The use of Cohort Size Shrinkage Index (CSSI) to quantify regional famine intensity during the Chinese famine of 1959-61 ======================================================================================================================== * Chunyu Liu * Chihua Li * Hongwei Xu * Zhenwei Zhou * L.H. Lumey ## Abstract There has been a growing interest in studying the causes and impact of the Great Chinese Famine of 1959-61. The Cohort Size Shrinkage Index (CSSI) is the most widely used measure to examine famine intensity and was used in at least 28 Chinese famine studies to date. We examined the potential impact of violations of three requirements for a valid CSSI measure: reliable information on cohort size by year of birth; a stable trend of cohort size by year of birth; and the absence of significant regional migration. We used data from the 1% China 2000 Census to examine the trend of cohort size over time and concentrated on the time window between 1950-70 to exclude policies and events with a large impact on birth trends other than the famine itself. Across China we established a significant difference in cohort size trends between pre-famine births and post-famine births, violating one of the main requirements for a valid CSSI measure. This leads to systematic differences in CSSI depending on what non-famine years are selected for comparison. At the province level, CSSIs estimated based on pre- & post-famine births tend to overestimate famine intensity at higher exposure levels and underestimate intensity at lower levels compared to CSSIs based on pre-famine births alone. This is problematic and demonstrates that the CSSI is not as robust an estimator of famine intensity as had been assumed previously. We recommend therefore that all CSSI should be based on pre-famine birth trends. Using data from Sichuan province, we demonstrate a less pronounced dose-response relation between famine intensity and tuberculosis outcomes using pre-famine based CSSI as compared to reported patterns based on pre- & post-famine based CSSI. We encourage researchers to re-examine their results of Chinese famine studies as local differences in cohort size of pre-famine and post-famine births may lead to significant discrepancies of CSSI estimation and change the interpretation of their findings. ## Introduction The Great Chinese Famine of 1959-61 (Chinese famine) arose from a combination of radical agricultural and economic policies and crop failures (1-5). The famine is associated with over 30 million excess deaths and ranks as one of the largest man-made disasters (6-9). The famine varied across regions and over time in China and a variety of demographic indicators has been used to quantify its intensity because reliable information on food consumption at the time of the famine is hardly, if ever, available (10-12). Among demographic indicators, the Cohort Size Shrinkage Index (CSSI) is the most widely used measure of famine intensity, especially in the fields of economics and health (**Figure 1**). It has been applied to different regions in China to examine potential dose-response relations between the degree of famine exposure and the number of births, deaths, and later economic and health outcomes in populations affected by famine (**Table S1**). The CSSI measure compares the observed cohort size of populations born in famine years to a normally expected cohort size estimated by interpolation or projection (13-16). A projected cohort size for famine births is usually estimated using the average cohort size of populations born both before and after the famine (pre- & post-famine births) (14, 17, 18). There is little consistency, however, in the way CSSIs were calculated, with a wide variety in the selection of data sources, estimating methods, administrative regions, and the choice of pre-famine, famine, and post-famine years included for study (**Table S1**). As noted, almost all studies combined pre- & post-famine births to construct a projected cohort size of famine births (**Figure S1**). ![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/01/01/2021.12.24.21268375/F1.medium.gif) [Figure 1.](http://medrxiv.org/content/early/2022/01/01/2021.12.24.21268375/F1) Figure 1. Chinese famine studies using cohort size shrinkage index (CSSI) Detailed information of listed studies is available in Table S1. For CSSI to be a valid measure of famine intensity across the country, reliable information will be needed on some important variables (15, 16): First, the size of regional birth cohorts, for instance from census data at the province or prefecture level; second, existing trends of cohort size over time in non-famine periods so that any deviations from existing trends in famine years can be documented; and third, the possible impact of regional migrations, if birth cohorts are classified by current residence rather than place of birth. To date, no publications on the Chinese famine have examined in sufficient detail if the quality of the information to construct a valid CSSI measure has been adequate for their research. Also, it remains unknown how robust the composite CSSI measure is to variations in its component parts. These include: First, variations in cohort size measures from the use of different data sources, including the 1% China census collected in 1982, 1990, and 2000, and the full province level census; second, disruptions in cohort size trends by events outside the famine period that make interpolations of famine births from pre- and/or post-famine births problematic; and third, misclassification of place of birth from the use of current residence in census records (7, 13-15, 17, 18). We will examine the robustness of CSSI measure to variations in their component parts using nationwide data. At the prefecture level within Sichuan province in China, we will examine the robustness of the CSSI famine intensity measure in more detail. A recent study in this province reported a dose-response relationship between early-life famine exposure and adulthood tuberculosis using CSSIs constructed based on pre- & post-famine births across prefectures (18). A