Jump to ContentJump to Main Navigation
Poverty and FaminesAn Essay on Entitlement and Deprivation$

Amartya Sen

Print publication date: 1983

Print ISBN-13: 9780198284635

Published to British Academy Scholarship Online: November 2003

DOI: 10.1093/0198284632.001.0001

Show Summary Details

(p.195) Appendix D Famine Mortality: A Case Study

(p.195) Appendix D Famine Mortality: A Case Study

Poverty and Famines
Oxford University Press

In this Appendix1 the size and pattern of mortality in the great Bengal famine of 1943 are studied. Mortality in the Bengal famine was a hotly debated issue during and just after the famine, and has, in fact, remained so. The pattern of mortality is worth studying also for the light it throws on the nature of the famine. The general features of the famine and its possible causation were studied in Chapter 6

D.1. How Many Per Week: 1,000, 2,000, 26,000, 38,000?

‘The Secretary of State for India’, wrote The Statesman, the Calcutta newspaper, on 16 October 1943,

seems to be a strangely misinformed man. Unless the cables are unfair to him, he told Parliament on Thursday that he understood that the weekly death‐roll (presumably from starvation) in Bengal including Calcutta was about 1000, but that ‘it might be higher’. All the publicly available data indicate that it is very much higher; and his great office ought to afford him ample means of discovery.2

Sir T. Rutherford, the Governor of Bengal, wrote to the Secretary of State for India on 18 October 1943, two days after The Statesman editorial:

Your statement in the House about the number of deaths, which was presumably based on my communications to the Viceroy, has been severely criticised in some of the papers. My information was based on what information the Secretariat could then give me after allowing for the fact that the death‐roll in Calcutta would be higher owing to the kind of people trekking into the city and exposure to inclement weather. . . . The full effects of the shortage are now being felt, and I would put the death‐roll now at no less than 2000 a week.3

Was this higher figure of 2,000 close to the mark?

The Famine. Inquiry Commission (1945a) noted that ‘from July to December 1943, 1,304,323 deaths were recorded as against an average of 626,048 in the previous quinquennium’, and the difference attributed (p.196) to the famine comes to a bit over 678,000.4 This would make the average weekly death‐roll in excess of 26,000 rather than 2,000.

The Famine Inquiry Commission went on to note that ‘all public health statistics in India are inaccurate’, and ‘even in normal times deaths are not fully recorded’. In rural Bengal deaths were reported by the village chowkidar (village watchman), in addition to his other duties, and he was ‘usually illiterate, and paid about Rs. 6 or Rs. 7 a month’. During the famine period, ‘in certain places the salaries of chowkidars were not paid and they deserted their posts to obtain work on military projects and aerodromes’, while ‘some of them died’.

The replacement of dead and vanished chowkidars was no easy matter and several weeks and months might elapse before successors could be found, during which deaths presumably went unrecorded. Further, in the height of the famine thousands of people left their homes and wandered across the countryside in search of food. Many died by the roadside—witness the skulls and bones which were to be seen there in the months following the famine. Deaths occurring in such circumstances would certainly not be recorded in the statistics of the Director of Public Health.5

Taking note of all this, the Commission arrived at the conclusion that ‘the number of deaths in excess of the average in 1943 was of the order of one million’—nearly all of it in the second half of the year.6 On this estimate the death‐roll in the second half of 1943 would seem to have been around 38,000 per week.

D.2 How Many in Fact?

No reason was given by the Commission for choosing the particular correction ratio that was used, except the thoroughly respectable one that it was arrived at ‘after due consideration of the available facts’ (1945a, p. 109). To this figure of one million deaths attributed to the famine of 1943, the Commission added the number of registered deaths in the first half of 1944 in excess of the previous quinquennial average without any correction. The reason for this asymmetry stemmed from the Commission's belief that there was ‘an unquestionable improvement in the collection of mortality statistics’ at the end of 1943 owing to efforts made by civil and military medical authorities (p. 109). The excess death registration for the first half of 1944 amounts to 422,371. Adding this to the estimate of one million for 1943, the Commission rounded off the mortality toll of the famine thus: (p.197) ‘about 1.5 million deaths occurred as a direct result of the famine and the epidemics which followed in its train’ (p. 110).

Dr Aykroyd, a distinguished nutrition expert, who was a member of the Commission and who in fact made the Commission's estimates of mortality, has recently stated (as was quoted in Chapter 6) that he now thinks ‘it was an under‐estimate, especially in that it took too little account of roadside deaths, but not as gross an under‐estimate as some critics of the Commission's report, who preferred 3 to 4 million, declared it to be’ (Aykroyd, 1974, pp. 77). Who were these critics and how did they arrive at their figures?

