Abstract
In accounts of the worldwide impact of the 1918 influenza pandemic, China remains a black hole of missing data. In the absence of systematically collected nationwide death statistics, scholars have used scattered and often impressionistic reports to suggest that the epidemic had only a mild impact in China and, in some cases, to raise the possibility that the epidemic originated in China. These works rely heavily on conclusions drawn from anecdotal reports of customs officers, a medical report from Canton, and uncritical use of Shanghai and Hong Kong crude death rates, which are shown herein to be seriously flawed or misstated. This article and its online supplement contribute to knowledge of the influenza epidemic in China by reassessing the available data on Hong Kong, Shanghai, and Canton and assessing hitherto neglected sources on seven majority-Chinese jurisdictions that enforced vital statistics reporting. The results refute the notion of a mild impact and show that the pandemic had an impact in most cases greater than that seen in Western countries like the United States and England and Wales.
In 1918 pandemic influenza swept around the world in two waves, an earlier “spring” wave that caused high morbidity and low mortality, and a “fall” wave carrying a virulent strain that caused high levels of mortality, as serious flu cases succumbed to deadly pneumonias. 1 In some localities, subsequent waves followed in early 1919 and late 1919 and early 1920. Worldwide mortality is estimated to have reached 30–50 million, making it one of the “most deadly events in human history” (Morens and Fauci, 2007: 1018).
In the many accounts of the worldwide impact of the 1918 influenza pandemic, China remains a black hole of missing data (Jordan, 1927: 223; Patterson and Pyle, 1991: 18; Johnson and Mueller, 2002: 112; Peckham, 2022). K. David Patterson and Gerald F. Pyle (1991: 18), citing anecdotal but no vital statistics data in making their estimate of total Chinese losses, acknowledged that the “uncertainty about mortality in China has a major impact on any global casualty estimates.” In the absence of systematically collected nationwide death statistics, scholars have used scattered and often impressionistic reports to draw conclusions about the severity of the epidemic in the whole of China and, in some cases, to raise the possibility that the epidemic originated in China. Robert Peckham (2022: 262–63) reviews a more recent literature, and laments that the same repertoire of patchy sources “is routinely churned over by foreign and Chinese scholars alike.” 2 However, he then proceeds to use those same sources, and makes no critical assessments of the widely cited sources that are shown here to be seriously flawed.
This article contributes to knowledge of the influenza epidemic in China by reevaluating sources that have long been accepted at face value and by presenting hitherto neglected vital statistics reports on Chinese and related East Asian populations. Almost all of these vital statistics pertain to Chinese populations that were subject to Japanese and European colonial administrations that enforced cause-of-death reporting by mandating burial permits. To obtain police-issued burial permits in these jurisdictions, household heads were required to obtain cause-of-death certificates from licensed practitioners (trained in either modern or traditional medicine); these reports then formed the basis for cause-of-death statistics compiled by police registrars. 3 Critical use of these sources puts the discussion of the Chinese case on a firmer empirical base, makes possible improved estimates of the numbers dying as a result of the pandemic, and refutes the notion that the impact of the 1918 influenza epidemic on the Chinese population was unusually mild (Ijima, 2003; Langford, 2005; Cheng and Leung, 2007). Instead, the data show that the pandemic had an impact in many cases greater than that seen in Western countries like the United States and England and Wales.
The possibility of a Chinese origin of the 1918 pandemic has long been rejected by most analysts (Patterson and Pyle, 1991: 8) but has had a recent revival. Work by Kennedy Shortridge (1997, 1999), Christopher Langford (2005), Dorothy Pettit and Janice Bailie (2008), and Mark Osborne Humphries (2013) has renewed speculation regarding a Chinese origin of the 1918 pandemic. These works, of which Langford’s is the most substantial, rely heavily on conclusions drawn from uncritical use of Shanghai and Hong Kong crude death rates (CDRs), 4 which are shown herein to be seriously flawed or misstated. 5 A separate article assesses the speculative claims by these authors that Chinese laborers recruited by the British and the French carried a virus to France during the First World War that seeded the 1918 influenza pandemic (Shepherd, 2022).
This article reassesses the four sources repeatedly relied on in previous studies and by advocates of a mild mortality impact—the widely cited Shanghai and Hong Kong death reports, anecdotal reports by customs officers, and a medical report on influenza in Canton’s (Guangzhou’s) missionary institutions—and serves as a cautionary tale for future researchers seeking to use these sources.
Presented here for the first time are mortality data showing the impact of the influenza pandemic on Chinese populations in seven additional jurisdictions: the British-administered Straits Settlements and the port of Weihaiwei 威海衛; the Chinese-administered city of Beijing and Huolu 獲鹿 county in Zhili 直隸 province; and three Japanese-administered territories, the Kwantung (Guandong 關東) Leased Territory and the South Manchuria Railway Zone (hereafter collectively termed the KLTZ), Kiaochow/Tsingtao (Jiaozhou 膠州/Qingdao 青島), and Taiwan. 6 In five of the seven jurisdictions, death reporting was mandatory, and vital statistics and census data were published at regular intervals (only Weihaiwei and Huolu did not make regular reports). The completeness and quality of the vital statistics reporting systems in each jurisdiction are assessed, attending to the indicators of underreporting and misclassification of causes noted by Samuel H. Preston, Nathan Keyfitz, and Robert Schoen (1972: 28–30) in their cross-national study of mortality patterns. Comparable rates are also calculated for Japan, Korea, Ceylon (Sri Lanka), England and Wales, and the United States, which serve as reference populations and baselines for comparison.
Instead of CDRs (which are subject to fluctuations in other infectious diseases), wherever possible the data presented here use the standard and preferred measure of epidemic excess, which calculates how much the 1918 cause-specific death rate for influenza and influenza-related causes (that is, the pneumonia, bronchitis, and influenza death rate, hereafter the PBI death rate 7 ) exceeds an expected rate based on an average PBI death rate for a set of previous years (see Table 1; for comparison, Table 2 provides the CDR measures). Measures are also calculated of the mortality in the fall months of 1918 when the epidemic peaked and of the unusually high young-adult mortality, a hallmark of the 1918 pandemic (Tables 3 and 4). Overall, measures of influenza-related epidemic excess mortality are calculated for a total of eight Chinese-majority jurisdictions and five reference populations; in six cases separate rates are calculated for colonial elite and colonial subject populations. In conclusion, the data show that the 1918 pandemic in China had an impact in many cases greater than that seen in Western countries like the United States and England and Wales and provide a basis for a new estimate of the numbers of Chinese dying in the pandemic. A fuller discussion of the methodology followed in measuring influenza-related epidemic excess mortality and constructing the four data tables is included in the Online Supplementary Materials. 8
Influenza-Related Death Rates and Epidemic Excesses, 1916–1920: East Asian Populations Compared.
Deaths from “infectious diseases” after subtracting deaths from notifiable infectious diseases are added to the deaths from PBI causes in Korea.
PBI deaths among Chinese in Tsingtao include those classified as “other infectious” and “generalized disease.” The excesses in Tsingtao are measured against the single year of 1917.
There was no compulsory registration of deaths in Shanghai, and the gross undercount of Chinese deaths renders the Shanghai data on Chinese unusable and the data on non-Chinese questionable.