re-analysis of this study showed that CSSIs estimated from pre-famine births alone have a different distribution than CSSIs from pre- & post-famine births combined (19). Using the latter can introduce an unrecognized bias in CSSI estimates which over-estimates the dose-response relationship between famine intensity and adult tuberculosis. On further examination, it became clear that the over-estimate was driven by post-famine cohort sizes that were significantly different from what was to be expected from pre-famine trends. This demonstrates that CSSIs calculated from interpolating pre- & post-famine births are sensitive to deviations in the trend of cohort size. We therefore need to examine the impact of the different calculating characteristics of CSSI on its mean and distribution and on the robustness of reported results in empirical studies (20). For our study, we used the 1% China 2000 Census as the principal data source to compare trends of cohort size for pre-famine births and post-famine births (19, 21). We constructed CSSIs based on cohorts born in different years (pre-famine, famine, and post-famine births) and compared CSSIs’ distributions at the province and prefecture levels. We examined the impact of migration on CSSI by comparing place of birth and current residence because this information at the province level is available in the 1% China 2000 Census but not in other census data. We further examined changes in CSSI by using different non-famine years (pre-famine years vs pre- & post-famine years) based on the tuberculosis surveillance study in Sichuan Province (15, 16, 18) and changes in CSSI from variations in the exact years used to define pre-famine, famine, and post-famine births, and showed how empirical results may change by using different CSSIs. These findings together show that because of significant changes in cohort size trends across China and in Sichuan, the choice of non-famine years is by far the most important constructing characteristic of CSSIs, influencing both its distribution and the results of empirical studies. ## Results ### Observed cohort sizes for births 1900-2000 **Figure 2A** shows the cohort size by birth year between 1900 and 2000 at the national level as documented by the 1% China 2000 Census. For individuals born until the late 1950s, a monotone relation is seen of higher mortality with advancing age. This shows as the smallest number of survivors among the oldest age groups. For individuals born thereafter, this relation no longer holds. The trend is interrupted in three periods that show marked declines in cohort size: for individuals born between 1958-62, 1971-80, and after 1990. In nearly all provinces, the same pattern is seen (**Figure S2**). Because of these irregularities, choosing an appropriate time window is essential when examining the cohort size trend and projecting cohort sizes for famine births. ![Figure 2.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/01/01/2021.12.24.21268375/F2.medium.gif) [Figure 2.](http://medrxiv.org/content/early/2022/01/01/2021.12.24.21268375/F2) Figure 2. Cohort size by birth year in China. A. Cohort size by birth year, 1990-2000. B. Cohort size trend from linear regression either based on either pre-famine births (1950-1957) or post-famine births (1963-1970). Data: 1% China 2000 Census, place of birth. ### Cohort size trends for births 1950-70 **Figure 2B** limits the time window to the period 1950-70, covering the periods used by the previous Chinese famine studies included in **Figure S1**. This selection of birth years avoids potential modeling artifacts arising from births between 1971-80 or after 1990 that showed marked declines in cohort size. **Figure 2B** shows two cohort size trends in this period, the first based on pre-famine births (1950-57) and the second based on post-famine births (1963-70). The cohort size trends are different as the first shows an increase with year of birth and the second does not. In most provinces, the pre-famine based cohort size trend is also different from the post-famine based trend (**Figure S3**). The difference in trends remained after excluding births in 1963 when cohort sizes increased with the end of the Chinese famine. ### Cohort size projections for famine births 1959-61 are dependent on the choice of non-famine years Cohort size projections for famine births depend on trends estimated from either pre-famine or post-famine births. Since these trends are different, the resulting trend from combining pre- & post-famine births as has been done in previous Chinese famine studies (**Table S1 and Figure S1**) can be different from either of the component trends. As an illustration at the province level, we compared the observed population cohort size for famine births of 1959-61 with the projected population sizes using either a) the observed pre-famine births (1950-57), or b) the observed pre- & post-famine births combined (1950-57 & 1963-70) (**Table S2**). The projected populations are higher than the observed populations in almost all provinces. The two methods also show significant differences at the province level, suggesting that estimates of the impact of the famine will depend on the choice of non-famine years. ### Cohort size shrinkage index (CSSI) is dependent on the choice of non-famine years The absolute value and the relative rank of CSSI at the province level also depend on the choice of non-famine years: pre-famine births (1950-57) vs pre- & post-famine births (1950-57 & 1963-70) (**Table S3**). The difference in pre-famine based and pre- & post-famine based CSSIs ranges from -16.2% in Qinghai province to 14.1% in Liaoning province. It is even more extreme for the five special regions, ranging from -60.6% in Xinjiang to 39.3% in Shanghai (**Table S4**). **Figures 3A-B** show the spatial distribution of pre-famine based and pre- & post-famine based CSSIs in China. Pre- & post-famine based CSSI tend to be higher than pre-famine based CSSI in the Midwest and lower in the Northeast and Southeast **(Figure 3C)**. The difference is not constant across the CSSI range, and at the middle of