The most quoted estimate—from the Anthropology Department of the Calcutta University—was based on a sample survey. The following estimates were released on 21 February 1944—much before the Famine Inquiry Commission had even been appointed:

The Anthropology Department of the University of Calcutta has carried out a sample survey of ten of the famine‐affected districts of Bengal. The statistics for eight districts have so far been tabulated. They cover eight hundred sixteen‐family units with a total membership of three thousand eight hundred and eighty. The total deaths in these groups during June–July 1943 and November–December 1943, has been three hundred eighty‐six or ten per cent during six months (i.e. 100 per thousand). As the death rate for Bengal does not exceed thirty per thousand per annum, i.e., fifteen per thousand for six months, the excess mortality (100–15) of eighty‐five per thousand, that is, eight and a half percent, has to be ascribed to famine and the pestilence that followed in its wake. As some areas in North Bengal were much less affected than Western or Central Bengal or the deficit areas of Eastern Bengal, some reduction has to be made to estimate the total mortality figures for Bengal. It will probably be an under‐estimate of the famine to say that two‐thirds of the total population were affected more or less by it. On this basis the probable total number of deaths above the normal comes to well over three and a half millions.7

The applicability of an excess mortality rate of 8 1 2 per cent to two‐thirds of the population of Bengal is, in fact, a piece of pure guesswork—and an illegitimate one at that, since the sample that was surveyed was chosen from the worst affected areas in Bengal. Later the leader of the group, Professor K. P. Chattopadhyaya, himself pointed out limitations of this estimate, and proposed a figure of 2.2 million for excess deaths in 1943. Adding the half a million excess deaths taken by the Famine Inquiry Commission for 1944, Chattopadhyaya came to a ‘minimum’ estimate of ‘total excess mortality’ equalling 2.7 million.8

Between Chattopadhyaya's figure of 2.7 million and the Famine Inquiry Commission's 1.5 million (not to mention the minute estimates (p.198) in contemporary official statements in London and New Delhi9), there remains a wide gap. The lack of evidence on the representative nature of Chattopadhyaya's sample renders it dubious and the arbitrariness of Commission's correction factor makes it difficult to evaluate their estimate also.10 But a more fundamental question concerns the time coverage of the mortality estimates. Both these figures cover up to June 1944. The acute starvation associated with the famine had ended around December 1943, even though ‘the death rate remained high throughout the greater part of 1944’ (Famine Inquiry Commission, 1945a, p. 1). When did the death rate, in fact, return to ‘normal’? The Famine Inquiry Commission did not answer this question.

It could not have. At the time the Report was submitted in 1945, the death rate had not yet returned to normal. When did it do so? This is clearly one of the first things to ascertain, since the forces of post‐famine epidemics to which the Commission refers in incorporating the excess deaths in the first half of 1944 in its total mortality estimate, went on raging for years.

For this, and indeed for any other year‐to‐year study, we have to rely on death registration data with suitable corrections. It is argued in the Census of India 1951, in its report on the ‘Vital Statistics of West Bengal: 1941–50’, that, while there are errors in registration, ‘under‐registrations are fairly uniform and do not take sudden leaps and bounds from year to year’ (vol. VI, Part 1B, pp. 1–2).11 While it seems most likely that the registration ratio did decline in 1943 and improved again in 1944, there seems to be little reason for assuming a radically different proportion of post‐1944 registration compared with pre‐1943 ratios.

For West Bengal, Jain's use of the reverse survival method yields an under‐registration of deaths of 33.9 per cent in 1941–50. This makes the actual mortality 51 per cent higher on the average than registered (p.199) mortality. I shall use this ratio of correction uniformly, though it should be noted that this would tend to underestimate famine mortality, since registration was especially bad in 1943—the year of the famine and of peak death even in terms of registration data. There is, thus, a downward bias in our estimation of famine deaths.12

In Table D1 numbers of the registered deaths for each year from 1941 to 1950 are given for West Bengal. The time pattern is one of monotonic decline except for the one severe jump upwards in 1943. In fact, despite falling each year after 1943, annual mortality did not return to the 1942 level even by the end of the decade. Since the number of deaths had tended to fall each year, the Famine Inquiry Commission's procedure of taking the average mortality in the previous quinquennium as the ‘normal’ mortality may understate excess mortality for the famine years. Instead, I have made two sets of estimates: estimate A, with the ‘normal’ being taken to be the average of the deaths in 1941 and 1942, and estimate B, with the 1942 death rate being taken as the ‘normal’. Estimate B yields, naturally, a higher series of ‘excess deaths’, which are presented for 1943–50 in Table D1 and Figure D1. However, even estimate B can be thought to be understating the magnitude of excess mortality, since the relevant comparison is not with the level in the prefamine year, but with the level to which the expected death rates would

Table D1 Recorded Deaths in West Bengal, 1941–50

Excess deaths








19412 Average


































Source: Based on death statistics from Census of India 1951 vol. VI, part 1B, Table 6.

Appendix D Famine Mortality: A Case Study

Fig. D1 Recorded Deaths During 1941–50 In West Bengal

have fallen in the post‐famine years but for the intervention of the famine.13

The numbers of excess deaths under assumptions A and B respectively for each year are given in Table D1. The ‘excess’ becomes negative for A from 1950 onwards and for B—it can be checked from later data—from 1951 onwards; this is so with a stationary total death norm, which—as discussed above—understates the levels of excess mortality.

