consumer's size;
hence, queues for manufactured goods are often worse than for foodstuffs.52

Mesa-Lago goes on to note that in the black market, where many of these
shortages can be made up, even essential goods are "unaffordable . for the
average citizen. An entire monthly wage could be spent on one pound of cof-
fee, a pair of shoes, a meter of fabric, and a couple of pounds of beef."53


530 Fertility Decline in Cuba: A Socioeconomic Interpretation

It is officially claimed that in revolutionary Cuba "aspirations for indi-
vidual progress are not satisfied as a function of income level, whose spread is
limited," and that "the accumulation of wealth is restricted and consumerism is
not encouraged."54 But our review of the historical evidence suggests that the
principal reason consumerism is not encouraged is the inability of the Cuban
economy to satisfy even some basic material needs and others acquired within
a given socioeconomic context. In Cuba materialism is very much alive, as the
background of successive waves of emigration shows-each wave of emigra-
tion increasingly drawing on those segments of Cuban society whose levels of
living have improved most with the achievements of the revolution but who
also expected more.
We believe, and have tried to show here, that the people of Cuba had and
presumably still have high aspirations for material well-being. Those aspira-
tions evolved under the shadow of the most consumer-oriented society in the
world, a society that for over 50 years influenced every aspect of Cuban life.
The promise of fulfilling those aspirations was at the center of the popular
support received by the revolution, and the promise was reinforced by the early
redistributive policies of the new regime. Many underprivileged Cubans be-
lieved that the revolution would create the necessary conditions to satisfy those
aspirations, and indeed the original intent of the revolutionary leadership was
to do so. One plausible response to frustrated national expectations is voluntary
restriction of family size by Cuban couples.

Low or negative economic value of children
Critics of our interpretation might argue that having children in Cuba is rela-
tively costless since the state provides many child-related services at no cost or
at subsidized prices. We take issue with this notion. Despite price supports and
free social services, raising a child in Cuba entails significant monetary as well
as other costs for parents. For example, only about 3 percent of Cuban children
of primary school age receive government scholarships covering all their
needs.55 Most parents are responsible for financial support of their children and
must find the means to provide them with what is not readily accessible through
the rationing system. Since 1976, most families have had to pay to place their
children in day care centers. Further, government policies that set very high
prices for "nonessential" goods discourage childbearing by raising the oppor-
tunity cost of having children. Women often must choose between having chil-
dren and wage employment, just as they do in market economies.
It can be further argued that a longer term effect of the structural changes
that took place in the economy of postrevolutionary Cuba was a reduction in
the economic value of children to parents. The virtual elimination of the private
sector has reduced the utility of children as a source of labor in family enter-
prises. In turn, the new economic system, by eliminating many of the eco-
nomic uncertainties with which the poor have to cope in Third World nations,
has reduced the value of children as a form of risk aversion. In Cuba, as in

Sergio Diaz-Briquets / Lisandro Perez

other socialist states, employment, or at least a minimum income, is virtually
guaranteed; health services are provided free or at low cost; essential goods, if
in short supply, are available at subsidized prices; and most of the population is
covered by a system of retirement pensions. Thus, children are no longer a
crucial source of old-age security.
In arguing that some of the changes instituted by the revolution have
caused a shift in parental assessments of the economic value of children, we
should add that for a substantial segment of the Cuban population such a shift
probably occurred much earlier than 1959. This relates to our point regarding
the need to take into account historical trends in Cuban fertility, as well as the
differential impact of the revolution on Cuba's prerevolutionary class system.
The moderately low fertility level prior to the revolution indicates that most
Cubans apparently did not view large families as economically advantageous
and did not rely on their children for old-age security or as insurance against
life's uncertainties. According to Mesa-Lago, in 1958 63 percent of the Cuban
labor force was covered "for old age, disability, and survivors insurance, while
all the labor force was covered against occupational accidents and diseases,
and female employees had maternity insurance."56 Clearly, then, the argument
that the changes instituted by the revolution caused the shift in assessments
regarding the value of children has applicability only to the social strata whose
welfare prior to the revolution depended on economic assistance from their
children. Nevertheless, by the 1970s such a shift, regardless of why and when
it occurred, was fairly generalized throughout the population and undoubtedly
contributed to the decline in fertility.