the range we see a cross-over between the estimates using different non-famine years. Pre- & post-famine based CSSI tends to be systematically larger than pre-famine based CSSI at higher average of both CSSIs and smaller at lower average of CSSIs (**Figure 3D**). Different choices of non-famine years therefore have opposite effects on CSSI estimates, depending on the magnitude of CSSI. The CSSI based on place of birth are highly consistent with CSSI based on place of residence (**Figure S4 and Table S5**). **Figure S5** shows that the mean values of both CSSIs are similar (37.4 vs 37.7) but that pre-famine based CSSI is spread more narrowly around the mean than pre- & post-famine based CSSI (with a SD of 9.8 vs 14.5 respectively). Dose-response relations between famine intensity at the province level and later economic and health outcomes will therefore appear to be more gradual when using the pre- & post-famine based CSSI because of its wider range used in modeling. ![Figure 3.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/01/01/2021.12.24.21268375/F3.medium.gif) [Figure 3.](http://medrxiv.org/content/early/2022/01/01/2021.12.24.21268375/F3) Figure 3. Pre-famine based cohort size shrinkage indexes (CSSI), pre- & post-famine based CSSI, and their difference at the province level A. Provincial CSSI (%) based on pre-famine births in 1950-57. B. Provincial CSSI (%) based on pre- & post-famine births in 1950-57 & 1963-70. C. Difference of provincial CSSI (%) between panel A and panel B. D. Provinces are ordered by the average of pre-famine based and pre- & post-famine based CSSIs. The solid line represents the linear regression of pre-famine based CSSI (filled circles) over the average of the two CSSIs. The dash line represents the linear regression of pre- & post-famine based CSSI (unfilled circles) over the average of two CSSIs. Each bar represents the difference of pre-famine based and pre- & post-famine based CSSIs for each province, and its color coding follows the panel C. Full names corresponding to province abbreviations can be found in ‘Materials and Methods’ section. Data: 1% China 2000 Census, place of birth. Nationwide pre-famine based and pre- & post-famine based CSSIs at the prefecture level are presented in **Table S6**. CSSIs at the prefecture level are based on place of residence as place of birth at this level was not available. It is therefore impossible to assess intra-province migration and its potential impact on CSSI. **Figure S6** shows spatial variations of residence based CSSIs at the province level and the prefecture level. We show variations based on pre-famine births (**Figures S6 A and D**), on pre- & post-famine births **(Figures S6 B and E**) and of their difference **(Figures S6 C and F**). Regardless of province level CSSI, the maps show residual CSSI variation comparing prefectures in a single province **(Figures S6 D-F)**. This is illustrated numerically by the wide range in prefecture based CSSIs within individual provinces (**Table S7**). ### Comparisons with the 1% China 1990 Census The 1% China 1990 Census has been widely used in previous studies although place of birth was missing from this collection (**Table S1**). For comparison, we therefore compared CSSIs at the province level based on this census to CSSIs based on either place of residence or place of birth as recorded in the 1% China 2000 Census. For all comparisons, the findings were highly consistent and inter-correlations exceeded .95 (**Table S5**). A highly similar cross-over between the estimates using different non-famine birth years can also be observed based on the 1% China 1990 Census (**Figure S7**). ### Applying pre-famine based and pre- & post-famine based CSSIs to the tuberculosis surveillance study in Sichuan Province We applied pre-famine based and pre- & post-famine based CSSIs to the tuberculosis surveillance study in Sichuan Province using the 1% China 2000 Census for Sichuan and the full Sichuan 2000 Census (18-20). Unlike CSSIs at the province level, pre-famine based CSSI at the prefecture level within Sichuan Province have a smaller mean and wider range compared to pre- & post-famine based CSSIs (**Table S8**). Using the 1% China 2000 Census, the pre-famine based CSSI dose-response estimate (IRR 0.21, 95% CI: -0.29, 0.71) is less than half the pre- & post-famine based CSSI estimate (IRR 0.52, 95% CI: -0.33, 1.36) (**Figure 4 and Table S9**). Using the full 2000 Sichuan Census (20), the pre-famine based CSSI dose-response estimate (IRR 0.38, 95% CI: -0.12, 0.88) was also less than half the pre- & post-famine based CSSI estimate (IRR 0.91, 95% CI: 0.19, 1.62). We estimated the difference of the two estimates 0.55 (95% CI: 0.06, 10.4) from the bootstrap method. Using different time windows for pre-famine years to construct CSSIs only leads to a minimal change in dose-response effect estimates (IRR 0.37, 95% CI: -0.12, 0.86 using 1949-57 as pre-famine years; IRR 0.40, 95% CI: -0.02, 0.82 using 1945-57 as pre-famine years). ![Figure 4.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/01/01/2021.12.24.21268375/F4.medium.gif) [Figure 4.](http://medrxiv.org/content/early/2022/01/01/2021.12.24.21268375/F4) Figure 4. Scatterplot of CSSI and incidence rate ratio (IRR) of tuberculosis among F1 across prefectures in Sichuan province Each prefecture is represented by a dot. The size of the dot is proportional to the inverse variance of the estimated IRR of each prefecture. The lines represent the meta-regression fits, and the shaded areas represent the 95% CIs. A. Data: 1% China 2000 Census, Sichuan Province, year of birth and place of residence. B. Data: Full Sichuan 2000 Census, age at the time of the census and place of residence. ## Discussion Over the last three decades, there has been a marked increase in the number of studies of the Chinese famine, including at least 28 studies that have used CSSI (**Figure 1**). Demographers have examined the famine’s impact on mortality, fertility, and the sex ratio (8, 9, 22-24). Political scientists described social factors related to the famine (1-6). Economists and medical professionals