Adding up until the excess mortality is eliminated yields a total of excess mortality owing to the famine of 648,000 for Assumption A and 784,000 for assumption B. If the turmoil of the partition of Bengal in 1947 and the displacement resulting from it make us reluctant to read the impact of the famine in the excess mortality figures beyond 1946, we (p.201) can be conservative and count the excess figures only during 1943–6.14 This yields a total registration excess mortality of 601,000 under assumption A and 673,000 under B.

If Jain's (1954) estimate of under‐registration in West Bengal during 1941–50 is applied uniformly, then these excess registration figures would have to be raised by 51 per cent to arrive at the actual excess mortality.15 This yields 908 thousand and 1.016 million respectively under A and B.

All of this relates to West Bengal only. The famine was at least as serious in East Bengal—later East Pakistan, now Bangladesh.16 Unfortunately, there is no ‘reverse survival’ estimate of under‐registration for East Bengal comparable with Jain's calculation for the West of Bengal. I have not, therefore, tried to make an independent estimate of famine mortality in East Bengal. However, the Census of Pakistan 1951 reports an estimate, viz. a figure of 1.714 million, ‘worked out from official statements, which as explained are largely estimates in the absence of reliable reports’.17 Added to my estimates for West Bengal, this yields 2.622 million and 2.730 million respectively, under assumptions A and B. Note that the East Bengal figures given in the Pakistan Census take account of deaths only up to 1944 and not up to 1946, as in our West Bengal estimates. Taking note of the facts that (1) the population of what became West Bengal was almost exactly a third of the population of undivided Bengal in 1941; (2) the registered number of deaths in West Bengal tended to be around a third of the total number of deaths in Bengal before 1943; and (3) in the famine year the number of registered deaths in West Bengal was again almost exactly a third of that in Bengal as a whole,18 if we feel bold enough to treat (p.202) famine excess mortality in West Bengal to be a third of that in undivided Bengal, then the total Bengal famine mortality works out as 2.724 million and 3.048 million respectively under assumptions A and B.

These figures are put together in Table D2. Since the Famine Inquiry Commission and K. P. Chattopadhyaya both gave excess mortality figures separately for the famine year 1943, the results of our calculation with blow‐up for Bengal are shown separately for 1943 also. It is interesting that Chattopadhyaya's over‐all estimate comes fairly close to those presented here, but the coincidence is accidental, since his figure refers to mortality in 1943 and in the first half of 1944 only. In fact, for 1943 as such the estimates given here are quite close to those of the Famine Inquiry Commission. The bulk of the difference in our respective total estimates arise from (1) the longer time coverage in my estimates (using, however, the same logic as employed by the Commission itself in attributing high post‐famine mortality to the famine), and (2) continued correction for under‐registration of deaths even beyond 1943 (using results of corrections through the ‘reverse survival’ method).

Table D2 Estimates of Bengal Famine Mortality

Excess mortality in 1943 (millions)

Total excess mortality due to the famine (millions)

Famine Inquiry Commission



K. P. Chattopadhyaya



Assumption A + Pakistan Census


Assumption B + Pakistan Census


Assumption A blown up for all Bengal



Assumption B blown up for all Bengal



Since there were several downward biases—as explained—built into the estimates presented here, we may be inclined to pick a figure around 3 million as the death toll of the Bengal famine. (It has also the merit of being a ‘round’ number—that arbitrary preference shown by our ten‐fingered species captivated by the decimal system.) But what emerges most powerfully from our analysis is not so much the largeness of the size of total mortality, but its time pattern—lasting for years after the famine. This was largely due to the epidemics associated with the famine, and to this issue I now turn.

(p.203) D.3 How Did They Die?

In December 1943, Bengal reaped a harvest larger than any in the past. Curiously enough, it was also the month in which the death rate in Bengal reached its peak in this century. The famine in the form of starvation had by then come largely to an end—starvation deaths seemed to have peaked around September and October that year. Cholera mortality reached its maximum in October and November. Malaria peaked in December, and continued in its elevated position through the next year and later. Smallpox reached its height in March and April 1944, and a greater height still one year later. The starvation phase of the famine had given way to the epidemic phase.

Table D3 presents the yearly time series of registered deaths from some of the principal causes. The sharp jump upwards in 1943 of cholera, malaria, fever, dysentery, diarrhoea, etc., can be easily seen. For seasonal reasons the impact of smallpox was not felt until the following year since it hits primarily in early spring. Taking the average mortality in 1941 and 1942 as the ‘normal’ mortality for each disease respectively, ‘excess mortality’ from each disease has been calculated for the period 1943–6. The last row presents the inter‐disease breakdown of excess mortality.

