IMPACT OF NEW AGRICULTURAL TECHNOLOGY ON THE ...

Journal

of Development

Economics

29 (1988) 199-227.

North-Holland

IMPACT OF NEW AGRICULTURAL TECHNOLOGY ON THE INSTABILITY OF FOODGRAIN PRODUCTION AND YIELD Data Analysis for Bangladesh and its Districts* Mohammad

ALAUDDIN

University of Melbourne, Parkoille, Vie. 3052, Australia Rajshahi University, Rajshahi, Bangladesh

Clem TISDELL University of Newcastle, Received

March

Newcastle,

1986, linal version

NSW 2308, Australia received July 1987

Recent studies claim that the ‘Green Revolution’ has increased foodgrain production and yield variability. Using Bangladeshi national and regional data, this paper provides intertemporal and cross-sectional evidence to the contrary. Intertemporal data indicate that on the whole the relative variability of foodgrain production and yield has fallen with the adoption of ‘Green Revolution’ technologies, Cross-sectional regional data further support this view. Using various proxies for technological change the results indicate that regions with higher rates of adoption of ‘Green Revolution’ technologies, have lower relative variability of foodgram production and yield.

1. Introduction Foodgrain production is central to the agricultural economy of Bangladesh. For the period 198&82 (that is, on average for the years 198s 81, 1981-82 and 1982-83) rice and wheat together occupied over 83 percent of gross cropped area [BBS (1984c, pp. 217, 218, 255)]. Rice is the most important crop in Bangladesh. During 198&82 about 80 percent of the total cropped area was planted to rice and this accounted for 93 percent of the total foodgrain production in the country. For centuries, Bangladesh was self-sufficient in food. In recent decades, however, it has become a net importer of food and its food import has been growing. Imports as a percentage of total available foodgrains for consumption have increased from less than two percent in the early 1950s to well over ten percent in recent years [Alauddin and Tisdell (1988)]. At the same time, agricultural produc*We are grateful to two anonymous referees for useful comments on an earlier draft paper and Kathy Renfrew and Janet Piper for statistical and computational assistance. 03043878/88/%3.50

0

1988, Elsevier Science Publishers

B.V. (North-Holland)

of the

200

M. Alauddin and C. Tisdell, Impact of the ‘Green Revolution’

tion in Bangladesh remains erratic because of considerable year-to-year climatic variability [BBS (1984c, p. 248)]. Fluctuations in domestic grain production are usually compensated for by fluctuations in grain imports [Alauddin and Tisdell (1988)] and this periodically strains the already scarce foreign exchange reserves of Bangladesh and makes development planning more difficult. Because of Bangladesh’s growing deficiency in its food supply, the government of Bangladesh has given top priority to the introduction of agricultural innovations for expanding food production. Modern seedfertilizer-irrigation technology was introduced in several phases. It commenced with the distribution of chemical fertilizers and the introduction of modern irrigation in the early 1960s. High Yielding Varieties (HYVs) of rice and wheat suitable for cultivation during the dry (r&i) season were introduced in the late 1960s. Subsequent years saw the introduction of HYVs of rice suitable for the wet (kharif) season. By 198&82, over 26 percent of the total foodgrain area had been brought under HYV cultivation compared to less than two percent during 1967-69 (average for the years 1967-68, 1968-69 and 1969-70). During this period the quantity of chemical fertilizers applied rose from just over 9 kgs of nutrient per hectare of gross cropped area to over 33 kgs. Nearly 70 percent of the dry season rice area is now under HYV cultivation. Practically all the wheat area is now under HYV cultivation and over 16 percent of the rainy season foodgrain area is under HYV cultivation. The proportion of the foodgrain area irrigated increased from about eight percent in 1969-70 to over 13 percent in 1980-82. In the light of these developments several pertinent questions can be asked. To what extent have these developments led to alterations in the variability of overall foodgrain production and yield? Is there any significant difference in production and yield variability between traditional and modern foodgrain varieties? Has irrigation had a stabilizing impact on production and yield fluctuations? All these questions are worth considering in the Bangladeshi context. In most LDCs, including Bangladesh, foodgrains are the most important wage goods and the income elasticity of demand for food is very high, probably of the order of 0.60 or higher [Johnston and Mellor (1961, p. 572); cf. Mahmud (1979, p. 65)]. In the short run, changes in relative food prices critically alter the real incomes of the low income earners. If technological change leads to greater variability in production and hence supply of foodgrains, their prices may become more unstable. In the LDCs, this can have important welfare consequences for low income earners and increase fluctuations in incomes received by grain producers [Mellor (1?78)]. Evidence from Bangladesh about changes in variability of production and yield of foodgrains prior to and following the introduction of the HYVs of cereals and associated techniques is outlined and investigated in this paper. After a brief literature review, we specify the methodology used for analyzing

M. Alauddin

and C. T&dell, Impact

oJ the ‘Green Revolution’

201

the data. Trends in variability in foodgrain production and yield for Bangladesh as a whole, as well as in the main regions (districts) of Bangladesh, are then discussed. District (regional) data are also used to examine how variability of foodgrain production and yield have altered with the introduction of new agricultural technology. Because rates of adoption and spread of the new technology by regions are uneven [see, for example, Alauddin and Mujeri (1986)], we investigate whether production and yield variability are systematically related to the rate of diffusion of new agricultural techniques.

2. Technological change and variability A brief literature review

of foodgrain production and yield:

Stability and adaptability of yields and production of crops have been discussed in the literature both from theoretical and empirical perspectives. Studies include those by Evenson et al. (1979), Finlay and Wilkinson (1963) and Tisdell (1983). Evenson et al. (1979) point to the need to draw a distinction between (a) stability of a genotype, that is, its changing performance with respect to environmental factors over time, and (b) adaptability, that is, its performance with respect to environmental factors that change across locations. Evenson et al. (1979) express concern that new High Yielding Varieties of crops could increase yield variability in developing countries and recommend more research into crops with a view to reducing such variability. Recent in-depth studies of Indian agriculture [Mehra (1981), Hazel1 (1982, 1984)] found evidence of increased instability in agricultural production following the introduction of modern agricultural technology.’ Parthasarathy (1984, p. A74) indicates a higher degree of variability following the ‘Green Revolution’ in the Indian state of Andhra Pradesh. He further claims that greater yield instability is positively associated with districts experiencing higher agricultural growth rates. While the studies by Hazel1 (1982) and Mehra (1981) are substantial, these studies suffer, in our view, from methodological limitations. First they assume arbitrary time-cutoff points. Secondly, they do not adopt a consistent rule for dropping observations for ‘unusual’ years from the data used. Probably both Mehra and Hazel1 were justified in dropping observations relating to 1965 66 and 1966-67 because of severe drought during those years. As Hazel1 (1982, p. 13) points out, ‘catastrophes of this kind are sufficiently rare and severe that they can be considered as separate phenomena from year to year fluctuations’. Mehra (1981, p. 10) argues that, ‘the mid-1960s witnessed two ‘Some earlier studies agricultural development

[e.g., Sen (1967), Rao (1975)] indicate a causal connection and instability in agricultural output growth in India.