But is there any firm evidence to link the poor performance of the econ-
omy, housing shortages, low consumption levels, and costs of children with
low fertility? We believe there is. We have already mentioned the changing
socioeconomic composition of successive waves of Cuban emigrants over the
years. Furthermore, of a sample of emigrants arriving in the United States in
1973 and 1974, 96 percent stated that one of the reasons they left Cuba was that
they saw "a 'lack of future' in terms of social and economic aspirations for self
and children."57
Further evidence is provided by an exploratory study recently published
in a Cuban public health journal. This study was conducted between December
1978 and June 1979 while its author was working in the sugarharvest with a
group of young males, all from Havana. Of the 15 married men in this small
sample, 13 indicated they did not want any more children because of housing
problems; 12 of them also said that they did not want any more children be-
cause their income was insufficient. Of the married or divorced fathers earning
150 pesos a month or less, 11 (out of 15, or 73 percent) stated that they did not
earn enough to satisfy the needs of their families.58 Finally, the few independ-
ent studies conducted in Cuba that have involved detailed field work in the
early 1970s have shown that the economic conditions were difficult even in
comparison to the conditions of the formerly underprivileged before the revolu-

Fertility Decline in Cuba: A Socioeconomic Interpretation

Realistic and pessimistic assessments of
economic development by the leadership
As we have noted earlier, public statements by the Cuban leaders, and espe-
cially the speeches of Fidel Castro, provide important insights into the state of
the economy. Such statements represent the principal means through which the
people receive information from the highest levels of government regarding the
course of the revolution. We have also noted that among the factors that con-
tributed to the optimism of the early revolutionary period were the naively
optimistic assessments and promises of economic development made by politi-
cal leaders at a time when favorable economic conditions prevailed. The down-
turns in the economy during the second half of the 1960s and after 1975 were
likewise reflected in the public statements of the leadership. The speeches of
Fidel Castro, in particular, reflected the deepening economic malaise, espe-
cially after the failure of the 10 million ton sugar harvest.
On 26 July 1970, for example, Castro delivered his "contradictions"
speech, in which he reviewed Cuban agriculture and industry and demonstrated
the seriousness of the economic situation. He also admitted to errors in the
management of the economy, indicating that "the leaders of this Revolution
have cost the people too much in our process of learning."6o
Thereafter, Castro's speeches on the economy and domestic conditions
in general never regained the optimistic tone that imbued his public statements
during the early years of the revolution. Even when the Cuban economy im-
proved during the first half of the 1970s, the often recklessly optimistic assess-
ments of Cuba's economic prospects were missing. The speeches made in 1973
and 1974 reflected a degree of optimism about the possibilities for economic
development afforded by high sugar prices, but it was a guarded optimism.
Starting in 1975, and concurrent with the economic decline, the speeches
by the Cuban leader, once again, took on a more somber tone, presenting
accomplishments, but also dwelling on failures and on the need to improve
conditions. i

The increasing availability of abortion
and contraception
The respective roles of abortion and contraception as proximate determinants
of fertility decline in Cuba, although not within the scope of the present discus-
sion, merit mention. They are additional and in a sense independent factors
apart from the crucial socioeconomic changes described here.
Restrictions on legal induced abortion, strictly applied in the years im-
mediately following the revolution, were progressively eased starting in 1964.
The legal abortion rate peaked at 69.5 abortions per thousand women aged
15-44 in 1974, and its subsequent tapering off can plausibly be attributed to
increased availability of modern contraceptive methods, particularly since the
mid-1970s, and the substitution of contraceptive practice for abortion.62
It seems possible that limited access to abortion and low availability of


Sergio Diaz-Briquets / Lisandro Perez

modern contraceptives in the late 1960s-when imports from the United States
had long ceased and local manufacture was at very low levels-may provide a
partial explanation for the relatively slow pace of fertility decline at that time,
despite the already worsening economic conditions. Certainly their subsequent
wide availability and the parallel sharp fertility declines are consistent with
arguments that access to abortion and contraception can accelerate if not trigger
fertility decline.


Our interpretation of the fertility decline in Cuba is in broad agreement with
explanations that assign dominant weight to social and economic change. A
specific point where we differ from most analyses concerning the causes of the
Cuban fertility decline is our feeling that the decline that followed the baby
boom has been influenced to a great extent by the poor performance of the
Cuban economy. A further distinguishing feature of our interpretation is the
assertion that economic, social, and political conditions have deeply affected
both the aspirations of the Cuban people for material well-being and the ex-
pected and actual satisfaction of those aspirations. The interplay of these aspi-
rations and perceptions largely laid the downward course of fertility. The
emphasis on these factors in explaining the Cuban fertility decline is par-
ticularly appropriate, we believe, if applied to that considerable segment of the
population that, prior to the revolution, was already well along in its demo-
graphic transition. It also has force in explaining the fertility behavior of for-
merly more marginalized social groups whose latent aspirations were
encouraged but not satisfied by political and economic changes resulting from
the Cuban revolution.