Before discussing the inter‐disease pattern of excess mortality, it is worth commenting on the absence of starvation as a major reported cause of death during that great famine. One reason for this peculiarity is that starvation was not typically used as a separate category in reporting deaths. This was due partly to the habit of using traditional categories in reporting causes of death, but also to the fact that typical starvation deaths show other identifiable symptoms at the final stages, and these proximate ‘causes’ tend to fit well into the traditional categories. For example, it is common to die of starvation through diarrhoea (indeed, ‘famine diarrhoea’ is a well‐known phenomenon) as well as dysentery—partly as a result of eating uneatable objects. Clearly, many of the deaths reported under ‘dysentery, diarrhoea and enteric group of fevers’ were, in fact, starvation deaths. The same holds for several other categories, including the general category of deaths owing to ‘fever’.19

Excluding ‘fever’, which is a diverse basket of diseases varying from influenza and measles to cerebro‐spinal fever and Kala‐azar, the ranking of the main diseases in terms of their contributions to excess mortality were (in decreasing order): malaria, cholera, ‘dysentery, diarrhoea and enteric group of fevers’, and smallpox. The nature of these ailments as well as direct accounts suggest that the explosive (p.204)

Table D3 Diseases and Deaths in West Bengal, 1941–6 Registrations

Dysentery, diarrhoea, and enteric group of fevers



‘Fever’ (excl. malaria)



Respiratory diseases other than TB




















1941–2 Average













































Excess: 1943–6









Share of total excess (%)









Source: Based on current registration data, reported in Census of India 1951, vol. VI, part 1B. Note that the ‘enteric group of fevers’ figure both under ‘fever’ and under ‘dysentery, diarrhoea, and enteric group of fevers’, but the overlap is quantitatively rather tiny.

(p.205) outbursts of epidemics during and immediately following the famine were affected not merely by starvation and malnutrition, but also by other factors, e.g. the impact of the famine on sanitary arrangements, water supply, and other civic amenities, exposure to vectors through movements in search of food, as well as inability to receive medical attention owing to destitution and a breakdown of public health facilities.20 In addition, infectious diseases can spread directly to people who may not have been affected otherwise by the famine. Epidemics do, of course, also have a rhythm of their own.21 Once an epidemic occurs, its echo effects may last for quite a few years.

The diseases unleashed by the Bengal famine had the dual characteristics of being both (1) epidemic diseases associated with previous famines, and (2) endemic diseases in the region. Malaria had been associated with Indian famines at least from the nineteenth century,22 and epidemics of cholera and smallpox had been observed in many previous famines, including the Bengal famine of 1770. Dysentery and diarrhoea are, of course, ‘peculiarly famine diseases’—as the Famine Inquiry Commission described them. The same applies to the mixed bundle called ‘fever’ other then malaria. But all these diseases were also endemic in the region. Malaria and fevers, which are sometimes difficult to distinguish,23 were the biggest killers in the pre‐famine days, followed at quite some distance by ‘dysentery, diarrhoea and enteric groups of fevers’, cholera, and smallpox in that order. In the sharing of famine mortality, the relative positions are not very different, with malaria and fever being followed at a substantial distance by cholera, ‘dysentery, etc.,’ and smallpox, in that order.

Perhaps the most interesting case is that of the dog that did not bark, viz. respiratory diseases including TB. These diseases killed many more in the pre‐famine period than any of the other group of diseases, with the exception of malaria and other fevers. But, remarkably, mortality from TB and from other respiratory diseases seem to have been hardly influenced by the Bengal famine (see Table D3). This experience is not unusual in the context of other Indian famines, in which TB and other respiratory diseases have not typically played a (p.206) prominent part, but there is something of a puzzle in this in a more general context. The linkage of TB and other respiratory diseases with malnutrition is well established (see Keys, 1950), and seems to be conceded even by those who dispute the influence of starvation as such on other diseases spread through infectious contagion (see, for example, Chambers, 1972, pp. 82–6).

Tuberculosis is, of course, slow to develop and is influenced more by chronic undernourishment than by a short period of severe starvation; this might suggest that the spread of tuberculosis would not be much enhanced by a famine. But famine‐induced movements and sanitary breakdowns may help in the expansion of the infection. More importantly, since tuberculosis and other respiratory diseases were already widespread in Bengal, it would be natural to expect that starvation during the famine would convert morbidity into mortality on a substantial scale. That this was not reported as having happened during and immediately after the Bengal famine thus does leave one with an interesting and important problem. Attributing this counter‐intuitive phenomenon comfortably to an assumed error of reporting is tempting, but this explanation would be convincing only with empirical evidence of the existence of such a bias in a large enough scale. Also, since TB and other respiratory diseases typically had rather undistinguished records in previous Indian famines as well, an ad hoc explanation for the Bengal famine of 1943 as such is not what is needed.

The Bengal famine killed mostly by magnifying the forces of death normally present in the pre‐famine period—a magnifying role that other famines had played in the past. The universality of this endemic‐to‐epidemic relationship is, however, seriously affected by the apparent inertness of TB and other respiratory disease. This inertness also seems to contrast quite sharply with the view taken of these diseases in the international literature on famine‐induced epidemics (see for example Keys, 1950, Foege, 1971, Chambers, 1972).

D.4 What Regional Distribution?

Excess mortality can be estimated separately for each district in West Bengal on the basis of the registration data presented in the Census of India 1951 (vol. VI, part 1B). These are presented in Table D4, with the ‘normal’ level of mortality being taken to be the average of the figures for 1941 and 1942. The percentage excesses for the famine year 1943 and for the period 1943–6 are presented separately, and the ranks in the two orderings of excesses are also given. The inter‐district variations are quite remarkable, even though for every district the excess is positive both for 1943 and for the period 1943–6.

