I thank the National Science foundation and the Department of Economics at the University of Wisconsin-Madison for providing support of the research on which this paper is based. I also wish to thank, for helpful comments on earlier drafts, Jeffrey G. Williamson and Morris David Morris. The helpful comments of two anonymous referees are also gratefully acknowledged. Any errors which remain are, of course, my responsibility.

¹ Anstey, Vera, The Economic Development of India, 4th ed. (London: Longmans, Green, and Co., 1952), pp. 143–48Google Scholar; Datta, K. L., Report on the Enquiry into the Rise of Prices in India (Calcutta: Superintendent, Government Printing, 1914), Vol. I, p. 21Google Scholar; Morison, T., The Economic Transition in India (London: John Murray, 1914), pp. 120–21Google Scholar; Morris, Morris David, “Towards a Reinterpretation of Nineteenth Century Indian Economic History,” Journal of Economic History, XXIII (December 1963), 612.Google Scholar

² Bhatia, B. M., Famines in India, 1860–1965, 2nd ed. (New York: Asia Publishing 1967), pp. 27–31Google Scholar; Chandra, Bipan, “Reinterpretation of Nineteenth Century Indian Economic History,” Indian Economic and Social History Review, V (March 1968), 50.Google Scholar

³ Dutt, Romesh C., The Economic History of India in the Victorian Age, 7th ed. (London: Routledge and Kegan Paul, Ltd., 1952), pp. 348–50, 536.Google Scholar

⁴ It does not hypothesize that shifts among food crops may have been so constrained. The crucial distinction is between a crop which can be eaten when harvested and one which must pass through the market to be converted to one which can be eaten.

⁵ Hurd, John, II, “Railways and the Expansion of Markets in India, 1861–1921” (paper presented at the Association for Asian Studies Annual Meeting, March 31, 1971, Washington D.C.).Google Scholar Hurd used data for 187 districts for which prices .were available in Prices and Wages in India.

⁶ Harnetty, Peter, “Cotton Exports and Indian Agriculture, 1861–1870,” Economic History Review, Series 2, XXIV (1971), 414–29.CrossRefGoogle Scholar

⁷ The extent of the coverage of these statistics is determined by availability. Statistics of the necessary kind are not available for Northwestern Provinces and Oudh until the middle of the 1880's; comparable figures for Bengal Presidency are available only after 1892–93. The summed series begins in 1875 because that is the first year for which total cultivation figures (as opposed to total assessed cultivation statistics) are available for Madras Presidency. No food grain statistics are available for Berar before 1870.

⁸ Krishna, Raj, “Farm Supply Response in India-Pakistan: A Case Study of the Punjab Region,” Economic Journal, LXXII (September 1973), 477–87.Google Scholar

⁹ Narain, Dharm, The Impact of Price Movements on Areas Under Selected Crops in India: 1900–1939 (Cambridge: Cambridge University Press, 1965).Google Scholar

¹⁰ For full results and discussion, see McAlpin, Michelle Burge, “The Impact of Railroads on Agriculture in India, 1860–1900: A Case Study of Cotton Cultivation” (unpublished Ph.D. dissertation, University of Wisconsin-Madison, 1973).Google Scholar

¹¹ It would also be possible to conclude that the data from the nineteenth century are so poor that no sensible results can be obtained with it. Certainly, there are numerous problems with the data. But to the best of my knowledge, there is nothing which is likely to impart any systematic bias to the acreage data. Price data are also unlikely to have any systematic bias. However, we must remember that prices given in the official reports are not likely to have prevailed everywhere in a district. The officially listed price of Rs. 4.2 per maund of jowār in Bellary district during the famine of 1877–78 may be correct for the district headquarters town, which was, after all, on a railroad. But it may be a gross understatement of the price of jowār in a village fifty miles from the district headquarters. In that village, the price of jowār may have been infinite—no grain for sale at any price. Because cotton is so much higher in value than jowār (per unit of weight), it is likely to have had a price that varied less (as a percentage) as distance from the railroads increased. Transport costs sufficient to double the value of jowār in fifty miles might only raise the price of cotton ten percent. If this is the case, then the relative price ratio may understate the cost of buying jowār in the villages in famine years and, in non-famine years, overstate its selling price.

¹² Or adequate rainfall. The districts along the west coast of India have little irrigation but over one hundred inches of rain a year is common.

¹³ Most of the farmers in all of the regions discussed in this article were small, independent farmers who were free to make their own decisions about what to plant. Decisions were not generally made by either landlords or moneylenders.

¹⁴ Great Britain, Parliament, Parliamentary Papers (command) 1880, Vol. LII cmnd. 2591 and Vol. LXXI cmnd. 3086 (1885), 3086–I (1881), 3068–II (1882), 3086-IV (1881), 3086-V (1881).