Lisandro Perez wishes to acknowledge the
support he received from the Joint Committee
on Latin American Studies of the Social Sci-
ence Research Council and the American
Council of Learned Societies.
1 Sergio Diaz-Briquets and Lisandro
Perez, "Cuba: The demography of revolu-
tion," Population Bulletin 36, no.l (1981);
Gerardo GonzAlez, German Correa, Margarita
M. Erraziriz, and Ratil Tapia, Estrategia de
desarrollo y transicion demogr6fica: el caso
de Cuba, vol.1 (Santiago, Chile: Centro Lati-
noamericano de Demograffa, 1978); Paula E.
Hollerbach, "Recent trends in fertility, abor-
tion and contraception in Cuba," International
Family Planning Perspectives 6, no.3 (Sep-
tember 1980): 97-106; and Barent F. Land-
street, Jr., Cuban Population Issues in Histor-

ical and Comparative Perspective, Latin
American Studies Program Dissertation
Series, Cornell University, 1976.
2 James T. Fawcett and Siew-Ean Khoo,
"Singapore: Rapid fertility transition in a com-
pact society," Population and Development
Review 6, no.4 (December 1980): 549-579.
3 For a view of prerevolutionary fertility
in Cuba see 0. Andrew Collver, Birth Rates in
Latin America: New Estimates of Historical
Trends and Fluctuations (Berkeley: University
of California, 1965); and Diaz-Briquets and
Perez, cited in note 1.
4 United Nations, Department of Eco-
nomic and Social Affairs, National Experi-
ence in the Formulation and Implementation of
Population Policy: Cuba, 1959-76, ST/ESA/

Fertility Decline in Cuba: A Socioeconomic Interpretation

SER.R./17 (New York: United Nations, 1977);
and Barent F. Landstreet, Jr., "Cuba," in Pop-
ulation Policies in the Caribbean, ed. Aaron
Lee Segal (Lexington, Mass.: Lexington
Books, 1975), pp. 127-158.
5 See Diaz-Briquets and P6rez, cited in
note 1; Hollerbach, cited in note 1; and Paula
E. Hollerbach, "Determinants of fertility de-
cline in post-revolutionary Cuba," in Fertilit
Decline in Developing Countries: Case Stud-
ies, ed. W. Parker Mauldin (forthcoming).
6 According to data provided by Gille for
selected countries and to estimates of age-spe-
cific rates for Cuba made by Catastis Cervera,
Cano, and Velazquez in the 1950s, fertility
levels in Cuba were about the same as or
somewhat higher than those of Australia, Can-
ada, New Zealand, Portugal, and the United
States whether fertility was measured by the
crude birth rate or by other more refined in-
dexes, such as the total fertility rate or the
gross reproduction rate:

CBR (1957) TFRa (1956)
Australia 22.9 3.22
Canada 28.2 3.66
New Zealand 25.1 3.40
Portugal 23.7 3.80
United States 25.4 3.58
Cuba 26.1 (1958) 3.60
aExcept for Cuba, TFRs were estimated from GRR

At this time, obviously, some of these coun-
tries were going through a baby boom. Halvor
Gille, "An international survey of recent fer-
tility trends," in National Bureau of Economic
Research, Demographic and Economic
Change in Developed Countries (Princeton:
Princeton University Press, 1960), pp. 17-35;
and Sonia Catasis Cervera, Pedro Cano, and
Elio Velhzquez, "Evoluci6n estimada de la fe-
cundidad en Cuba: 1900-1950," Estudios de-
mogr6ficos, Centro de Estudios Demo-
grificos, Universidad de la Habana, Serie 1,
no.5, 1975.
7 According to the 1981 version of the
Population Reference Bureau's World Popula-
tion Data Sheet, only 24 developing countries
had total fertility rates in 1980 comparable to
that of Cuba in the late 1950s. Some of these
countries, such as Argentina and Uruguay,

had had lower fertility rates than Cuba during
the 1950s.

8 Alfonso Farn6s Morej6n, "Cuba: tablas
de mortalidad estimadas por sexo, period
1955-70," Estudios demograficos, Centro de
Estudios Demograficos, Universidad de la
Habana, Serie 1, no.8, December 1976.

9 Diaz-Briquets and P6rez, cited in note 1,
p. 6. This interpretation is based on assess-
ments and statistical data found in, among
other sources: Steven E. Beaver, Demo-
graphic Transition Theory Reinterpreted
(Lexington, Mass.: D. C. Heath and Com-
pany, 1975); Cuban Economic Research Proj-
ect, A Study on Cuba (Coral Gables: Univer-
sity of Miami Press, 1965); Jorge I.
Dominguez, Cuba: Order and Revolution
(Cambridge: The Belknap Press of Harvard
University Press, 1978); International Bank for
Reconstruction and Development (World
Bank), Report on Cuba (Baltimore: Johns
Hopkins University Press, 1951); Carmelo
Mesa-Lago (ed.), Revolutionary Change in
Cuba (University of Pittsburgh Press, 1971);
and Lowry Nelson, Cuba: The Measure of a
Revolution (Minneapolis: University of Min-
nesota Press, 1972).