Note: PBI = pneumonia, bronchitis, and influenza.
All-Cause Crude Death Rates and Epidemic Excesses, 1916–1920: East Asian Populations Compared.
The excesses in Tsingtao are measured against the single year of 1917. A substantial decline in gastrointestinal causes of death from 1917 to 1918 contributed to the decline in the CDR among Japanese in Tsingtao.
There was no compulsory registration of deaths in Shanghai, and the gross undercounting of Chinese deaths renders the Shanghai data on Chinese unusable and the data on non-Chinese questionable.
Epidemic Excess of Deaths (PBI or All Causes) during the Fall Epidemic Wave, 1918: East Asian Populations Compared.
Deaths from “infectious diseases” are added to the deaths from PBI causes in Korea.
The excesses in Kiaochow/Tsingtao are measured against the single year of 1917.
There was no compulsory registration of deaths in Shanghai, and the gross undercount of Chinese deaths renders the Shanghai data on Chinese unusable and the data on non-Chinese questionable. The Shanghai all-cause measures for non-Chinese include only October and November, which were the peak months; the outbreak of smallpox in December 1917 reduces the value of that month as a reference.
As influenza deaths were reported by month only from 1918 in Hong Kong, the excess rates use 1923–1924 as a baseline. Monthly breakdowns of pneumonia deaths were not reported in these years.
Note: PBI = pneumonia, bronchitis, and influenza; PI = pneumonia and influenza.
Increases in the Young Adult Proportion of Deaths (PBI or All Causes), 1918: East Asian Populations Compared.
Deaths from “infectious diseases” are added to the deaths from PBI causes in Korea.
No reports of deaths by age are included in the reports from Shanghai (Chinese), Hong Kong, Beijing, and Weihaiwei for the years in question. There was no compulsory registration of deaths in Shanghai, and the gross undercount of Chinese deaths renders the Shanghai data on Chinese unusable and the data on non-Chinese questionable.
The excess in Tsingtao is measured against the single year of 1917.
The Mortality Impact of the Pandemic According to Langford
In his 2005 article Langford attemps to assess the extent to which China experienced the influenza pandemic of 1918–1919. He concludes that the impact of the epidemic in 1918 on both the foreign and the Chinese populations in China was mild compared to those recorded in Britain and the United States. Inspired by Shortridge, 9 he speculates that this may have been due to earlier circulation of a mild precursor virus that conferred immunity on populations resident in China (whether Chinese or foreign) when the virulent strain appeared in the fall of 1918. He further hypothesizes that Chinese laborers carried the precursor virus to Europe during the First World War where it evolved into the virulent strain that spread around the globe in the fall of 1918 (Langford, 2005: 473, 475, 491–92). Both Pettit and Bailie (2008: 22) and Humphries (2013: 70–71) premise similar arguments on the claim that China suffered less severely in the pandemic. What evidence justifies these conclusions?
The task of measuring the impact of the influenza pandemic in China is not an easy one. The country inherited by the Republic of China in 1912 had for years been weakened by aggressive foreign powers. Colonial enclaves dotted China’s coastal regions (among them Hong Kong, Shanghai, Taiwan, Kiaochow/Tsingtao, and Kwantung). By 1918 central government authority, already weak when the Republic of China was founded, had been weakened further by the warlord era that had begun in 1916. Modernization plans were stymied, and there were almost no vital statistics reporting systems in any locality under Chinese government control (Yip, 1995: 103–4). Thus, almost no statistical data on the epidemic have come to light from Chinese government sources (Beijing and Huolu are exceptions), so it is not surprising that Langford exclusively uses English-language sources to make his case.
Langford relies heavily on the limited statistical information available from two foreign-controlled enclaves, the Shanghai International Settlement and the colony of Hong Kong. He restricts his analysis of the Shanghai and Hong Kong data to simple comparisons of CDRs for the epidemic years to prior years and to monthly reports of influenza deaths (which do not include the many deaths attributed to pneumonia that resulted from epidemic influenza). Langford makes no attempt to assess data quality and makes little use of the cause-specific data that are available. He then proceeds to cull impressionistic reports from the Chinese Maritime Customs and Post Office, and a medical report on missionary institutions in Canton, for evidence of the epidemic’s impact. This section critiques Langford’s use of each of these sources in turn. 10 The next section then proceeds to present mortality data documenting the influenza epidemic drawn from seven additional Chinese-majority jurisdictions.
Shanghai
The reports from the Shanghai International Settlement cover that part of Shanghai administered separately from both the Chinese city and the French Concession. In comparing the CDRs of 1918 to those for the years 1913–1917, Langford finds a decline rather than an increase in mortality in the epidemic year 1918, especially for the Chinese population (where the CDR of 12.8 per 1,000 was only 88 percent of the 1913–1917 average) but also for the non-Chinese (“resident foreign”) population (where the CDR of 16.5 per 1,000 was only 95 percent of the 1913–1917 average). It is questionable to conclude, as Langford (2005: 478–79) does, that this indicates “that the outbreak may have been less severe” among the Chinese population without first acknowledging that death reporting for the Chinese population of Shanghai was certainly much less complete than for the non-Chinese population. Observers knowledgeable of the living conditions of Shanghai’s Chinese populace will immediately ask how the CDRs reported for the Chinese population in every year from 1913 to 1920 could be considerably lower than those for the non-Chinese population, despite the latter’s higher standards of living and privileged access to health care (cf. Nakajima, 2004: 125–26). The obvious answer is reporting deficiencies. Yet Langford’s overall conclusion relies heavily on the Shanghai data, which he characterizes as “extremely strong” (Langford, 2005: 486), but which are shown below to be seriously incomplete, especially in their coverage of the Chinese population.
In 1918 the population of the International Settlement was estimated (there was no census) at 21,000 foreigners (around 10,000 Europeans and Americans, 10,000 Japanese, and 1,000 Indians) and 659,000 Chinese (Annual Report of the Shanghai Municipal Council, 1918: 98A; Feetham, 1931, 1.51–52). The disparity in quality of the vital statistics reporting on the two populations is evident in the sources. The annual Health Officer’s Reports of the International Settlement give detailed reports on the deaths of the small foreign community broken down by age and by thirty-three causes of death up through 1922. The same reports give no age breakdown for the Chinese deaths and use a truncated cause-of-death classification of six named infectious diseases plus a catchall “all causes” category. 11 This should not be taken to imply that deaths from infectious disease were carefully reported; there was in Shanghai no compulsory registration of deaths, nor any legal obligation to report cases of notifiable infectious diseases (e.g., cholera, smallpox). Case reporting depended on the voluntary cooperation of medical practitioners qualified in Western medicine, who served only a fraction of the Chinese population. The simple death statistics presented for the Chinese population reflected a general lack of diligence in collecting data on the large Chinese population in the years surrounding the 1918 epidemic that affected not just the detail with which the data were reported, but also the completeness and quality of reporting (Annual Report of the Shanghai Municipal Council, 1918: 98A–102A, 1923: 100; Stanley, 1918: 6–9; Feetham, 1931, 2.48).