¹⁵ The Famine Commission's Report suggests that ten percent should be enough for seed grain and wastage. The figure may be low for seed grain. The estimate of 460 pounds of grain per adult per year may be somewhat high. Numerous estimates from the twentieth century place per capita per day consumption of grain near one pound. This would imply a yearly need of 365 pounds. However, the estimate of 460 pounds may be closer if we include pulses (lentils) as well as grain. See Sinha, R. P., Food in India (Oxford: Oxford University Press, 1961), pp. 17–18Google Scholar; Mukerji, Karunamoy, Agriculture: Famine and Rehabilitation in South Asia (Santiniketa, Visva-Bharati, 1965) p. 157.Google Scholar

¹⁶ Moreland, W. H., India at the Death of Akbar: An Economic Study (Delhi: Atma Ram and Sons, 1962), pp. 226–28.Google Scholar These figures are from an earlier period than the one with which we are concerned. Pack bullock technology is unlikely to have changed between 1600 and 1800. Moreland's estimate of the carrying capacity of a bullock may be somewhat high. If so, the argument I am making is strengthened.

¹⁷ Quoted in Datta, K. L., Report on the Enquiry into the Rise of Prices in India (Calcutta: Superintendent, Government Printing, 1914), Vol. I, p. 180.Google Scholar

¹⁸ Morison, T., “The Instability of Prices in India Before 1861,” Royal Statistical. Society Journal, LXV (September 1902), 518–19.Google Scholar

¹⁹ Datta, Report, p. 21.

²⁰ North Western Provinces, “Information Regarding the Cultivation of Cotton in the North-Western Provinces,” Selections from the Records of Government, North Western Provinces, (Allahabad, 1862), p. 68.Google Scholar

²¹ This refers to the distribution of rainfall through the agricultural year. See K. L. Datta, Report, pp. 451–53 for comparative data on the variance of total rainfall for parts of India and other agricultural regions of the world. See also Mann, Harold, Rainfall and Famine: A Study of Rainfall in the Bombay Deccan, 1865–1938 (Indian Society of Agricultural Economics: Bombay, 1955), especially pages 16–30.Google Scholar

²² It is not clear whether this “good” should mean average or above average.

²³ Data on long term storage costs for the nineteenth century do not appear to exist. Grain was stored in pits in the ground that were sealed with mud. The Government of India Grain Storage Research and Training Institute in tests found that grain so stored deteriorates because (1) water seeps in and grain moisture content rises, (2) moisture migration in grain mass during the winter months condenses on the top and surface or grains resulting in damage from moulds, heat, and caking. I would hazard a guess that under such conditions, no usable grain would be left after four or five yean.

²⁴ This model is a “safety-first” type. However, it is not clear to me whether it is best described formally as one where the decision rule is “max μ subject to ” as in Pyle, and Turnovsky, “Safety-First and Expected Utility Maximization in Mean-Standard Deviation Portfolio Analysis,” Review of Economics and Statistics, LII (1970), 75–81)Google Scholar or as a lexicographic safety-first rule as in Boussard, and Petit, (“Representation of Farmers' Behavior Under Uncertainty with a Focus-Loss Constraint,” Journal of Farm Economics, XXXXIX [November 1967], 869–880)Google Scholar and in Roumasset (“Choice of Technique by Lexicographic Safety First for Peasant Agriculture and Reswitching of Risky Techniques,” Discussion Paper #72–3, Institute of Economic Development anl Research, School of Economics, University of the Philippines, 1972 [mimeo]). Decision on which specification of the model would prove most useful can, I think, be postponed until additional research has provided enough data to make quantitative work on the model productive.

²⁵ These results are obtained by using a binomial distribution with the number of trials equal to 30, the number of disasters equal to zero, and varying the probability of disaster from .025 to .05.

²⁶ While calamitous famines occurred over major portions of India every 25 to 30 years during the nineteenth century, below normal crops occurred in smaller areas with much greater frequency. Those years in which crops were below normal but starvation and/or need for relief were rare presumably represent those where storage behavior provided adequate insurance. The calamitous famines were caused by combinations of harvest failures which were too large or too close together to be successfully coped with by storage. A complete list of shortages after 1860 is available in Bhatia, Famines in India: 1860–1965.

²⁷ This statement is not intended to imply that they did not prepare for (or worry about) successive crop failures. Rather, they might anticipate using different survival techniques such as migration into, the hills or into nearby non-famine areas. Also, the costs of storage of grain to last two years may have been prohibitive.

²⁸ Generally speaking, cotton cannot be grown in black cotton soil oftener than one year in three if fertility is to be maintained without heavy fertilization. No such restrictions apply to the cultivation of jowār or bafra. This restraint can be safely ignored, I think, because only three areas (the province of Berar and the districts of Wurdah in Central Provinces and Broach in Bombay Presidency) approached or exceeded this limit of one-third of land under cotton. Even these areas may not have been as close to maximum possible specialization as they appear because the total cultivation figures I used to get percentages of land planted with cotton were net of current fallows.