between

202


drought years 1965-66 and 196667 of such unusual severity as to significantly alter the variance of any period in which they are included, thus casting doubts about the validity of their conclusions’. On closely examining the foodgrain production data presented by Sawant (1983, p. 476), one can identify the two worst years during the period 196768 to 1977-78 which corresponds to the second period designated by Mehra (1981) and Hazel1 (1982). In 1972-73 Indian foodgrain production dropped by over eight million metric tons (eight percent) from the previous year’s production. It was even worse in 1976-77 when the decline was ten million metric tons (over eight percent) from the production of 1975-76. Apart from 1965-66 and 1966-67, no other year between 1950-51 and 1977-78 saw such an absolute decline in foodgrain production in India. The seriousness of the problem can also be seen if one considers the yields of the two major foodgrains (rice and wheat) which together constitute over 70 percent of all foodgrains during the period 1967-68 to 1977-78 [Hazel1 (1982, p. 13)]., Using data from Joshi and Kaneda (1982, p. A3), one can identify a few bad years in terms of yields per hectare. During 1972-73 wheat yield fell by 109 kgs per hectare (over eight percent) from the 1971-72 level of 1,380 kgs. In 1973-74 it dropped by another eight percent. For rice, 1972-73 was not as bad as it was for wheat. But during 1974-75 rice yield dropped to 1,045 kgs per hectare from the previous year’s 1,151 kgs (over nine percent). The worst year for rice during the second period was 1976-77 when rice yield fell to 1,088 kgs from 1,235 kgs in 1975-76 (over nine percent). Overall, of course, 1972-73 and 1976-77 remain the two worst years during the second period. To be consistent one would have expected these two years to have been dropped by Mehra and Hazel1 from the analysis of the second period. In that case, one could end up with a different picture to that suggested by Mehra (1981) and Hazel1 (1982). It seems likely that the findings of both Mehra (1981) and Hazel1 (1982) are sensitive to changes in time-cutoff points and to their decisions to delete certain observations. This gains some support from a recent study by Hazel1 (1985). In that study, Hazel1 compares instability in world cereal production between two periods, viz., 1960-61 to 1970-71 and 1971-72 to 1982-83, and also examines instability in cereal production in different regions of the world, e.g., in South Asia and in India. When comparing the instability of cereal production between the two periods for India, he does not drop observations for 1965-66 and 196667. Nor does he drop the observations for 1972-73, 1976-77 or 1979-80 when total foodgrain production fell by a huge 22 million metric tons, i.e., about 17 percent [Sawant (1983, p. 476)]. When no observations are dropped from either period, one finds that the coefficient of variation of cereal production in India decreases by 29 percent [Hazel1 (1985, p. 150)] during the second period (1971-72 to 1982-83) as compared to the first (1960-61 to 1970-71) whereas the earlier studies by


203

Mehra and Hazel1 indicated a rise in the coefficient of variation. A recent study [Jain et al. (1986, p. 80)] extending Hazell’s analysis to 1983-84 and without dropping any observations from either period finds that the period of new technology (1967-68 to 1983-84) is associated with a lower production and yield variability compared to the earlier period (1949-50 to 1966 67). However, other limitations of the Hazel1 study apply to the Jain et al. study. Thus the assumption of arbitrary cutoff points and inconsistency in deletion of observations, can lead to conflicting results. Mehra (1981) claims that irrigation reduces yield instability. Our analysis of Mehra’s data [Mehra (1981, p. 29)] for fifteen Indian States supports her claim. Coefficients of correlation between the ranks of both standard deviation and coefficient of variation of yield, on the one hand, and the percentage area irrigated were 0.1096 and -0.4714, respectively. Further analysis of Mehra’s data indicates the standard deviation of yield has a tendency to increase with percent area under HYV and fertilizer application (rank correlation coefficients respectively are 0.4688 and 0.4528) and to decline with higher intensity of cropping (rank correlation coefficient: -0.0783). Coefficient of variation of yield shows a tendency to decline with increased use of modern technology. The coefficients of rank correlation with percent area under HYV, fertilizer application and cropping intensity respectively are -0.1500, - 0.1179 and -0.5584. This is consistent with the hypothesis that modern technology has tended to reduce overall yield variability rather than increase it, a matter which will now be specifically explored using Bangladeshi national and regional data.

3. Methodological

framework of analysis

Linear trend lines* were fitted to relevant Bangladeshi data (foodgrain production and yield) for two periods: 3 194748 to 1968-69 (period of traditional technology) and 1969-70 to 1982-83 (period of modern technology) for Bangladesh as a whole and for the districts or regions4 shown in fig. 1. Variability of production and yield is measured in terms of their yearly deviations from the estimated trend values. Let *We fitted semilog trend also but the results were not any different in terms of R2 and t-values. 3National data on all HYVs are available since 1967-68 but distinct level data on all HYVs are available only from 1969-70. To provide a common basis of comparison hetween national and regional results, we have taken 1969-70 as the dividing line. As there was little penetration of the new technology in 1967-68 and 1968-69, the time-cutoff point we have adopted in this er is unlikely to prejudice overall findings. pa? For the purpose of our analysis, we consider the 17 districts which were used in Bangladesh until the later part of the 1960s. At present, Bangladesh is divided into more than 60 administrative districts which have been created from the 17 districts.

l.D.E.-

D


204

LEGEND INTERNATIONAL

.

DlSTRlCT CAPITAL

SO”RCE:

sO”ND*RI

v

z

-

BOUNDARY CITY ADAPTED

9) FROM

BBS

(1982)

890

Fig. 1. Map of Bangladesh

X,=a+b,+u,,

900

showing

y

910

the 17 districts

considered

920

e

in this paper.

(1)

where X, denotes the dependent variable (production or yield), t is time and u, is the random disturbance term with zero mean and variance s2. Then absolute variation around the trend can be measured by the standard error of estimate (SE) of the relevant variable which is given as

M. Alauddin and C. T&dell, Impact of the ‘Green Revolution’

205

SE=[(X,-QZ/(n-2)]+, where X, is the estimated trend value of X, and n is the number of observations. Variance is defined as the square of standard error of estimate. The relative variability, the coefficient of variation5 (CV), is then defined as

CV=(SE/X) x 100,

(3)

X being the mean of X,. Since we wish principally to consider inter-district variability in foodgrain production, relative variability is more appropriately measured by the coefficient of variation than the standard error of the estimate or the unexplained variation. Because different districts have different mean levels of production and yield (see table 5), standard deviations are likely to be higher for the districts with absolute values of those variables. A comparison of inter-district standard deviations, therefore, would be of little value. Let us, therefore, concentrate on an inter-district analysis of relative variability. One way of examining the implications of production and yield variability is to calculate probability of these variables falling five percent or more below the trend for each year for each district. One difficulty with such an exercise is the unknown distribution of possible production and yield outcomes because of a single observation for each variable in one year. Since the variance of production and yield around the trend can be estimated for the entire period, it is possible to obtain average probabilities on the assumption of a constant variance for all years during the period. These probabilities can be obtained as follows [see Hazel1 (1985, p. 149)]. Let Xi, = Xi + ut, where Xi is the overall mean for the ith district, Xi, is the trend value of X for the ith district in year t within a period and U, is the deviation from the mean in year t. Then the probability of a shortfall of five percent or trend more below can be estimated from the Pr[u,/SE < -O.O5X/SE] where SE is the standard error of estimate. Assuming u, is approximately normally distributed the derived probability can be obtained from the tabulated values of cumulative normal distribution. While this approach, adopted by Hazel1 (1982, 1985), is useful in making a comparative analysis of variability across districts, it suffers from two limitations. First, production or yield stability or lack of it is typically measured around an arbitrarily assumed trend line depicting a hypothetical path for change in certain variables over a time interval of specified duration. The choice of mathematical function is likely to influence any conclusions SThe definition of coefficient of variation employed in tables 1, 4 and 5 does not conform to the standard definition used in statistical literature. As we have measured deviations around the trend values instead of the actual mean value, it is perhaps more appropriate to term it coefficient of ‘unexplained’ variation rather than coeflicient of variation.


206

about the pattern of growth and instability of output and yield [cf. Ray (1983), Rudra (1970)]. Secondly, variability is assumed to remain constant over a period of time for a district. The methodology used here, which is similar to that used by Hazel1 (1982, 1985) does not provide information about the changing behavior of variability with the passage of time. Ray (1983) also suffers from this limitation. We adopt the approach of Hazel1 (1982) for two reasons: First, this method has not been applied to Bangladeshi data. It has been applied to Indian data and it is useful to have comparison between India and Bangladesh. Secondly, our objective is to examine the behavior of variability between districts. We first consider Bangladeshi national time series data and then focus our attention on regional data looking at trends over time and interregional (cross-sectional) differences. 4. Observed behavior of variability 1947-48 to 1968-69 and 1969-70 to 198283:

An analysis of national time series data

Variability of foodgrain yield and production for Bangladesh as a whole can now be examined. Table 1 sets out average production and yield of various food crops for Bangladesh as a whole for 194748 to 1968-69 (Pgriod 1) and 1969-70 to 1982-83 (Period 2). In the first period there was little penetration of HYV technology whereas the second period corresponds to significant adoption of HYV technology. Table 1 also sets out the respective coefficients of variation and the probabilities of a five percent (or greater) fall in yield and production below the trend. Intertemporal comparison shows significant rises in production and yield of all varieties of foodgrains. While rice production rose significantly, wheat production registered a phenomenal increase between the two periods. Foodgrain yield increased by 22 percent. The real increase in yield would be higher when one takes into account the increase in cropping intensity from 130 percent to 146 percent over the same period. An important way in which the new varieties have added significantly to foodgrain production is through extra cropping in each year and hence increased productivity on already cultivated land [Alauddin and Tisdell (1986)]. Furthermore, table 1 indicates a substantial decline in relative variability of foodgrain production and yield as between periods. Also for all foodgrain crops, except wheat, the probabilities of a five percent or greater fall in production and yield below the trend have fallen markedly. In general, irrigated food crops show greater declines in variability compared to rainfed ones. This is confirmed by the fact that reductions in the coefficients of variation have been much greater for rabi foodgrains (boro rice and wheat taken together) than for kharif foodgrains (aus and man rice taken together). This suggests that irrigation has reduced instability. It is interesting to note

1219 939 1197 2142 1441 1107 2041 1228

B. Yield (kgslha) Rice 1005 AU.5 863 Aman 1049 Boro 1191 Wheat 622 Kharif 994 Rabi 1125 All food 1003

2

21.29 8.81 14.11 79.85 131.67 11.37 81.42 22.43

42.22 38.56 16.56 360.90 1161.76 22.37 410.29 46.60

T/, Change

9.15 10.74 10.36 44.01 41.18 9.08 43.57 9.16

Period

6.74 9.13 8.93 15.09 41.96 8.41 16.64 6.85

- 26.32 - 15.02 - 13.80 -65.71 1.90 - 7.36 -61.80 - 25.29

of variation (%) _-__ Period 2 % Change 1

Coefficient

___

1

29.26 32.10 31.46 45.46 45.18 29.09 45.42 29.26

Period

1

Probability fall below

22.93 29.19 28.77 37.03 45.26 27.60 38.21 23.27

Period

2

-21.63 -9.07 -8.55 - 18.54 0.18 -5.12 - 15.87 - 20.47

% Change

of a 5% or greater trend (%)

Foodgrain

6.97 4.35 - 37.58 23.64 12.51 -47.08 8.23 5.86 -28.81 27.12 19.66 -27.51 8.48 6.77 - 20.24 27.79 22.99 - 17.27 15.28 7.98 -47.76 37.18 26.56 - 28.56 13.18 15.13 14.75 35.24 37.03 5.08 7.45 5.96 - 19.92 25.08 20.08 - 19.94 15.20 6.47 - 57.45 37.11 21.97 - 40.80 6.98 4.40 -36.99 23.70 12.77 -46.12 __-__“Khar$ refers to summer and rainy rice crops viz., aus and aman rice while rabi refers to dry season food crops viz., boro rice and wheat. ‘All food’ is the total rice and wheat crops. Source: Adapted from BBS (1976, pp. 1-2, 4-7, 8-10, 12-15, 2&29; 1979, pp. 1688171; 1980, pp. 2tL-25, 3G31, 33-34, 3637, 4652; 1982, pp. 232, 235-238, 240-241; 1984a. pp. 39, 42; 1984b, pp. 249-252, 255); BRRI (1977, pp. 89); World Bank (1982, tables 2.5 and 2.6). Local aman includes transplant and broadcast aman. High yielding varieties of aus, aman and boro rice also include pajam varieties.

12336 2979 7010 2346 429 9990 2776 12765

8674 2150 6014 509 34 8164 544 8708

Rice AUS Aman Boro Wheat Kharif Rabi All food

~~

values ___ 1 Period

(000 m tons)

Period

Average

___

Table

in average quantities, coefficients and probabilitity of a 5 percent or greater fall below the trend: production and yield, Bangladesh, 1947-69 (Period 1) and 1969-70 to 1982-83 (Period 2).”

A. Production

Crops

Changes

208


that the production and yield of wheat became more variable during the second period. This is probably in part because only a little over a third of the area under wheat cultivation is irrigated compared to 90 percent for Bore rice. Therefore, the stabilizing impact of irrigation is likely to be much less prominent for wheat than ,it is for Bore rice. The behavior of overall variability in foodgrain production and yield presents a contrasting picture to that for Indian agriculture [cf. Mehra (1981, p. 18)]. One can employ the Hazel1 variance decomposition method [Hazel1 (1982)] to identify the apparent sources of change in production variance following the introduction of the new agricultural technology. While we have some reservations about this method [see Alauddin and Tisdell (1986)], since it is similar to that of Minhas and Vaidyanathan (1965), the method has been widely used despite its mechanistic nature. We applied it using table 1 in order to assess the relative importance of various components of change [Hazel1 (1982, table 6, p. 20, and eq. 13, p. 21)]. However, the results were statistically insignificant and indicate that overall production variance only increased by 20 percent between the periods analyzed. Furthermore, the change was numerically too small to enable a robust analysis of components of change to be completed. Examination of the time-trends of variability within a region may be considered using an alternative approach. The degree of variability can be measured in terms of deviations from a moving average of a specified period (say five years) of the relevant variables. The variance of a variable for any year is estimated as its observed variance for the five-year period up to and including the year under consideration. So the variance itself is a moving value. This enables one to get a series of values of absolute and relative measures of variation and identify particular phases during which variability tends to increase, decrease or remains more or less stationary. Using this approach and applying it to Bangladeshi time-series data, we briefly examine the changing behavior of production and yield variability over time. Table 2 sets out the absolute and relative variability of production and yield of all foodgrains. Observe the following: First, there seems to be little time trend in both absolute and relative measures of variability. Secondly, absolute production variability seems to increase initially and then appears to stabilize, albeit with occasional minor fluctuations. Relative production variability increases at first and then seems to decline over time. Thirdly, both absolute and relative measures of yield variability show initial increases and then tend to stabilize if not decline. If one plots the data for absolute and relative production and yield variability changing phases of their behavior can be examined. On the basis of such plots (not reported here), we can identify (visually) two distinct phases (Phase 1: 1949-50 to 196G61; Phase 2: 1961-62 to 1980-81). (Later we divide the phases differently. We do so on the basis of specific changes in


209

Table 2 Absolute and relative variability of production grains based on five yearly moving averages: 1982-83.” Production

variability

and yield Bangladesh

of all food1947-48 to

Yield variability

Year

Absolute (000 m tons)

Relative (%)

Absolute (hgs/hectare)

Relative (%)

1949-50 195&51 1951-52 1952-53 1953-54 195455 1955-56 1956-57 1957-58 1958-59 1959-60 196&61 1961-62 1962-63 1963-64 196465 1965-66 196&67 1967-68 1968-69 1969-70 1970-7 I 1971-72 1972-73 1973-74 1974-75 1975-16 l97G77 1977-78 1978-79 1979-80 198g-81

363.38 228.48 464.06 463.62 698.33 769.65 760.15 706.23 884.55 993.62 1166.77 1075.25 787.10 715.23 755.77 758.78 668.90 827.60 999.71 950.07 769.37 906. IO 978.34 838.90 1236.46 1028.09 1034.79 1216.24 880.8 I 1095.95 539.90 686.43

4.93 3.05 6.10 6.06 9.36 10.00 9.8 I 9.44 Il.55 11.97 13.63 12.22 8.27 7.23 7.51 7.55 6.32 7.68 9.01 8.45 6.79 8.19 8.79 7.63 10.91 8.77 8.28 9.41 6.60 7.97 3.78 4.73

47.65 48.21 33.90 28.13 42.56 74.19 73.40 74.57 83.50 78.71 101.94 94.59 73.30 60.14 59.09 63.97 46.62 47.67 50.02 49.75 44.12 58.62 65.15 64.08 84.05 75.04 66.52 74.85 54.79 66.68 34.06 42.01

5.16 5.23 3.72 3.12 4.83 8.14 7.95 8.18 8.94 7.97 10.10 9.23 6.77 5.42 5.30 5.81 4.13 4.22 4.41 4.38 3.88 5.26 5.82 5.73 7.36 6.43 5.43 6.00 4.3 1 5.14 2.57 3.15

“Absolute variability is measured by standard variability by coefficient of variation. Source: Based on data from sources mentioned

deviation in table

and relative I.