have compared early-life famine exposure with later economic achievements and health (10, 15, 18, 25-27). Such studies have improved our understanding of underlying causes and of long-term impacts of this man-made disaster. A robust and valid indicator of famine intensity at the local level is needed however to compare study findings and interpret study results (10, 11, 28). The Cohort Size Shrinkage Index (CSSI) is the measure that has been the most widely used for this purpose. To construct a valid CSSI, reliable information will be needed on the size of cohorts born before, during, and after the famine (15, 16). There should also be no disruptions in overall cohort size trends from non-famine sources (15, 16, 20). And current place of residence, as collected in the census, should be an adequate indicator of place of birth. We established that CSSI measures of famine intensity across China were extremely similar for different sources of birth cohort size, including the 1% China 2000 Census, the 1% China 1990 Census, and the full Sichuan 2000 Census. CSSI estimates were also not affected by using place of residence rather than place of birth as both available in 1% China 2000 Census. And CSSI estimates were robust to small variations in the definition of pre-famine, famine and post-famine years as seen in published studies. Contrariwise, CSSI measures of famine intensity were highly sensitive to the choice of non-famine years. When selecting pre-famine births, post-famine births, or both groups combined to project famine births, all selections lead to different estimates of CSSI. Specifically, combining pre- & post-famine births to construct CSSI leads to different estimates of famine intensity compared to selecting pre-famine births. This is problematic and demonstrates that the CSSI is not as robust an estimator of famine intensity as had been assumed previously (13-16, 18). In the following, we discuss in turn our CSSI findings using different sources of cohort size of pre-famine, famine, and post-famine births; the impact on CSSI of changing cohort size trends over time; potential misclassification of place of birth from using census data on current residence; and urban-rural famine differentials. ### Combining sources to determine pre-famine, famine, and post-famine cohort size In Chinese famine studies to date, different waves of the China population census between 1980 and 2000 have been used to construct CSSIs at multiple administrative levels (**Table S1**). Available studies suggest that the population census data as collected in post-famine era, especially in 1990 and 2000, are fit for this purpose (13-16, 18, 29, 30). We cannot think of any reason why the collected information on date of birth and residence, and additionally on place of birth in the 2000 census, might be systematically biased. Representation in the 1990 and 2000 census for those still alive at the time will also be in essence complete and especially is not conditional on place or date of birth. We therefore used the most recent publicly available data from the 1% China 2000 Census to collect the needed information for estimating famine intensity (5, 7, 15, 16). On the use of the 1% China 2000 Census as our main data source and not the 1% China 1990 Census, we note that our study findings are highly consistent across census years (**Table S5**). In both census waves, we found the same systematic differences between pre-famine based CSSI and pre- & post-famine based CSSI (**Figure 3D and Figure S7**). For our analysis of CSSI at the prefecture level in Sichuan province from different sources, we note that for the 1% China 2000 Census information was available on year and month of birth. This could lead to minor differences between findings from the full Sichuan 2000 Census where year of birth was calculated from the difference between reported age at the time of the census and the year of census (18, 20). ### Changing cohort size trends and how these affect CSSI estimates We found that major social and political events can have a big impact on the trend of cohort size. Most importantly, the famine itself is likely to have affected subsequent cohort trends for births and deaths (7-9). As another example, family planning pilot programs introduced in some regions from the 1960s and nationwide in the 1970s most likely contributed to the cohort size decline observed for individuals born after these years (22, 31, 32). For Chinese famine studies, it is critically important therefore to select a suitable time window for observation of a stable cohort size trend that can be used to project famine related changes. We think this is only possible with a focus on pre-famine births as further explained below. Chinese famine studies to date (**Table S1**) have failed to recognize that the cohort size trends among pre-famine births and post-famine births differ substantially (**Figure 2B and Figure S3**). This difference can be illustrated by comparing the observed and the projected cohort size for famine births based on either pre-famine births or the commonly used pre- & post-famine births (**Table S2**). The trend change also leads to different CSSI estimates. Unfortunately, there is no constant difference in CSSI estimates over the CSSI range as the measures intersect: pre- & post-famine based CSSIs tend to be systematically larger than pre-famine based CSSIs at higher mean CSSIs and systematically lower at lower mean CSSIs **(Figure 3D**). In our view, this shows a systematic bias of the commonly used CSSI based on pre- & post-famine births. Previous Chinese famine studies have ignored the change in cohort size trend and have used the average cohort size of pre- & post-famine births to generate CSSIs (**Table S1 and Figure S1**). To avoid the problems identified above with changing trends, we projected cohort sizes for famine births by extrapolating the trend of pre-famine births in 1950-57. This approach was also used in the earliest demographic study of excess mortality associated with the Chinese famine (8) and the Ukraine famine of 1932-33 (33). ### Changing trends