Introduction
The birth rate in Okinawa Prefecture has continued to be Japan's highest, although@it has been subject to the same national tendency towards decrease. The factors behind@this, however, appear to have seldom been studiedl). Prior studies in ethnology, etc.have confirmed that in the southern @part of Okinawa Honto (main island of Okinawa) , a f~Lmily is structured along the father-eldest son line as its axis, and the patrilineal family prevails. It has been viewed that this indigenously Okinawan family formation norm may influence marriage and fertility behaviors and hence the birth rate. This study attempts to analysize Okinawa's high birth rate mainly through gender-preference indexes.
Analytical Framework
The level of fertility in Okinawa is high compared with other prefectures. Although the differences among prefectures tend to be reduced over time, the total fertility rate (TFR) in Okinawa has always been the highest since 1973 when Okinawa was returned to Japan. The TFR in 1993 in Okinawa is 0.5 point higher than the national average (Figure 1), and, thus, there is a large difference with many other prefectures. Compared with the birth rate by age of mother for the whole Japan, the figures for Okinawa are different at the ages 30 and above (Figure 2) .
We can consider that the norm abcut family formation in this area has some connections with this difference. Families in this area are composed of the patrilineal descent group called "Munch~ (sib) " . "Munch~" which has the family formation norms with a son preference is often
found in the South of the Okinawa Island. But such norms are generally found in atherparts of Okinawa with some variations.
First, we would like to examine the relationships of the marriage type and the birth@to such norms of family formation. It is easy to understand them if we compare the "Munchti" with the "D5zoku" which were found in many other areas of Japan'). Thus, we explain the basic principles
of the two family systems affecting demographic behaviors.
"D6zoku" : This is based on the patrilineal descent, but the biological filiation is not necessarily regarded as important. Mutual recognition is needed, but the blood relationship is not always necessary. The first priority is given to the continuation of "le (the traditional Japanese family) " . "Munch~" : Patrilineal (Male) descent line takes priority and rigid adherence to the biological relationship is found. Only the successor of the family is allowed to have "To-to-me (a mortuary tablet)", who should have been designated as such (Figure 3) . The biological continuation of patrilineal descent is regarded as more impor-tant than the succession of the "family" .basic principles of family formation suggest ?.What do the differences between the two And how do they influence the demographic behaviors ' In the case of "D5zoku" systems, giving priority to the succession of the "fam-ily" , it is normal for the parents without any daughters to adopt a son. In Okinawa where male descent line is the basic principle, daugh-ters cannot succeed the family andthey marry out. If the couples have no sons, they will adopt a son of the htisband's brother. But under this norm women were expected to give birthto at least one son and, still now, women aged 30 and above are expected tc do so and feel some pressure for
that.
With regard to the effect of the family formation norms on demographic behaviors, the following four hypotheses are proposed. H I : The patterns of marriage regarding the postnuptial residence will show the differences. In
Okinawa where the male descent line is important and daughters cannot succeed their family, the proportion of the uxarilocal marriages in which newly married couples live with the bride's family may be low and, in contrast, the proportion of virilocal and neolocal marriages might be higher than in other prefectures where the succession of the family is highly regarded. H 2 : If Okinawans have a strong social norm to have at least one son, it can be considered that women in Okinawa continue to give births until they can have a son. Thus, the number of children increase. H 3 : Sex ratio at birth is biologically fixed, and in our modern society where the human rights
are respected we cannot artificially select the sex of a child. But the couple can make a choice about the number of children considering the sex composition of their children. If they want to have a son at all costs, "the youngest child's sex ratio" for women who have finished child-bearing will be higher than the normal value. H 4 : As a result of the H 2 and H 3, the sex composition of their children will be a criterion for deciding whether they will have another child or not. Thus, the parity progression ratio will differ by the sex composition of children. Therefore, if the hypotheses, H 1-H4 are empirically supported, we can ascertain that the family formation norms in Okinawa have a large effect on the demographic behaviors and that the sex-selective fertility behavior leads to a high level of fertility in Okinawa. We can also confirm that the category of regions with the same family formation norm constitutes a region -unit for the analysis cf the fertility transition. The family formation norms can be the taol for an efficient analysis of factors of fertility transition.
Data and Methods
Data
l. "Fertility Survey in Selected Rural Areas in 1979", Institute of Population Problems ; the data for 210 couples and their 875 children in the South area of Okinawa Island (plural deliveries and foetal deaths are excluded) .
2. "Vital Statistics". Department of Statistics and Information. Ministry of Health and Welfare.
Method
As for the Hypotheses 1, because we cannot get such statistics at present, we will confirm the effect of family formation norms on the marriage patterns by the data on the number the uxorilocal marriages.
To test the H 1, the number of marriages by type of postnuptial residence (virilocal, uxorilocal or neolocaD can be used as an index. Although we collect no information on the type of marriage in postwar vital statistics, we can make use of the data from prewar vital statistics for 1904-1938. The H 2, H 3 and H 4 will be examined using the survey data. The H 2 will be examined using the survey data on the number of the children ever-born and the ideal number of the children by age among married women. As for the H 3, the level of "the sex ratio of the youngest child" can be used to ascertain whether the couples had decided to finish childbearing by the last child's gender. The H 4 will be tested using the fertility behavior indices about the preference for the child's sex ; for example, the parity progressian ratios by children's sex composition. And if the parity progression ratias are different by the sex combination of children, it is supposed that their
parents have a special preference for their child's sex')
Results
As for the marriage patterns regarding H 1, the following results are obtained from Figures 4
and 5.
