¹ Until the late 1930's, agriculture directly provided employment to nearly 70 percent of the Taiwanese male labor force and contributed over 50 percent of the island's production of goods by value. Foreign trade was dominated by agriculture. Without exaggeration, it can be said that the economic development of colonial Taiwan was the development of its agriculture.

² The data used in analyzing output growth are official government statistics. The quality of the production data varies with the crop and with the period. However, agricultural production data are reasonably reliable and fairly complete by the late 1900's. Crop production during the colonial period was estimated as the product of crop area and estimated crop yield. The crop data before 1906 are not usable primarily because of the extensive underreporting of cultivated land before the completion of the land cadastral survey. As the results of the cadastral survey became known between 1903 and 1906, sharp rises occurred in the reported cultivated area. During the colonial period, crop yields were estimated by the crop reporter with the assistance of experienced farmers in his district. That this method yields reasonably accurate results is perhaps best exemplified by the fact that, although in the 1930's the Japanese conducted sample cutting surveys to determine the yield of rice, rice production statistics were nevertheless still based upon the yield estimates of the crop reporters. Evidently the information from the two sources was sufficiently i n agreement to make a change in the method of estimation unnecessary. Underreporting of crop production, either due to the underestimation of crop area or yield, most likely did occur during the colonial period, but it probably was not sufficiently widespread to call into question the general reliability of the production data. The acceptance of crop production data as usable by the late 1900's does not imply that all data after this date are equally reliable. The Japanese statisticians, at times, had voiced dissatisfaction with the crop statistics and had suggested ways to improve them. Cf. , Taiwan, Government-General, Bureau of Colonial Development, A Simple Explanation of the Present Agricultural Statistics in Taiwan (Taipei, 1924). Since at least some of these suggestions must have been adopted, the quality of the crop statistics must have improved over the years. While we suspect the data collected during the early period up to the late 1920's to be less reliable than that collected later, the available evidence does not indicate that the earlier data are biased in a systematic manner or that they are so seriously deficient (as are the data on the period before 1906) that they are useless. Until this conclusion is refuted by further research, we must consider the data collected in the 1910's and 1920's usable for analysisGoogle Scholar.

³ The ratios of cultivated land to agricultural population during the colonial period are as follows:

While land was never in abundant supply, Taiwan, even during the 1930's and 1940's, enjoyed a more favorable land-labor relationship than either Japan or Mainland China.

⁴ Estimates of crop area are, however, slightly less reliable than those for the cultivated area, primarily because of the uncertainty attached to the crop area of minor crops. So far, land is treated as an homogeneous input. In reality, of course, there exist large variations in the quality of the land used as input to agriculture. To know that variation in quality exists among different plots of land does not make it easier to take into account quality differences in estimating land as an input to agriculture. Until more information is available and the art of estimation improves, total crop area probably still provides the best measure of land as an input to agriculture. It should also be pointed out that double cropping does not necessarily bring land of the same quality into cultivation. It is well known that land is less productive when used the second time. Thus, double cropping has the same effect as bringing into cultivation marginal land, as in both cases the additional land input is inferior.

⁵ See Hsieh, S. C. and Lee, T. H., Agricultural Development and its Contributions to Economic Growth in Taiwan, JCRR Economic Digest Series: No. 17 (Taipei, 1966), p. 24; andGoogle ScholarHo, Y. M., Agricultural Development of Taiwan 1903–1960 (Nashville: Vanderbilt University Press, 1966), Table 12Google Scholar.

⁶ The basic problem with the JCRR labor input index is that it is essentially an index of crop area. Small sample surveys were conducted by the government in Taiwan to investigate the production of a number of major crops—rice, sweet potatoes, and sugar cane. From these surveys, estimates of labor days required per unit of crop area were obtained. The government- and the university-operated experimental farms also conducted studies on a number of minor crops, which provided information on the labor input required per unit of crop area to produce these crops. By multiplying the “required” labor input per unit of crop area (obtained from the surveys and experiments) and the actually recorded crop areas, JCRR arrived at its estimates of labor inputs in terms of worker-days. Several objections can be raised immediately. First, the sample farms used in the survey were not randomly selected. Rather, the sample was composed of peasants who were capable of keeping records. Coincidentally, they are probably also the better farmers, whose farming techniques and labor input per unit of land cultivated may be quite different from the average. It is not certain in which direction the sample results are biased. (My feeling is that the bias is upward.) This bias is present, even to a greater degree, when the data from the experimental farms were used. Second, with the exception of rice, only one observation of “required” labor per unit of crop area (often from postwar years) was available. In the case of rice, annual observations of labor input per hectare were available only for the years between 1937 and 1943, inclusive. So the annual labor input per hectare of rice cultivated for the years 1911 to 1936 was also based on a single observation of “required” labor input. Thus this method of estimation assumed a constant land-labor ratio over a period of thirty or forty years. The resulting index, therefore, reflects only the movement of crop areas and not the movement of labor inputs. Thirdly, by basing its estimate of labor input on crop area statistics alone, JCRR ignored the labor supply conditions. The complete disregard of the supply conditions led to some obviously incorrect results. The influenza epidemics of 1915, 1918, and in particular of 1920, drastically reduced the growth of agricultural population. In fact, agricultural population experienced a slight decline after the 1920 epidemic. This led, as one would expect, to labor shortages in agriculture. But by neglecting all factors other than crop area this very significant change in labor availability escaped detection by the JCRR labor input index.