10 From the early twentieth century on,
the public health literature of prerevolutionary
Cuba was replete with references to the "so-
cial problem" of abortion. Conversations with
physicians presently residing outside of Cuba
confirm that although abortions were illegal,
for all practical purposes they were available
on demand and at a very low cost for those
who could not afford a steep fee. Many physi-
cians, as well as less skilled abortion practi-
tioners, even offered their services through
home visits. Some of the physicians inter-
viewed, who were probably more sensitive to
the abortion issue than most of the general
population, told one of the authors of this pa-
per that the issue of abortion did not provoke
the kind of heated controversies that are com-
mon in the United States. It is certain that
some groups, such as the Catholic Church,
were opposed to it; however, the influence of
the Church on the vast majority of Cubans was

11 The estimated provincial crude birth
rates in 1958 were as follows:


Sergio Diaz-Briquets / Lisandro Perez

Pinar del Rio 28.4
Havana 19.6
Matanzas 21.6
Las Villas 23.0
Camagiiey 25.5
Oriente 33.9
Total Cuba 26.1
SOURCE: Gonzdlez et al., cited in note 1, Table

12 Personal communications.
13 It is also true that fertility has remained
at relatively stable levels for decades in Argen-
tina and Uruguay, two countries that histor-
ically had lower fertility than Cuba. However,
these countries have opposed family limitation
efforts since they adhere to a pronatalist
14 Landstreet, cited in note 1; Hollerbach,
articles cited in notes 1 and 5; Gonzalez et al.
and Diaz-Briquets and P6rez, both cited in
note 1.
15 Part of the fertility increase in the early
1960s may have resulted from the making up
of marriages (and births that were postponed)
during the immediately preceding turbulent
years. However, it is unlikely that these fac-
tors accounted for more than a limited portion
of the baby boom.
16 The extent to which enforcement of the
laws against induced abortion contributed to
the fertility increase cannot be ascertained for
obvious reasons. In conversations with physi-
cians who performed abortions in Cuba at the
time, one of the authors was told that for a
period of time the prevalence of abortion de-
clined since practitioners began to be more
careful in order to avoid detection and prose-
cution. Eventually the incidence of abortion
began to rise, reaching within a relatively
short time, according to these informers, a
level not very different from that of earlier
years. The probable impact of a shortage of
contraceptives is even more uncertain, but is
not likely to have been very significant, in part
because a reduction in imports of contracep-
tives was probably eventually made up by an
increase in their local manufacture.
17 Juan P6rez de la Riva, "La population
de Cuba et ses problemes," Population 22,
no.l (January-February 1967): 102.


18 Claes Brundenius, "Measuring income
distribution in pre- and post-revolutionary
Cuba," Cuban Studies 9, no.2 (July 1979):
29-44. The shift in the distribution of income
between 1953 and 1962 estimated by Brun-
denius is in basic agreement with that inde-
pendently derived by Arthur MacEwan in his
Revolution and Economic Development in
Cuba (New York: St. Martin's Press, 1981),
19 For reviews of these developments see:
Nelson P Vald6s, "The radical transformation
of Cuban education," in Cuba in Revolution,
ed. Rolando E. Bonachea and Nelson P. Val-
d6s (Garden City, N.Y.: Anchor Books, 1972),
pp. 422-455; Martin Carnoy and Jorge
Wertheim, "Cuba: Economic change and edu-
cation reform," World Bank Staff Working Pa-
per no.317, Washington, D.C., 1979; Comi-
si6n Econ6mica para America Latina
(CEPAL), Cuba: estilo de desarrollo y politi-
cas sociales (Mexico City: Siglo Veintiuno
Editores, 1980); Ross Danielson, Cuban Medi-
cine (New Brunswick, N.J.: Transaction
Books, 1979); and Sergio Diaz-Briquets, The
Health Revolution in Cuba (Austin: University
of Texas Press, forthcoming).
20 Archibald R. M. Ritter, The Economic
Development of Revolutionary Cuba: Strategy
and Performance (New York: Praeger, 1974),
p. 106. See also Carmelo Mesa-Lago, The
Economy of Socialist Cuba: A Two Decade
Appraisal (Albuquerque: University of New
Mexico Press, 1981).
21 For example, an excessive amount of
livestock was slaughtered and brought to mar-
ket in response to increases in demand. This
led to a deleterious shrinkage in the size of
herds, with detrimental effects on their even-
tual maintenance and growth.
22 For a classic study of the complex and
diverse motivations of the emigrants for leav-
ing Cuba during the first few years of the revo-
lution, see Richard R. Fagen, Richard A.
Brody, and Thomas J. O'Leary, Cubans in Ex-
ile: Disaffection and the Revolution (Stanford:
Stanford University Press, 1968).
23 Montaner, in a "nonacademic" but
very insightful book on the Cuban revolution,
shows that these views were widely held by
the more politicized social groups. These ideas
have been encountered by anyone familiar