agricultural technology rather than on the basis of observed breaks in the patterns of variability.) During Phase 1 standard deviations of both production and yield show a strong tendency to increase. During Phase 2 there is no tendency for them to increase. Overall there seems to be some tendency for absolute variability to rise. This might be due to some uncommon natural and political factors such as severe drought and flood and the War

210

M. Alauddin and C. Tisdell, impact of the ‘Green Revolution’

of Liberation in the early 1970s. In any case, one would anticipate a rise in standard deviation on account of the substantial rise in the absolute level of production. One can identify two similar phases in the behavior of relative variability of production and yield. However, there is a difference in that relative variability has declined to a lower average value. To facilitate quantitative comparisons, we estimated regression lines involving indices of relative variability of production and yield (ICI/P and ICI/Y) and time (T) for the three periods: Phase 1, Phase 2 and the entire period. These are set out in part A of table 3. During Phase 1 both yield and production variability show a strong tendency to increase. The strong explanatory power and high r-value lend clear support to this claim. However, there is dramatic change in the behavior of relative variability when one considers the equation relating to Phase 2 and the one that relates to the entire period. Even though the signs seem to indicate a declining trend over time one should note the poor quality of the estimates both in terms of R2 and t-values. In view of this one needs to be wary. There is no clear simple downward trend but a fortiori no upward trend is apparent. In terms of the above analysis of absolute and relative production and yield variability, there is little evidence that variability increased during Phase 2 compared to Phase 1. Indeed if one makes allowance for the unprecedented drought and flood of 1972 and 1974 respectively and political factors like the War of Liberation of 1971, Phase 2 seem to be associated with a somewhat greater production and yield stability than Phase 1. There is no evidence that the introduction of new agricultural technology has overall increased instability of national foodgrain production and yield in Bangladesh. Nevertheless so far we have divided the entire period into two phases only on the basis of mathematical considerations as depicted by the plots mentioned above. Even though the second phase (1961-80) can be broadly identified with the period of new technology and some elements of it (e.g., chemical fertilizers and modern irrigation) were introduced in the early 1960s the new technological package in Bangladesh did not assume any real significance until the later part of the 1960s when HYVs of rice and wheat were introduced. Wide scale adoption did not start until well into the 1970s. Thus, if one were to make intertemporal comparison of trends in variability in foodgrain production and yield in the pre- and post-‘Green Revolution’ periods, a priori considerations would suggest that any of the years 1967-68 through 1969-70 is likely to provide a more clearcut dividing line. The following discussion concentrates on an examination of behavior of production and yield variability of Bangladeshi foodgrains by shifting the dividing line between the two phases to the later part of the 1960s as may be more logical from technological considerations. As the ‘Green Revolution’ is a continuous process a dividing line based on a single year may not

M. Alauddin

and C. Tisdell, Impact of the ‘Green Revolution’

211

Table 3 Time trends in absolute and relative variability observed mathematical ground and technological 1949-50 to 1980-81.” A. Observed mathematical

of production considerations:

and yield on Bangladesh

ground

No.

Equation

RZ

r-value

Period

1. 2. 3. 4. 5. 6. 7 8. 9. 10. 11. 12.

ISDP=50.086+1.120T lSDP= 23.802+ 6.354T ISDP= 57.253 +0.717T ISDY=81.625+0.018T ISDY = 43.682 + 7.8747 ISDY=82.102-O.lSlT ICVP=91.539-0.293T ICVP=45.884+9.0667 lCVP=95.200-0.625T ICVY=112.375-1.064T ICVY=64.261+9.167T ICVY= 104.139-0.909T

0.2828 0.9161 0.0787 O.OQol 0.7726 0.0027 0.0116 0.8811 0.0487 0.0999 0.7240 0.0727

3.440 10.451 1.240 0.040 5.823 0.221 0.593 8.609 0.960 1.824 5.122 1.188

Entire Phase Phase Entire Phase Phase Entire Phase Phase Entire Phase Phase

R2

t-value

B. Technological

period 1 2 period 1 2 period 1 2 period 1 2

considerations Equation

Dependent variable

Year

Phase

1.020 1.040 0.980 0.680

1949-66 1949-67 1949-68 1949-69

1 1 1 1

0.2699 0.2461 0.2522 0.4289

2.110 1.900 1.840 2.600

1967-80 1968-80 1969-80 197G-80

2 2 2 2

0.440 -0.135 -0.560 -0.975

0.0045 0.0005 0.0092 0.0304

0.270 0.090 0.410 0.770

1949-66 1949-67 1949-68 1949-70

1 1 1 1

-1.192 - 1.862 - 2.866 -4.826

0.0531 0.1055 0.2009 0.4740

0.820 1.140 1.590 2.850

1967-80 1968-80 1969-80 1970-80

2 2 2 2

Intercept

Coeflkient

ICVP ICVP ICVP ICVP

78.548 79.205 80.329 82.746

1.364 1.248 1.061 0.679

0.0613 0.0602 0.0507 0.0236

ICI/P ICVP ICI/P ICVP

142.834 146.060 155.464 193.450

-

2.443 2.563 2.907 4.280

ICVY ICI/Y ICVY ICVY

102.020 105.280 107.826 110.455

ICI/Y ICVY ICVY ICI/Y

112.372 130.466 157.915 212.136

I =Index of variability standard deviation and production and yield.

_

with 197677= 100. SD and CV respectively represent coefficient of variation. P and Y respectively refer to

realistically depict the behavior of variability between the two phases. Therefore, a series of dividing lines starting with 1967-68 as the beginning of the second phase are considered. The least squares regression estimates of the trend lines for the different phases based on the new dividing lines are

212


presented in part B of table 3. For brevity we report those that relate to relative production and yield variability. A comparison of the behavior of relative production variability (coefficient of variation) between the two phases clearly indicates that during the first phase it shows a weak rising trend while a relatively stronger tendency to fall is observed during the second. The relative variability of yield shows a weak declining tendency in the first period and a relatively stronger tendency to fall in the second. The statistical quality of the estimates in both the cases seem to show general improvement with a forward shift in the dividing line. A comparison of the estimates set out in part B with those in part A of table 3, indicates that the rise in variability during the 1949-60 period is much stronger than in the case of the redefined first phases. By the same token, the fall seems to be stronger in the newly defined second phases compared to the 1961-82 period. Thus, irrespective of how one draws the dividing line between the two phases, that is, on observed mathematical grounds or on a priori considerations the period associated with the new technology can in no way be identified with a period of rising variability in foodgrain production and yield. If anything, with the forward shift in the time cut-off points both measures of variability show an increasingly stronger tendency to fall, that is, as the ‘Green Revolution’ becomes more firmly established. A number of factors, apart from those suggested elsewhere in this paper may have contributed to the declining relative variability of yield after the introduction of HYVs. In the beginning, experimental stations often test and release a wide range of varieties. Some of these prove to have higher variability under field conditions than is apparent under experimental conditions and are discontinued. Others may initially be applied outside the regions ecologically most suited for them. Thus general learning about the ecological suitability and appropriateness of introduced varieties to particular areas takes place over time. In addition, individual farmers become more familiar with the environmental and husbandry requirements of new varieties so they can improve their cultural practices. This is an individual learningby-experience phenomenon. Both of these experiental factors will tend to reduce yield variability with the efflux of time.

5. Interregional

and intertemporal

analysis of variability

In this section, production and yield variability of all foodgrains are examined for each district and then compared for time periods and between districts. Secondly, variability of foodgrain yield and production are examined within districts and compared for different seasons and varieties of crops and between districts.

