Mortality responses to rice price fluctuations and household factors in a farming village in central Tokugawa Japan

Abstract

Drawing data from the annual local population registers called shumon-aratame-cho (SAC), this study examines the patterns and covariates of mortality by sex and life stage in the farming village of Nishijo in central Japan in 1773–1869. The discrete-time event history analysis of mortality in the village offers three major findings. First, mortality responses to rice price fluctuations varied greatly by residents' sex and life stage. Second, coresident kin were important for survival throughout the life course, though the magnitude and nature of the mortality effects of coresident kin differed across sex and age groups. Third, household resources and positions that individuals occupied within a household influenced chances of survival especially in adult and elderly years.

Keywords:

• Mortality
• Rice price
• Household factors
• Tokugawa Japan
• Discrete-time event history analysis

Acknowledgements

An earlier version of this article was presented at the 23rd Annual Meeting of the Social Science History Association, Chicago, November 19–22, 1998. The authors are grateful to Akira Hayami for being instrumental in making the original population registration data available. We also thank Satomi Kurosu, Shuma Morimoto, and Shoko Hirai for their assistance in creating the data file for multivariate analysis.

Notes

1 Saito (1999) reports that according to one source (Ogashima, 1894), there were 28 famines in the Tokugawa period, and according to another (Saito, 1966), there were 61 famines from 1600 to 1900.

2 According to Hayami (1972), this medical doctor was the son of the village mayor and largest landholder in Nishijo. Born in 1818 and named Risaburo, he was apprenticed to a doctor in the neighboring village of Suga in 1835. In 1839, Risaburo went to Kyoto for training in modern medicine. Returning to Nishijo in 1845, he formed his own household and practiced medicine in Nishijo until his death in 1863.

3 For details on the nature of rice prices and Tokugawa agrarian economy, see Feeney and Hamano (1990) and Tsuya and Kurosu (2000). For specifics of these five rice price series in the Nobi region, see Iwahashi (1981, pp. 207–212, 1988) and Yamasaki (1983, pp. 235–238).

4 Not all births and infant deaths were recorded in SAC registers. Only those children who survived from birth to the subsequent registration were entered into the registers. Consequently, many who died before the first registration after birth were excluded and never came under observation.

5 Out of 45 cases of mysterious disappearance from the SAC records in Nishijo, 64% (29 cases) occurred in the period from 1773 to 1789. The remaining cases of mysterious disappearances are scattered over the period from 1790 to 1869.

6 In addition to chronological age (i.e., age according to the Gregorian calendar) and SAC age, there is also the traditional Japanese method of counting age. As in the rest of East Asia, a child is considered 1-year-old at birth, and an additional year is added each New Year's Day thereafter. Consequently, if counted by the traditional Japanese method, most newborns, if they survived, appeared in population registers as aged 2 sai, even though they might only be in their second day of life. If population registration were conducted each New Year's Day (which was rarely the case), traditional Japanese age (in sai) minus one would equal SAC age.

7 For details of discrete-time event history analysis, see Allison (1984) and Yamaguchi (1991, Chapter 3). For details of logistic regression, see Aldrich and Nelson (1984) and Retherford and Choe (1993, Chapter 5). Campbell and Lee (1996) conducted a similar discrete-time event history analysis of mortality in a rural area in Liaoning Province during the Qing Dynasty in China using Eight Banner household registers. Bengtsson (1993, 1995) was the first to bring together the life cycle of demographic events within households with short-run variations in economic conditions.

8 According to Guilkey and Murphy (1993), when records are repeated over five times, the problem of intercorrelation of observations becomes serious and affects the estimation results.

9 The formula was independently discovered by White (1980) and is known in the econometrics literature as White's method.

10 Putting it differently, we can also consider “time” to be an individual's age in his/her life course.

11 Assuming that the effects of rice price variations may differ according to household and individual characteristics, we also estimated the model by including interactional terms between rice price and other covariates. However, none of these terms was found significant, and the explanatory power of the model did not improve significantly by inclusion of an interactional term. (In fact, it was often reduced.)

12 The cut-off ages of 30 and 40 for females and males, respectively, were chosen because a large majority of the residents were married by those ages. If we were to impose younger ages as cut-off points, we would introduce a serious early marriage bias, since those who appeared after that age would be considered as “marital status unknown.”

13 Life expectancy at age 1 (e ₁ ) by sex in other Tokugawa villages estimated by previous studies is listed below. (The figures for Hida villages are the unweighted averages of estimates by Jannetta and Preston (1991) for 20-year periods; those for Fujito and Nishikata are the unweighted averages of estimates by Hanley (1974) for 5-year periods.)

Table