Thank You, Akira Hayami! The Xavier Database of Historical Japan

e-ISSN: 2352-6343 DOI article: https://doi.org/10.51964/hlcs11113 © 2021, Kurosu, Takahashi, Dong This open-access work is licensed under a Creative Commons Attribution 4.0 International License, which permits use, reproduction & distribution in any medium for non-commercial purposes, provided the original author(s) and source are given credit. See http://creativecommons.org/licenses/.

"Thank you, Francisco Xavier!" wrote Akira Hayami (1928Hayami ( -2019, the founder of Japanese historical demography in 1979. Without St. Francisco de Xavier who first brought Christianity to Japan in 1549, the Shumon-Aratame-Cho (SAC; religious affiliation investigation registers) would never have started. SAC was the "chance by-product of the Tokugawa Bakufu's fear and loathing of Christianity" (Hayami, 1979) which serves as the main source for Japanese historical demography. This article introduces the source, the construction, as well as the impact of the first digitized data set in Japanese historical demography. The data set was named after Xavier who was one of the founders of the Society of Jesus. The construction of individual-level longitudinal data based on household registers, SAC and similar Ninbetsu-Aratame-Cho (NAC) for early modern Japan has opened up arrays of possibilities for investigating the demographic behavior of commoners in the Tokugawa period . Japanese historical demography has come a long way since Akira Hayami's application of the method of "family reconstitution" to Japanese household registers in the late 1960s. Hayami's lifetime collection of materials on historical demography from his earlier offices including Keio University, International Research Center for Japanese Studies (Nichibunken), and Reitaku Tokyo Center, are now hosted as the "Reitaku Archives" at Reitaku University, organized and maintained by the Population and Family History Project (PFHP) headed by Satomi Kurosu. This article first introduces the sources used for the Xavier Database and the studies of Japanese historical demography in general. Particular attention is paid to how Hayami's collection of historical records, those which he called "treasure of humankind", were transcribed and put into a format with linked annual individual/household information. Next, we discuss the process of coding and construction of the Xavier Database, the first systematic database Hayami initiated in late 1980s. While there are numerous datasets and sources available across Japan by now, the sources for northeastern communities used for the Xavier Database are known to be the most detailed. We review the Xavier data, focusing on three villages (Niita, Shimomoriya, Hidenoyama) and one town (Koriyama) in the current Fukushima prefecture. Finally, we discuss how the studies based on the dataset advanced our understanding of people's lives through the analysis of behavior and organization of individuals, married couples, and households, as well as their influence on the studies of historical demography and family history.
Two primary sources used for the Xavier Database as well as studies of historical demography in Japan are SAC and NAC. 1 Both of them are household registers and include basic information such as an individual's name, age, and relationship to the household head. A household can include not only kin members but also non-kin and servants. Notably, households can be registered even without any resident (e.g., actual residents left for work migration).
SAC was a religious investigation initiated around 1638 by the Tokugawa government as a measure to prevent the entry and spread of Christianity (Cornell & Hayami, 1986;Hayami, 1979). The quality, dates of compilation, and availability of SAC vary depending on the village and region. Some SACs are based on de jure information and contain excess numbers of elderly persons, who, for example, outmigrated and possibly died elsewhere but are still listed in their households of origin. NAC preceded SAC and was a type of population register in nature similar to the contents of SAC, except that NAC excludes information on religious affiliation of individual villagers (Hayami, 2001). Meanwhile, NAC tends to record detailed information on population de facto. However, in most cases, it is not easy nor practical to distinguish between the two because the two sources are often indistinguishable and are indiscriminately compiled resulting in the title of the documents Shumon-Ninbetsu-Aratame-Cho.
It should be noted that unlike records of elites in other societies, both SAC and NAC documented "commoners". They include peasants, fishermen, merchants, etc., who were the majority of the population in 17th-to early 19th-century Japan. Tokugawa period was a highly stratified society. The elite bureaucrats and administrators during the Tokugawa period (samurai class), as well as the members of the Imperial court whose residency was segregated, were recorded separately from the commoners. 2 The documents available today are copies of the SAC/NAC, kept in the hands of village officials after submitting them to local lords. The officials kept the register copies in order to add annotations for changes of individual vital events (e.g., birth, death, marriage, service) that would occur until the next survey. These annotations provide valuable information of vital records. However, the level of details in such registers differs by local government practices. For some han (domains, administrative units governed by daimyo, territorial lords), SAC was not always done every year. For other domains, only those after certain ages were registered (e.g., after age 15 for Maeda domain, after age 8 for Kishu and Hiroshima domains). The more detailed listings include origins and destinations of migrants with reasons (e.g., marriage, adoption, service), as well as household landholdings. According to Hayami (2001, p. 25), some of the best sources, in terms of quality and length (continuing more than one century with very few years missing in-between) come from the villages in Nihonmatsu domain (in current Fukushima prefecture), and this is where the very first attempt to systematically construct a database was made.

