Professor Sen's version

Professor Sen presents his version in support of his claim that this famine and many others were not caused by a decline in food availability, and that the FAD approach to famine analysis is in- appropriate. He says that there was at least 1 I % more food available in Bengal in 1943 than there had been in 1941 when there was no famine, so the famine could not have been caused by a decline in food availability.(For example, Sen, op cit, Ref 2, 1977b, pp 42, 53;1980 P 80.) Instead, it occurred because of a change in the distribution of existing food supplies arising from wartime conditions, particularly inflation. This meant that some groups of the population received higher incomes and ate more, leaving little for the rest of the population. At the same time, others had insufficient money to buy food, so they starved.

A complete list of the explicit causal hypotheses he puts forward is shown below.(Sen, op cit it, Ref 2, 1977, pp50, 51 1981 a, pp 75-78.)

1) Demand factors related to inflation raised the rice price in 1942.

2) An uneven expansion of income and purchasing power.

3) Impoverishment of occupational groups not directly affected (from March 1943 on).

4) The change from the stable prices of 1914-1939 to an era of more rapidly rising prices.

5) Speculative withdrawal and panic purchases were encouraged by administrative chaos (especially between December 1942 and March 1943, but also up to November).

6) . . . demand forces were reinforced by an "indifferent" winter crop and by vigorous speculation and panic hoarding from March to November 1943'.

7) The prohibition of the export of cereals from other provinces.

8) The policy of removing boats from areas threatened with Japanese invasion. (Sen, op cit, Ref 2,1984, p 461; 1980b, p 619.)

9) The policy of removing excess grain stocks from areas threatened with invasion. (Ibid.)

His arguments tend to the conclusion that some groups of people ate more grain than usual, leaving less for the rest of the population.

Sen ignores the Hindu-Muslim conflict which most commentators consider a serious factor (eg Dutt op cit, Ref 8; Ghosh, op cit, Ref 8 and N.S.R. Rajan, Famine in Retrospect, Pamda Publications Bombay, India, 1944. Also, N. Mansergh ed The Transfer of Power, 1942-7, Vol III, HMSO London, UK, 1971; N. Mansergh, The Transfer of Power, 1942-7, Vol IV, HMSO, London, UK, 1973, p 358). It was even claimed by a leading politician that Bengal had been deliberately starved out by other provinces' which refused to permit the export of grain (P. Moon ed, Wavell: The Viceroy's journal, OUP, Oxford, UK, 1973, p 239). He also gnores the sabotage of the railways which were bringing in grain.

The examination of Sen's theory will consider first, whether there was in fact a shortage. Then his explanatory hypotheses will be examined - all of them, as there is considerable disagreement as to which he considered essential. Finally, government policy during the famine will be discussed.


Was there a shortage?

Sen's argument depends on his analysis of the production figures showing that there was no shortage. If either his figures or his analysis of them are shown to be wrong, his whole argument collapses. In this section it will be shown that the production figures are so unreliable that they can give no support to his argument. His production figures came from the Famine Commission, which is at pains to show how unreliable they are. They are based on a crop forecast, not even a post-harvest estimate, and are based on subjective estimates of the areas planted and the probable yield.

As the estimates were based on area planted rather than harvested, and as yields were based on estimates made during the growing season they would not have made full allowance for the effects of the cyclone between estimation date and harvest date. It is difficult to see why Sen should quote Blyn's estimates as though they provided independent evidence (G. Blyn Agricultural Trends in India 1891- 1947: Output, Availability and Production, University of Pennsylvania Press, Philadelphia, PA, USA, 1966). They are based on the same poor data and they are even more aggregated.

For instance, the following is the method followed in the province of Bengal. Each Circle Officer (a gazetted revenue officer with jurisdiction over three or four thanas' [i.e. 400 square miles (Famine Commission, op cit, Ref 2, 1945a, p 7)] ascertains from personal inspection and by questioning other local officers and cultivators, the relation which the area under the crop bears to the normal acreage of that crop in that area, this normal acreage being determined in accordance with certain instructions. The Circle Officers send their estimates to the Subdivisional Officer who, after making such corrections as he considers necessary, either from his own knowledge, experience and observations or by enquiry, sends a consolidated estimate for the subdivision to the District Officer. The latter, in his turn, makes such modifications as he thinks necessary on the basis of his own experience and information obtained from the District Agricultural Officers and other sources, and forwards the district estimate to the Director of Agriculture. Clearly, acreage estimates prepared in this manner cannot be accurate.'(Famine Commission, op cit, Ref 2, 1945b, pp 44-5.)