A contemporary League of Nations report questioned the accuracy of death reporting for the large Chinese population within the settlement, and found the death rates for the Chinese population to be “incredibly low” because so many Chinese deaths were going unreported (White, 1923: 84–85). The Health Officer’s Report in 1923 included a critique of the death reports for the Chinese population as “only approximate, for there is no compulsory registration of deaths in Shanghai,” and noted that “Chinese deaths are recorded by Health Inspectors after daily investigation, and only a few [are] certified by qualified medical practitioners” (Annual Report of the Shanghai Municipal Council, 1923: 100). The report’s discussion of Chinese deaths goes on to add that “this death-rate does not give an accurate index of the health of the Chinese in the Settlement, for when the Chinese suffer from chronic or incurable disease they often leave the Settlement for their native villages and there await death. . . . In Shanghai, where over-crowding, destitution and precarious modes of living are common, and where Tuberculosis is prevalent among the lower classes, the real death-rate is undoubtedly higher than the recorded death-rate” (Annual Report of the Shanghai Municipal Council, 1923: 100). The 1923 annual report notes that the recorded death rate for Chinese in Hong Kong, where an ordinance made registration of deaths compulsory, was 25.47 per 1,000 (where the sick also commonly left the colony for home villages in China); the report also illogically claims that in Shanghai “probably not more than 15 percent of deaths among Chinese escape record,” despite a recorded death rate for Chinese of only 10.3 per 1,000, less than half the Hong Kong rate (Annual Report of the Shanghai Municipal Council, 1923: 100).
The 1930 Health Officer’s Report gave more evidence of serious defects in death reporting for the Chinese population when it began to list separately deaths classified as “exposed corpses,” and revealed that this category accounted for 37 percent of total Chinese deaths (Annual Report of the Shanghai Municipal Council, 1930: 122).
In explanation, it has been customary for ages for the Chinese of certain classes to place “unwanted” bodies on vacant ground and the records of the Shanghai Public Benevolent Cemetery, the principal organization which collects such bodies for burial, show that the larger proportion are collected from the outlying sections of the Eastern and Western districts. While it is obviously impossible to give a detailed analysis of the causes of these deaths, nevertheless observations made by officers of the Department show that many of the bodies are those of beggars and indigents, a type unfortunately all too common in all large cities. The remainder consist mainly of “unwanted” bodies, comprising stillborn children, infants, and those of transient residents who happen to die here without relatives or funds. (Annual Report of the Shanghai Municipal Council, 1930: 123)
Based on these revelations there can be no doubt that the reports of deaths in the Chinese population of the International Settlement are so seriously defective that they cannot be relied on as a measure of influenza-related mortality and that the heavy reliance placed on them by Langford and the literature in general is misplaced.
Although the CDRs reported for the foreign population were likely based on more complete reporting than for the Chinese population, it should be reiterated that there was no compulsory registration of deaths in the International Settlement. Langford (2005: 479) acknowledges in the Shanghai case the probability of “errors in the estimation of crude death rates,” but suggests measures of relative changes over time may control for such errors within populations. It seems doubtful that relative measures are sufficient to compensate for the much greater underreporting of Chinese deaths, which was likely to worsen in epidemic years.
The simple ratios of annual CDRs for all causes used by Langford are imperfect indicators of the severity of excess influenza mortality in any case, as offsetting improvements in other cause categories mask increases in deaths resulting from influenza. In the reference years used by Langford, 1913–1917, there were significant outbreaks of scarlet fever (1917), cholera (1914), and smallpox (1913–1915, 1917). The absence of such outbreaks made 1918 a relatively healthy year in Shanghai, despite the flu (Annual Report of the Shanghai Municipal Council, 1922, 113A). It was the decline in deaths due to these other infectious diseases that caused the 1918 CDR to fall below the average of previous years, not the less severe outbreak of influenza fallaciously posited by Langford (2005: 478–79). Thus Shanghai’s 1918 CDR for the better-reported non-Chinese population fell by 7.88 per 10,000 over the average for 1916–1917, though the influenza-related death rate rose by 8.87 per 10,000 (Tables 1 and 2).
A distinctive characteristic of the 1918 epidemic was a milder spring wave followed by the emergence in the fall months of a virulent influenza strain that caused a spike in influenza-related deaths. The Health Officer’s Report states that the 1918 outbreak of influenza commenced at the end of May and subsided by the end of June, then reappeared in the last week of September until the second week in November; it was the second, fall wave that caused the most deaths (Annual Report of the Shanghai Municipal Council, 1918: A108). Deaths from influenza alone among Chinese showed two peaks, one in June (152 deaths) and one in October–November (247), which account for 95 percent of the reported annual total (418) (Langford, 2005: 476). Monthly totals of PBI deaths are available only for the non-Chinese population; the proportion of the annual PBI deaths in October and November 1918 rose to 37.8 percent, more than twice the 1916–1917 average of 16.3 percent, confirming the severity of the fall wave in Shanghai (Table 3). As in many of the jurisdictions impacted by the pandemic, the non-Chinese population of Shanghai experienced an increase in 1918 of the proportion of deaths among young adults (Table 4). In the spring (March and April) of 1919, the virulent strain returned to Shanghai, resulting in reported death rates for influenza and influenza-related causes (non-Chinese only) that were even higher than those reported in 1918 (Table 1).
In conclusion, there is no reliable information documenting the death rates for the Chinese population of Shanghai in the period 1916–1920. The reporting on the non-Chinese population is better, but, given the lack of mandatory reporting, still incomplete. Nevertheless, the increases in the influenza-related death rate, the peak in the fall death rate, and the elevated mortality of young adults conform to patterns observed around the globe in 1918, and cast doubt on the hypothesis that foreigners in Shanghai enjoyed any special immunity.
Hong Kong
The second population whose death reports Langford relies on is that of Hong Kong. In Hong Kong, he finds increases in CDRs in 1918 and 1919, but that these increases were much lower for the Chinese than the non-Chinese populations, suggesting a weaker epidemic impact among the Chinese. However, he relies on the 1918 Chinese CDR that is miscalculated in the original report as 24.5 when it should have been 29.59 (discussed below); when corrected, the ratio of the 1918 CDR to the average CDR for 1913–1917 rises considerably (from 1.09 to 1.32), undermining Langford’s interpretation of a low epidemic impact. The ratios comparing the 1918 CDR to previous years for both the Chinese (corrected to 1.32) and non-Chinese (1.56) populations in Hong Kong are higher than or comparable to the ratios Langford calculates for the United States (1.33), England and Wales (1.20), and Ceylon (1.21) (Langford, 2005: 478).
Analysis of the Hong Kong data begins by first assessing the quality of the Hong Kong death reports. As mentioned above, death registration among Chinese in Hong Kong was mandatory and thus is considered to have been much more complete than in Shanghai. However, underreporting of deaths among infants and young children in Hong Kong was an acknowledged problem, as there was severe under-registration of births (and thus of infant deaths) among the Chinese population (White, 1923: 72–73; Hong Kong Medical and Sanitary Reports, 1916: 15; Report on the Census of the Colony for 1921, 1921: 11–12).