Fertility Decline in Cuba: A Socioeconomic Interpretation

with Cuba's historical record. Carlos Alberto
Montaner, Secret Report on the Cuban Revo-
lution (New Brunswick, N.J.: Transaction
Books, 1981), p.254.
24 Statement made by Ernesto Guevara as
chairman of the Cuban delegation at the Fifth
Plenary Session of the Organization of Ameri-
can States Special Meeting of the Inter-Ameri-
can Economic and Social Council, 8 August
1961, Punta del Este, Uruguay; in Rolando E.
Bonachea and Nelson P. Vald6s (eds.), Che:
Selected Works of Ernesto Guevara
(Cambridge, Mass.: The MIT Press, 1969),
p.28. In 1980, even without taking into ac-
count the effects of two decades of persistent
inflation, Cuba's per capital income was well
under half the level predicted by Guevara for
this year. All the other promises failed to ma-
terialize as well.
25 "Informe del Dr. Regino Boti, Minis-
tro de Economia a la Primera Reuni6n Nacio-
nal de Producci6n," Obra revolucionaria,
no.30, 26 August 1961, pp. 18-19. Shortly af-
ter these pronouncements were made, these
items began to be strictly rationed. Rationing
continues to this day.
26 "Discurso de Fidel Castro, aniversario
de la ley de reform agraria," Obra revolu-
cionaria, no.3, 27 May 1960, p.11; and "Dis-
curso de Fidel Castro," Obra Revolucionaria,
3 June 1960, p.11.
27 Diaz-Briquets and Perez, cited in note
1, p. 15.
28 For a discussion of the family-related
policies of the Cuban government, especially
the Family Code of 1975, see Lisandro P6rez,
"The family in Cuba," in The Family in Latin
America, ed. Man Singh Das and Clinton
Jesser (New Delhi: Vikas Publishing House,
29 Hollerbach, cited in note 1.
30 Maruja Acosta and Jorge E. Hardoy,
Urban Reform in Revolutionary Cuba, An-
tilles Research Program Occasional Paper
no.l, Yale University, 1973, p.8.
31 Junta Central de Planificaci6n, La si-
tuaci6n de la vivienda en Cuba y su evolucion
perspective (Havana: Editorial Orbe, 1976).
32 Calculated from data in Junta Central
de Planificacion, cite in note 31, pp. 54-63,
and official data on annual housing construc-

33 Sergio Roca, "Housing in socialist
Cuba," in Oktay Ural (ed.), Housing, Plan-
ning, Financing and Construction: Proceed-
ings of the International Conference on
Housing, Planning, Financing, Construction,
Miami Beach, Florida, 2-7 December 1979
(New York, Pergamon Press, 1980), vol.1, p.
34 Laura Bergquist, "Behind the Cuban
exodus," GEO 2 (August 1980): 14; Fred
Ward, Inside Cuba Today (New York: Crown
Publishers, 1978), pp. 8-9; Fred Ward, "In-
side Cuba today," National Geographic 151,
no.1 (January 1977): 41; and Helga Silva,
"Cubans dislike daily hardships but praise
effects of Revolution," The Miami Herald, 7
August 1979, p. 12-A.
35 Republic of Cuba, National Committee
of Statistics, Anuario estadistico de Cuba,
1978 (Havana, no date), Table 8.
36 Republic of Cuba, cited in note 35; and
Diaz-Briquets and P6rez, cited in note 1, p. 18.
As with the marriage rate, interpreting trends
in the divorce rate is complicated by legaliza-
tion of consensual unions.
37 Jorge Hernbndez et al., Estudio sobre
el divorcio, Centro de Informaci6n Cientifica
y T6cnica, Universidad de la Habana, Havana,
1973, p. 67.
38 MacEwan, cited in note 18, p. 161.
39 Mesa-Lago, cited in note 20, pp.
40 Mesa-Lago, cited in note 20, p. 3.
41 MacEwan, cited in note 18, p. 181.
42 Mesa-Lago, cited in note 20, p. 35.
43 Ritter, cited in note 20, chapter 5, pp.
44 Ritter, cited in note 20, especially pp.
183-187; and Mesa-Lago, cited in note 20, pp.
45 Soviet assistance has been massive,
some 17 billion dollars between 1961 and 1979.
Some observers claim that in 1978, a relatively
prosperous year, Soviet assistance accounted
for between one-fourth and one-third of gross
national product. Alberto Recarte, Cuba:
economic poder (1959-1980) (Madrid: Al-
ianza Editorial, 1980), p. 153; and Lawrence
H. Theriot and JeNelle Matheson, "Soviet
economic relations with non-European
CMEA: Cuba, Vietnam and Mongolia-