In 1942 the estimates would have been particularly bad because parts of Bengal, notably those hit by the cyclone, were on the verge of an insurrection, and the army was busy, burning villages etc. The collectors of the statistics did not know the normal acreage or yields, only people's estimates of the deviation from the norm, so there was a substantial error in the estimate of total production as well as an unknown aggregation bias. This was particularly important for the December 1942 crop, as the damage was limited to certain areas.

As the cyclone, tidal waves, flooding and disease which caused so much damage had never occurred before, there was no experience to guide anyone. Previous famines had been due to drought in other areas of Bengal. It appears that no attempt was ever made to check these crop forecasts against anything else, or to amend them in the light of experience, so it was not known if they even indicated the direction of the change. Checking would in any case have been difficult as only a proportion of the crop was marketed and the marketing system was not monitored.

Desai provides a useful review of the agricultural and other statistics of this period, and his rigorous use of them is exemplary. (R.C.Desai, Standard of Living in India and Pakistan,1931-2 to 1940-41, Popular Book Depot, Bombay, India,1953.) He compares the official estimates of agricultural surveys with the results of scientific surveys carried out by Mahalanobis. He shows that the discrepancies are large, with survey estimates being between 47% and 153% of the official estimate. The discrepancies also vary from year to year, with the sample estimate of the jute crop being 2.6% above the official estimate in 1941, and 52% above it in 1946. (With jute, where exports provided a check, the sample proved correct.) Since there was no sample survey of the rice crop until after the famine, we do not know how inaccurate the 1942 forecast was. On the basis of Mahalanobis's survey after the famine, the Famine Commission revised the estimates for all previous years, putting them up by 20%. This changed the level of the estimate, but there was no way of correcting for error around this estimate with hindsight. Another form of bias arises from subjective eye estimates of the prospective yields. Mahalanobis found that, even with scientific sampling methods on a mature crop, enumerators tended to select the best fields and the best areas of damaged fields. (P.C. Mahalanobis, Recent experiments in statistical sampling in the Indian Statistical Institute', Philosophical Transactions of the Royal Society, Part iv, pp 326-378, 1946. G.R. Allen gives examples of this bias when subjective methods are used (Agricultural Marketing Policies, Blackwell, Oxford, UK, 1959, pp 164-6).) This gives a large upward bias when much of the crop is damaged, as in 1942. The bias would be much worse with untrained observers using subjective methods. The quality of the data was also bad, even in the better organized studies:

. . . the apathy of the administrators and the peculiar difficulties in which statistical work has to be carried out in India has to be experienced in order to be properly appreciated. I may perhaps quote one concrete example. In 1939 the Government of Bengal decided to prepare a complete record, plot by plot, of the land sown with jute. After these records were prepared the Government arranged to have certain portions checked by permanent Government officers. The primary records, when checked, were found so unreliable that the Bengal Government ordered all the records to be destroyed.'(ibid.)

There is also a more serious form of bias - the scale of incompetence and corruption was so vast that virtually every administrator and politician had cause to want evidence suppressed or altered. There is some indication that pressure was brought on statisticians to do this . . . the political pressure resulting in an inaccurate census, the hint of the virtual suppression of an unpalatable report', (Elphinstone commenting on Mahalanobis, 1946, p 374). Mahalanobis, op cit, Ref 24, stated that The average administrator in India expects the scientific or statistical technician to supply evidence or proof in favour of what the administrator thinks to be right, rather than to give independent advice on objective grounds. Intellectual dishonesty, to which Major Elphinstone has referred, would in such circumstances be an actual advantage in securing promotion in official posts. This is why I have never favoured the idea of the Statistical Institute being run as a Government department or under predominating Government control'. It was also strongly rumoured at the time that the Indian government had deliberately printed only a few copies of the Famine Commission report, to limit the circulation of its criticisms (Accrete, op cit, Ref 9). Bhatia mentions that it was decided not to publish the evidence to the Famine Commission after proof copies had been run off (B.M. Bhatia, Famines in India, Asia Publishing House, Bombay, India,1967).

Sen makes a great deal of the fact that the rice crop in Bengal was recognized to be indifferent rather than exceptionally bad'.(Quoting Document No 265, p 357 in Mansergh, op cit, Ref 18.) In fact, the document he quotes stated that there was both cyclone damage in certain areas and an indifferent crop in Bengal generally. The combined effect was seen as being exceptionally serious. This is a clear misrepresentation of what is in the source document. We must conclude from this that the statistics are so bad that one cannot confidently say that the true production lay within 50% of the official estimate in any one year, if, like Sen, we rely entirely on official crop figures. However, Sen did not rely on the absolute figures. He relied on his assessment that production for 1943 was at least 11 % higher than that for 1941. Since both figures could be 50%, out, and the bias is not random, it is quite possible that the 1941 crop could be three times that of the 1943 crop (again if one relies entirely on crop figures). This means that the margin of error of his statement is of the order of 3000 %. Sen's production figures are so unreliable that they can give no support to his thesis, a thesis which relies entirely on these figures. Si nce the effect of a Type I error, accepting his thesis when it is wrong, is to cause or worsen a famine, then it must be rejected.