Hong Kong’s General Registration Office was established by ordinance in 1896 for registration of both births and deaths. To obtain a burial permit, the regulations required registering a death in the district within which the death occurred at death registration offices typically located within local police stations (Hong Kong Sanitary Report, 1922: 14–15). 12 Chinese deaths were reported by registration areas, all of which were on Hong Kong Island or the Kowloon Peninsula. The figures thus do not include deaths from the New Territories (Hong Kong Medical and Sanitary Reports, 1918: 21, Table 1; Hong Kong, 1918). The annual Medical and Sanitary Reports estimated the mid-year populations of Chinese and non-Chinese in the colony each year. The Chinese population in 1918 included 93,400 in the New Territories and 454,600 in the “death-registration area” of Hong Kong Island, Kowloon, and the “population afloat” (referring to the Tanka/boat people) (Hong Kong Medical and Sanitary Reports, 1918: 9).
Because Chinese deaths were only reported for the area of Hong Kong excluding 266 square miles of the New Territories that were “outside the [Sanitary] Board’s jurisdiction” (Hong Kong Medical and Sanitary Reports, 1918: 7), the corresponding “population at risk” used in the reports was that of the death-registration area (454,600). The death rates for Chinese reported in the annual Medical and Sanitary Reports were calculated as the ratio of registered deaths to the Chinese population of the death registration area (cf. Hong Kong Medical and Sanitary Reports, 1917: 9, 14, 1919: 9, 14). However, the 1918 death rate for all causes among Chinese (13,450 deaths) is given in the text of the 1918 Medical and Sanitary Reports as 24.5 per 1,000 (Hong Kong Medical and Sanitary Reports, 1918: 14), which is erroneously calculated on a Chinese population denominator of 548,000 that includes the New Territories. When calculated correctly using the population of the death registration area (454,600) (as was done in all previous years) the 1918 CDR is 29.59 per 1,000 (the rate is correctly reported in three sources: Hong Kong, 1918: 18; Hong Kong Vital Statistics, 1918: 28; Hong Kong Blue Book for the Year 1918, 1919: 245 [rate for Chinese and foreign populations combined]). Langford unwittingly accepts the mistaken, inconsistently calculated rate when he compares Hong Kong’s 1918 CDR to other years (Langford: 2005: 478). 13
Reported PBI deaths in 1918 for the Chinese population of the death-registration area (454,600) in Hong Kong totaled 3,326, an excess of 1,178 deaths over the 1916–1917 average (2,148). As Table 1 shows, the 1918 excess in the influenza-related death rate over the baseline rate for non-epidemic years (1916–1917) is 22.7 per 10,000. This is a conservative estimate, given that Chinese deaths, especially among infants and young children, were underreported, and causes of death were more likely misassigned to vague catchall categories in the Chinese than in the foreign population.
The cause-specific measure of the epidemic excess indicates that the impact of the influenza pandemic in Hong Kong was much greater than previous accounts have suggested. First, this is because unwary scholars have repeated the error in reporting the irregular CDR among Chinese in 1918 caused by the miscalculation of the population at risk in that year’s Medical and Sanitary Reports. Second, use of the CDR instead of the cause-specific PBI rate masks the true impact of the influenza epidemic. This is because the CDR comprehends all causes, and is affected not just by fluctuations in influenza-related deaths, but also by the rise and fall of deaths attributed to other causes. For example, in 1918 not just influenza but a severe outbreak of cerebrospinal meningitis in the spring (957 deaths) and a large number of injuries (782 deaths compared to 184 in 1917, primarily due to the Happy Valley Racecourse fire in February) raised the Chinese death total, while a significant decline in smallpox (26 deaths compared to 549 in 1917) offset some of the increase. Nevertheless, the substantial increase in the CDR in 1918 of 53.7 deaths per 10,000 may reflect some influenza-related deaths not captured due to misassignment of cause (Table 2). These fluctuations in unrelated causes of death demonstrate why CDRs cannot reliably measure the impact of influenza outbreaks. Langford made no use of the cause-specific reports available for both Chinese and non-Chinese populations of Hong Kong. 14
The 1922 Sanitary Report includes for the first time reports of deaths by month (without regard to nationality) for influenza (but not for pneumonias or respiratory causes) for the years 1918–1922 inclusive. The high proportion of influenza deaths in the fall months of 1918, October to December (232 of an annual total of 405, or 57.3 percent), confirms the impact of the fall wave in Hong Kong (Hong Kong Sanitary Report, 1922: 64). However, the exclusion of influenza-related respiratory deaths (pneumonia, bronchitis) results in a low fall influenza death rate of 4.96 deaths per 10,000 (Table 3). That no reports of deaths by age are made available in the Hong Kong Medical and Sanitary Reports until 1921 prevents analysis of differential impacts by age.
In sum, the Hong Kong data on death rates and deaths by cause are more reliable than the data available for Shanghai in the 1918 period. However, the Hong Kong data are still known to underreport the number of deaths, especially among infants and young children, in the Chinese population, and the quality of the assignment of causes to deaths is likely to be inferior to that for the non-Chinese population and inferior as well to the quality of diagnosis for most Japanese-administered territories discussed below.
Trends in influenza-related deaths show an increase in 1918 of 22.7 per 10,000 (slightly higher than the increase for non-Chinese of 20.6), and confirm a significant, if moderate, impact of the epidemic on Hong Kong’s Chinese population, thereby controverting the notion that the pandemic had a milder impact on the Chinese than the non-Chinese population and that the Chinese population enjoyed a special immunity. Thus neither the Shanghai nor the Hong Kong data can sustain a claim that the impact of the 1918 influenza epidemic on the Chinese population was peculiarly mild.
Chinese Maritime Customs Trade Reports as a Source on the Epidemic
Apart from the few quantitative reports from Shanghai and Hong Kong, the English-language sources on the influenza epidemic in China used by previous studies are restricted almost entirely to qualitative reports containing estimates of prevalence and mortality, sometimes based on selected clinic populations, but often merely impressionistic. To extend coverage to all of China, Langford’s 2005 article contributes a systematic review of the Returns of Trade and Trade Reports (hereafter, the Trade Reports) published by the Chinese Maritime Customs. 15 These annual reports of trade from treaty ports spread across China sometimes contain references to epidemics, presumably when they have a serious impact on economic activity. However, most Trade Reports never ventured beyond the narrowly economic, and were never intended to provide reports on health. Most were authored by Westerners who had limited contact with the Chinese communities in which they were stationed. Thus failure to mention the influenza epidemic in these reports cannot be taken as evidence of an insignificant outbreak, as Langford (2005: 484) implies. The Chinese Maritime Customs published a parallel series, the Medical Reports, that compiled reports on local health conditions by medical doctors posted in each port between 1871 and 1910. Unfortunately, publication of this annual series was discontinued around 1910. 16 Another parallel series of annual reports, the British Consular Trade Reports, often ranged into general political and social conditions in the treaty ports (collected as part of Irish University Press Area Studies Series, 1971–1972). However, the publishing of individual reports for each treaty port or consular outpost was discontinued around 1916, because of the First World War. The successor reports published by the Board of Trade’s Department of Overseas Trade (the earliest I have seen is Fox, Rose, and Brett, 1920, which covers 1919) abandon port-by-port reporting for a general report on British trade in China as a whole. Thus, by the arrival of the 1918 influenza epidemic, two key Western-language sources reporting on local conditions throughout China had been discontinued. Langford is thus left with the Trade Reports as his primary source, which he supplements with provincial-level reports of the Chinese Post Office.