There are some differences between the two rankings. Malda, which ends up as the most affected district over‐all, was one of the less affected (p.207)

Table D4 Excess Mortality in West Bengal: Breakdown by District


Average mortality, 1941–2 (‘normal’)

Excess mortality, 1943

Excess mortality, 1943–6

Percentage excess, 1943

Percentage excess, 1943–6 (annual)

Excess rank, 1943

Excess rank 1943–6

Intensity class according to Bengal Govt Revenue Dept

Intensity class according to Bengal Govt Dept of Industries



























































West Dinajpur
















































































Source: Based on Census of India 1951, vol, VI, part 1B.

(p.208) districts in the famine year itself. Similarly, Midnapur, which was most affected in the famine year, ends up in a somewhat moderate position for the whole period. The pattern of the epidemics that followed the famine re‐ordered the districts in terms of mortality. However, the two rankings are not unrelated, and the value of Spearman's rank correlation coefficient is 0.60, which offers no problem in rejecting the null hypothesis that the two rankings are independent.

What is perhaps of greater interest is the fact that the Bengal government's diagnoses of the relative severity of the famines in the different districts differed quite substantially from the excess mortality rankings for 1943–6 as well as for 1943 itself. A five‐category classification of the subdivisions was issued by the Revenue Department in 1944, and a four‐category classification by the Department of Industries in the same year.24 Putting together the classification of the subdivisions within each district, I have presented a broad three‐class partitioning in Table D4 reflecting the two official views of ‘degree of incidence of famines’. Both put Malda—ultimately the most affected district—in the lowest category of incidence. The two did the same to Murshidabad and Birbhum, but in fact both the districts had a high incidence of excess mortality in 1943 as well as in the period 1943–6. On the other hand, 24‐Parganas, which neighbours Calcutta, and from where many destitutes trekked into Calcutta at the height of the famine,25 was put in the highest category of incidence in both the official lists, despite being only moderately placed in the excess mortality rankings for the famine year as well as the post‐famine period.26 Since relief operations were strongly influenced by these diagnoses, the discrepancies are of a certain amount of practical interest.

Finally, a remark on the excess mortality in Calcutta is worth making. Most people who died in Calcutta from starvation and from related diseases in the famine year were destitutes who had moved into Calcutta in search of food; the regular residents of Calcutta were protected by various public and semi‐public schemes of food distribution (p.209) (see Famine Inquiry Commission, 1945a). Based on this observation, it has been frequently stated that the residents of Calcutta escaped the famine.27 This is largely true as far as starvation is concerned, but in the epidemics that were induced by the famine, Calcutta had its own share of casualties, reflected by the excess mortality figures after 1943, i.e. after virtually all the famine destitutes from elsewhere had left or been repatriated.

D.5 Which Occupation Category?

The death registration figures do not specify occupational backgrounds. We can, however, surmise something about probable death rates by examining the rates of destitution of different income groups. These were computed on the basis of a sample survey conducted by Mahalanobis, Mukherjea and Ghosh (1946), already used in Chapter 6 above, and are presented in Table D5 (taken from Table 6.7 above). In the second column the destitution rates are added up with the transition to the occupation of ‘paddy husking’ – a typical destitution syndrome for rural women with children. On this basis it would appear that the

Table D5 Destitution Rates of Different Occupation Categories in Bengal: January 1943–May 1944

proportion of destitution

Proportion of destitution and transition to paddy husking

Peasant cultivation and share‐cropping



Part‐time agricultural labour



Agricultural labour



Non‐cultivating owners









Husking paddy








Profession and services



Non‐agricultural labour



Other productive occupations



Source: See Table 6.7 above.

(p.210) most affected groups were fishermen, transport workers, and agricultural labourers. In terms of absolute numbers, agricultural labourers as an occupation group were dominant.

One of the few direct surveys of the occupational basis of famine mortality was presented by Mukerji (1965) for five villages in the Faridpur district in East Bengal; the survey was conducted in 1944. The results are presented in Table D6. In these villages the highest mortality category is agricultural labour. The importance of agricultural labour among the famine victims is brought out also by the survey of destitutes in Calcutta conducted in 1943 by T. Das (1949).

Our information on this crucial aspect of famine mortality is limited and somewhat haphazard. And we have virtually no information at all on the occupational composition of post‐famine mortality in the epidemics.

Table D6 Distitution in Five Surveyed Villages in Faridpur

Occupation on 1/1/43

Proportion of destitution (%)

Proportion being ‘wiped off’ during 1943 (%)

Peasant cultivation and share‐cropping



Agricultural labour






Petty trader



Crop‐sharing landlord



Priest and petty employee



Office employee












Source: See Table 6.8 above.

