Conclusions and Discussion

There exists an extensive literature on the effects of short-term economic stress, usually measured by fluctuations in grain prices, on vital rates in pre-industrial Europe. According to Galloway (1988), the strength of the link between grain price fluctuations and mortality seems to vary across societies and periods, whereas fertility tended to be highly sensitive to grain price variations throughout most of pre-industrial Europe.

Our study of two agrarian villages in the northeast of Japan based on micro-level multivariate analysis offers new insights on demographic responses to short-term economic stress in pre-industrial rural Japan. We found that first marriage and

out-migration were most responsive to short-term economic stress, though the magnitude and nature as well as the type and reason of the responses varied between these two demographic behaviours.

The probability of first marriage for both men and women decreased significantly in times of economic stress, and this was especially the case in the nineteenth century. Further, the nuptiality response to short-term economic stress was not immediate, but had a time lag of 2—4 years. Among different types of first marriages, virilocal marriages were most sensitive to economic stress, with some time lag. This suggests that virilocal marriages, in which a man brought the bride into his household, were sensitive to local economic conditions. This also implies that in virilocal marriages, local economic conditions a few years prior to the marriage mattered most, probably because the decision-making power regarding inheriting sons' marriages in pre-industrial Japanese agrarian households was in the hands of their parents/ families and planning for such a family matter was long term, starting a few years prior to actual marriage.

Migration was also highly sensitive to changes in local economic conditions, and this was especially true for male out­migration. The probability of male out-migration declined significantly when local economic conditions deteriorated. The response of male out-migration to short-term economic stress was immediate and lasted for a few years. Male out-migration declined under economic stress primarily because a large proportion of it was labour migration. Since a large part of labour migration was destined to other localities within the Nihonmatsu domain, the deterioration of the local economic conditions meant reduced labour opportunities in the domain as a whole, thus decreasing opportunities for out-migration for employment. Looking at out-migration by motivation, service migration and absconding both appear to have been highly sensitive to short-term economic stress, but the directions of the effects were opposite: when local economic conditions worsened, service migration decreased but absconding increased, and this was true for both sexes.

Mortality responses to short-term economic stress were limited mostly to two gendered-life-stage groups: females during infancy and childhood, and men in adult and elderly years. The responsiveness of mortality of baby girls was especially high and immediate. Moreover, at times of economic hardship, the survival of baby girls also depended on the resources and wealth of their households. When local economic conditions deteriorated, the mortality of female infants was significantly reduced if they lived in wealthy households. The mortality of adult and elderly men was also sensitive to economic stress, and this was especially true for elderly men.

As to why mortality responses were largely limited to opposite sexes at the two ends of the life cycle—female infants and elderly men—we speculate that the economic and social sensitivity of their mortality in the context of physiological robustness may constitute part of the explanation. Being physiologically more robust than male infants, female infants were nonetheless socially and economically

the most vulnerable sub-population in pre-industrial rural Japan.

We have also found that household resources and wealth influenced all four demographic behaviours under consideration, although the nature, direction, and magnitude of the demographic effects differ by sex, life stage, and type of behaviour. Migration was strongly influenced by household landholding, while the responses of first marriage, mortality, and marital fertility to landholding were more limited. Household landholding significantly and consistently lowered the likelihood of out-migration of both sexes, and this was the case regarding migration for all reasons: labour migration, migration due to marriage, and absconding.

The probability of first marriage among inheriting sons and daughters increased significantly with their households' landholding, although the degree of increase was not always linear. As virilocal marriages constituted a large majority of all male marriages, male first marriages were generally sensitive to household landholding. On the other hand, uxorilocal marriages being only a minority, female first marriages as a whole were not as responsive to landholding. Because marriages in pre-industrial Japan were largely family and household enterprises (Tsuya and Kurosu 2000⅛), it is understandable that household resources and wealth, as measured by landholding, increased the likelihood of the first marriages of inheriting sons and daughters.

Household landholding also increased the probability for a married woman to have a recorded female birth (but not a male birth). This implies that those women living in households with more resources and wealth could afford to keep a baby girl without resorting to infanticide.

This study also discerned temporal trends and differentials in some of the four demographic behaviours under consideration. Especially notable were changes that occurred during the last sub-period of 1840—70 in female demographic behaviours. During this sub-period, female child mortality declined sharply, the likelihood of a recorded female birth increased precipitously, and female age at first marriage, while increasing steadily, went up especially notably. As a result, women's mean age at first marriage increased from 12.5 years in 1716—59 to 16.5 years in 1840—70. Altogether, these findings provide yet another piece of demographic evidence to support the general discussion of the improvements in female socio-demographic status that occurred during the last decades of Tokugawa Japan (Hayami and Kito 2004).

The relationships between short-term economic stress and demographic behaviours in pre-industrial societies are complex, dynamic, and situational. Clearly, more studies are needed to account for the dynamic mechanisms underlying these relationships. Our study on two agrarian villages in northeastern Tokugawa Japan offers some observations that might help to clarify some of this complexity. Our next agenda is to analyse these relationships in comparative perspective, by taking individual demographic, household, and life-course characteristics into account.