Langford tabulates the brief and often sketchy references to influenza outbreaks (or their absence) in the Trade Reports from a total of forty-five treaty ports in the years 1918 and 1919 (Langford, 2005: 480–81, 496–500). In both years, reports from sixteen ports made no reference to influenza. Langford finds references to influenza (usually no more than one or two sentences of anecdotal impressions) in the reports of twenty-nine ports (twenty-one in 1918, thirteen in 1919 [of which seven were also mentioned in 1918], plus two based on the Chinese Maritime Customs Decennial Reports 1912–21). The reports mentioning influenza come from all the coastal provinces, and from the interior (e.g., Szechwan, Hupeh) as well, suggesting that influenza spread throughout China in 1918. Outbreaks described as serious in 1918 include those at seven ports (Mukden, Wanhsien, Shasi, Wenchow, Kongmoon, Sanshui, and Szemao), and three describe villages in the vicinity as being hard hit (Lungchingtsun, Yingkou, and Hankow). Langford found Trade Reports suggesting the epidemic was “relatively mild” in twelve ports in 1918; however, his assessments, based on the references he provides, are open to question. He classifies as “mild occurrences” reports that mention the mild spring wave but which comment that the fall wave was worse (Antung, Chungking, Nanking, Soochow). Ports such as Ningpo, Chinwangtao, and Dairen, for which no mention of influenza is made in the Trade Reports, are reported in other sources to have experienced severe outbreaks (Sugg, 1919; “Influenza plays havoc in province of Chihli,” 1918; “Influenza epidemic at Dairen,” 1918), undermining the inference that absence of evidence from the Trade Reports is evidence of absence when it comes to influenza outbreaks.
Out of twenty-one district (usually provincial) reports included in the annual reports of the Chinese Post Office for 1918, Langford finds references to the influenza epidemic in nine reports, and in eight of these the comments suggest a serious epidemic, often affecting postal personnel (contra Langford 2005: 484, 502; Report on the Working of the Chinese Post Office for 1918, 1919). The reports of nearly every province are dominated by tales of banditry, political unrest, and military operations interfering with postal work, which suggests that outbreak of disease was only one of several concerns. Included are reports of serious influenza outbreaks in China’s northwest (Shensi, Kansu, Sinkiang), areas not covered in the Trade Reports, providing more evidence that influenza spread throughout China in 1918. It is noticeable that the Chinese Post Office reports give a more consistent picture of a severe epidemic than do the Western-authored Trade Reports.
On the basis of his review of the Trade Reports and Post Office reports, Langford comes to two non-controversial conclusions: that influenza was widespread in China in 1918–1919 and that the severity of the epidemic varied from locality to locality (Langford, 2005: 486).
Langford (2005: 487–88) finds that the impressionistic accounts in the Trade Reports tend to imply a heavier impact on the Chinese than the foreign populations, though not without some ambiguity. Overall, he favors a conclusion that the impact of the epidemic in 1918 on both the foreign and the Chinese populations in China was mild (at least in some areas) compared to those recorded in England and Wales and the United States. 17 He then speculates that the moderate impact may have been due to earlier circulation of a mild precursor virus that conferred immunity on populations resident in China (whether Chinese or foreign) and that protected them when the virulent fall wave of 1918 appeared (2005: 491–92). (The mild nature of the precursor virus means its presumed circulation cannot be documented.) Langford posits a connection between the hypothesized mild precursor virus and the virulent fall 1918 strain in the Chinese workers sent to France during the First World War, an argument critically assessed in a separate article (Shepherd, 2022).
The Canton Delta Region
Claims that influenza mortality in 1918 in the southern city of Canton was low (Langford, 2005: 475, 488; Shortridge, 1999: 385) are based on qualitative reports and selective hospital case statistics presented by Dr. William Cadbury (1918, 1920). Cadbury cites no quantitative data on the general population, understandably as Canton’s first (and failed) attempt to inaugurate death registration came only in 1921 (White, 1923: 80; Yip, 1995: 103).
Cadbury presents an account of the epidemic in Canton based on case reports he collected from Western doctors and school personnel representing a total of nine institutions, primarily Christian-run hospitals and schools, except for the local branch of the Chinese Post Office, from which he obtained qualitative data only (Cadbury, 1920: 5–7). Cadbury presents case data showing that the autumn wave was the most severe, resulting in the greatest number of cases (552 in his sample, compared to 98 in June and 74 in December–January), with bronchial symptoms leading to pneumonia in the most severe cases (25 in total), and causing the largest number of deaths (13) (Cadbury, 1920: 6). There is no discussion of how representative a sample drawn from elite schools and missionary hospitals might be of the general Chinese population.
Cadbury (1920: 15–16) ends his article discussing the mortality impact of the epidemic and cites reports of high mortality from the United States, the Philippines, and India. Based on his observations, he concludes that “in Canton the disease seems to have been very much less malignant than in the United States, although there were rumors of entire families being wiped out, and of 500 deaths having occurred in one block of a city street. At the Canton Hospital there were no deaths in June, and only four among 27 cases in October. Rumors indicate that the disease was much more fatal in certain outlying districts of Canton” (Cadbury, 1920: 15, italics added). Cadbury’s 1918 hospital report contains similar qualifying language: “The mortality in Canton . . . was not great, but in some of the surrounding towns and villages it is reported to have been very serious” (Cadbury, 1918: 67). Thus in both reports Cadbury refers to the possibility of much more severe outbreaks in the vicinity (outside the purview of missionary institutions), but he took no steps to verify these reports. Fortunately, we have additional contemporary reports that document the epidemic’s impact in the neighborhood of Canton.
The accounts in the Trade Reports for Kongmoon (Jiangmen 江門), Samshui (Sanshui 三水), and Lappa (拱北), all treaty ports in the Canton Delta region, confirm a severe impact in Canton’s “outlying districts.” The account of the fall wave in Kongmoon refers to a virulent form of influenza and many deaths, and the account of Samshui reports that influenza “claimed many victims” (Langford, 2005: 497). The Chinese Maritime Customs Decennial Reports 1912–21 (1924) entry for Lappa (the Portuguese name for the island of Wanzai 灣仔, opposite Macau) reports 270 cases of influenza and 120 deaths in 1918 in Macau, a much higher case fatality rate (44 percent) than experienced by Cadbury’s patients (15 of 552 or 2.7 percent) (Langford, 2005: 500). (For additional comments on the situation in Macau see the Online Supplementary Materials.) Newspaper accounts report influenza “raging” elsewhere in Guangdong (Lungmen and Conghua districts) and in Guangxi (“Influenza raging in Kuangsi, large number of deaths among Chinese,” 1918; “Influenza raging in Kuangtung,” 1918). Scholars who have cited Cadbury to support the proposition that the epidemic had only a mild impact in South China (Langford, 2005: 475, 488; Shortridge, 1999: 385; Cheng and Leung, 2007: 361–62) have ignored Cadbury’s own qualifications of his findings and the readily available evidence on nearby communities from the Trade Reports.
In conclusion, careful examination of the available evidence undermines the claim that southern Chinese were only mildly affected by the pandemic and enjoyed some special immunity. That these southern populations enjoyed no special immunity and were seriously affected receives strong support from the better-quality vital statistics data available from the Straits Settlements and Taiwan, discussed below.