D.6 Famine Mortality as Magnified Normal Mortality

Peculiarities in the pattern of famine mortality compared with normal mortality have been a subject of discussion for a long time. A supposedly lower impact of famines on women is one of the ‘regularities’ that has received some attention in India. Sir Charles Elliot, Famine Commissioner of Mysore in 1876 and Census Commissioner of India for the 1881 Census, summarized the general belief regarding nineteenth‐century (p.211) Indian famines: ‘all the authorities seem agreed that women succumb to famine less easily than men’.28

Was this the case with the Bengal famine? Das (1949) found, in his survey of destitutes in Calcutta in September 1943, that ‘for every dead woman there were nearly two dead men’ (p. 93). In its Report the Famine Inquiry Commission referred to Das's findings—then available in unpublished form—and also noted that there was a higher proportionate increase in male deaths compared with female deaths in 1943.29 The Commission referred to the contrary result from Mahalanobis's survey of 2,622 families which found a higher percentage of mortality among women, but went on to comment on the ‘considerable irregularity’ in the various subdivisions covered in the survey.

The sex breakdown of pre‐famine ‘normal’ mortality given by the average of 1941 and 1942 as well as that of the excess mortality in 1943 and in the period 1943–6 are all presented in Table D7, based on registration data. The ratios seem remarkably stable through the famine. While the proportion of men in excess mortality in 1943 is a bit higher than in the pre‐famine average, the difference is small, and over the larger period of famine mortality the proportionate breakdown of the excess is just the same as for the pre‐famine average.30

There may, of course, be biases in the registration system, but this should apply to registrations both before and during the famine. In fact, it is more likely that there was a serious bias in Das's sample survey of destitutes in Calcutta which contained a large proportion of families that had ‘lost their male earning members’, and this bias would be reflected in the results of the survey, which asked respondents to recall which members of the family had died.31 To what extent this type of observation bias was present also in the accounts of the nineteenth‐century famines, I do not know, but certainly as far as the 1943 famine is concerned there is little need for going into the rather contrived explanations32 that have been proposed to explain the supposed contrast of sex ratios.


Table D7 Excess Mortality of Men, Women, Children, and the Old: West Bengal

Average mortality, 1941–2

Excess mortality, 1943

Excess mortality, 1943–6


Percentages of total


Percentages of total


Percentages of total















Children below 5







Old people above 60







Source: Based on current registration data, reported in Census of India 1951, vol. VI, part 1B. Note that ‘men’ and ‘women’ include figures for all ages, and ‘children below 5’ and ‘old people above 60’ include those for both sexes.

(p.213) Das (1949) also noted a much higher proportion of deaths among children, and opined that ‘this will certainly cripple the next generation of the Bengalees’.33 The Anthropology Department of Calcutta University had reported a similar bias in its press statement in 1944.34

Is this borne out by the registration data? The answer seems to be no. The data are given in Table D7. The proportion of children below five in average mortality in the immediate pre‐famine period was 29 per cent, and that is also the percentage of children in excess mortality in the famine year (1943) as well as in the four‐year period of famine mortality (1943–6).35 The extraordinarily high level of mortality of children is, of course, an excruciating problem, but that is a characteristic not only of famine mortality but also of normal mortality in the absence of famine in this part of the world.

Table D7 also presents the mortality figures for the old people, those above sixty. Once again the proportions of famine mortality mirror the pattern of normal mortality.

I end this section with a final observation dealing with the monthly pattern of death at the height of the famine. Table D8 presents the monthly death registrations during June–July of 1943–4, when mortality was at its highest, and also the average monthly registrations in the preceding quinquennium.36 The similarity between the two monthly patterns is striking.37 This is brought out clearly by figure D2 as well. (For the benefit of the blind, I note that regressing monthly mortality y in the famine period on normal pre‐famine monthly mortality x, by least‐squares, yields a very high value of r 2. The estimated regression function, in fact, is y = 3,175x − 122,535, with r 2 having the convincing value of .95.) The famine seems to have worked by magnifying the forces of mortality each month, heightening the peak mortality relatively more. (p.214)

Table D8 Mortality by Months During July 1943–June 1944 Compared With Previous Quinquennial Average

Deaths during 1943–4

Quinquennial average deaths: 1938–42





































Appendix D Famine Mortality: A Case Study

Fig. D2 Monthly Pattern Of Recorded Mortality Before and During the Famine

(p.215) D.7 Concluding Remarks

While it is not possible to say at all precisely how many people were killed by the Bengal famine of 1943, there is evidence that an estimate of around 3 million would be closer to the mark than the figure of 1.5 million arrived at by the official Famine Inquiry Commission (and widely quoted in later works). The difference is largely due to:

  1. 1) continued high ‘excess mortality’ for several years after the famine, caused by famine‐induced epidemics the impact of which the Commission considered only for 1943 and the first half of 1944;

  2. (2) underestimation by the Commission of the actual extent of under‐registration of deaths in official records.

Both these contrasts largely reflect differences between the data available to the Commission and those available now. Apropos(1), the Commission, working in late 1944 and early 1945, could hardly have gone beyond the first half of 1944 in its mortality coverage. Apropos (2), the Commission chose to use an arbitrary correction for under‐registration, not having any way of estimating it directly or indirectly. In contrast, we can use the results of ‘reverse survival’ exercises based on Census data of 1951 vis‐à‐vis those of 1941 and the results of a direct sample survey held in 1948. There is thus no quarrel, only a very substantial difference in the respective estimates based on current information (see Section D.2).