Sergio Diaz-Briquets / Lisandro Perez

Cuba," in Joint Committee Print, Soviet Econ-
omy in a Time of Change: A Compendium of
Papers Submitted to the Joint Economic Conm-
mittee, Congress of the United States (Wash-
ington, D.C.: US Government Printing Office,
10 October 1979), vol.2, p. 560.
46 Mesa-Lago, cited in note 20, pp.
47 Mesa-Lago, cited in note 20, p. 36.
48 Mesa-Lago, cited in note 20, p. 162.
49 On the basis of official statistics and
population figures it can be estimated that 28
percent of Cuban families in 1958 owned tele-
vision sets and about 70 percent owned radios.
It is important to emphasize that this was over
20 years ago, and that the cost of producing
those items then was much higher than it is at
present. In 1958 there were 132 radios and 54
television sets per thousand population; in
1978 the corresponding figures were 151 and
50. By many consumers, if not by the major-
ity, some of these goods were considered ne-
cessities and almost everyone hoped to have
them. Juan F. Noyola, a Mexican economist
highly sympathetic to the Cuban revolution,
stated in 1959: "In a country like Cuba, where
a large proportion of the population has an in-
come level that allows it to enjoy some of
life's conveniences and modern technology,
these articles [referring to such appliances] are
absolute essentials." Juan F. Noyola, La
economic cubana en los primeros ahos de la
revoluci6n 3y otros ensayos (Mexico City:
Siglo Veintiuno Editores, 1978), p. 70. Castro
himself, recognizing the national appetite for
household appliances, promised on one of his
many public appearances that "families would
not have to do without radios, televisions and
refrigerators." Fidel Castro, "Exposici6n de
Fidel Castro por el canal 2," Obra revolu-
cionaria, no.3, 27 May 1960, p. 24.
50 Diaz-Briquets, cited in note 19, pp.
48-49, 191.
51 Mesa-Lago, cited in note 20, pp.
52 Mesa-Lago, cited in note 20, p. 160.
53 Mesa-Lago, cited in note 20, p. 162.

54 Direccion de Demograffa, Comit6 Es-
tatal de Estadisticas (Republica de Cuba) and
Centro Latinoamericano de Demografia
(CELADE), Cuba: el descenso de la fecun-
didad, 1964-1978, San Jose, Costa Rica, June
1981, p. 71.
55 Comite Estatal de Estadisticas (Re-
piblica de Cuba), Anuario estadistico de
Cuba, 1979, Havana, Table 10, p. 213.
56 Mesa-Lago, cited in note 20, p. 169.
57 Alejandro Portes, Juan M. Clark, and
Robert L. Bach, "The new wave: A statistical
profile of recent Cuban exiles to the United
States," Cuban Studies 7, no.1 (January 1977):
58 Guillermo GonzAlez P6rez, "Algunas
consideraciones sobre aspects de la fecun-
didad susceptibles de ser estudiados en los
hombres: studio exploratorio," Revista
cubana de administration de salud 7, no.l
(January-March 1981): 77-85.
59 See in particular Ernst Halperin, The
Taming of Fidel Castro (Berkeley: University
of California Press, 1981), pp. 141-151,
229-230; and the four books based on the field
work conducted by Oscar Lewis and his col-
laborators in Cuba (all published by the Uni-
versity of Illinois Press, Urbana): Oscar
Lewis, Ruth M. Lewis, and Susan M. Rigdon,
Four Men, 1977; Four Women, 1977; Neigh-
bors, 1978; and Douglas S. Butterworth, The
People of Buena Ventura: Relocation of Slum
Dwellers in Postrevolutionary Cuba, 1980.
60 Fidel Castro, "Speech at the mass rally
in celebration of the 17th anniversary of the
attack on the Moncada," Granma, 27 July
1970, p. 8.
61 Fidel Castro, "Informe central rendido
al Primer Congreso del Partido Comunista de
Cuba, 17 de Diciembre de 1975," Granma re-
sumen semanal, 28 December 1975, pp. 2-8;
and Fidel Castro, "Informe central rendido al
Segundo Congreso del Partido Comunista de
Cuba, 17 de Diciembre de 1980," Granma re-
sumen semanal, 28 December 1980, pp. 6-16.
62 Hollerbach, cited in note 1.