can also affect CSSI distributions In many current Chinese famine studies (**Table S1**), CSSI has been incorporated in regression models as a continuous measure of famine intensity. Changing cohort size trends of pre-famine births and post-famine births may not only affect CSSI point estimates but also their distributions. This changes how the study results will be interpreted. Using data from the tuberculosis surveillance study in Sichuan Province (15, 16, 18), we demonstrated that the dose-response relation between CSSI and the IRR for tuberculosis is much weaker using pre-famine births as the reference compared to using the combined pre- & post-famine births as the reference (**Figure 4**). This holds true both when using data from the 1% China 2000 Census and the full Sichuan 2000 Census. Our findings illustrate how the choice of non-famine years determines the nature of a possible dose-response relationship. ### Misclassification of place of birth At the province level, the 1% China 2000 Census includes information on place of birth. We found that CSSI constructed based on place of birth is highly comparable to CSSI based on place of residence (**Figure S4**). This is consistent with our finding that inter-province migration among births 1950-70 was as low as 6%. At the province level, place of residence in 1990 or 2000 was a sufficient indicator of place of birth for the purposes of our study. At the prefecture level, information on place of birth was not available. Therefore, the degree of intra-province migration is unknown and also its potential to bias studies that rely on place of residence as a proxy for place of birth. Large intra-province migrations from rural to urban areas in specific regions are likely among the 1950-70 birth cohorts: initially from rapid industrialization which led to a large-scale transfer of labor from agriculture to industry and from rural to urban locations in 1957-60 (34); then possibly in reverse following production declines and the Chinese government decision to send workers back to their rural homes in 1961 to help increase food production (3); then again from rural to urban with the development of the urban economy (35). These migrations may have been mitigated to some extent however by a variety of government policies that were designed to control migration (36-39). For famine studies at the prefecture level, collecting additional information on migration over the life-course will therefore be important. ### Urban-rural famine differentials There was a substantial rural-urban difference in famine intensity because government policies prioritized food security in urban areas (7-9). As an example, urban population deaths increased from 9.2 to 13.8 per thousand between 1958-60, showing an increase of close to 50% (40). By contrast, rural population deaths increased from 12.5 to 28.6 per thousand, showing an increase of well over 100% (40). The provincial famine intensity was mainly determined by the situation in rural areas because over 80 percent of Chinese population at the time lived in rural areas (8, 41). When we excluded urban populations from our analyses of the 1% China 2000 Census, we found our results did not change. ### Strengths and Limitations Our study has many strengths. We used census data to construct robust measures of famine intensity; demonstrated that combining pre- & post-famine births to construct CSSI can lead to unrecognized bias in estimating famine intensity and can exaggerate potential dose-response relations as empirically demonstrated by the tuberculosis study; and avoided misclassification of place of birth at the province level. These strengths provide the basis for our recommendations for the analysis of famine studies in China. There are also some potential limitations. Census data only include individuals who live long enough to be included and this could be a problem if the famine had a long-term effect on mortality. In the one study we know on the question, this does not seem to be the case (38). We also established in simulations that the CSSI as a measure of famine intensity is highly robust to potential changes in mortality among famine births. And potential misclassifications of place of birth could arise if pregnant women migrated during the famine and gave birth in another provinces. By regulation, a child’s place of birth was registered however according to mother’s Hukou status (42). ### Suggestions for future studies While we advocate the use of CSSI in Chinese famine studies, results should be presented with their limitations in mind. Because of different pre-famine and post-famine cohort size trends, the use of pre-famine based CSSI will generally be the most appropriate. Current high-quality studies should therefore be re-analyzed with this in mind and divergent findings compared and discussed (14-16, 18, 43, 44). For the use in future studies, we provide data of CSSI at both the province and prefecture level in supplementary materials (**Table S6**). This information is essential to identify provinces and prefectures across China where famine intensity was highest. These will be the most suitable locations for future famine studies. We suggest that the focus should first be on studies in one or two of the most famine-severe provinces, including Sichuan and Anhui provinces (**Figure S6 & Table S7**). Prefecture based studies should be attempted later, as there can still be substantial variations in famine intensity within province. These studies will generate improved estimates of famine intensity at the local level as will be needed to establish potential dose-response relationships between famine exposure and its long-term impact. ## Conclusions Chinese famine studies to date have failed to recognize that cohort size trends among pre-famine births and post-famine births can differ substantially. This is problematic as the construction of CSSI as a valid measure of famine severity relies on a stable cohort size trend. We demonstrate that CSSI is not a robust estimator of famine intensity as had been assumed previously. It can overestimate famine