While the gigantic size of excess mortality attributable to the famine is of a certain amount of interest, the time pattern of mortality is of possibly greater relevance. Very substantially more than half the deaths attributable to the famine of 1943 took place after 1943. The size of mortality did not return to the pre‐famine situation for many years after the famine, and the epidemics of malaria and other fevers, cholera, smallpox, dysentery, and diarrhoea that sprung up during and immediately after the famine went on raging for a long time (see Tables D1 and D3 and Figure D1). This has obvious implications for health policy.

Regarding the regional pattern of famine mortality, the relative importance of different districts changed quite a bit between the starvation phase and the later epidemic phase (see Table D4). What is perhaps of greater interest is that the official diagnoses of the relative severity of the famine in the different districts differed substantially from the pattern emerging from the ‘excess mortality’ calculations, both for the starvation phase and for the later epidemic phase (see Section D.4). Since government relief and rehabilitation work was based on these official diagnoses, the contrasts were of practical import.

Information on the occupational pattern of mortality is very limited, but some general impressions emerge from a broadly based 1944 survey (p.216) covering the occupational pattern of destitution, and two local 1943 surveys directly going into deaths related to occupations (see Section D5). In absolute terms, the most severe incidence of famine mortality during the famine itself fell almost certainly on the class of agricultural labourers. Their relative incidence was high too, but that applies also to other groups like fisherman, transport workers and non‐agricultural labourers in rural areas. In Chapter 6 the nature and causation of the observed occupational pattern of destitution were analysed, relating them to the positions of the different groups in the structure of production and exchange in the economy.

Regarding the diseases that took most of the toll, they had the dual characteristics of being both (1) endemic diseases in the region, and (2) epidemic diseases in past famines (see Section D.3). Gigantic as the famine was, it killed mostly by adding fuel to the fire of disease and mortality normally present in the region. This possibly explains why the seasonal pattern of famine deaths even during the actual famine and its immediate aftermath was essentially the normal seasonal pattern—just linearly displaced severely upwards (see Table D8 and Figure D2). The sex and age patterns of famine mortality also seem to show remarkable similarity with the normal pattern of mortality in pre‐famine Bengal (see Table D7).


(1) This Appendix draws heavily on Sen (1980b), written in memory of Daniel Thorner.

(2) ‘The Death‐Roll’, editorial, The Statesman, 16 October 1943. See also Stephens (1966). Ian Stephens was the editor of The Statesman, a British‐owned paper, which distinguished itself in its extensive reporting of the famine and its crusading editorials.

(3) Letter to Mr. L. S. Amery, no. L/E/8/3311; document no. 180 in Mansergh, (1973), vol. IV, pp. 397–8. The earlier communication referred to by Rutherford is document no. 158 in the same volume.

(4) Famine Inquiry Commission (1945a), p. 108. For the year as a whole the difference came to 688, 846.

(5) Famine Inquiry Commission (1945a), p. 109. See also Census of India 1951, vol. VI, part IB, pp. 1–2.

(6) Famine Inquiry Commission (1945a), pp. 108–9.

(7) Reprinted in Ghosh (1944), Appendix G.

(8) Chattopadhyaya and Mukherjea (1946), p. 5.

(9) There is something puzzling about the official statements on the minute size of mortality. Lord Wavell records in his ‘journal’ on 19 October 1943, when he became the new viceroy, that the outgoing viceroy, Lord Linlithgow, confessed to him that ‘in July he expected that deaths in Bengal might be up to 1,000,000 or 1 1 2 million, and that we looked like getting off better than he had thought possible’ (Wavell, 1973, p. 34). Presumably the government had meanwhile persuaded themselves that the situation was incomparably better than had been ‘thought possible’!

(10) Aykroyd (1974) is candid in acknowledging the arbitrariness of his estimate: ‘at all events, the figure of 1.5 million deaths is in the history books, and whenever I come across it I remember the process by which it was reached’ (p. 77).

(11) It is perhaps also worth remarking that, for India as a whole, the ratio of registered deaths to the estimated number of deaths obtained by using the ‘reverse survival method’ for 1941–50 by S. P. Jain (1954) is 0.73, while the same method had yielded a ratio of 0.74 for 1931–40. See Jain (1954), p. 44. The estimates for earlier decades are of Kingsley Davis: 0.74 for 1931–40, 0.72 for 1921–30, and 0.70 for 1911–20.

(12) A substantial net migration from East to West Bengal during the late 1940s would also tend to underestimate the actual death rate during 1941–50, and thus underestimate the under‐registration of deaths, thereby underestimating famine mortality.

(13) Note that the absolute number of deaths went on falling through the decades, despite the increase in the size of the population, which failed to increase only during the immediate famine years; see Census of India 1951, vol. VI, part 1B, pp. 2–4.