Figure 1
Original "Ninbetsu-    Figure 2). Hayami recalled that it was an "innovation" as it finally allowed researchers to track what was happening to the individuals and households "longitudinally". And, indeed, BDS was a breakthrough as a method for compiling detailed longitudinal information. It also opened up the opportunity for anyone with knowledge of contemporary Japanese to understand the content of these household registers, which otherwise would be hard to decipher without extensive training in early modern calligraphy. The annual information of a household member (name, relationship, age) is transcribed on the left panel of a BDS. Any movements or changes of status, including birth, death, marriage/adoption, and name changes, are annotated with a symbol and described at the bottom of the BDS. Any household members currently not residing in the household are placed in the right panel of the BDS with information of their locations and the reasons for not residing (for example, working as servant at other village/household). At the very right column of a BDS, the number of horses and the landholdings of the household are entered. SAC/NAC of the same household in the consecutive years are matched and transcribed following the previous year. Organized in this way, both the information of individuals and households can be tracked for the entire observation period.
The linkage of households from one year to the next was done by matching the order of appearance of the households and checking names and ages of household members. Once the BDS was written, individuals were given unique identifiers. When the linking of individual information was done manually this was done by coders; for example, matching and giving the same ID to an individual who left the household of origin and entered another household via marriage/service. Individuals were matched as long as the move took place within the village. Thus, BDS became a valuable source of longitudinal demographic information that follows intriguing life histories of commoners -like reading a biography.
Because it was before the computer revolution, most of Hayami's initial work was done manually extracting information from BDS to construct individual life courses and family. The ITS (individual tracing sheet, see Figure 3) was Hayami's invention for tracking individual movements. And, the FRF (Figure 4) follows the famous "family reconstitution form" of the French historical demographer, Louis Henry. These forms are still kept at the Reitaku Archives, although they are taken over by computer calculation and are no longer used. While Hayami has been transcribing and organizing materials that he and his group collected all over Japan, his research concentrated in the areas of Suwa (Hayami, 1973) and Nobi (Hayami, 1992) -central Japan. He rarely used BDS of the Nihonmatsu domain. Instead, Saeko Narimatsu singlehandedly transcribed the original NACs to BDS. With the help of research assistants, she linked individual information from SAC/NAC to BDS and further cross-checked qualitative information of the area from local documents. Based on this laborious work, she published two books on Shimomoriya (Narimatsu, 1985) and Niita (Narimatsu, 1992), which are must-reads for later researchers studying the area. Map 1 shows the original data collection and ongoing database construction efforts spreading nationwide in Japan, as well as the area of the Xavier Database introduced in this article. The steps described above are also shown in Table 1 in relation to the figures and maps. The Xavier Database is one of the first series of Japanese historical demography datasets constructed by Akira Hayami and the members of the Eurasia Project Japan (EAP-J 1995(EAP-J -2000 3 and is the most detailed and vigorously used in the last decades. 4 Details of Xavier construction are discussed in the next section.