intensity at higher exposure levels and underestimate intensity at lower levels. As an alternative for Chinese famine studies, we recommend CSSI based on pre-famine births alone. ## Materials and Methods ### Main Data Previous Chinese famine studies have used population census data collected from the 1980s onwards to construct cohort size shrinkage indexes (CSSIs) for selected regions at different administrative levels (**Table S1**). Our study used the 1% China 2000 Census ([https://international.ipums.org/international/](https://international.ipums.org/international/)) (21), because it provides more detailed demographic information compared to other census data. Relevant to our analyses, this sample includes birth year, place of birth at the province level, and place of residence at both the prefecture and province levels (30). By contrast, the 1% sample of China’s 1990 census does not provide information on place of birth. In the 1% China 2000 Census we can therefore compare place of birth at the time of the famine with place of residence at the 2000 censes for individuals who were then still alive. This comparison provides information on inter-province migrations over time. We examined the cohort size of selected populations by year of birth. In selecting the time window of 1950-1970, we aimed to define a relatively stable period in terms of birth trends that was only interrupted by the famine events. The beginning of the period corresponds to the establishment date of the People’s Republic of China in 1949 and the end by the introduction of family planning policies across China in the 1970s. Our nationwide study included the 29 province level regions and 340 prefectures. The regions are: Sichuan (SC), Anhui (AH), Guizhou (GZ), Hunan (HuN), Henan (HeN), Qinghai (QH), Jiangsu (JS), Ningxia (NX), Guangxi (GX), Gansu (GS), Shandong (SD), Yunan (YN), Hubei (HuB), Fujian (FJ), Hebei (HeB), Zhejiang (ZJ), Jiangxi (JX), Guangdong (GD), Liaoning (LN), Shanxi (SX), Shaanxi (SaX), Jilin (JL), Inner Mongolia (IM), Heilongjiang (HLJ), Beijing (BJ), Tianjin (TJ), Shanghai (SH), Xinjiang (XJ), and Tibet (TB). Before the 1980s, Hainan was part of Guangdong province, and Chongqing (CQ) was part of Sichuan province. Five province level regions were grouped separately. These included the 3 municipality cities, Beijing, Tianjin and Shanghai which received preferential food supplies during the famine and where early family planning policies were introduced starting in the 1960s (45), and two less famine-affected provinces, Xinjiang and Tibet, to which government policies encouraged active immigration (46). ### CSSI construction from expected cohort size in the absence of famine Previous Chinese famine studies have generally calculated CSSI as ![Graphic][1] for regions at different administrative levels, where *N*famine is the observed average cohort size of famine births and *N*non-famine is the average cohort size of the combined pre- & post-famine births. This assumes that the average cohort size of pre- & post-famine births represents the expected cohort size in the absence of famine. We have suggested however that using the average cohort size of pre- & post-famine births may introduce bias in constructing a CSSI because the cohort size of post-famine births shows irregular patterns (19). In addition, taking averages alone may overlook underlying trends of cohort size (20). To address these concerns, we compared linear extrapolations of populations born in different non-famine years (pre-famine births vs pre- & post-famine births) to project expected cohort sizes at the province level in the absence of famine. Linear models were used as these appeared to correspond adequately with the monotone increase in cohort size in the births years 1950-1957 preceding the famine. More refined including exponential models with quadratic terms showed no significant improvement in fit. We defined the birth years 1959-61 as famine years, 1950-57 as pre-famine years, and 1963-70 as post-famine years. We excluded births in 1958 or 1962 from analysis because the beginning and the end of the Chinese famine differed by province and births in these years could easily be misclassified on exposure status. Two sets of CSSI were constructed: the first based on linear extrapolation of pre-famine births (1950-57) and the second on the interpolation of pre- & post-famine births (1950-57 & 1963-70). The distributions of the two CSSI sets were compared. In all tables and figures, provinces were ordered by the average CSSI from the two sets. Spatial variations of the two sets of CSSI were mapped at both the province and prefecture levels. Results based on the 1% China 2000 population census were compared with findings based on the 1% China 1990 population census. ### Applying CSSIs to tuberculosis surveillance study in the Sichuan province The tuberculosis surveillance study in the Sichuan province collected information on clinically diagnosed and laboratory-confirmed active tuberculosis cases diagnosed between 2005 and 2018 classified by sex, age at diagnosis, year of diagnosis, and residential prefecture (47). Following previous analytic approaches, a mixed-effects meta-regression was run to evaluate associations between the famine intensity and tuberculosis risk in famine cohort (born in 1958-62) at the prefecture level in Sichuan province (18-20). For famine births, we fitted linear models to examine the relation between famine intensity as represented by CSSI on the log of the ratio of the observed vs expected non-famine incidence rates. We modeled multiple sets of CSSI using a) different combinations of non-famine years (pre-famine years vs pre- & post-famine years) at prefecture level in Sichuan province and b) either the 1% China 2000 Census for Sichuan province or the Full Sichuan 2000 Census. The data and code used for our study are publicly available at Github repository [https://github.com/qu-cheng/TB_famine](https://github.com/qu-cheng/TB_famine). Detailed analytical methods can be found elsewhere (18). ## Data Availability All data and code necessary to reproduce the main