213

M. Alauddin and C. Tisdell, Impact 01 the ‘Green Revolution’

5.1. Intertemporal comparison yield variability

of district level foodgrain

production

and

Table 4 sets out selected characteristics of foodgrain production and yield on regional basis prior to and following the ‘Green Revolution’ in Bangladesh vary considerably between districts. The districts of Chittagong, Noakhali, Mymensingh, Rangpur, Bogra and Dinajpur have experienced large production increases well in excess of the national increase, whereas there was virtually no increase in average foodgrain production for Faridpur and Khulna. A similar picture emerges in respect of average yield. As for production and yield variability, significant declines occurred in many districts and, in fact, variability rose in only three districts, namely Comilla, Noakhali and Sylhet. Sylhet only had a slight increase in yield variability. The probability of production falling five percent or more below trend fell in all but four of the 17 districts and the probability of annual yield falling five percent or more below trend declined in all but one district, Sylhet, where the increase was very small. Overall there have been substantial declines in the probability of yield falling five percent or more below the trend. On the whole, the regional data support the hypothesis that a reduction in overall yield variability has coincided with the introduction of new technology. 5.2. Behavior of variability

in the period of new technology

As can be seen from table 4, significant inter-district differences exist in the variability of foodgrain production ranging from 6.10 percent for Dhaka to 17.33 percent for Sylhet. Faridpur, Chittagong Hill Tracts, Dinajpur, Pabna, Comilla, Khulna and Kushtia have production variability towards the upper limit of this range. Turning to the question of overall yield variability between districts, one can see that it ranges between 3.88 percent in Mymensingh to 11.64 percent in Khulna. Chittagong, Dhaka and Comilla, Dinajpur, Jessore, Rangpur, Pabna and Sylhet have relative yield variabilities towards the lower bound of the range. Significant inter-regional differences in the variability of seasonal and varietal yields can be observed. Table 5 sets out average production and yield of foodgrains by season and variety for the 17 districts for the period 1969-70 to 1982-83. Also presented are the corresponding coefficients of variation. Significant inter-regional variations in overall yield are noticeable ranging from 849 kgs in Faridpur to 1,800 kgs in Chittagong. Seasonal yield differences also exist. Chittagong has the highest yields during the kharif season (1,580 kgs) as well as during the rabi season (2,672 kgs). Faridpur has the lowest yield during the kharif season (723 kgs) while Sylhet has the lowest yield during the rabi season (1,064 kgs). Faridpur, Khulna, Kustia, Pabna, Rangpur, Rajshahi, Mymensingh, Jessore and Dinajpur have lower rabi season yields than the remaining seven districts.

4

Dhaka Mymensingh Faridpur Chittagong Chittagong HT Noakhali Comilla Sylhet Rajshahi Dinajpur Bogra Rangpur Pabna Khulna Kushtia Jessore Barisal

508.0 1252.2 464.4 400.3 71.8 358.5 648.0 770.2 575.1 415.8 317.1 697.6 284.8 492.8 195.9 413.7 813.3

(000 m tons)

Period

1

724.3 2135.2 463.3 711.9 100.8 626.7 935.2 1093.0 776.8 656.5 509.3 1181.3 400.9 509.0 246.9 557.5 1021.0

Period

2

42.58 70.52 - 0.665 77.84 40.39 74.81 44.32 41.91 35.07 57.89 60.61 69.34 40.77 3.29 26.03 34.76 25.54

% Change

17.30 11.42 17.88 17.24 18.80 11.99 10.23 13.40 18.62 20.42 19.24 13.19 16.40 25.18 16.23 13.78 16.98

Period

1

6.10 6.13 13.88 6.60 15.08 12.24 13.09 17.33 10.16 13.60 8.36 8.23 14.29 15.07 13.89 8.65 10.01

Period

2

- 64.73 - 46.32 - 22.39 -61.70 - 19.80 2.04 27.92 29.33 - 45.46 - 33.38 - 56.52 - 37.61 - 12.84 -40.16 - 14.42 - 37.25 -41.05

% Change

38.63 33.07 38.97 38.59 39.51 33.83 31.25 35.46 39.44 40.33 39.74 35.24 38.02 42.11 37.91 35.95 38.44

Period

1

20.64 20.73 35.94 22.45 36.99 34.13 35.13 38.63 31.14 35.64 27.50 27.16 36.32 36.99 35.94 28.17 30.85

Period

2

- 46.57 - 37.32 - 7.78 -41.82 -6.378 0.89 12.42 8.94 -21.05 - 11.63 - 30.80 - 22.93 - 4.47 - 12.16 - 5.20 -21.64 - 19.75

% Change

in average quantities, coefficients of variation and probabilitity of a 5 percent or greater fall below the trend: Foodgrain production and yield, Bangladesh, 194748 to 1968-69 (Period 1) and 1969-70 to 198283 (Period 2).” ___ Probability of a 5% or greater Average values (000 m/ton Coefficient of variation (%) fall below trend (%)

A. Production

Districts

Changes

Table

Yield

(kgs/ha)

Source:

1026 1005 833 1141 1147 953 1023 1077 952 1008 1024 967 875 1084 884 987 1044

Based on data


9.

from sources

1306 1225 1013 1134 1060 1055 1111

1307 1289 842 1800 1415 1269 1362

mentioned

27.39 28.26 1.08 57.76 23.36 33.16 33.14 21.73 19.64 19.64 27.54 26.68 15.77 4.61 19.91 6.89 6.42 in table

1.

10.92 8.26 13.33 13.41 12.99 11.75 9.09 7.61 12.71 13.39 14.84 10.34 14.74 16.88 14.82 11.65 13.70

5.20 3.88 9.15 4.78 10.31 10.32 7.86 7.78 7.99 4.81 6.58 7.18 8.09 11.64 10.09 5.50 9.81

- 52.34 - 53.03 - 31.37 -64.37 - 20.57 - 12.16 - 13.58 2.19 -37.14 -64.09 -55.64 - 30.53 -45.09 -31.05 -31.88 - 52.82 -28.37

32.71 27.26 35.39 35.46 35.05 33.54 29.12 25.56 34.72 35.46 36.80 31.39 36.73 38.36 36.80 33.40 35.76 16.82 9.87 29.20 14.75 31.38 31.42 26.24 26.01 26.56 14.92 22.39 24.32 26.84 33.36 31.03 18.14 30.50

-48.58 -63.79 - 17.49 - 58.40 - 10.47 -6.32 -9.89 1.76 - 23.50 - 57.92 -39.16 - 22.52 - 26.93 - 13.03 - 15.68 -45.69 - 14.71

216


Table 5 Foodgrain

production

and yield: Mean variety, Bangladesh

Mean production District

Rabi

A. Production Dhaka Mymensingh Faridpur Chittagong Chittagong HT Noakhali Comilla Sylhet Rajshahi Dinajpur Bogra Rangpur Pabna Khulna Kushtia Jessore Barisal

values and coefficients of variation districts, 1969-70 to 1982-83.

level

Coefficient

by season

of variation

and

(%)

Kharif

Local

HYV

Rabi

Kharif

Local

HYV

247.5 606.4 101.5 219.9 21.4 125.4 266.5 397.0 137.0 56.9 75.3 113.3 91.0 39.7 59.9 58.5 140.7

476.8 1528.8 362.3 492.0 79.4 501.3 668.7 695.9 639.8 599.6 434.0 1068.0 309.9 469.2 187.7 499.0 880.3

433.2 1337.6 375.4 272.4 45.9 358.1 522.7 849.7 634.5 518.6 346.9 918.6 313.1 424.2 155.6 437.1 774.1

291.1 797.5 88.3 439.6 54.9 268.1 412.4 243.3 142.3 137.9 162.4 262.6 87.7 84.7 92.0 120.4 246.9

12.97 20.94 33.01 9.50 23.36 23.84 14.71 25.54 17.81 52.20 28.29 52.34 37.14 21.41 24.54 35.21 29.57

8.01 7.00 16.31 10.20 16.00 15.54 16.63 15.48 11.79 13.66 7.42 5.67 18.55 16.86 17.95 11.80 14.40

9.03 7.72 15.69 13.00 28.76 13.74 17.81 22.11 13.37 17.03 10.55 9.48 17.09 18.34 20.24 12.15 17.22

9.52 10.72 37.26 10.53 13.84 19.25 15.50 33.87 20.59 41.41 22.72 36.48 31.59 20.07 16.63 13.37 23.53

2339 1988 1913 2672 2234 2545 2178 1604 1867 1890 2154 1922 1818 1816 1777 2018 2516

1064 1132 732 1580 1287 1137 1187 1186 1049 1169 1223 1178 892 1098 938 1002 1019

978 1014 731 1285 948 985 1049 1160 1020 1085 1100 1085 874 1039 825 920 947

2805 2615 3008 2564 2367 2421 2387 2499 2609 2263 2391 2436 2674 2411 2370 2593 2539

8.55 10.82 9.57 10.48 12.80 12.02 7.16 12.78 7.93 8.10 11.37 11.55 8.80 10.13 22.62 10.85 8.27

6.39 7.24 9.97 8.23 11.58 14.07 11.12 8.01 8.58 5.30 7.03 6.03 9.87 13.21 12.90 6.99 13.46

6.34 5.13 9.44 8.56 12.87 12.39 9.91 9.83 9.51 6.64 6.64 7.56 9.27 13.28 14.18 6.96 15.04l

10.66 9.68 12.47 12.13 17.62 13.59 12.82 13.53 12.96 8.62 18.61 12.52 12.12 20.45 15.99 Il.26 11.93

(000 m tons)

B. Yield (k&ha) Dhaka Mymensingh Faridpur Chittagong Chittagong HT Noakhali Comilla Sylhet Rajshahi Dinajpur Bogra Rangpur Pabna Khulna Kushtia Jessore Barisal Source:

Based on data from sources

-mentioned

in table 1.