COLLECTION AND TRANSCRIPTION OF REGISTERS
Meanwhile, there are numerous village records entered during and after the time of EAP-J for both longitudinal (red dots in Map 1) and cross-sectional data (orange dots in Map 1) types used in historical demography and family studies. Also, there are plenty of understudied BDS (green dots in Map 1) and microfilms (blue dots in Map 1), now hosted at the PFHP. Since longitudinal data from various locations were not in the same format as the data of the Xavier Database, they were not integrated into one database during the time of EAP-J. Cross-sectional data include those with only one-or twoyear SAC records. While those cross-sectional data have not been utilized for longitudinal research in Japanese historical demography, it has a potential for investigating regional variations as well as fertility dynamics (e.g., Drixler, 2013;Kurosu, 2008 (Hayami 2001, pp. 74-79).

Map 1 Xavier Data and Collections of Japanese Historical Demography Sources
Source: Reitaku Archives, Population and Family History Project, Reitaku University.
Note: See Table 1 for the explanation of the colored dots. Table 1 The

different steps of database construction in relation to the map and figures
Step Notes: While steps 1-4 were applied to make forms from various sources that remain to digitize, steps 5 and 6A are specific to our construction of the Xavier Database, the subject of this article. Basic statistics of the Xavier Database are reported in Table 2. (Kurosu, 2020).

As of now, Reitaku Archives include microfilmed or paper-copied documents of original SAC/NAC for about 1,870 villages/towns (Step 2). Among them, about 470 villages/towns (about 9960 village/townyears) transcribed in the format of BDS (Step 3). A meta-database of these materials has recently been made available online (limited use) to search the geographic location and type of historical records held at PFHP
No official documentation exists as to what motivated Akira Hayami to initiate and how he pursued the construction of the Xavier Database. However, according to the detailed and yet complicated codebooks, as well as the memories of staff members who worked on the project over decades, the digitization was done with a series of trials and errors. What started in the 1980s remains an evolving process. In this section, we explain how the Xavier Database was transcribed and coded, turned into relational databases (DB2) for separate villages, and finally utilized for variable construction and analysis.
The Xavier data are based on household registers, both NAC and SAC in the northeast Japan. While this article mainly discusses four communities of Nihonmatsu domain (NACs), the Xavier Database also includes villages of the eastern-mountainous area of Fukushima 5 , Aizu, and the village Yanbe in the current Yamagata prefecture (SACs). Numerous studies used the BDS (step 4 in Table 1) of Aizu and Yanbe with some researchers' own data construction efforts (e.g., Hayami & Okada, 2005;Kinoshita, 2002;Okada, 2006). Also, since the NAC of Koriyama-shimo-machi, the northern town of Koriyama, has too many missing years for longitudinal research, we only include Koriyama-kami-machi, the southern town, in this article.

5
There is yet another village, Sasahara (NAC) in Nihonmatsu. As the size of the village was very small, it was conveniently used to test SQL runs but is not included in this article.

CONSTRUCTION OF THE XAVIER DATABASE
The first trial for the construction of a large longitudinal dataset started in the 1980s. Based on BDS, coders extracted information and organized them by handwriting into three paper forms, named A5, A6 and A7: household events and information (A5), individual age (A6), and individual events and life course (A7, see Figure 5)