The Impact of the Pandemic in Seven Additional Chinese-Majority Jurisdictions
The flawed and misstated sources relied on by Langford and many other scholars in support of the thesis that the impact of the 1918 influenza pandemic on the Chinese population was unusually mild (the Shanghai death reports, the misstated Hong Kong death rate, the anecdotal reports by customs officers, and Cadbury’s medical report on influenza in Canton’s missionary institutions) have now been critically assessed and their limitations identified. Upon examination, none of these sources provide reliable evidence supporting a mild mortality impact, and scholars must not take them at face value.
We next introduce what previous discussions have largely or entirely ignored: seven Chinese-majority populations for which measures of influenza-related epidemic excess mortality can be derived from quantitative reports and which are for the first time presented together here. In five of the seven jurisdictions death reporting was mandatory, and vital statistics and census data were published at regular intervals (only Weihaiwei and Huolu county did not make regular reports). Just as we have in the cases of the mortality statistics from Shanghai and Hong Kong, it is incumbent on any analysis using data from these seven jurisdictions to subject their death reporting systems to similar scrutiny. To conserve space (and spare impatient readers), an abbreviated introduction to each jurisdiction’s vital statistics sources and a brief report of the results will be given here, reserving detailed quality assessments of each jurisdiction’s vital statistics for the Online Supplementary Materials, where there is also a general note on cause-of-death reporting. Comparison among the jurisdictions is made easy in Tables 1–4, where consistently calculated measures of the mortality impacts of the epidemic are presented for all the jurisdictions covered (including reference populations). The tables are organized geographically from north to south for the East Asian jurisdictions. We first treat the mortality data from two additional British-administered jurisdictions, the Straits Settlements and Weihaiwei; then two Chinese-administered jurisdictions, Beijing and Huolu county; and then a series of three Japanese-administered jurisdictions, the KLTZ, Kiaochow/Tsingtao, and Taiwan.
Straits Settlements
Highly reliable censuses and vital statistics reports are available from the British crown colony of the Straits Settlements in the years surrounding the 1918 epidemic. The Chinese population of 432,764 (52 percent of the total Straits Settlements population; see Table 2) was made up largely of sojourners who originated in South China (70 percent were born there) and remained in constant contact with China’s southeastern coast (Nathan, 1922: 95, 227). 18 Thus the impact of the epidemic on the Straits Settlements Chinese provides a good window onto the statistically undocumented impact on their compatriots in the home districts of Southeast China and any alleged immunities they might have enjoyed.
The 1918 excess in the influenza-related death rate over the baseline rate for non-epidemic years (1916–1917) in the Straits Settlements is 65.3 per 10,000 (Table 1) for all nationalities. The increase in the all-cause CDR for Chinese alone was an even higher 102.8 per 10,000 (Table 2). The influenza-related excess death rate of 65.3 per 10,000 greatly exceeds that for the United States (40 per 10,000) and England and Wales (33 per 10,000). The pattern of the impact of the influenza epidemic in 1918 in the Straits Settlements matches that from societies around the world: a virulent fall wave that to an unusual degree fell on young adults (Tables 3 and 4). In sum, the severity of the epidemic among the Straits Settlements Chinese makes it highly unlikely that their home communities in Southeast China, with which they were in constant contact, could have been protected by an immunity that was not shared with those in the Straits Settlements. (Additional details on the sources for the Straits Settlements are included in the Online Supplementary Materials.)
Weihaiwei
On the northern side of the Shandong Peninsula in 1918 sat the British-administered territory of Weihaiwei, a naval port acquired in 1898 in the rush to counter the Russian acquisition of Port Arthur in Liaodong across the Bohai Strait and the German acquisition of Kiaochow/Tsingtao. The 1918 Colonial Office report for Weihaiwei states: “The epidemic of influenza was universal throughout the Territory in a very virulent form. In one family of eight six died of the disease, and nine hundred deaths were reported. But it is certain that many deaths occurred which were never reported. It is not possible to give accurate mortality statistics, as registration of deaths is not compulsory in the Territory” (Weihaiwei, 1918: 4). Despite the undercount, nine hundred influenza deaths out of Weihaiwei’s 1918 population of 147,177 (Report on the Pandemic of Influenza, 1918–19, 1920: 386) corresponds to an influenza-related death rate of 61.1 per 10,000, a high level that exceeds the PBI death rates reported from the United States (56) and England and Wales (57) (Table 1).
Beijing
Beijing was the one city under Chinese administration that registered deaths during the period of the influenza epidemic (Gamble, 1921; Campbell, 1995). The challenges presented by the registration system and its improving but unsystematic classification of causes of death are detailed in the Online Supplementary Materials. The respiratory disease death rate in 1918 is estimated at 96.8 per 10,000, an increase of 23.1 per 10,000 over the 1916–1917 rate (Table 1). The all-cause CDR actually fell in 1918 by 10.6 per 10,000, despite the influenza epidemic, suggesting that, as was the case for other populations (e.g., Japanese in the KLTZ and Kiaochow/Tsingtao, Europeans in the Straits Settlements, non-Chinese in Shanghai), 1918 in Beijing was an otherwise healthy year (Table 2).
The finding of a significant mortality impact from the influenza epidemic accords with two reports from medical journals of heavy influenza mortality in Beijing. E. T. Hsieh reported that the “recent epidemic appeared in Peking on October 6, 1918, and soon spread all over the city. It was prevalent also in Tientsin [Tianjin], Paotingfu, and many cities along the railroad lines. This proved to be a serious epidemic, with a high percentage of deaths” (Hsieh, 1918: 129). A second unsigned notice in the China Medical Journal relays journalistic reports that “in Peking fully 50% of the Chinese have been affected and the mortality has been heavy” (“Influenza,” 1918: 608). A journalist’s report of November 9, 1918, noted the ravages of influenza in Beijing on both the native and foreign populations, and the unusual number of funerals parading through the streets (“Flu hits Peking hard,” 1918). These reports of a significant epidemic impact are consistent with the findings from the quantitative data from Beijing. (Additional details on the sources for Beijing are included in the Online Supplementary Materials.)
