Dependent variables

This study has four general dependent variables: mortality, marital fertility, first marriage, and out-migration. In the analysis of mortality, the dependent variable is a dichotomous variable measuring whether an individual died or not within the next one year from an NAC registration to the immediately succeeding registration.’⁷ Out of the 6,543 individuals appearing in the population registers in the two villages, there were 2,839 deaths (’,479 male and ’,360 female deaths) recorded in the NAC registers.

ages 55 to 74—we estimate the logistic regression model for males and females separately.

In the analysis of marital fertility, a ‘birth’ is a birth recorded in the NAC register, rather than an actual birth. Hence, we analyse, strictly speaking, not marital fertility but marital reproduction. Our dependent variable is a dichotomous variable indicating whether a married woman of reproductive age (15—49) records a birth within the next one year (until the subsequent NAC registration).¹⁸ To 2,127 women aged 15—49 appearing in the population registers in the two villages, there were 2,620 births (1,346 male births and 1,274 female births) recorded and matched. Given the possibility of sex differentials in infant mortality as well as of sex-selective infanticide, our analysis of marital fertility also differentiates births by their sex.

Providing that the focus of our analysis is on marital fertility/reproduction, we need to clarify the definition of a woman ‘at risk’ of having a (recorded) birth within marriage. We restrict the analysis to married women who were alive and present in the village throughout a ‘current’ year (one year until next registration), and had their husbands residing in the village at the beginning of the current or previous year.¹⁹ Put differently, we exclude women who died during a current year and those who did not have their husbands living in the village from the beginning of the previous year.

Turning to the analysis of first marriage, our dependent variable is a dichotomous variable measuring whether or not a never-married individual experiences first marriage within the next one year. Being influenced by local customs, marriage and its registration in pre-industrial Japan were largely contextual; the timing (year) of first marriage therefore had to be inferred from an entry of a new household member between two consecutive registrations and concomitant changes in relationships among household members. Hence, we measured the timing of first marriage by comparing household members between two consecutive registrations and, if changes occurred, taking the difference between the year of birth and the year of marriage as defined above.

Measurement of the timing of first marriage is straightforward for individuals who were under constant observation from birth. However, a problem arises for those who had already been present when the records began, or for those who first appeared in the records some time after birth. Many women especially moved into or out of the villages due to marriage and marital disruption whereas only a small proportion of them remained in their native villages throughout their lifetimes (Tsuya 2000). Given the high geographical mobility of women associated with marriage, it is unwise to limit the analysis of first marriage to women who lived in the villages throughout their adult years, for doing so renders the data too small and selective. Thus, we used a less conservative definition of ‘first marriage’ than the usual: if marriages were observed for the first time for individuals who were under age 50 and had neither a spouse nor a child listed when they first appeared in the registers, those marriages were regarded as ‘first marriages’.

Our analysis of first marriage looks at ages 10—49 for males and 5—49 for females.²⁰ Our multivariate analysis is also confined to individuals who were

residing in the villages prior to their ‘first marriage' as defined above.²’ The data for our analysis of first marriage consists of ’,’69 single males and ’,072 single females who were at risk of experiencing first marriage, with 805 recorded male first marriages and 8’4 female first marriages.

Furthermore, our analysis of first marriage distinguishes three types of marriages based on individual movements annotated in the NAC records: intra-village virilocal marriages (called ‘virilocal marriages' hereafter); intra-village uxorilocal marriages (called ‘uxorilocal marriages'); and marriages accompanying out-migration from the villages (called ‘marry-out' marriages).

Finally, with regard to the analysis of out-migration, our dependent variable is a dichotomous variable indicating whether or not a resident of Shimomoriya or Niita moved out of the village within the next one year.²² In our analysis, out-migration refers to the observed movement of any persons (both legal and other residents) out of the villages to other communities. Thus, ‘out-migration' includes both the movements of individuals whose legal domicile was in one of the two villages and the return migration of persons whose legal domicile was elsewhere. To 6,543 individuals who appeared in the population registers in the two villages, 4,230 out-migrations (2,489 male and ’,74’ female) were recorded.

The NAC registers in Shimomoriya and Niita annotated in detail information on people's movements across the village boundaries, including: (’) the name of destining village and the name of the household head of destination for out- migrants and those of origin for in-migrants; and (2) the reasons for migration. Based on the first piece of information, we can differentiate the out-migration of ‘natives', individuals whose legal domicile was in one of the two villages, from the migration of ‘non-natives' whose legal domicile was elsewhere. We therefore examine migration of natives and non-natives separately.

Using the information on reasons for migration, out-migration of natives can be further differentiated into movements due to: marriage, adoption, service, change of legal domicile, absconding (i.e. illegal disappearance from the village), and other.²³ Among these, service (hoko) is by far the most common reason for male out-migration with 68% of such movements being service-related.

their reasons were mostly marriage (41%) or service (35%), with absconding constituting the third-largest category (12%). Our analysis of out-migration of native females therefore is divided into these three types of movements. To avoid the estimation bias caused by increasingly selective populations who survived to very old age, we restrict our analysis of out-migration to individuals under age 75.

4.2