Table 5 indicates that the aggregate regional production of HYV foodgrains is relatively more variable than that of the local varieties with the exception of the districts of Chittagong, Comilla, Kushtia, and Chittagong Hill Tracts where HYV production is relatively less variable than local variety foodgrain production. Rabi foodgrain production is relatively more variable than khariffor all districts except Chittagong and Comilla where the

M. Alauddin and C. Tisdeil, Impact

of the ‘Green Revolution’

217

opposite is the case. When compared with the relevant column in table 4, the overall annual relative variability of foodgrain production and yield is seen to be generally smaller than that of foodgrains in different seasons (rabi and kharif) and of different varieties (local and HYV). Similar comparative variations can be observed for varietal and seasonal production variabilities. In general Dhaka, Mymensingh, Chittagong and Rajshahi show relatively lower values of variability. HYVs show highest variability in Dinajpur, closely folowed by Faridpur, Rangpur, Pabna and Sylhet and lowest Dhaka followed by Chittagong, Mymensingh, Jessore and Comilla. Table 5 also sets out the relative yield variability of foodgrains for different varieties and seasons. In the majority of districts, foodgrain yields during the kharif season are less variable than those during the rabi season. However, for seven districts viz., Noakhali, Khulna, Faridpur, Comilla, Rajshahi, Pabna and Barisal rabi season yields are less variable than those during the khar$ season. Local variety yields are relatively less variable than those of the HYVs for all districts except Barisal. Overall yields are less variable than the component yields (cf. table 3). Again this may be due to aggregation and lack of perfect correlation between components. But there are a few exceptions. Comilla, Rajshahi and Barisal have lower rabi yield variability than the overall one. In Rangpur, kharifseason yields are less variable than the overall yield. What the preceding discussion seems to indicaie is that on a seasonal basis HYVs show greater relative variability than local varieties but not on an annual basis. This may be due to averaging and other factors. Thus the fact that HYV yields are more stable than traditional varieties on an annual basis but are more variable within seasons could be due either (a) HYVs are more stable than traditional varieties in one season but not in the other or, (b) HYVs have more negative correlations between seasons than do traditional varieties. An important contribution of the ‘Green Revolution’ seems to have been to moderate annual relative production and yield variability. However, it does not appear to have reduced seasonal or varietal variabilities either at all or to some extent.6 To what extent is the observed behavior of foodgrain production and yield variability between districts in Bangladesh related to the introduction of new technology by districts in Bangladesh? To what extent do cross-sectional data suggest that there is a relationship between the adoption of new technology and variability of food production and yield? Let us consider this 6This seems to be generally true of relative production and yield variability Coefftcients of variation of production and yield of kharifand rabi HYVs considered for districts in Bangladesh indicated greater relative production and yield variability to all HYVs taken together. That is, on an annual basis HYVs show lower relative than on a seasonal basis.

of HYVs. separately compared variability

218


matter bearing in mind that correlation between factors does not necessarily imply a causal relationship between them. Indicators of adoption of new technology by districts are set out in table 6 and some indicators of the 1969-70 scenario are stated. Table 6 indicates that cropping intensity (ZNTN), use of fertilizer and irrigation (FRTHEC and PRFAI) tend to go up as the percentage of foodgrain area under HYV (PRHYHF) increases. These aspects come into sharper focus when one plots various indicators of variability against the surrogates for adoption of the new technology and intensity of cropping. Fig. 2 plots coefficients of variation of production and yield (UPROD and CVYZELD) against percent area under HYV foodgrains. In order to make quantitative comparisons of variability across districts we estimated four regression lines with percent area under HYV foodgrains as the explanatory variable and each of the other two as a dependent variable. These are illustrated by fig. 2. Even though the signs seem to indicate a decline in variability with increase in the percent area under HYV foodgrains, one should note the poor quality of the estimates both in terms of explanatory power and statistical significance of the coefficients. In view of this one needs to be wary. While the evidence does not indicate a strong tendency for variability to decline with percent area under HYV, it may rule out any upward tendency in variability with the percentage area sown to HYVs. Fig. 3 depicts the behavior of the same measures of. variability across districts in relation to the intensity of cropping. The relationship of relative variability of foodgrain production and yield to cropping intensity is similar to the one for relative production and yield variability and percent area under HYV foodgrains. The estimated regression lines are illustrated in fig. 3. The estimated relationships are poor both in respect of R2 and t-values. One, therefore, should not place too much significance on these regression lines. Despite this it could be said with a fair degree of confidence that neither overall production nor overall yield variability shows any tendency to increase with increase in the intensity of cropping. If anything there is a tendency for these variabilities to decline with increases in the incidence of multiple cropping. The above evidence suggests that the new agricultural technology has not caused increased variability of foodgrain production and yield on a regional and annual basis. If anything the new agricultural technology has had an ameliorating influence on such variability.

6.Observed behavior of variability: An analysis of underlying factors In our analysis we find that districts with a greater percentage of foodgrain area under HYV and irrigation and using a greater amount of chemical fertilizer per hectare and making more intensive use of land tend to show

219


Table 6 Selected

indicators

A. Indicators

of technoloeical _I

District _____ Dhaka Mymensingh Faridpur Chittagong Chittagong HT Noakhali Comilla Sylhet Rajshahi Dinajpur Bogra Pabna Rangpur Khulna Kushtia Jessore Barisal B. 1969-70

of spread 1969-70

of new technology 1969-70 to 1982-83 scenario: Bangladesh districts.”

and

of the

chanee

Cropping intensity (%)

% Food area (irrigated)

% HYV area in food

Fertilizer (kg/hec)

146.0 169.3 157.4 148.7 150.7 152.3 161.3 136.4 127.8 141.3 157.3 152.8 180.7 124.2 132.5 135.9 139.8

16.4 16.4 4.4 21.6 13.4 10.4 14.2 26.3 12.8 5.9 9.1 6.5 4.6 5.4 18.3 5.7 7.7

19.6 19.4 6.0 45.6 36.6 25.5 27.0 12.5 8.8 11.7 18.1 8.8 11.6 8.6 17.4 9.6 11.0

34.3 19.5 6.6 59.0 13.7 20.2 41.4 10.0 16.3 17.3 34.3 16.5 10.9 8.7 29.3 14.9 11.3

scenario

District

YLD6970 (kgs/ha)

IRRI6970


1143 1152 891 1673 1482 1193 1313 1422 1170 1093 1055 1141 985 1118 817 974 1176

8.1 13.7 2.3 16.3 14.5 3.6 9.9 22.5 15.3 5.4 5.0 2.6 3.0 3.6 6.8 2.7 2.7

(%)

FERT 6970 (kgs/ha) 15.8 8.2 1.9 51.4 11.7 11.6 17.8 6.0 7.0 9.5 17.3 5.3 5.4 6.6 10.9 6.0 5.8

“Figures in part A are based on the 1969-70 to 1982-83 averages, YLD6970, IRRI6970, FERT6970 respectively refer to foodgrain yield, percent of foodgrain area irrigated, and amount of fertilizer (nutrients) per hectare of gross cropped area in 1969-70. Source: Based on data from sources mentioned in Table 1 and EPBS (1969, pp. 4@41), BBS (1979, pp. 162, 212, 166167) BBS (1982, pp. 206, 209, 213), BBS (1984a, pp. 31, 33).