CODING AND DATA ENTRY
The codes in each table were elaborate and complicated. The origin of the Xavier data predated easy access to computers and even preceded the Japanese word processor for data entry (Ono, 1993). It was vital to convert the information written in BDS into numbers. Therefore, it required detailed codebooks. For example, there are more than 260 codes for the relationship to the household head A7(i) distinguishing kin relations, sex, blood or marriage/adoption relationship. Only trained coders could manually, using paper and pencil, convert detailed information of BDS into thousands of sheets that consist of sheer numbers. While the codes made data entry faster, the steps involved in coding cost enormous time and manpower, as well as an increasing number of errors as more steps are required (Ono, 1993).
The forms A5, A6 and A7 were then digitized manually using a special computer program. As of today, the coding of 7 villages and 2 towns has been done 6 but the completion of the data entry was accomplished only for the four communities discussed in this article while the rest was, unfortunately, only partly completed.
The database construction that started in Keio University in the early 1980s was carried over to Nichibunken (Kyoto), with Hayami's move in the late 1980s. It was at this time when the Xavier data were finally put to use. Yoshihiko Ono, associate professor at Nichibuken at the time, converted almost all the files into a relational database separating tables into time-constant and time-dependent tables using DB2 on an IBM platform. With the help of an international EAP member, George Alter, a group of EAP-J members 7 headed by Yoshihiko Ono studied SQL and database management and started the construction of variables for international comparison. It was first tested on a small village, Sasahara, and then on the two villages of the Xavier Database (Shimomoriya and Niita) which became the basis for the Japanese contribution to the international collaboration of EAP (Tsuya & Kurosu, 2004, 2010c. Constructing variables for the EAP model required laborious processing. We used DB2 SQL to write long commands to construct each variable, export them to CSV files, and import and merge them in STATA for village-based analysis. This required advanced statistical and data management skills. It also required high levels of concentration and patience, as we had to go back-and-forth to the BDS and complicated codebooks whenever an inconsistency arose. The inconsistency could be due to the SQL program, data entry errors of coders, or data entry errors of BDS transcribers. It could even be due to the original documents -mistakes of village officials in Tokugawa period. With all these laborious and time-consuming processes, we were not able to add more than two villages during the entire EAP international collaboration (Kurosu, 2016). In the late 2000s, however, we started to add another village and town from the Xavier Database to our analysis: Hidenoyama and Koriyama. Table 2 shows a summary of the database for the four communities as well as the information available in those communities. Altogether they cover the span of 162 years with a relatively small number of missing years. In addition to its long coverage, the Xavier Database provides detailed information about demographic characteristics and reason for individual "entrance" under observation (due to birth or immigration) and the month/year of death and other "exits". This is the case even for Koriyama despite the heavy in-and out-migration of the town. Socioeconomic indicators, which are not often recorded in SAC in other places, are also abundant. The two Xavier villages (Niita and Shimomoriya), in particular, are among the best quality historical population panel data available in East Asia (Dong, Campbell, Kurosu, Yang, & Lee, 2015a).
That said, we need to be careful about one problem inherent in "annual" household registers: the omission of events that happened between registers. NACs of these communities were enumerated annually at the beginning of the third lunar month 8 , an important point to bear in mind for demographic calculations. The timing of marriage, for example, cannot be determined as clearly as many European

THE XAVIER DATABASE: POPULATION AND SETTINGS
parish registers. If one comes into a household with an annotation of enduke (marriage), the person and his/her partner married sometime between the two register years. An even more serious problem for certain demographic analysis is the omission due to infant death. Those who were born and died between the two registers may not be recorded.
Map 2 shows the location of the four communities. Shimomoriya and Niita were almost totally agricultural with an annual average population size of about 300 and 450, respectively. Hidenoyama, located about 3 km from the center of the growing market town of Koriyama, became a suburb to this town with an annual average population of about 280. Within the 160 years of observation, however, the area has gone through major famines in the 1780s and 1830s, as well as various epidemics and local disasters. The population trend in Figure 6 reflects the hardship the villages went through. In particular, the two agricultural villages, Shimomoriya and Niita, started to lose their population before the Tenmei famine in the 1780s and did not recover until the 19th century. Such constant decline is not observed in Hidenoyama. As for Koriyama, being a post town 9 with diverse economic activities, the local population increased over the years from 800 to about 2,600 inhabitants while commercial sectors developed. Koriyama was formally designated as a town (machi) in 1824 and served as the economic as well as the political center of Asaka County. While populations of neighboring villages stagnated, Koriyama experienced a stable population growth. Although it was affected by famines, the population of Koriyama soon recovered because of both in-migration and natural increase. Thus, the Xavier Database shows different population trends between Koriyama and neighboring villages, suggesting an interesting contrast of demographic dynamics between the rural and urban communities. Post towns, Shukuba-machi in Japanese, were constructed along the major routes/streets in the Tokugawa period. They provided lodgings for public officials, who were forced to periodically travel between their domain and Edo with their vassals; as well as rest for travelers, who were observed more frequently as traveling became more popular throughout the country during this period. Commercial sectors also developed in these towns and catered to the needs of commoners and neighboring villagers (Kurosu, Takahashi, & Dong, 2017).