The JCRR index of agricultural labor force and Ho's index of gainfully employed in agriculture are quite similar. Ho's index is adopted as a starting point for estimating labor input to agriculture primarily because Ho converted his estimates of fainfully employed to male equivalents and his estimates are described in some etail, so that one is able to evaluate them.

⁷ Some of the technological innovations, however, may have been labor-saving, which would of course partly offset the effects of the labor-using innovations. It is believed that labor-saving innovations were, however, in the minority.

⁸ Rodenticide made of yellow phosphorus was used in 1926 and the use of rat traps was widely extended in 1936. Cf. Horng, W. H., A Report of the Taiwan Field Rat Control Campaign, Plant Industry Series: No. 17, JCRR (Taipei, 1959), p. 1. Compared with what. was. spent on fertilizer, however,, expenditures. for. these. items constituted a negligible share of working capitalGoogle Scholar.

⁹ , Ho, in Agricultural Development of Taiwan, pp. 52–56, used animal energy as his proxy measure of fixed capital input. He estimated that fixed capital input (in index form with 1901 = 100) declined from a peak in 1910 (204.34) to 182.64 in 1920, 166.55 in 1930, and 128.00 in 1940. As a measure of energy input to agriculture, Ho's estimates are unsatisfactory because they implicitly assume that animals are worked the same number of days each year. (For the postwar years, Ho also failed to account for the introduction of machine energy.) As a measure of fixed capital input, Ho's estimates are grossly'inadequate because of his restrictive coverage. Farm implements, farm structures, and irrigation facilities—all'important parts of agricultural capital—are ignored. Of these, possibly the most important are irrigation facilities.Google Scholar

¹⁰ The methodology used in this section was developed by ProfessorSolow, Robert M. in “Technical Change” and the Aggregate Production Function,” Review of Economics and Statistics, XXXIX (08. 1957),. 312–20. The usual assumptions are made:, (1) the production function is linearly homogeneous; (2) perfect competition exists, so factors are paid their marginal products; and (3) there is neutral technological changeCrossRefGoogle Scholar.

¹¹ Rent surveys conducted during the colonial period reported rent to be in the neighborhood of 40 to 50 percent or total harvest and therefore also lend support to the estimates of a1 of Alternatives I and II. See, for example, Shigeto, Kawano, A Study of Taiwan's Rice Economy (Tokyo: Yuhikaku Ltd., 1941), p. 325, and Taiwan, Government-General,Google ScholarBureau of Colonial Development, Basic Agricultural Survey, No. 39Google Scholar.

¹² A more direct method is to estimate L, the rate of productivity change. For various periods and for various aggregate input measurements, L is estimated and the results are presented in the following tabulation:

The results, of course, confirm what is shown to the text; that is, productivity increase was insignificant until the 1920's and thereafter became a very important factor in explaining output increases.

¹³ Several reasons exist to explain the declining productivity (that is, the “over explanation” of output increases) during the early years. Possibly, productivity was constant during the early years but is shown to have declined because input measurements (specifically, labor and fixed capital) were upward biased. Perhaps a more satisfying and more convincing explanation is that during the first quarter of the twentieth century, the “state of arts” in Taiwanese agriculture remained approximately stable, while the quality of two of the inputs to agriculture (land and labor) deteriorated.

In estimating productivity, the two major inputs to agriculture—land and laborare assumed to be homogeneous when in fact they are not. The most fertile land is cultivated first; only when land becomes scarce is marginal land of inferior quality used. The same holds true for labor. Children and women are used in the field only when labor supply is short. The additional land and labor that were brought into the production process during the first quarter of the twentieth century were noticeably inferior in quality. By 1910 the quality of additional land input deteriorated because only marginal land was available. The quality of additional labor input deteriorated primarily because a series of epidemics between 1915 and 1920 both weakened the population and brought a larger number of women into the labor force to replace the ailing male workers. While, an attempt was made to account for the difference between male and female labor by expressing labor input in terms of male equivalent, it is doubtful that the simple adjustment used was adequate. Without technological change in a sufficient amount to compensate for the inferior inputs, productivity (calculated by assuming homogeneous inputs) declined.