Even if they were correct, Sen's production figures would not be sufficient to show that there was no shortage as they do not allow for stocks. The amount of food available can be defined as current production plus imports plus stocks. The Famine Commission argues that it was not normal to start eating the December crop until March, both because of the need for normal stocks and because the rice is not palatable for some months after the harvest so there was a three month carryover. (See Famine Commission, op cit, Ref 2, 1945 pp 179-99, Professor Hussein's minority report; Government of India, Report on the Marketing of Rice in India and Burma, Government of India Press, Calcutta, India,1942.) The very poor crop of December 1940 meant that the rice ran out earlier than usual and people started eating the crop of December 1941 as soon as it was harvested. This means that consumption in 1941 was well above the adjusted current supply of rice' (This is: production plus imports, ignoring stocks, rather than what would normally be considered as supply.) quoted by Sen, while the consumption of 1942 was well below it. Since stocks were used up at the beginning of 1942, stocks were low at the end in spite of a good harvest. This means that the amount of grain available for consumption in 1943 was nearer to the adjusted current supply' than it was in 1941. Even if the adjusted current supply' had been 11% higher in 1941 than in 1943, as Sen claims, the amount actually eaten was much lower.

This, if true, is fatal to Sen's thesis. He rejects all the arguments of the Famine Commission on carryover. Indeed, he makes the very serious allegation that: Later the facts were squared with theory by "revising" the facts, by introducing mythical variations in the unobserved item called "the carry-over from previous years"'(Sen, op cit, Ref 2, 1977, p 75). He provides no supporting evidence. Instead, he states that there was no carryover from year to year, or if there were, which he denies, there was a larger carryover into 1943 than into 1941. I cannot accept Sen's suggestion that we should ignore the carryover on the grounds that we have no statistical data on the stock position. (Ibid, pp 42, 55.) Still less can I accept his assertion that variations in carryover were mythical - surely there is no country that does not aim at a substantial carryover and it seems beyond belief that exactly the same amount was carried over from a famine year as from a year of bumper harvests. It might be mentioned that all the explanations Sen gives for the famine are equally unquantified (mythical?) variations in equally unobserved (non-existent?) items. The difference is that where the Famine Commission gives ten pages of argument and facts in support of their carryover explanation, Sen gives only a sentence or two in his explanations.

I shall not recapitulate the arguments put forward by the Famine Commission and others to show that there was a much reduced carryover into 1943. Instead, I shall show that Sen's own figures destroy his case. Table 1 is based on exactly the same figures as Sen's, except that I have arbitrarily chosen a carryover of 8 500 000 tons at the beginning of 1941. This shows that a surplus of nearly one year's supply in January 1939 turns into a deficit of one million tons in 1943 - enough to cause the famine. Even if they were correct, Sen's production statistics would not show that there was no shortage in 1941. It is necessary that his assumptions on carryover should also be true. Not only are they extremely unlikely and contrary to all the evidence, but his own figures show them to be wrong. This section refutes his thesis.

Table 1. Stocks of rice in Bengal, 1939-1943
(Thousand tons)


Stocks at this date

+ Aman crop

+ Boro crop

+ Aus crop

+ Imports

Minus seed

Minus Consumption
1 Jan 1939 8500 0 201 0 182 434 3977
1 June 1939 4472 0 0 1758 300 11 3977
1 Nov 1939 2542 7805 0 0 0 0 1591
1 Jan 1940 8756 0 194 0 100 426 4017
1 June 1940 4607 0 0 1525 258 13 4017
1 Nov 1940 2360 5178 0 0 0 0 1607
1 Jan 1941 5932 0 203 0 100 468 4057
1 June 1941 1710 0 0 2250 232 14 4057
1 Nov 1941 121 8876 0 0 -2 0 1623
1 Jan 1942 7372 0 206 0 0 453 4098
1 June 1942 3027 0 0 1649 0 14 1098
1 Nov 1942 565 6024 0 0 0 0 1639
1 Jan 1943 4949 0 218 0 100 520 4138
1 June 1943 609 0 0 2390 100 17 4138
1 Nov 1943 -1056            

Source: Based on figures in the Famine Inquiry Commission (FIC), see text, op cit, Ref 2, p 215.
Notes: Opening stock is an arbitrary assumption. Consumption per head is the mean current supply less seed for the years 1928-42. This is below the mean for the first half of the century. A population growth rate of 1 % has been assumed(Sen,1977, p 40). The remaining 64 000 tons of 1943 imports are assumed to have come in November and December. No allowance is made for the unrecorded exports by road and country boat in 1942. These would have increased the deficit substantially.