²⁹ I draw two distinctions. One is food crops versus non-food crops. You can eat food crops. Non-food crops must be sold and the revenues used to purchase food. I make this sharp distinction because we are interested in crisis years when food consumption is threatened. Second is cash crops versus crops grown for home consumption. A lower risk is attached to growing a food crop as a cash crop than a non-food crop. Price risk is avoided, since the food crop can, if necessary, be consumed without interaction with the market. Because of the lower risk, growing a food crop which could be marketed (like a particular type of wheat) could come about more easily than growing a non-food crop for sale.

³⁰ Unfortunately, that literature provides almost no data on the quantity of grain stored on a year-to-year basis.

³¹ Hyderabad Assigned Districts, Report on Settlement and Survey, 1877–78, p. 5, (hereafter HADBSS).

³² HADBSS, 1879–80, p. 3; 1882–83, p. 4.

³³ Hyderabad Assigned Districts, Report on the Administration of the Hyderabad Assigned Districts, 1878–79, p. 4.

³⁴ HADHSS, 1884–85, p. 4.

³⁵ The price of jowār was 2.4 to 2.6 rupees per maund in 1877–78 and 1878–79. In 1884–85 the price was 1.3 to 1.5 rupees per maund. The opportunity to market surpluses before the construction of a railroad was probably limited to occasions when an army was in the vicinity, when people from a famine district migrated to a non-famine one, or when the Banjaras passed close by. The phenomenon of farmers who will market a surplus, indeed are anxious to market it, but who do not allocate their land on the basis of lagged relative prices is not unique to India. Mayhew, Anne, “A Reappraisal of the Causes of Farm Protest in the United States, 1870–1900,” Journal of Economic History, XXXII (June 1972), 471Google Scholar, posits similar behavior on the part of American farmers in the Middle West before 1870. “These descriptions of farming in the Middle West prior to 1870 are not descriptions of farmers who depended on markets for disposal of their produce; rather, they are descriptions of farmers who were quick to take advantage of opportunities to market goods as they arose. … Similar descriptions of pre-1870 farming operations can be found in many sources and all indicate a willingness to take advantage of opportunities to sell produce, but they do not provide a picture of farmers who devoted most of their time and land to growing crops they planned to sell.”

³⁶ Central Provinces, Annual Report on the Administration of the Central Provinces, 1869–70, pp. 85–86.

³⁷ Hyderabad Assigned Districts, Cotton Report of the Hyderabad Assigned Districts, 1878–79, p. 3.

³⁸ Cassels, Walter R., Cotton, an Account of its Culture in the Bombay Presidency (Bombay, 1862), p. 164.Google Scholar

³⁹ The acreage patterns presented in McAlpin, “Impact of Railroads,” are also consistent with the storage model. The minimal changes in percentages of land devoted to different crop groups until late in the period 1860–1900 are precisely what the model would predict. Expansion of cotton cultivation was greatest in Berar where farms were relatively large and where the railroad early connected the area to secure grain producing regions.

⁴⁰ It would be desirable to estimate supply equations with stored grain made an explicit factor. This has not been done (and probably cannot be done) because of the lack of data. The only data on grain storage I have been able to locate covers less than ten years for six districts. There is scattered qualitative evidence but it is not consistent enough or plentiful enough to be used in any meaningful way.

⁴¹ Government of India, Railway Department (Railway Board), History of Indian Railways Constructed and in Progress, Corrected up to 31st March 1945 (Simla: Government of India Press, 1947), pp. 95, 102.Google Scholar

⁴² Government of India, Census, 1901, XV A, pt. 3, p. 4.

⁴³ See Gadgil, D. R., The Industrial Evolution of India in Recent Times, 1860–1939, 5th ed. (Bombay: Oxford University Press, Indian Branch, 1971), pp. 26–27Google Scholar, for a discussion of the effects of the famine in Madras. The famine lasted for two years in the districts cited. Absolute loss of population was: Cuddapah, 230,156; Kurnool, 235,881; Bellary, 185,480; Anantapur, 141,366. After the famines of 1896–97 and 1899–1900, none of these districts showed a decline in population, although the increases in population between the 1891 and 1901 censuses were small.

⁴⁴ In the absence of any statistics on agriculture for the 1870's from Northwestern Provinces and Bengal, both major grain regions, I doubt that it is possible to determine whether India was actually short of requirements during the 1876–78 famine, or whether there was enough grain if it had been better distributed.

⁴⁵ K. L. Datta, Report, Vol. II, pp. 186–87. See McAlpin, “Impact of Railroads,” for earlier cotton prices.

⁴⁶ As yet, the transition period between lack of price responsiveness in the nineteenth century and price responsiveness in the twentieth century has not been adequately explored. There are problems of obtaining consistent data to permit detailed work on the period between 1885 and 1915. When such data is obtained, it may be possible to make more explicit the mechanisms by which constraints on peasant behavior were eased.