(14) Note, however, that the strictly monotonic decline of the number of deaths continued right through 1947 (see Table D1). The death rate per thousand also underwent a strictly monotonic decline, since a declining number of deaths with an increasing population size implies a strictly monotonic fall of the death rate. Excess mortality figures beyond 1946 have, however, been ignored to avoid overestimating famine mortality, by biassing the procedures in the opposite direction.

(15) This may be compared with the Famine Inquiry Commission's correction of recorded excess mortality in 1943 of 688,846 to one million, which amounts to a correction factor of 45 per cent. (For some inexplicable reason the Commission notes the correction ratio to be ‘some 40 per cent’—p. 109.) For 1944, however no correction was made by the Commission. A ‘pilot survey’ conducted by government of Bengal in 1948 found the correction factor to be 46.4 per cent (see Chaudhuri, 1952, p. 9).

(16) See Famine Inquiry Commission (1945a, pp. 114–15).

(17) Census of Pakistan 1951, Chapter III, p. 30. The arbitrary nature of this estimate is emphasized, and reference is also made to the fact that, ‘according to popular belief, however, the deaths from famine in East Bengal were between two and two and a half million’.

(18) The number of registered deaths in 1943 was 624,266 for West Bengal and 1,873,749 for Bengal as a whole (see Famine Inquiry Commission, 1945a, p. 108, and Census of India, 1951, vol. VI, part 1B, p. 21).

(19) Compare the problem of interpreting the large number of deaths from lethal scurvy during the Irish famine of 1845–6.

(20) See Famine Inquiry Commission (1945a) on these disruptive consequences of the famine and on the large‐scale trekking of destitutes in search of food. See also Ghosh (1944) and Das (1949).

(21) See Bailey (1957). In fact, because of the spread effects of epidemics, the Bengal famine may also have contributed to deaths outside Bengal, especially in Orissa and Bihar. See Famine Inquiry Commission (1945a), pp. 104–5. See also Census of India 1951, vol. XI, part I, p. 41.

(22) See the Reports of the Indian Famine Commissions of 1898 and 1901. Also the findings of S. R. Christophers regarding the nineteenth‐century famines, quoted in Famine Inquiry Commission (1945a), p. 122.

(23) On this see the Report of Indian Famine Commission of 1898.

(24) Quoted in Mahalanobis, Mukherjea and Ghosh (1946), pp. 11–14.

(25) A sample survey of the destitutes in Calcutta conducted in September 1943 revealed that nearly 82 per cent of the destitutes surveyed came from this one district (see Das, 1949, p. 58).

(26) Deaths occurring in Calcutta of people normally residing in 24‐Parganas should, in fact, be attributed to the 24‐Parganas itself. This correction would tend to raise somewhat the excess mortality rates of the 24‐Parganas. The required corrections are difficult to estimate because of lack of precise data on ‘normal residence’ of those dying in Calcutta during the famine and post‐famine years. But rough breakdowns would seem to indicate that the relative position of the 24‐Parganas would not change drastically, especially for the period 1943–6. The contrast between the reality and the official perception will still hold, and the importance of being close to Calcutta in having one's distress officially observed will not disappear.

(27) E.g., ‘In the end not a single man died of starvation from the population of Greater Calcutta, while millions in rural areas starved and suffered’ (Sir Manilal Nanavati's note, Famine Inquiry Commission, 1945a, p. 102).

(28) For this and other observations, see Census of India 1911, vol. I, part I, appendix to Chapter VI; and also Das (1949), pp. 93–6.

(29) Famine Inquiry Commission (1945a, pp. 110–11). The Department of Anthropology had noticed the same, and referred to it as ‘a very sinister and significant feature’ of the Bengal famine (see Ghosh, 1944, Appendix G, p. 183).

(30) The male population exceeded the female population in Bengal, and the recorded death rate per unit of population was higher for women in every year during the decade 1941–50 through the famine (see Census of India 1951, vol. VI, part 1B, Tables 7 and 8, pp. 29–30).

(31) Das (1949), p. 93.

(32) My favourites are some of those proposed by Mr. W. C. Bennet, C. S.: ‘Women find employment as maid‐servants in the houses of rich men when men have no work to look for’; ‘women possess ornaments of value which they may dispose for their own benefit whenever necessary’; ‘the woman in a Hindu family always keeps the household stores, and has no scruple in availing herself of the advantage it gives her’ (see Census of India 1911, vol. I, part I, appendix to Chapter VI, pp. 220–2).

(33) Das (1949), pp. 91–2.

(34) See Appendix G in Ghosh (1944).

(35) The Famine Inquiry Commission (1945a) noted a decrease in the number of deaths for infants under one month, but attributed this to a decrease in the number of births as well as to a reporting bias (p. 109). Adjustments for this group would not affect the total proportions of children in excess mortality by very much.

(36) The data come from Famine Inquiry Commission (1945a), p. 213.

(37) Cf. Jutikkala and Kauppinen's (1971) observation regarding ‘catastrophic’ and ‘normal’ mortality in pre‐industrial Finland (1749–1850): ‘The figures suggest that the seasonal distribution of deaths did not differ significantly between “catastrophic” and “normal” years’ (p. 284).