Other evidence

When working with such unreliable statistics, it is wrong to depend on a single set of statistics. Instead, economists demand confirmation from other statistics, other data. This other evidence does not support Sen. Many people gave warnings of the famine, warnings which conflicted with the official production estimates. As early as December 1942, after the cyclone and before the aman' crop had been harvested, the trade was talking of the worst crop in 20 years. (Famine Commission, op cit, Ref 2, p 33) Traders bought up any stocks they could in Bengal and they went into the neighbouring provinces of Orissa and Bihar to buy grain and standing crops. They were prepared to smuggle the rice into Bengal if trade restrictions prevented them from doing it legally. The trade had its own way of estimating supplies (including stocks) and did not rely on official estimates. In this case, they were so certain of their estimates that they invested all they could borrow, and of course they made a lot of money as a result. The Bengal government ignored their warnings.

Bhatia, quoting from the unpublished evidence to the Famine Commission, states that public men and organizations had warned the government. (Bhatia, op cit, Ref 26, p 35.) Sen himself quotes pressure, from a businessman much involved in rice trading', to increase imports by one million tons as late as October 1943. (Sen, op cit, Ref 2, 1977, p 54 quoting from Document 174 in Mansergh, op cit, Ref 18, p 390.) This would of course have been against the businessman's interests if he had large speculative stocks.

During the famine, it became increasingly apparent that the shortage was much more serious than the production statistics suggested. The total failure of all the government's intervention measures to bring down the price is particularly significant. The government's attempts to break the Calcutta market' by dumping large quantities on the market failed, both because it proved impossible to buy or seize any large quantities and because what was put on the market vanished without a ripple. During the period of free trade with Bihar and Orissa, 91000 tons was imported and dumped on the market (and some was bought illegally and smuggled in). The effect was to push up prices sharply in Bihar and Orissa, but there was no noticeable effect on prices in Calcutta. (Famine Commission, op cit, Ref 2, p 532.) This suggests that 91 000 tons was a large proportion of any surplus in Bihar and Orissa, but was a very small amount in relation to the Bengal deficit. (Calcutta would consume only 21-24 thousand tons of rice and 15 000 tons of wheat a month (Famine Commission, op cit, Ref 2, p 203).) The house-to-house search for stocks showed only that the stocks were much lower than expected.


Conservative estimates?

Sen states repeatedly that his estimates are both conservative and reliable. He says that his calculations are based on a careful tally on food availability in Bengal'. He talks of presenting the results of a food supply calculation, taking into account local production and trade, choosing - wherever the data permit - an assumption as unfavourable to 1943 as possible'. He concludes that Current availability of food was at least 11 per cent higher than in 1941, when there was nothing remotely like a famine.'(Sen, op cit, Ref 2,1984, p 461.) Elsewhere he says: This is most certainly an over-estimate for 1941 vis a vis 1943, but this is an acceptable bias as it favours the thesis we are rejecting', To bias the figures as much as possible against 1943. . .'. (Sen, op cit, Ref 2,1977, p 40.) He may also be interpreted as claiming a much greater accuracy for them than is justified, because he frequently quotes different secondary sources as giving much the same estimate of total production or import needs.(See, for example, Sen, op cit, Ref 2, 1977, Pp 53-4.) Since these secondary sources are all based on the same official production estimates, no added confidence is given. His scathing comments on those who consider that the famine was caused by shortages emphasize the impression that he is totally confident of his figures. In fact, the figures he gives are not in any sense conservative. I have shown the output figures to be unreliable. The import figures are no more reliable than such figures usually are, and in addition they fail to cover trade by road and country-boat - for these, he uses the Famine Commission guesses, and not a conservative figure. (Note that the Famine Commission assumes, and Sen accepts, an identical unrecorded net import in 1941, a year of shortages and recorded net imports, and 1942, a year when Bengal had a surplus and the rest of India a shortage and when Bengal had substantial recorded net exports.) His conservative adjustments consist of making a slight allowance for unrecorded wheat imports, an alteration of a fraction of 1% of the total. Again, he makes much of choosing a 1 % population growth rate instead of 0.46%, which makes a difference of 1% when he uses it for comparing 1941 with 1943. These conservative adjustments do not make any noticeable improvement to the accuracy of the aggregate figures he uses. To state that such estimates are conservative is a misstatement of perhaps 30%, and to state that they are the result of a careful tally is a further misstatement. This, in itself, is enough to cast doubt on all the evidence he presents in favour of his thesis.

Previous | Return to main | Next