Figure 6
Population trends of Niita, Shimomoriya, Hidenoyama, and Koriyama Many advances in the use of the data occurred during the Japanese Eurasia Project. Flat files for analysis based on the two villages, Niita and Shimomoriya, were first made specifically for the international comparison of mortality in EAP , and successively for marriage and reproduction Tsuya & Kurosu, 2010c). Japanese contributions to these volumes were based on Niita and Shimomoriya (Tsuya & Kurosu, 2004, 2010c. Modifying the models used in the international collaboration of EAP, Tsuya and Kurosu produced more detailed studies of mortality and reproduction (Tsuya & Kurosu, 2005, 2010b. These studies brought new findings about how individual demographic behaviors were influenced by socioeconomic status of households (indicated by landholdings), household context (measured by the presence of co-residing kin), and status in the household (head or closeness to household-head). Further, applying the EAP models but going beyond the topics covered by the EAP international comparison, specific topics for Japan were investigated utilizing the information available in the Xavier Database. These studies emphasized the hierarchy of gender and closeness to household-head in the stem family system; for example, the differential stress caused by the deaths of heads vs. fathers on women and children (Tsuya & Kurosu, 2002), household context and migration and survivorship (Tsuya & Kurosu, 2010a, leaving home, remarriage, divorce, and adoption (Kurosu, 2004(Kurosu, , 2007(Kurosu, , 2011(Kurosu, , 2013. These studies used event history analysis and revealed how the principle of the stem family mediated individual as well as household choices. Later on, similar models for mortality and marriage were tested with a larger sample by including the village of Hidenoyama (Tsuya & Kurosu, 2017, 2020. Further, including the town of Koriyama, we started to systematically compare patterns of marriage between village and town populations (Kurosu, Takahashi, & Dong, 2017). The EAP model was also tested with more rigorous statistical analysis to investigate the association between types of post-marital residence and reproduction . These studies demonstrated "northeastern characteristics" of marriage and reproduction influenced by the norm of "primogenitor" (succession by first child regardless of sex) that distinguished the area, going even beyond rural and urban differentials.
Researchers who shared knowledge and methods for working with the Xavier Database flourished in the Japanese Eurasia Project. They used various methods (both descriptive and statistical) to empirically tackle issues of family history which had been confined to institutional or household-based approaches.
Comprehensive studies of Koriyama and Niita appeared in two Ph.D. dissertations using the Xavier Database and complementary local historical records (Hirai, 2008;Takahashi, 2005) 10 . Takahashi (2005) was one of the first who systematically studied a booming town, Koriyama, and indicated the importance of understanding migration as well as domanial population policy in the demographic model. The urban graveyard theory was examined but found not to be applicable for this town. Hirai (2008) studied the dataset of Niita from a comprehensive approach to individual life courses and households and found that households became more resilient and stable towards the 19th century.
The longitudinal data of the Xavier Database also offered dynamic approaches to "old" questions about household structure, continuity, and succession. Okada applied the studies of household cycles to Nihonmatsu villages and asserted a modified version of Hammel-Laslett in order to show the clear 10 Yet another dissertation by Okada (2006) demonstrated the developmental cycle of households in Aizu domain proposing a modified Hammel-Laslett model. She did not directly use the Xavier Database but BDS of Aizu villages from which the Xavier Database was constructed.