I.D.&

E


01

I

1

5

10

PERCENTAGE

15 OF

WV

Fig. 2. Top - Coefficient of variation area, Bangladesh districts, 1969-70 foodgrain production vs. percentage 1982-83.

20 FOODGRAIN

25

30

35

40

1

1

45

50

It

AREA

of overall foodgrain yield vs. percentage of HYV foodgrain to 1982-83. Bottom - Coeffkient of variation of overall of HYV foodgrain area, Bangladesh districts, 1969-70 to

221


I

I

T

I

I

I

190

200

*

t 120 INTENSITY

130

140 OF

150

160

170

160

CROPPING

Fig. 3. Top - Coefiicient of variation of overall foodgrain yield vs. intensity of cropping, Bangladesh districts, 1969-70 to 1982-83. Bottom - Coeffkient of variation of overall foodgrain production vs. intensity of cropping, Bangladesh districts, 1969-70 to 1982-83.

222


lower relative variability of foodgrain production and yield. Thus the new technology seems to have a moderating impact on overall foodgrain production and yield variability. The introduction of biological and related innovations made it possible to replace traditional rice cultivation by a whole range of new technologies. This has important implications in terms of higher productivity per unit area as well as greater choice and flexibility. Cultivated land that was once left fallow for a significant part of each year (for example, during the dry season) is now used for crops such as wheat and dry season rice. New technologies have enabled the incidence of multiple cropping to increase. An increase in the incidence of multiple cropping can reduce the overall (annual) variability of production and a fortiori reduce the coefficient of variation of production. In Bangladesh, the introduction of new technology has permitted the use of land for a number of crops of foodgrains over the whole year, e.g., its use for kharif (summer and wet season) foodgrains (aus and aman rice) and rubi (dry season) foodgrains (bore rice and wheat). Furthermore, within one season, land can be allocated between local and HYVs of the same foodgrain crop, e.g., local and HYVs of uus rice. When technological innovations reduce constraints on the use of a non-renewable resource like land and allow the cultivation of two or more crops instead of only one crop during a year on the same plot of land or different varieties of the same crop during the same crop season, the chance of a total crop failure considered annually can be reduced. The situation is akin to diversification of portfolios as a hedge against uncertainty [cf. Markowitz (1959)]. However, we do not wish to give the impression that it is the increase in incidence of multiple cropping alone that has reduced relative variability of yields, nor that irrigation alone is the only factor involved in this. The combined package of new technology has contributed to it and the use of many elements in the package are highly correlated. Thus, in the previous section, variables such as the index of multiple cropping, extent of irrigation, might best be regarded as proxies for the introduction of a whole bundle of new technologies. To illustrate the correlation issue: the incidence of multiple cropping in Bangladesh is, for instance, closely associated with the availability of irrigation, and expansion of irrigation enables greater human control to be exerted over the growing conditions of crops. Consequently both elements may add to stability. It is also conceivable that the more ready availability of supplementary inputs such as fertilizers and pesticides with the expansion in the market for these due to the ‘Green Revolution’ has made it easier for growers to stabilize their production. Variations in the use of such inputs can be more finely tuned to changing environmental conditions. However, we are aware that fine-tuning does not always lead to greater stability in agricultural production. Also, as pointed out earlier, experiential factors may make for a decline in relative variability of yields following the

223


introduction of HYVs, namely in the rejection of risky varieties after early use, the appropriate locational-use of varieties by trial-and-error and in the development of and learning about appropriate cultural practices for the varieties adopted. Much more research is required to apportion the role of each of these factors in reducing yield instability. At this stage, it is pertinent to raise some critical questions about our analysis. To what extent has the spread of HYVs themselves been influenced by the conditions preceding the ‘Green Revolution’? To what extent has land potentially suitable for HYV cultivation influenced the adoption and diffusion of new technology? Is it possible that lower relative variability and higher yields and the introduction of modern technology are all positively correlated? To what extent have the yields of initial years induced the adoption of HYVs? Let us address these questions. We do not have detailed data for all these variables prior to the ‘Green Revolution’. But given the fact that HYVs were just beginning to be introduced in the late 1960s it would not be too unrealistic to assume that the information relating to 1969-70 is reasonably indicative of the technological scenario obtaining just prior to the introduction of the new technology. Table 6 presents data on overall foodgrain yield (YLD6970), percentage area irrigated in foodgrains (IRR16970) and quantity of chemical fertilizers per hectare of gross cropped area (FERT6970) for the year 1969-70. A comparison of district data for the three technological proxies for 196070 (i.e., YLD6970, IRR16970 and FERT6970) with those of the entire period (cf. part A and part B of table 6) indicates that the districts show much the same rankings in the later years as in 1969-70. To see the impact of the 1969-70 values of these variables on the spread of HYVs in later years, we have estimated three regression lines with percent area under HYV foodgrains as the dependent variable and each of the 1969-70 technology proxies as an independent variable. These are presented as eqs. (4), (5) and (6). PRHYVF=

-26.4170+0.0377YLD6970

(R’=O.5557,

t =4.331),

PRHYVF=

11.0559+0.79601RR16970

(R’=O.2056,

t= 1.970), (5)

(R2 =0.6755,

t= 5.588). (6)

PRHY VF= 8.2411+0.7957FERT6970

(4)

The estimates of eqs. (4) and (6) can be considered reasonably good tits from a statistical point of view. The quality of the estimates clearly indicates considerable impact of the 1969-70 foodgrain yield and fertilizer application on the adoption of HYVs in later years. Eq. (5) implies that percent of foodgrain area irrigated does not have a high explanatory power. But the coefficient is significant at the five percent level and has a positive sign. The high absolute values of the coefficients in eqs. (5) and (6) indicate stronger

224


impact on HYV adoption of irrigation and fertilizer application in the earlier years. Thus the conditions of the initial years are on the whole associated with the spread of the HYVs in subsequent years. To examine the relationship between average foodgrain yield for the entire period (MEANYLD) and relative yield variability (CVYIELD) we estimated eq. (7). The overall average yield has only a weak explanatory power and the coefficient is far from being highly significant. The negative sign, however, indicates that higher average yields tend to be associated with lower relative variability. CVYIELD=

(R’=0.1226,

12.4767-0.0039MEANYLD

t= 1.448).

(7)

From the above analysis, it is still possible that districts with higher initial yields had initially lower relative variability. If this were so, it might give rise to circular causation [Myrdal (1968)]. But our evidence7 indicates that there is little systematic relationship between initial yield and initial variability of yield and production in districts and the rate of adoption of new technology. However, there is a slight positive relationship. Possibly in districts with inherently low relative variabilities adoption of HYVs seemed less risky to farmers. This and other factors such as the extent of the land area potentially suitable for HYV cultivation and increased usage of complementary inputs, e.g., chemical fertilizers and irrigation provided a hospitable environment for a more rapid adoption of HYVs in those districts relative to others. Furthermore, farmers in districts with higher yields at the beginning of the ‘Green Revolution’ were likely to have been familiar with better crop care and modern farm management techniques (‘learning-by-doing’ factors) than farmers in districts with lower yields. Apart from technological factors, socioeconomic factors like the distribution of land might have also affected the diffusion of HYVs. Recent studies [e.g., Alauddin and Mujeri (1986), Hossain 71t is interesting to see if production and yield variability during the earlier period (CVPI and CP’YZ) have a relationship with those of the subsequent years (CVP2 and CVY2) or on the extent of adoption of HYVs (PRHYVQ. To examine this, we estimated eqs. (A) to (D). Eqs. (A) and (B) which show the relationship between variabilities during the pre- and post-‘Green Revolution’ periods do not indicate any strong impact of the variability of the earlier period on those of the latter. Eqs. (C) and (D) estimate the relationship between the production and yield variability of the pre-‘Green Revolution’ period and the percent area under HYV foodgrains. They do not show any statistically signiticant relationship between the two sets of variables. Thus eqs. (A) to (D) do not provide strong support for the hypothesis that a positive association exists between the variability of the earlier period on those of the latter or between the extent of HYV adoption. But a fortiori there is no evidence to suggest that they are negatively related. Cl’P2=8.1356+0.1955CW~ CVY2=2.6477+0.4086CVYI PRHYVF=26.0626-0.5220CVPI PRHYVF=23.9697-0.5212CVYI

(R*=0.0428, (R*=0.1959, (R’ = 0.0326, (R’=0.0146,

t=0.819), t=1.912), t = 0.471), t=0.471).