Huolu County
The sole original contribution to be found in the recent literature on the pandemic from the People’s Republic of China is that by Hao Hongnuan (2015). Hao discovered in the Hebei Provincial Archives a Huolu county report documenting the number of cases and deaths from the 1918 epidemic. Hao’s research documents the accumulating newspaper and other reports of serious epidemic outbreaks in November 1918, and the growing realization that the symptoms matched those reported from abroad for the deadly “Spanish” flu. The newspapers also noted the booming business for coffin shops and pharmacies (Hao, 2015: 66–67). Alarmed by the crisis, the Zhili provincial governor issued an order dated December 19, 1918, directing each county to compile reports on deaths from the October and November outbreaks 感冒時疫 (Hao, 2015: 67). To date, the report from Huolu (modern Luquan 鹿泉 district of Shijiazhuang 石家庄 city) is the only Zhili county report to come to light. The Huolu report was compiled and sent to the provincial governor in May 1919 (Hao, 2015: 67). The report contains two tables detailing the number of flu cases and the number of deaths, one organized by age and one by ward. Deaths in October and November 1918 totaled 1,945 out of a county population estimated (from a 1914 survey) at 262,049, resulting in a death rate of 74.2 per 10,000. Cases were ten times the number of deaths, at 19,917. As was typical of the 1918 epidemic, young adults aged twenty-one to forty years accounted for a high proportion (47 percent) of the total deaths (915 of 1,945) and a high proportion (46 percent) of total cases (9,150 of 19,917). Unfortunately, we know little of the procedures by which these numbers were collected. Deaths, which require burial permits and are marked by public funerals and burials, are much easier to document than case totals, which are much more difficult to verify (many suffer quietly at home, making no contact with medical or other officials). We do know that efforts to reform and strengthen the Zhili police, along Japanese lines, including responsibility to maintain population registers, were initiated in the late Qing and continued by Republican administrations, which suggests ward police had the capacity to compile the report (MacKinnon, 1980: 155–63; Reynolds, 1993: 162–69; Duara, 1988: 60–61). But this is a one-time report rather than a series of annual reports, and thus we are not able to calculate the epidemic excess, nor are we able to ensure that all PBI deaths were included in the totals. Nevertheless, this remains a rare and valuable source on the epidemic’s impact on an entire county and, like Weihaiwei, deserves to be included among our quantitative sources.
Three Japanese-Administered Chinese-Majority Jurisdictions
We are able to calculate measures of influenza-related epidemic excess mortality for both the majority-Chinese and the minority-Japanese populations living in three Japanese-administered jurisdictions that enforced the reporting of deaths, the KLTZ, Kiaochow/Tsingtao, and Taiwan. In addition to vital statistics reports, the Japanese sources on the KLTZ, Kiaochow/Tsingtao, and Taiwan also include valuable data on clinic visits which reveal increases in respiratory complaints during the epidemic and differences between Japanese and Chinese in access to medical care. The Online Supplementary Materials also include notes on Wataru Ijima’s (2003) and Akira Hayami’s (2015) earlier assessments of the epidemic in Japan’s colonies, neither of which used the primary vital statistics sources available for the KLTZ, Kiaochow/Tsingtao, and Taiwan.
The KLTZ
Japanese administration of the KLTZ in Northeast China began in 1905, following Japanese victory in the Russo-Japanese War. The Japanese-administered area covered 1,300 square miles at the southern end of the Liaodong Peninsula (including the port cities of Port Arthur and Dairen) and one hundred square miles of railway zone (which included Japanese-administered districts in large cities like Mukden [Shenyang 瀋陽] and Changchun 長春) (Report on Progress in Manchuria, 1929: 53; Buck, 2000: 73–74, 77; Matsusaka, 2001: 71, 86–87, 174–75). The KLTZ encompassed a sizeable Chinese population of 599,231 and a Japanese population of 129,693 (Table 2). Highly reliable censuses, vital statistics, and clinic visit reports are available from the KLTZ in the years surrounding the 1918 epidemic (see the Online Supplementary Materials for details).
In 1918 the death rate for influenza-related causes for the Chinese population showed increases of 65 per 10,000 over the 1916–1917 averages, which greatly exceeds those of the United States (40 per 10,000) and England and Wales (33 per 10,000) (Table 1). The lower increase of 25 per 10,000 for the Japanese colonialists reflects their greater access to public health services (documented in the clinic visits reports) and the higher standard of living which helped protect them from the worst effects of the epidemic. The concentration of influenza-related deaths in the fall (Table 3) and the increase in the proportion of deaths among young adults aged 20–39 (Table 4) fit the global pattern typical of the 1918 influenza epidemic. (Additional details on the sources for the KTLZ are included in the Online Supplementary Materials.)
Kiaochow/Tsingtao
The leased territory of Kiaochow and its port city of Tsingtao were under Japanese administration from November 1914 to the end of 1922. In 1918 the population included 180,944 Chinese and 25,894 Japanese (Table 2). Several indicators suggest that the vital statistics reporting system was still developing in 1918, resulting in underreporting and problematic assignments of causes of death; these challenges are detailed in the Online Supplementary Materials. In 1918 the increases in the death rate for influenza-related causes are a modest 24.2 per 10,000 for the Chinese and only 1.8 for the Japanese (Table 1). These results (likely affected by misclassification of causes) suggest a moderate epidemic impact, but two additional indicators provide evidence of a more serious epidemic impact. First, the clinic visit reports confirm large increases in visits for respiratory complaints in 1918 for both Japanese and Chinese. Second, the all-cause excess CDR for Chinese in 1918 is a high 75 per 10,000, and the very high excess CDR in the fall of 1918 for Chinese of 103.6 per 10,000 (Tables 2 and 3) surely reflects the impact of the virulent fall strain of influenza. Hayami (2015: 48) cites a December 1, 1918, newspaper report of an “epidemic that devastated villages near Qingdao (Tsingtao).” So, despite the flaws in the reporting of causes of death, there are important indicators that suggest a serious impact among the Chinese. (Additional details on the sources for Tsingtao are included in the Online Supplementary Materials.)
Taiwan
The censuses and vital statistics reporting systems in Taiwan have a verified reputation for a high degree of quality and completeness (Barclay, 1954; Shepherd, 2011). 19 Seized by the Japanese in 1895, Taiwan had its first of seven modern censuses in 1905, when annual publication of vital statistics data began, continuing through the early 1940s. Though governed as a colony of Japan, Taiwan retained its close trade connections to China’s southeast coast, and likely shared any immunities (as Langford hypothesizes) present in the coastal populations of China. A second census followed in 1915, and Taiwan’s population in 1918 stood at 3,420,604 Taiwanese and 152,092 Japanese. The PBI death rate in 1918 exceeded the 1916–1917 average by 62.36 per 10,000 for Taiwanese and 35.54 per 10,000 for Japanese in Taiwan (comparable to the 1918 PBI excess for Japanese in Japan of 33.3), reflecting both a serious impact and the privileged conditions enjoyed by the Japanese colonialists (Table 1). The concentration of PBI deaths in the fall of 1918, and the increase in the proportion of PBI deaths suffered by young adults fits the global pattern of the fall 1918 epidemic (Tables 3 and 4). Annual reports of clinic visits also show marked increases in both respiratory and infectious disease complaints in 1918 for both Taiwanese and Japanese in Taiwan. (Additional details on the sources for Taiwan are included in the Online Supplementary Materials.)
Reference Populations: Japan, Korea, Ceylon, England and Wales, and the United States
In order to provide measures by which to judge the relative severity of the pandemic across all jurisdictions, this section addresses the pandemic impact in several reference populations: Japan, Korea, Ceylon, England and Wales, and the United States. To ensure comparability, we compute for these non-Chinese populations the standard cause-of-death measures used throughout this article and present them in Tables 1–4. And as for the Chinese jurisdictions, we make quality assessments of each jurisdiction’s vital statistics, which can be found in the Online Supplementary Materials.