findings of this study are available at the Github repository: [https://github.com/chunyu-yes/CSSI\_famine\_severity](https://github.com/chunyu-yes/CSSI_famine_severity). ## Data Availability All data produced in the present work are contained in the manuscript. [https://github.com/chunyu-yes/CSSI\_famine\_severity](https://github.com/chunyu-yes/CSSI_famine_severity) ## Footnotes 1. C.L. and C.L. contributed equally to this work. 2. To whom correspondence may be addressed. Email: lumey{at}columbia.edu ### Author contributions C.L., C.L., and L.H.L designed research; C.L., C.L., H.X., Z.Z. and L.H.L performed research; C.L., C.L., H.X., Z.Z. and L.H.L analyzed data; C.L., C.L., and L.H.L drafted the paper; H.X. and Z.Z. provided technical support; and C.L., C.L., H.X., Z.Z. and L.H.L revised the paper. C.L. and C.L. contributed equally to this work. The authors declare no competing interest. ## Acknowledgments The authors are particularly grateful to the Minnesota Population Center for providing the data of the 1% China 1990 and 2000 Census, the National Bureau of Statistics in China for originally producing the census data, and the research team at Berkely University for making the data of tuberculosis surveillance study in the Sichuan province publicly available. ## Footnotes * Author affiliations updated * Received December 24, 2021. * Revision received December 30, 2021. * Accepted January 1, 2022. * © 2022, Posted by Cold Spring Harbor Laboratory The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission. ## References 1. 1. D. L. Yang, “The political economy of the Great Leap Forward” in Calamity and reform in China: State, rural society, and institutional change since the Great Leap Famine. (Stanford University Press, 1996), pp. 42–67. 2. 2. G. H. Chang, G. J. Wen, Food availability versus consumption efficiency: Causes of the Chinese famine. China Econ. Rev. 9, 157–165 (1998). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S1043-951X(99)80012-1&link_type=DOI) 3. 3. J. Y. Lin, D. T. Yang, On the causes of China’s agricultural crisis and the great leap famine. China Econ. Rev. 9, 125–140 (1998). 4. 4. J. Y. Lin, D. T. Yang, Food availability, entitlements and the Chinese famine of 1959–61. Econ. J. 110, 136–158 (2000). 5. 5. S. Cao, The Deaths of China’s Population and Its Root Cause during 1959-1961 (In Chinese). Chin. J. Popul. Sci. 1, 14–28 (2005). 6. 6. G. H. Chang, G. J. Wen, Communal dining and the Chinese famine of 1958–1961. Econ. Dev. Cult. Change. 46, 1–34 (1997). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1086/452319&link_type=DOI) 7. 7. A. Garnaut, The geography of the Great Leap famine. Mod. China 40, 315–348 (2014). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1177/0097700413507425&link_type=DOI) 8. 8. X. Peng, Demographic consequences of the Great Leap Forward in China’s provinces. Popul. Dev. Rev., 639-670 (1987). 9. 9. Z. Zhao, A. Reimondos, The Demography of China’s 1958-61 Famine. Popul. 67, 281–308 (2012). 10. 10. C. Li, L. H. Lumey, Exposure to the Chinese famine of 1959–61 in early life and long-term health conditions: a systematic review and meta-analysis. Int. J. Epidemiol. 46, 1157–1170 (2017). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/ije/dyx013&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=28338900&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) 11. 11. C. Li, E. W. Tobi, B. T. Heijmans, L. H. Lumey, The effect of the Chinese Famine on type 2 diabetes mellitus epidemics. Nat. Rev. Endocrinol. 15, 313–314 (2019). 12. 12. L. H. Lumey, A. D. Stein, E. Susser, Prenatal famine and adult health. Annu. Rev. Public Health 32, 237–262 (2011). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1146/annurev-publhealth-031210-101230&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21219171&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000290776200014&link_type=ISI) 13. 13. X. Meng, N. Qian, The long run health and economic consequences of famine on survivors: Evidence from China’s Great Famine. IZA Discussion Papers, 2471 (2006). 14. 14. C. Huang, Z. Li, M. Wang, R. Martorell, Early life exposure to the 1959–1961 Chinese famine has long-term health consequences. J. Nutr. 140, 1874–1878 (2010). [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToibnV0cml0aW9uIjtzOjU6InJlc2lkIjtzOjExOiIxNDAvMTAvMTg3NCI7czo0OiJhdG9tIjtzOjUwOiIvbWVkcnhpdi9lYXJseS8yMDIyLzAxLzAxLzIwMjEuMTIuMjQuMjEyNjgzNzUuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 15. 15. H. Xu, L. Li, Z. Zhang, J. Liu, Is natural experiment a cure? Re-examining the long-term health effects of China’s 1959–1961 famine. Soc. Sci. Med. 148, 110–122 (2016). 16. 16. H. Xu, Z. Zhang, L. Li, J. Liu, Early life exposure to China’s 1959–61 famine and midlife cognition. Int. J. Epidemiol. 47, 109–120 (2018). [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) 17. 17. S. Song, Prenatal malnutrition and subsequent foetal loss risk: Evidence from the 1959-1961 Chinese famine. Demogr. Res. 29, 707–728 (2013). 18. 18. Q. Cheng et al., Prenatal and early-life exposure to the Great Chinese Famine increased the risk of tuberculosis in adulthood across two generations. Proc. Natl. Acad. Sci. U.S.A. 117, 27549–27555 (2020). [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoicG5hcyI7czo1OiJyZXNpZCI7czoxMjoiMTE3LzQ0LzI3NTQ5IjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjIvMDEvMDEvMjAyMS4xMi4yNC4yMTI2ODM3NS5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 19. 19. C. Li, Z. Zhou, L. H. Lumey, Early-life exposure to the Chinese famine and tuberculosis risk: Unrecognized biases from different measures of famine intensity. Proc. Natl. Acad. Sci. U.S.A. 118, e2102809118 (2021). [FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6NDoicG5hcyI7czo1OiJyZXNpZCI7czoxODoiMTE4LzE2L2UyMTAyODA5MTE4IjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjIvMDEvMDEvMjAyMS4xMi4yNC4yMTI2ODM3NS5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 20. 