Uxorilocal marriage rate for
the whole Japan gradually de-
clined while that for Okinawa
remained at the lowest level
among 47 prefectures during the
35 years between 1904 and 1938.
Such social norms regarding
postnuptial residence remains
more powerful in Okinawa than
in areas with the "D6zoku" sys-
tem. Although there were some
changes caused by the revision
of Civil Code after World War
II , the regional differences in
marriage patterns have rarely
been examined. In Okinawa,
uxorilocal marriages have rarely
been found. Instead, the viri-
10cal marriage rate is the high-
est in Japan. This is ane af the
effects the family formation norms about the marriage pattern. In fact, I was convinced of the
fact that the uxorilocal marriage avoided as much as possible during one interview survey in K
town in the South of Okinawa in 1992.
Concerning the H 2, H 3 and H 4 about the fertility behavior, I got the following results :
a ) Among married women aged 35 and above, the ideal and the actual numbers of children are
approximately the same in Okinawa. This means that they rationalized their actual behavior, but
the ideal number of children among married women aged 20-34 is high. Thus, in comparison with
the Japanese average the norm about the number of children had a great effect on fertility
behavior (Table 1).
b ) The sex ratio at birth is normaly 105-106. In the case of Okinawa samples the sex ratio for
the total (870 children) is 105 and it means normal. The youngest child's sex ratio of married
women aged 40-49 (106 couples) who finished childbearing is "I41" which is higher than that of
those aged 30-39 (107) . This means that they wanted a boy for their youngest child.
c ) If there is a social norm to have many children in Okinawa, the proportion of married
women having only daughters will be reduced. But out of 210 married women, the number of those with the combination of both male and female children is 91. Among 106 married
women aged 40-49 those with the combination is 76, and only two of them have only@daughters. One of them gave birth to seven daughters, and another one still desired to have another child. 104 out of 106 women have at least one son.
d ) Because the special desire for sons can be seen in the fertility behavior discussed in
c) . I am going to examine the parity progression ratios by sex cornposition of children in Okinawa. About 60% of married women aged 40-49 had more than 6 children. Accordingly, in the case of the progression from the first birth to the second, the norm about the number of children has a greater effect than the sex preference for children. Thus, the sex of the first child does not influence the next birth. Table 2 shows the parity
progression ratios in relative terms by the sex
composition of children for the parity of two to four by age of married women. Among those with the parity of two, the proportion progressing to the third differs by the sex composition in the following order : Iflf>Imlf> Imlm (m and f mean male and female).
As for the progression from the third to the fourth, 3f> Im2f> 3m> 2mlf. As for the progression from the fourth to the fifth, 4f> Im3f> 4m> 3mlf> 2m2f>. In the case of 2m, 3m and 4m combinations, those women progress to the next parity due to the preference for balanced sex composition. But in the 2f, 3f and 4f cases, those women want a son, and the proportion of giving birth to the next child is 100~6. There is a very high probability that they will continue giving birth until they get a son. Those women with many children but only one son are also likely to continue giving birth to have their second son. It is considered as an ideal behavior to continue giving birth until they have at least two sons. These indices show that the son preference is stronger than the preference for balanced sex composition of children, and that the norm for a large family is little or only potentially affected by the preference for balanced sex composition. Throughout the process of family formation the fertility behavior is guided by son preference.
Summary and Conclusion
In the South area of Okinawa Island, the norm about the family size is originallyset at a higher level than in Japan as a whole. To add to this, the principle of patrilineal descent have a pronatalist effect on fertility behavior. In other words, when Okinawa women have their second and the subsequent births, they show the behavior suggestive of the son preference. But, at first, the family size norm precedes the son preference and hey have some children any way. In this process the following four patterns are found.
a ) The couples which have already at least two sons (including two sons only) reveal the preference for balanced sex composition of children and they continue childbearing hoping for a
daughter within the limit of their ideal number of children.
b ) The couples who have balanced sex composition of children (at least one child each of both sexes) will continue giving birth within the limits of ideal family size determined by the norm.
c ) The couples which have only daughters will continue giving birth until they get a son.
d ) In contrast to the case b) , there are many couples which have one son and some daughters but continue giving birth for another son. In this case the couples may expect to have more children than the ideal. At the final stage of childbearing "the coordinative mechanism" to change the sex composition of children is often operative in the direction to have more sons. This is the reason why the sex ratio of the youngest child among married women aged 40-49 tends to be higher than 105. Before they finish childbearing, they would like to adjust the sex combination of their children. Fertility level cf married women in Okinawa is especially high in the ages 30 and
above compared with the average for Japan as a whole. In Okinawa where the norm about family formation is based on the principle of patrilineal descent, we can clarify the distinctive fertility behavior : differentiated by the sex composition of
children . Through all the process of family formation the fertility behavior guided by the son preference is found. Especially at the finaJ stage of childbearing, the coordinative mechanism gives priority to the sons preference over the family size norm, resulting in a large number of children than the
ideal. As a consequence, this mechanism has contributed to the high level of fertility in this area. In conclusion, the family formation norms found in Okinawa have a very large effect on the demographic behaviors including the distinctive patterns of fertility determined by the sex composition of children as well as the high level of fertility. The category of the regions where the same family formation norms are found can be used as a unit of analysis for determinants of fertility transition. The set of family formation norms as a cultural factor can be an efficient
tool for the analysis of fertility transition.