¹⁴ .Thes e are by no means the only agricultural institutions established by the government. Some of the others include tea inspection stations, plant inspection stations, sericulture improvement stations, and veterinary serum production offices. See the Agricultural Association of China, Manual of Agricultural Institutions in Taiwan (Taipei, 1961)Google Scholar.

¹⁵ The most significant advances in crop, improvement made during the colonial period were in connection with sugar cane and rice. Major new varieties of sugar cane were introduced in 1902 (Rose Bamboo); 1917 (Java thin stalk: POJ36, POJ105, POJ161), 1926. (Java thick stalk: POJ2725), and 1937 (F108). Rice seed was improved initially through seed selection, that is, eliminating inferior varieties from planting and propagating those with high yield. The major breakthrough came in the 1920's, with the successful introduction of the higher-yielding Japonica varieties of rice (commonly called Ponlai rice). The first Ponlai variety, Nakumura, was released for planting in 1924, followed in 1926 by the release of perhaps the most popular rice variety ever grown in Taiwan, the Taichung No. 65.

¹⁶ , Kawano, Taiwan's Rice Economy, p. 15.Google Scholar

¹⁷ The estimated average annual consumption of commercial fertilizer per unit area (Kg/Ha) in rice cultivation for 1922–24 and 1930–32 are as follows:

Source: Computed from Taiwan Government-General, Bureau of Colonial Development, Taiwan Agricultural Yearbook, various issues.

¹⁸ , Taiwan, Provincial Government, Department of Agriculture and Forestry, Taiwan Sugar Statistics, No. 2 (Taipei, 1948), p. 143.Google Scholar

¹⁹ The pao-chia system was a scheme which held the village and familial groups responsible for the conduct of their members. Cf. U.S. Office of the Chief of Naval Operations, Civil Affairs Handbook, Taiwan, OPNAV 50 E-12 (06 1944), pp. 77–79. The pao-chia system was also used primarily as a means of disseminating information. The police, however, played a more direct role, acting as the local government until the 1920's. Consequently, the police system was given a wide variety of tasks to perform, among which were such extension services as the promotion of higher-yielding rice seeds and new farm techniques and the planting of windbreaks.Google ScholarCf. “Police and Economic Development in Taiwan during the Japanese Period,” Bank of Taiwan Quarterly, Vol. V, No. 4 (03. 1953), pp. 253–73Google Scholar, and Myers, Ramon H. and Ching, Adrienne, “Agricultural Development in Taiwan under Japanese Colonial Rule,” Journal of Asian Studies XXIII (08. 1964), 565Google Scholar. Even after it was removed from extension work in the 1920's, the police department was still occasionally called on to persuade reluctant farmers to adopt new farming techniques.

²⁰ , Taiwan, Provincial Government, Department of Agriculture and Forestry, The Reorganization of Farmers' Associations in Taiwan (Taipei, 1950), p. 2.Google Scholar

²¹ Ibid.

²² Ibid., p. 4.

²³ Ibid., p. 12.

²⁴ Farmer participation, for all practical purposes, was compulsory, and financially the SAU's were supported jointly by the government, the Farmers' Association, and individual contributions.

²⁵ In a credit survey conducted in 1933, 87 percent of the 37,543 farm families interviewed had outstanding debts. A smaller survey in 1940 showed similar results. Cf. Taiwan, Government-General, Bureau of Colonial Development, Survey of Agricultural Credit, Basic Agricultural Survey No. 33, and Survey of Agricultural Credit, Basic Agricultural Survey No. 43.

²⁶ By extending the organized money market to rural Taiwan, rural credit conditions were greatly affected. In the 1933 credit survey, it was found that 34 percent of the rural credit was provided by the organized money market (development bank, commercial banks, and the Agricultural Cooperatives). By the time of the 1940 survey, the organized money market provided over 50 percent of rural credit. These same surveys show that with the expansion of the organized money market in rural Taiwan, rural interest rates also declined. Of the outstanding rural debts surveyed in 1933 only 39.7 percent carried an interest rate of less than 10 percent. By 1940 loans financed at an interest rate of less than 10 percent formed 89.7 percent of the outstanding debt. It is significant that with credit more accessible, farmers were more willing to borrow for productive purposes. In 1933, 49.75 percent of the outstanding debts were incurred for productive purposes; by 1940 this share had increased to nearly 64 percent.

²⁷ Besides these, many other rural institutions were introduced. In particular, the irrigation association should be mentioned. This organization made it possible for the government not only to regulate irrigation and practice complex rotational schemes but it also helped to accumulate sizable sums of money in support of government irrigational activities.

²⁸ , Taiwan, Provincial Government, Department of Agriculture and Forestry, The Reorganization of Farmers' Associations in Taiwan, p. 20.Google Scholar

²⁹ It is interesting to note that under government assistance and promotion, the irrigated area increased rapidly during the 1920's an d 1930's.

³⁰ As late as 1940, there were still some 2500 Japanese agricultural and forestry technicians, representing nearly two-thirds of the total number, working in Taiwan. Cf. Barclay, George W., A Report on Taiwan's Population to the JCRR (Princeton: Office of Population Research, Princeton University, 1954), p. 63. An even larger number of Japanese held administrative posts both in the rural institutions and in government departments that dealt with agriculture. Without these Japanese personnel, agricultural transformation in Taiwan would have been made more difficultGoogle Scholar.

³¹ The Colonial Government collected information on the daily wage received by agricultural hired laborers (male) in seventeen administrative districts, although riot all of the districts reported every year during the period, 1910–1942. Of these seventeen districts, three (Taipei, Taichung, and Tainan) were particularly important and also had the most complete data. The money wage index was constructed by weighting the wage rates reported in Taipei, Taichung, and Tainan by the number of people in each of these districts who reported their occupation as agriculture in the 1930 Census.

³² Regressing agricultural real wage index, agains t time, t, gives the results:

(a) Agricultural money wage deflated by the agricultural price index.

(b) Agricultural money wage deflated by the index of wholesale prices in major cities.

³³ Gleason, Ralph N., Taiwan Food Balances 1935–54, JCRR Food and Fertilizer Series No. 5 (Taipei, 1956).Google Scholar

³⁴ Ibid., p. 7.

³⁵ The decline in the daily per capita availability of rice during the late 1930's and the early 1940's (Table 5) may be slightly exaggerated. These were the years when the Colonial Government introduced some very stringent controls over rice collection which may have had the effect of encouraging farmers to underreport production.

³⁶ For the years 1925–1939, data on freight and insurance are available, and the adjusted export surplus (accounting for freight and insurance) shows the following:

The adjusted figures show export surplus to be sufficiently large not to alter the conclusion that there was a net diversion of production value from Taiwan since the mid-1910's.

³⁷ The presence of a large export surplus should not be taken as evidence that colonization was the most “profitable” course Japan could have adopted. To evaluate the “profitability” of colonization, it is necessary to consider the opportunity costs of the Japanese capital, labor, and technical and entrepreneurial talents employed in Taiwan. This is, of course, a much different and a more difficult problem than the one being analyzed in this article.

³⁸ They were the Taiwan Seito, Daninippon Seito, Meiji Seito, and Ensuiko Seito. See , Kawano, Taiwan's Rice Economy, p. 164Google Scholar.

³⁹ This regulation divided the sugar cane growing areas into supply regions, and each region was assigned to a specific refinery. Without the permission of the government, cane produced in one supply region could not be transported outside the region or be used other than for sugar manufacturing. Cf. Chen, Cheng-Siang, Taiwan: An Economic and Social Geography, Research Report No. 96, Fu-min Geographical Institute of Economic Development (Taipei, 1963), p. 312. Farmers were theoretically allowed to switch from growing cane to growing other crops. In practice, however, this freedom was greatly limited. Soil conditions and the availability of water (controlled by the government-managed irrigation associations) usually made sugar cane the most suitable crop for cultivation. This regulation not only insured the modern sugar refineries of a steady supply of sugar cane but also helped to accelerate the decline of primitive sugar millsGoogle Scholar.

⁴⁰ This was the Teikoku Seito, the fifth largest sugar company in Taiwan. Cf. , Kawano, Taiwan's Rice Economy, p. 182. Generally only a weak connection existed between the price of sugar and th e price of cane. The correlation coefficient of the price of rice and the price of cane was 0.9201, and that of the price of cane and the price of sugar was 0.5306.Google ScholarCf. Yu, R. J., “Research in the Relative Prices of Rice and Sugar in Taiwan,” Bank of Taiwan Quarterly, Vol. V, No. 4 (03. 1953), p. 50Google Scholar.

⁴¹ Grajdanzev, Andrew, Formosa Today (New York: Institute of Pacific Relations, 1942), p. 99.Google Scholar

⁴² These figures are calculated from Taiwan, Provincial Government, Bureau of Accounting and Statistics, Taiwan Province: Statistical Summary of the Past 51 Years (Taipei, 1946), pp. 996 and 1004Google Scholar.