RESEARCH WITH THE XAVIER DATABASE
cyclical change of Japanese households as well as succession strategies (1998,2000,2002). The stem family orientation of the peasant family became clear, particularly among land-owning peasants. Hirai identified characteristics of the continuity of households spanning over a century (e.g., 2003, 2006b).
This series of studies on individual life courses and household cycles written in both English and Japanese became the core achievements of the Japanese Eurasia Project (Ochiai, 2006b) and affected the fields of historical demography and family history and sociology in various ways. First, the novelty of findings sustained by detailed and statistically sound analysis painted clear and fresh pictures of commoners' life courses and families in the 18-19th centuries. For example, universal marriage was followed by flexible divorce and remarriage. Early marriage did not necessarily bring an early start of childbearing. Fertility was extremely low because of the practice of sex-selective infanticide, and adoption compensated for demographic constraints when sons or daughters were lacking. These demographic behaviors were in accordance with peasants' strategy of keeping the stem family intact. These new findings were surprises to family sociologists influenced by the western tradition of family modernization, which emphasizes monolithic and developmental changes. They also made demographers and historians reconsider the importance of households and the stem family system in understanding the contrasting population dynamics in Japan compared with other countries (e.g., Ochiai, 2009;Saito, 1998Saito, , 2000.
Studies based on Xavier data opened possibilities for broadening research into two directions: longitudinal and regional approaches. First, within about 160 years of observation, a general improvement of climate and development of cash crops as well as sericultural industries in the region boosted development of the regional economy. Various improvements in demographic behaviors are observed, particularly related to females: increase in the female age of marriage (Kurosu, 2012a;Kurosu, Takahashi, & Dong, 2017;Tsuya & Kurosu, 2014), decline of female infant mortality (Tsuya & Kurosu, 2004), and increase of reproduction (Tsuya & Kurosu, 2010b, 2010c. Also, homogenization of age at marriage and shortening of length between marriage and divorce were associated with increasing household size and complexity making households more stable towards the end of Tokugawa period. Hirai calls the process "emergence" of Japanese traditional family that emphasizes its continuity and the stem family household (Hirai, 2003(Hirai, , 2006a(Hirai, , 2006b(Hirai, , 2008. Second, the characteristics of northeastern population and family were clarified empirically in the comparative framework of regional variation in Japan (e.g., Kurosu, Tsuya, & Hamano, 1999;Ochiai (Ed.), 2006a;Okada & Kurosu, 1998;. While fertility was generally low in early modern Japan, it was very low in the northeast (Takahashi, 2005;Tsuya & Kurosu, 2010b, 2010c. The survival rate tends to be lower than other studied areas of Japan (Tsuya & Kurosu, 2004). Marriage was early, universal, and more flexible than in the rest of Japan (Kurosu, Tsuya, & Hamano, 1999). Labor migration before marriage delayed marriage in central Japan, but in the northeast migration took place after marriage, making marriage a safety net for households to have them return for sure after service (Nagata, 2001(Nagata, , 2004. Researchers argue that the strategy was intended to increase the working-age population in households to overcome environmental hardships in the northeastern region (Hayami, 2009;Hayami & Kurosu, 2001). These arguments, however, are only based on limited village studies of the northeast, Nishijo in central, and Nomo in the southwestern Japan 11 , and await further investigation.
This article introduced one of the major databases of Japanese historical demography, called the Xavier Database, which includes individual-level longitudinal data of populations in three villages and one town of the current Fukushima prefecture between 1708 and 1870. We discussed the long and complicated process used by the founder of Japanese historical demography, Akira Hayami, and his colleagues to collect, transcribe, code, and finally put the content of the historical population registers into a database. We also reviewed the content and studies that flourished domestically and internationally using these data in the last two decades.