Women in Agriculture:
Counting the Labor Force
in Developing Countries

Ruth B. Dixon

The theme of woman as invisible worker has captured the
imagination of a number of critics of standard labor force statistics (e.g.,
Boserup 1970, 1975; Youssef 1977; Perez-Ramirez 1978; Blumberg 1979; Fong
1980a; Palmer and von Buchwald 1980; UN Secretariat 1980; Wainerman and
Recchini de Lattes 1981). Not only has domestic labor-a predominantly fe-
male activity in all societies-been excluded from censuses and national ac-
counts such as the gross national product (Jain et al. 1979; International Center
for Research on Women 1980; Rogers 1980), but even women's work in the
fields has eluded the collectors of rural labor force statistics. In some countries,
female unpaid family helpers in agriculture are systematically excluded, result-
ing in a consistent undercount of the agricultural labor force and an underesti-
mate of the proportion female. In other countries, women appear and disappear
in large numbers from one survey to the next. In 1954, the Algerian census
counted 981,000 women agricultural laborers (37 percent of the farm labor
force) but in 1966 only 23,000 (2 percent of the total). The census of India
counted 49 million women in agriculture in 1961 but only 26 million in 1971, a
drop from 36 to 20 percent of the total. And Brazil's 761,000 women in the
agricultural labor force in the 1950 census grew to 3 million in a 1976 house-
hold survey, up from 7 to 21 percent of the total agricultural labor force.
Problems of definition and procedure plague the collection of labor force
statistics in almost all situations, in industrialized and developing countries
alike, for young and old, male and female (see, for example, the discussion of
the complexities of measuring economic activity and under- and unemploy-
ment in UN Department of Economic and Social Affairs 1967:339-396). Yet
women (and children) in farm work are disproportionately undercounted in
most population censuses for a number of reasons, some relating to the strict
definitions of economic activity proposed in international guidelines (ILO
1976) and some relating to the selective application of these guidelines, as


Women in Agriculture: Counting the Labor Force in Developing Countries

when farm wives and children (especially students) are simply assumed not to
be economically active. Since the extent of undercounting is not consistent
across countries or even within countries over time, census-based labor force
statistics form a shaky basis for analyzing international trends and variations.
The examples just cited suggest that the observed differences are often statisti-
cal artifacts rather than real behavioral patterns.
A quite different picture of female labor force participation emerges
from rural time-use studies that count the labor contributions of all household
members, male and female, adult and child. By measuring the duration, reg-
ularity, and intensity of work throughout the year, these surveys describe the
flow of productive labor (apart from domestic work) through the days and
seasons rather than the stock of workers at one point in time (Fong 1980b; FAO
1978). As a consequence, they generally reveal a far higher rate of participation
by women and children. In Egypt, for example, where the 1960 census counted
women as only 4 percent of the total agricultural labor force, a detailed rural
labor record survey indicated that about one-quarter of all nondomestic produc-
tive work in farm households was done by women (ILO 1969:27). And in a
sample of eight villages in Sharquiyya Governate considered representative of
the diverse cropping regions of the Nile Delta, women performed 44 percent of
adult family farm labor and 27 percent of the work of hired hands (calculated
from Richards and Martin 1981: Tables 2, 17).
Although the growing number of time-use surveys offers a wealth of
information on household patterns of labor allocation within particular commu-
nities and farming systems, most cannot be generalized to larger populations.
At the aggregate level, then, we are left with large-scale national censuses or
surveys of often dubious repute. Even here, however, it is possible to compare
the results of different approaches to counting the stock of workers. For exam-
ple, the effects of changes in definition or practice from one census or survey to
the next can be assessed quite readily for some countries. Counts from national
population censuses and labor force surveys can also be compared with esti-
mates from the International Labour Office that attempt to standardize labor
force series in each country for changes in data collection methods over time.
And, in turn, estimates from the ILO can be compared with counts obtained for
a number of countries from a quite different source-the censuses of agricul-
tural holdings coordinated by the Food and Agriculture Organization of the
United Nations. Finally, the ILO-FAO comparisons form a basis for deriving
new estimates of the sex composition of the farm labor force for all countries,
reflecting the contrast between the generally more restrictive ILO approach and
the more inclusive approach of the FAO.