20. Q. Cheng et al., Reply to Li et al.: Estimate of the association between TB risk and famine intensity is robust to various famine intensity estimators. Proc. Natl. Acad. Sci. U.S.A. 118, e2103254118 (2021). [FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6NDoicG5hcyI7czo1OiJyZXNpZCI7czoxODoiMTE4LzE2L2UyMTAzMjU0MTE4IjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjIvMDEvMDEvMjAyMS4xMi4yNC4yMTI2ODM3NS5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 21. 21. M. P. Center, Integrated Public Use Microdata Series, International: Version 7.3. Minneapolis, MN: IPUMS. Available at [https://doi.org/10.18128/D020.V7.2Deposited](https://doi.org/10.18128/D020.V7.2Deposited) 24 September 2019. 22. 22. J. S. Aird, Population studies and population policy in China. Popul. Dev. Rev., 267–297 (1982). 23. 23. B. Ashton, K. Hill, A. Piazza, R. Zeitz, “Famine in China, 1958–61” in The population of modern China. (Springer, 1992), pp. 225–271. 24. 24. J. Banister, “Mortality” in China’s changing population. (Stanford University Press, 1987), pp. 78–120. 25. 25. Z. Cheng, R. Smyth, Does Childhood Adversity Affect Household Portfolio Decisions? Evidence from the Chinese Great Famine. ResearchGate [Preprint] (2021). [https://www.researchgate.net/publication/352244671\_Does\_Childhood\_Adversity\_Affect\_Household\_Portfolio\_Decisions\_Evidence\_from\_the\_Chinese\_Great\_Famine](https://www.researchgate.net/publication/352244671\_Does\_Childhood\_Adversity\_Affect\_Household\_Portfolio\_Decisions\_Evidence\_from\_the_Chinese_Great_Famine) (accessed 6 November 2021). 26. 26. W. Long, G. G. Tian, J. Hu, D. T. Yao, Bearing an imprint: CEOs’ early-life experience of the Great Chinese Famine and stock price crash risk. Int. Rev. Financial Anal. 70, 101510 (2020). 27. 27. X. Feng, A. C. Johansson, Living through the Great Chinese Famine: Early-life experiences and managerial decisions. J. Corp. Finance 48, 638–657 (2018). 28. 28. C. Li, E. W. Tobi, B. T. Heijmans, L. H. Lumey, Reply to ‘Early-life exposure to the Chinese Famine and subsequent T2DM’. Nat. Rev. Endocrinol. 16, 125–126 (2020). 29. 29. W. Lavely, J. Lee, W. Feng, Chinese demography: the state of the field. J. Asian Stud. 49, 807–834 (1990). 30. 30. W. Lavely, First impressions from the 2000 census of China. Popul. Dev. Rev. 27, 755–769 (2001). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1728-4457.2001.00755.x&link_type=DOI) 31. 31. D.L. Poston Jr., B. Gu, Socioeconomic development, family planning, and fertility in China. Demography, 531-551 (1987). 32. 32. I. Attane, China’s family planning policy: an overview of its past and future. Stud. Fam. Plann. 33, 103–113 (2002). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1728-4465.2002.00103.x&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=11974414&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000174953800009&link_type=ISI) 33. 33. O. Rudnytskyi, N. Levchuk, O. Wolowyna, P. Shevchuk, A. Kovbasiuk, Demography of a man-made human catastrophe: The case of massive famine in Ukraine 1932-1933. Can. Stud. Popul. 42, 53–80 (2015). 34. 34. A. J. Jowett, The demographic responses to famine: the case of China 1958–61. GeoJournal 23, 135–146 (1991). [PubMed](http://medrxiv.org/lookup/external-ref?access_num=12317880&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) 35. 35. K. H. Zhang, S. Song, Rural–urban migration and urbanization in China: Evidence from time-series and cross-section analyses. China Econ. Rev. 14, 386–400 (2003). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.chieco.2003.09.018&link_type=DOI) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000188613900002&link_type=ISI) 36. 36. Y. Liu, Research on Population Migration in Gansu Province from 1959 to 1960 (In Chinese). Dangshi Yanjiu Yu Jiaoxue, 90–101 (2017). 37. 37. R. Li, Population Migration during the Great Leap Forward Period and the Hard Times (In Chinese). Popul. Sci. China 4 (2000). 38. 38. S. Song, Does famine have a long-term effect on cohort mortality? Evidence from the 1959–1961 Great Leap Forward Famine in China. J. Biosoc. Sci. 41, 469–491 (2009). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1017/S0021932009003332&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=19302727&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) 39. 39. Q. Li, An Investigation of the Phenomenon of Farmers Flowing to Cities in the 1950s (In Chinese). 21st Century (2005). 40. 40. R. Li, Preliminary Dissolving Analysis of Population Mortality in Hard Times (In Chinese). Popul. Res. 25, 43–49 (2001). 41. 41. T. Gørgens, X. Meng, R. Vaithianathan, Stunting and selection effects of famine: A case study of the Great Chinese Famine. J. Dev. Econ. 97, 99–111 (2012). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jdeveco.2010.12.005&link_type=DOI) 42. 42. J. Zhu, The household registration policy of yesterday, today and tomorrow (In Chinese). Newsletter About Work in Rural Areas, 17–17 (2010). 43. 43. X. Meng, N. Qian, P. Yared, The institutional causes of China’s Great Famine, 1959– 1961. Rev. Econ. Stud. 82, 1568–1611 (2015). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/restud/rdv016&link_type=DOI) 44. 44. C. Huang, C. Guo, C. Nichols, S. Chen, R. Martorell, Elevated levels of protein in urine in adulthood after exposure to the Chinese famine of 1959–61 during gestation and the early postnatal period. Int. J. Epidemiol. 43, 1806–1814 (2014). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/ije/dyu193&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=25298393&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F01%2F01%2F2021.12.24.21268375.atom) 45. 45. F. Yang (2004) Historical research on family planning of contemporary China (Ph.D. thesis in Chinese). (Zhejiang University, Zhejiang, China). 46. 46. M. Sabit, Y. Mamat, Analysis of Population Changes and Its Causative Factors in Xinjiang in the Last 50 Years (In Chinese). J. Arid Land Resour. Environ. 4, 114–119 (2008). 47. 47. T. Li et al., Evidence for heterogeneity in China’s progress against pulmonary tuberculosis: uneven reductions in a major center of ongoing transmission, 2005–2017. BMC Infect. Dis. 19, 1–11 (2019). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/s12879-019-4277-8&link_type=DOI) [1]: /embed/inline-graphic-1.gif