11
Databases for Nishijo (NOBI) and Nomo (NOMO) were constructed during the Japanese Eurasia Project, and also gave opportunities for developing sociological and/or statistical investigations using historical records (e.g., Nakajima, 2016;Nakazato, 2004).

CONCLUSION
During the last decade, new attempts are being made into at least four directions. First, the two villages of the Xavier Database, Niita and Shimomoriya, are harmonized and used for several comparative studies with other East Asian historical populations from northeastern China, Taiwan, and Koreaa collaborative effort sometimes referred to as the East Asia Project or EAP II (Dong et al., 2015a). The standardization and harmonization of different East Asian population datasets, an effort led by Hao Dong in collaboration with each relevant research team, have enabled detailed, systematic comparative studies about similarities and differences in population dynamics and family histories within East Asia (e.g., Dong, 2016;Dong, Campbell, Kurosu, & Lee, 2015b;Dong, Kurosu, & Lee, 2019;Dong, Manfredini, Kurosu, Yang, & Lee, 2017). Second, although still being at the experimental stage, comparative studies of early modern and post-modern Japanese families reveal some interesting resiliency or continuity of Japanese family. These studies bridge Tokugawa and contemporary Japan applying the same model for divorce and marriage to Xavier data and contemporary survey data (Kurosu & Kato, 2018;Tsuya & Kurosu, 2019). Third, there is a new effort coding and adding more information to the original Xavier Database from records of migration and land-lease. In the last few years, we identified geographic locations of 5,000 migration records from the Xavier Database. We can track where migrants came from and went to and the reasons for migration such as service, marriage, or adoption. This will add a spatial dimension to the longitudinal analysis (Kurosu, 2020;Kurosu, Takahashi, & Nagaoka, 2017;Nagaoka, Kurosu, & Takahashi, 2020). Other unique information of the Xavier Database not used until now is land lease/rent and livestock. We added more details of the land transactions (from whom, to whom) for the village of Niita. This has started to show us how land transactions were associated with social mobility and equality as well as demographic patterns of the villages (Arimoto & Kurosu, 2020). We are also constructing a database on livestock based on the records of horses and cattle in BDS. This will make additional useful information about agricultural conditions and the peasant economy related to population dynamics (Takahashi, 2018). Finally, efforts led by Satomi Kurosu with the cooperation of Hao Dong have been undertaken at PFHP to integrate various data collections and sources to advance larger-scale, longitudinal research of family behavior and population history in early modern Japan 12 . The comparative and interdisciplinary approach applied to records of thousands of lives promises new understanding of our history and the resilience of people to socioeconomic and environmental changes.
As the foundation for all, the Xavier Database still inspires us with interesting research topics and possibilities going beyond the field of historical demography and the family. We are extremely grateful for the inexhaustible research opportunities provided by the database.
We are grateful to Akira Hayami (1928Hayami ( -2019 for leaving us amazing sources and datasets, and for his leadership, insights, and guidance in the studies of historical demography. Part of this article is based on the paper we co-authored with Hayami and presented at the 2019 Annual Meeting of the Social Science History Association in Chicago, IL, "Constructing Individual-Level Longitudinal Data for Japanese Historical Population: Challenges and Opportunities". We would like to thank Saeko Narimatsu and all the current and former project members who contributed to collecting, transcribing, and organizing old documents as well as entering and cleaning the data sources. We would like to thank Shoko Hirai, Hideki Nakazato, Aoi Okada, and Atsushi Nagaoka, researchers who worked with Xavier data; Toshiko Mochida, Hirofumi Ohnuma, and Yuriko Kikuchi, research assistants at PFHP for preparing this draft. 12 This project is supported by "Social and geographic mobility and life course: Studies using multigenerational panel data" Grants in Aid for Scientific Research (B), Japan Society for the Promotion of Science (KAKEN 19H01569).