Conceptual problems in counting
agricultural workers
Concepts and methods of reckoning labor force participation based on contem-
porary Western experience have proven inadequate when applied to developing
countries, where workers are more likely than their Western counterparts to be


Ruth B. Dixon

self-employed rather than wage earners, to work seasonally rather than year-
round, to be underemployed rather than formally unemployed, and to engage
in a fluid or sporadic pattern of diverse and shifting economic activities (Du-
rand 1975; Standing 1977; Mueller 1978; Blacker 1978, 1980). Moreover, the
boundary between domestic production for the household's own consumption
and economic activity for sale or exchange is less clearly drawn in developing
countries, especially in rural areas, and especially among women (Boserup
1975; Boulding 1976; Fong 1980a).
These difficulties are compounded in the agricultural sector, where sub-
sistence farmers may sell very little of their produce, where unpaid labor on
their own land alternates with wage or exchange labor on another's, where
children may regularly tend animals and women grow foodstuffs in their
kitchen gardens or process crops in their compounds but not work in the fields,
and where trade or small crafts are added to agricultural work in a seasonal mix
of household activities. Indeed, the conceptual distinctions between persons
who are economically active and inactive, and between agricultural and non-
agricultural occupations, can become hopelessly blurred, particularly in the
case of women (and children). Efforts to sharpen the distinctions by enforcing
a strict (i.e., more Western) definition of labor force participation inevitably
result in a poor description of economic activity in the agricultural sector.
Consider, for example, this day in the life of Soherey Devi, a Bihari
woman of northern India:

First . I must wash the pots and sweep. Then I go to collect wood and cow
dung for fuel and grass for the bullock. If there is food I cook a mid-day
meal. . In the afternoon I must go again to collect grass for the bullocks, and
then if there is food I prepare the evening meal. If I am needed I work in the
fields too. I must plant the paddy, spread the fertilizer, turn over the earth around
the maize, and help in the harvest (Stokes 1975:219).

Would Soherey Devi be counted as a member of the agricultural labor
force? This depends, among other considerations, on the particular definition
of economic activity used, on whether she is asked about her secondary as well
as primary occupation, on the timing of the survey during the agricultural cycle
and the length of the reference period, and on whether the interviewer actually
asks her about what she does rather than assuming that she is "just a house-

What is economic activity?
According to the United Nations, economically active persons are "all the
persons of either sex who furnish the supply of labor for the production of
goods and services during the time-reference period chosen for the investiga-
tion" (Blacker 1978:47).' Planting, cultivation, and harvesting in anticipation
of profit, preparation of products for sale, care of livestock, and repair of farm
equipment are included; excluded are "household duties such as the prepara-

542 Women in Agriculture: Counting the Labor Force in Developing Countries

tion of food and the care of chickens and other livestock which are used for
consumption instead of for exchange" (Boulding 1976:318).
Despite UN recommendations, the definition of economic activity varies
greatly from country to country; and it is in the classification of female work
that the application of various standards seems most capricious. Countries such
as Turkey and Thailand count almost all farm women as unpaid family helpers,
resulting in an enumerated agricultural labor force that is half female. But if
unpaid workers are excluded, women are reported as only 5 percent of Turkish
farm labor and about 20 percent in Thailand. Other countries apply the defini-
tion of economic activity selectively so that female (but not male) unpaid fam-
ily helpers are excluded. This practice explains the wide fluctuations in the
reported female labor force in three North African countries. The 1966 census
of Tunisia, specifically excluding 250,000 female family helpers, reported a
farm labor force that was only 2 percent female. Added back in, women be-
come 38 percent of the total in 1966, the same proportion as was reported in the
1956 census. Similar observations can be made for Algeria and Morocco.
Between these two extremes, whether women in farm families are de-
fined as economically active or inactive depends in part on a necessarily arbi-
trary decision as to where the household ends and the farm begins, in spatial
terms, or where housework ends and production begins, in economic terms.
How, for example, does one classify such borderline activities as collecting
water and fuel, tending kitchen gardens, processing and storing crops, or feed-
ing chickens? (In one effort at clarification, a Fiji survey specified that tending
more than ten chickens was an economic activity, but fewer was housework
[see Blacker 1978:48].) Why, critics ask, should the production of raw
foodstuffs be regarded as economic activity while their preparation and cook-
ing, which is necessary for their consumption, is not?

Let us take, for example, the chain of processes leading to the production of a
loaf of bread: the harvesting of the wheat, the threshing and winnowing of the
grain, the milling or pounding of the grain into flour, the kneading of the flour
into dough, and the baking of the dough into bread. Where, it may be asked, in
this series of actions does economic activity begin and end? I suggest that in
practice the answer is determined not by the intrinsic nature of the operation but
by the point at which it is performed by "housewives"-i.e., by female unpaid
family workers (Blacker 1980:72).

One solution to the dilemma created by the artificial statistical dicho-
tomy imposed upon a natural continuum would be to classify as economic
production all household work such as cooking, sewing, and child care, recog-
nizing its function of reproducing the labor force and of substituting for goods
and services that might otherwise be purchased (e.g., prepared foods, ready-
made clothing, hired housekeepers). Although this solution would satisfy those
who wish to see the value of women's domestic production recognized in na-
tional accounts statistics, recourse to it would guarantee that virtually all able-
bodied women (and men and most children above a certain age) would be
economically active, with little if any variation over time and space. Important

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs