1. Controversies over Colonial Era Yield per Acre
(1) G. Blyn's Research
Whether Britain's domination of India brought an increase in agricultural productivity is among the great controversies in Indian historical economic research.
One of the most systematic works of statistical research on changes in agricultural productivity in the colonial era is George Blyn's Agricultural Trends in India, 1891-1947: Output, Availability, and Productivity (Philadelphia, 1966). Statistics on agricultural productivity in the colonial era start with Agricultural Statistics of India, first published in 1884, for acreage planted for major crops, and Estimate of Area and Yield of the Principal Crops in India, which begins in 1891, for the per-acre yields. The district-level statistics were published under the title Season and Crop Reports.
Estimates of agricultural output by the Indian government, which form the core of these statistics, are based on the following means of calculation: "a given crop's output" = "district standard (or average, normal) yield per acre" x "crop acreage" x "seasonal conditions." The first item, the "standard," is decided at the district level for each crop. There is no information on the precise specifications, but the standard yield of a crop is hypothesized to be that produced at under normal weather conditions in a given period of time on land of average quality. It was decided that the district standards would be adjusted on the basis of the results of crop-cutting experiments conducted every five years but, as will be described below, there were periods during which the five-yearly experiments were not conducted. Crop acreage means area planted regardless of whether or not the crops were harvested. The information is based on reports of village officials but there is general agreement that the figures were accurate. The "seasonal conditions" are figures for output which district officials have indicated for a given year in terms of percent or in terms of anna. The standard is a village's average yield under normal weather conditions, and the information is based on the judgments of village officials stated in reports to the districts. As will be seen, major doubts about the accuracy of these figures have emerged.
Blyn showed that, according to these statistics, yearly production growth in food crops from 1891 to 1947 was no more than 0.11 percent, meaning a per capita decline in output. The growth rate was high in the first half of this period and low during the second half. Production of nonfood crops rose considerably but there was great variation among different types of nonfood crops.
(2) The Research of Alan W. Heston and Criticism of His Work
Alan W. Heston attempted to correct Blyn's interpretation regarding the per capita fall in output of food crops through research examining changes in agricultural productivity in the Bombay Presidency. The standard yield per acre in Bombay's Presidency did not undergo revision throughout the long period from 1897 to 1946, so that changes in output per unit area for this period are the results of changes solely of seasonal conditions. Heston claims that this seasonal condition is a by-product of the revenue system and cannot be understood to reflect real overall output. Starting from this point, he points out that movements of seasonal conditions do not correlate with precipitation or other yearly trends in the weather. The reason why seasonal conditions show decline during the earlier period is administrative reasons while the causes of decline in the latter period are political pressures and nationalist movements. Moreover, there is also a bias in the standard yield per acre, particularly from 1886 to 1897 when it was set at a high level. The official administrators were not sufficiently knowledgeable about the true yield, and desired to exaggerate the crop yield and thus minimize the apparent burden of taxation. Heston concludes that there is no evidence to back either the view that yield per acre fell in the period 1886 to 1947 or the opposite view, that it rose ("Official Yields per Acre in India, 1886-1947," Indian Economic and Social History Review, 10:4 ).
Quite a few counterarguments have been advanced against Heston's claims. For example, Sumit Guha recently made the following criticisms. First, responding to Heston's assertion that seasonal conditions, since they do not correlate with precipitation trends, do not accurately indicate true output, Guha claims that there is some degree of correlation. Further, Guha argues, precipitation is only one of several factors with an effect on a year's harvest and therefore, even if there is no correlation, it cannot be claimed that seasonal conditions do not reflect the true output. Because prices for agricultural products had risen, the proportionate share of agricultural products in land taxes had already fallen to a very low level by 1907, so there was definitely no administrative necessity to lower the seasonal conditions. As to the effect of 'no-rent' campaigns, even if the land taxes of the landlords had been lowered, the tenants who mounted the no-rent campaigns would have realized no benefit. Therefore, Ashwani Saith argues, Heston's claim that the tenants' movement brought about a lowering of harvest indicators is unfounded. They also criticize Heston's contentions on other grounds, one being that there were regions which, like Madras, experienced rises in per acre production, and that political factors therefore cannot explain the outcomes.
Guha also compares results of crop-cutting experiments for some regions, and argues once again that per acre production of food crops declined (Sumit Guha, ed., Growth, Stagnation or Decline? Agricultural Productivity in British India, Delhi, 1992).
2. Changes in Output of Paddy Per Acre in Tamilnadu: 1870s - 1950s
(1) From the 1870s to 1918
Even Blyn's research indicates that food crop yields per acre rose on the whole in the Madras Presidency in South India. This trend is shown in his main source, Season and Crop Reports. Guha criticizes this finding. Using a comparison of data from Madras districts' crop-cutting experiments, he argues that per-acre production of paddy declined in this region. He used three sets of crop-cutting experiment data in his comparisons. The first consisted of the crop-cutting experiments carried out in districts where land settlement operations were being conducted in the 1870s and 1880s (hereafter '1870 data'); the second resulted from the experiments conducted by the Indian Council of Agricultural Research (ICAR) in 1944-49 (hereafter '1944-49 data'); and the third was harvest data from the results of the crop-cutting experiments published in Season and Crop Reports from 1955 to 1957 (hereafter '1955-57 data').
Since there are suspicions that the standard yields per acre and the harvest indicators may not be accurate, Guha's attempt to use data from crop-cutting experiments is certainly appropriate. However, at least three more sets of experiments were conducted in the Madras Presidency between the 1870 crop-cutting experiments and the 1944-49 crop-cutting experiments. These were a five-year experiment conducted in 1901/02 (hereafter '1901 data'), a five-year experiment conducted from 1906 to 1911/12 (hereafter '1911 data'), and a 1917 experiment (hereafter '1917 data'). Because these three sets of crop-cutting experimental data and the 1870 crop-cutting experiments carried out at the time of the land settlement operations -- in other words, the four crop-cutting experiments conducted before 1917 -- used basically the same methods, it is possible to compare their results to estimate to some extent changes in paddy productivity.
As shown in Table 1, the results of the four crop-cutting experiments indicate that yield per acre rose in each of the nine districts of Tamilnadu between 1901 and 1917. Data from 1870 for two districts does not exist, but in five of the remaining eight districts we also know that per-acre yields increased between the 1870s and 1917.
As mentioned earlier, the standard yield per acre was revised according to results of the crop-cutting experiments carried out every five years. In fact, the districts' standard yields per acre were revised following these experiments, whereas in the period from 1919 to 1954 when the experiments were not conducted, there were no revisions in the standard yield and the same figures were published continuously from 1918. Season and Crop Reports did not list figures for the standard yield per acre before 1906, but in 1892 the Madras government used numbers of the same nature as the standard yield per acre in its calculations. These figures are taken from a variety of results whose statistics have been judged and, especially, from land settlement reports. These values for the 1892 standard yield per acre and the later standard yield per acre values correlate with the above results, and indicate that productivity rose in all districts, suggesting a rise in paddy production per acre.
(2) From 1919
There are major difficulties with the estimates of paddy production per acre from 1919 for Madras. As previously mentioned, crop-cutting experiments were not conducted by the Madras government from 1917 through 1954. ICAR did conduct crop-cutting experiments between 1944 and 1949, but the methods were much different from those used previously. Whereas ICAR experiments used random sampling to select fields, previous experiments had entrusted selection to the discretion of the officials. The defects of the previous sampling method were recognized by officials of the Revenue Department. Reports of the Revenue Department state that average assessments on the land selected for crop-cutting experiments were higher than the average for the Presidency as a whole. There was, in other words, a tendency to conduct experiments on land yielding above-average harvests, so the results could not be taken as directly representing average yields for the entire Presidency. Thus testing results prior to 1918 exhibit a bias, so that it is not appropriate to compare them to 1944-49 results based on random sampling to discover changes in output. It has to be acknowledged that Guha's use of a comparison of the 1944-49 and 1870 crop-cuttings to conclude that yield per acre fell is inappropriate.
If the 1944-49 data is incompatible with earlier results, what indicators can we use to study changes in output? As already mentioned, for the period from 1919 to 1954, when the standard yield per acre was not revised, there is just one set of time series data for seasonal conditions. The same situation holds for the Bombay Presidency for the period from 1907 until India achieved independence. However, as noted earlier, there are doubts about the reliability of seasonal conditions. For this reason, we should make comparisons with other data and investigate the reliability of figures for seasonal conditions.
To begin with, comparing the 1944-49 crop-cutting data with the appropriate Season and Crop Reports data for 1945-49, that is, yield per acre calculated from standard yield per acre and seasonal conditions, shows that they approximate one another (Table 2). Though in many districts, values for the former are higher than for the latter, with one exception, the differential is less than ten percent. Further, the 1953-54 standard yield per acre, especially values for the districts in paddy regions, calculated from seasonal conditions closely match the results of the 1955-57 crop-cutting experiments. As previously mentioned, the 1944-49 ICAR experiments were conducted at fields selected through random sampling, and therefore those figures should closely approximate actual output. We do not have sufficient information regarding means of sampling for the 1955-57 data, but since we can assume that the same random sampling practices as in 1944-49 were employed, I believe that the results closely corresponded to the actual harvest. In brief, though there is definitely a differential between seasonal conditions and the actual harvest for the 1940s and 1950s, both show changes occurring in almost the same way and at nearly the same degree.
These kinds of comparisons allow us to infer that the seasonal conditions listed in Season and Crop Reports rather closely reflect the actual harvests in the Tamilnadu region. If this inference is correct we can conclude that output of paddy per acre in Tamilnadu stagnated in the 1920s and 1930s, fell in the mid-1940s, and recovered to reach its prewar level in the 1950s. Other estimates of changes from the 1920s using data suggested by other descriptive materials roughly correspond to this inference.
In summary, we can infer that paddy output per acre in Tamilnadu rose on the whole from the 1870s through the 1910s, stagnated during the 1920s and 1930s, fell drastically during the 1940s, and revived during the 1950s.
However, we must acknowledge that there are problems with these estimates. Our assumption that pre-1919 data is mutually comparable rests on the assumption that all of the tests overestimate real average output because they did not use random sampling. The degree of bias may have varied by year. Thus, it is difficult to know to what extent it is possible to make comparisons using pre-1919 crop-cutting data. Our estimates of harvests from 1919 are based on the close movements of the harvest indicators for 1944-49 and the 1950s and on calculations of harvest outputs derived from crop-cutting experiments. However, there are grounds to suppose that results of these crop-cutting experiments were communicated to the Madras government before it announced the seasonal factors and that the seasonal factors were modified to correspond to the experiments' data prior to their publication. If that is the case, it undermines the reasons for believing that the seasonal conditions for Tamilnadu generally reflect real changes. It is likely that coming research will enable us to confirm our conclusions.
Table 1:Yields of Rice per Acre in Tamilnadu as Obtained by Crop-Cutting Experiments(lbs.)
Table 2:The Rice Harvest per Acre(lbs.)
University of Tokyo, Institute of Oriental Culture