Conclusion
Based on a flawed analysis of CDRs from Shanghai and Hong Kong, Cadbury’s report on missionary institutions that ignores the Chinese population in the Canton Delta, and anecdotal accounts included in trade reports submitted by customs officials posted in treaty ports in China, Langford argues that the impact of the influenza epidemic in 1918 on both the foreign and Chinese populations in China was mild compared to that recorded in Britain and the United States. He speculates that the moderate impact may have been due to earlier circulation of a mild precursor virus that conferred immunity on populations resident in China (whether Chinese or foreign) that protected them when the virulent fall wave of 1918 appeared. The shortcomings in the Shanghai and Hong Kong death reports documented here undermine Langford’s claims of a mild impact, and evidence based on vital statistics newly gathered from seven additional majority-Chinese populations shows that the impact of the 1918 influenza epidemic was in many cases greater than the impact on the populations of the United States and England and Wales. Langford’s hypothesis of Chinese immunity from a precursor virus can thus be rejected. (Additional evidence of the impact of influenza on the Chinese laborers in France, and on the populations in the provinces from which the laborers were recruited, confirms this conclusion; see Shepherd, 2022).
As shown in Tables 1 and 2, in seven populations—Japanese in Korea, the KLTZ, Kiaochow/Tsingtao, and Taiwan; non-Chinese in Shanghai; and Chinese in the KLTZ and Beijing—when the 1918 death rate is compared to the average death rate in previous years, the influenza-related excess is greater than the crude all-cause excess. In these cases, the lower-than-expected CDRs reflect the absence in 1918 of non-influenza-related infectious disease outbreaks common in earlier years, and confirm that 1918 in many places was an otherwise healthy year, excepting the influenza pandemic. These examples demonstrate how misleading the widespread use of the crude all-cause death rate (which leaves uncontrolled the fluctuations occurring in other infectious diseases) can be as a measure of the influenza-related epidemic excess.
Returning to Langford’s precursor virus hypothesis, the inclusion in our tables of the Japanese colonial territories of Korea, the KLTZ, Kiaochow/Tsingtao, and Taiwan (and their clinic-visit data), and of Hong Kong and the Straits Settlements, show that the privileged class of colonialists consistently benefited from their higher standards of living and better access to medical treatment compared to the colonized populations, which suffered more severely from the 1918 epidemic. Class, rather than immunity, better explains why death rates among the colonial elites were lower than those among the colonized, and often lower than those of the non-elite masses in the home countries.
The four tables provided here contribute for the first time consistently calculated death rates circa 1918 from seven Chinese populations (comprising more than 5.8 million people) in addition to the reports from Shanghai and Hong Kong relied on by previous studies, and much of it is of higher quality. In the Chinese populations of the KLTZ, Taiwan, and the Straits Settlements, influenza-related excess mortality in 1918 was 62–65 per 10,000, considerably higher than the excess in the United States (40 per 10,000) and England and Wales (33 per 10,000) but comparable to that of Ceylon (65.6 per 10,000). The single-year influenza death rates of 61–74 per 10,000 from Weihaiwei and Huolu indicate similarly high levels of mortality. The modest influenza-related excess death rates in 1918 for Chinese in Hong Kong, Beijing, and Tsingtao (22.7, 23.1, and 24.2 per 10,000, respectively) may be lower than the true rates because of undercounts and lower-quality reporting of causes of death. It should be remembered that the calculation of the epidemic excess (which subtracts the baseline average from previous years) is a conservative measure of the impact of the epidemic strain, which, to the extent it supplants and not just supplements the previous non-epidemic strain, also accounts for deaths in the baseline in epidemic years. While the lack of Chinese national vital statistics from 1918 is still to be lamented, the addition of new data from these seven Chinese-majority populations from both north and south goes some way to illuminating China’s “black hole of missing data.”
Previous quantitative estimates of the mortality impact of the 1918 influenza epidemic on the mainland Chinese population have relied on questionable methodologies to make up for the lack of national-level statistics. Patterson and Pyle (1991: 18), citing only anecdotal accounts and on the basis of an estimated mainland Chinese population of 400–475 million, substitute analogy for the missing vital statistics to conclude that “flu death rates of 10–20 per thousand, quite reasonable for a poor country but well below estimates for Indonesia or India, would indicate a range of 4.0–9.5 million dead.” Ijima (2003: 108–109, table 7.4) bases his estimate on his calculation of influenza death rates from Hong Kong and the KLTZ, which he calculates as 2.5 and 3.2 per 1,000, respectively. But these calculations are seriously flawed: First, Ijima uses only deaths classified as influenza and not the PBI totals that capture the full impact of the epidemic, nor does he use an excess deaths methodology; second, he repeats the mistake of using the Hong Kong population number that includes the New Territories instead of excluding them; and third, in the case of the KLTZ he combines figures for the Japanese and Chinese populations. The result is a low estimate of 1–1.28 million total deaths for all of China. Noticing that mortality rates in Taiwan and Japan were much higher, Ijima speculates that the “extensive human traffic” between Taiwan and Japan spread the disease in these jurisdictions but does not explain why that factor would not also apply to the KLTZ and the busy port of Hong Kong.
The KLTZ, Taiwan, the Straits Settlements, Weihaiwei, and Huolu show death rates in the range of 60 per 10,000 that we can use to produce a new estimate. An influenza-related (PBI) epidemic excess mortality of 60 per 10,000 for a mainland Chinese population estimated at 440,000,000 (Perkins, 1969: 16; Ho, 1959: 79, 86) implies a total of 2.64 million influenza-related deaths in 1918. This is much higher than Ijima’s 1–1.28 million, but also much lower than Patterson and Pyle’s “guestimate” of 4–9.5 million dead. Future estimates of the worldwide impact of the 1918 influenza epidemic would do better to use an estimate grounded in the new data uncovered here.
In a discussion lamenting the dominance of patchy, anecdotal, and impressionistic sources in accounts of the pandemic in China, Peckham (2022: 263) concludes that “ultimately, morbidity and mortality figures for China are pure guesswork given the absence of sources.” But as shown here, vital statistics sources for seven majority-Chinese populations were not absent; they were neglected. For decades the literature on the 1918 influenza epidemic in China has relied on the unreliable Shanghai death reports, the misstated Hong Kong CDR, and Ijima’s miscalculations as the primary quantitative measures of the epidemic’s mortality impact. Because these sources suggested an unusually mild impact compared to Western countries, they gave birth to speculation about Chinese immunity to the disease, the power of traditional Chinese medicine to limit the disease (Cheng and Leung, 2007), and a Chinese origin for the virus (Shortridge, 1999; Langford, 2005). The new data presented here show not a mild but rather a serious impact of the pandemic on the ethnic Chinese population of East Asia that conforms to the picture we see in populations around the globe: heavy mortality in the fall of 1918 and which was particularly concentrated in young adults. Rather than an exception, the experiences of Chinese populations in East Asia prove to be what was usual and ordinary for 1918: influenza mortality of unprecedented proportions.
Supplemental Material
sj-docx-1-mcx-10.1177_00977004231189278 – Supplemental material for Reassessing the Mortality Impact of the 1918 Influenza Pandemic in China
Supplemental material, sj-docx-1-mcx-10.1177_00977004231189278 for Reassessing the Mortality Impact of the 1918 Influenza Pandemic in China by John Robert Shepherd in Modern China
Footnotes
Acknowledgements
I am grateful to those who have given me feedback or encouragement along the way: Bill Laveley, Matt Sommer, Dorie Solinger, the editors of Modern China, and anonymous reviewers. I am also indebted to the Interlibrary Loan Department of the University of Virginia Library.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Notes
Author Biography
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