(A) (W (C) (D)


(1980)] suggest that land concentration of modern technology. More equitable than in others seems to have favored foodgrains.

225

can be an impediment to the spread distribution of land in some districts the adoption of modern varieties of

7. Concluding remarks Recent studies [e.g., Hazel1 (1982), Mehra (1981)] suggest that the ‘Green Revolution’ has a destabilizing impact on production and yield of foodgrains in India. As pointed out earlier in this paper, their findings might have been influenced by the assumption of arbitrary cutoff points and lack of consistency with regard to deletion of observations for ‘unusual’ years within a particular phase. Examination of both national and district-level data from Bangladesh indicates that the ‘Green Revolution’ may (in contrast to Hazell’s findings for India) have reduced relative variability of foodgrain production and yield. Districts with higher adoption rates of HYVs and associated techniques seem to have lower relative variability. Furthermore, the probability of production and yield falling a certain percent below the trend seem to be lower for high HYV adoption districts than those with lower adoption rates. For Bangladesh as a whole, intertemporal analysis indicates falling relative variability of foodgrain production and yield, and this is also true for most districts in Bangladesh. It we examine inter-district or interregional data, we find a tendency for relative variability of foodgrain yield and production to fall with greater proportionate use of HYV and with the magnitudes of the proxies for adoption of modern technology. This may indicate that modern technology (‘Green Revolution’) has had a moderating impact on the relative variability of yields. We find that those regions with inherent lower relative variability of yields have not been more ready to adopt new technologies and so we can rule out this possible source of circular causation. Our evidence indicates that the ‘Green Revolution’ has not in practice been a source of increased relative variability of foodgrain yield and production. Bangladeshi experience indicates that on the contrary the ‘Green Revolution’ may have reduced such variability. In this respect the findings for Bangladesh differ from Hazell’s (1982) for India. One should, however, avoid generalizing either the Indian or the Bangladeshi experience. Rice is the predominant foodgrain in Bangladesh. In contrast, India’s foodgrain production includes other important crops.8 Rice “During 1969-70 to 1982-83, rice constituted nearly 97 percent of total foodgrain in Bangladesh. In India on the other hand the corresponding figure for rice during 1977-78 was 44.8 percent [Hazel1 (1982, p. 13)]. Other important food crops in India same period were wheat (25.6 percent), Jowar (10 percent), Maize (6.4 percent) and percent).

production 1967-68 to during the Bajra (5.4

226


seems to be a low risk crop and its coefficient of variation has declined in many countries since the 1960s [Hazel1 (19831. Rice is grown throughout the year and in all parts of Bangladesh. Failure of rice in one season can be cushioned. Whereas other crops e.g., wheat, barley, millets are grown only in one season and there is little scope to cushion the effects of higher variability in one season by yields in another. This partly explains why wheat shows greater variability compared to rice in Bangladesh. Furthermore, regions in India are much more diverse in terms of climatic and geophysical characteristics than in Bangladesh. More analysis and consideration of such issues is called for both to isolate the factors which have caused Bangladeshi experience to differ from the Indian and to identify the fundamental influences on variability which to date appear to have been mostly considered in a mechanistic fashion. We need, however, to qualify our findings in two respects. First, our analysis is based on official data whose reliability and accuracy is not beyond question [see, for example, Boyce (1985), Pray (1980); see also Alauddin and Tisdell (1987)]. But as there are no other comprehensive sources of information, we had to use oflicial data keeping their limitations in mind. Secondly, the statistical quality of some of our estimates while adequately casting doubts on earlier views or thesis about trends in foodgrain variability, are not suflicient to establish the opposite views or thesis. Nevertheless (taking account of the evidence, statistical or otherwise), our results suggest for Bangladesh that the ‘Green Revolution’ has reduced, and is continuing to reduce, as it proceeds, the relative variability of foodgrain production and yield.

References Alauddin, M. and M.K. Mujeri, 1986, Growth and change in the crop sector of Bangladesh: A disaggregated analysis, Journal of Contemporary Asia 16, 5473. Alauddin, M. and CA. Tisdell, 1986, Decomposition methods, agricultural productivity growth and technological change: A critique supported by Bangladesh data, Oxford Bulletin of Economics and Statistics 48, 353-372. Alauddin, M. and C.A. Tisdell, 1987, Trends, projections and sustainability of Bangladeshi food production: An alternative viewpoint, Food Policy 12. Alauddin, M. and C.A. Tisdell, 1988, New agricultural technology, sustainable food production: Bangladesh’s achievements, predicaments-and prospects, in: CIA. Tisdell and P. %laitra, eds., Technological change, development and the environment: Socio-economic perspectives (Routledge, London). BBS, 1976, Agricultural production levels of Bangladesh 1947-72 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1978, Statistical pocket book opf Bangladesh, 1978 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1979, Statistical year book of Bangladesh, 1979 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1980, Year book of agricultural statistics of Bangladesh, 1979-80 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1982, Statistical year book of Bangladesh, 1982 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1984a, Monthly statistical bulletin of Bangladesh, March 1984 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1984b, Monthly economic situation of Bangladesh, August 1984 (Bangladesh Bureau of Statistics, Dhaka).

M. Alauddin and C. Tisdell, Impact

of the ‘Green Revolution’

221

BBS, 1984c, Statistical year book of Bangladesh, 1983-84 (Bangladesh Bureau of Statistics, Dhaka). BBS, 1985, 1983-84 year book of agricultural statistics of Bangladesh (Bangladesh Bureau of Statistics, Dhaka). Boyce, J.K., 1985, Agricultural growth in Bangladesh, 1949-50 to 198e-81: A review of evidence, Economic and Political Weekly 20, A31-A43. BRRI, 1911, Workshop on ten years of modern rice and wheat cultivation in Bangladesh (Bangladesh Rice Research Institute, Joydevpur). EPBS, 1969, Statistical digest of East Pakistan, no. 6 (East Pakistan Bureau of Statistics, Dacca). Evenson, R.E., J.C. O’Toole, R.W. Herdt, W.R. Coffman and H.E. Kaufman, 1919, Risk and uncertainty as factors of crop improvement research, in: J.A. Roumasset, J.M. Boussard and I. Singh, eds., Risk, uncertainty and agricultural development (Southeast Asian Regional Center for Graduate Study and Research in Agriculture and the Agricultural Development Council, Laguna, Philippines) 249-264. Finlay, K.W. and G.N. Wilkinson, 1963, The analysis of adaptation in a plant breeding programme, Australian Journal of Agricultural Research 14,142-154. Hazell, P.B.R., 1982, Instability in Indian foodgrain production, Research Report no. 30 (International Food Policy Research Institute, Washington, DC). Hazell, P.B.R., 1984, Sources of increased instability in Indian and U.S. cereal production, American Journal of Agricultural Economics 66, 302-311. Hazell, P.B.R., 1985, Sources of increased variability in world cereal production since the 196Os, Journal of Agricultural Economics 36, 145-159. Hossain, M., 1980, Foodgrain production in Bangladesh: Performance, potential and constraints, Bangladesh Development Studies 8, 39-10. Jain, H.K., M. Dagg and T.A. Taylor, 1986, Yield variability and the transition of the new technology, in: P.B.R. Hazell, ed., Summary proceedings of a workshop on cereal yield variability (International Food Policy Research Institute, Washington, DC) 11-85. Johnston, B.F. and J.W. Mellor, 1961, The role of agriculture in economic development, American Economic Review 51, 566593. Joshi, P.K. and H. Kaneda, 1982, Variability in yield in foodgrain production, Economic and Political Weekly 11, A2-A8. Mahmud, W., 1919, Foodgrains demand elasticities of rural households in Bangladesh: An analysis of pooled cross-section data, Bangladesh Development Studies 1, 59-10. Markowitz. H.M.. 1959, Portfolio selection - efficient diversification of investments. Cowles Foundation Monograph no 16 (Wiley, New York). Mehra, S., 1981, Instability in the Indian agriculture in the context of the new technology, Research report no. 25 (International Food Policy Research Institute, Washington, DC). Mellor, J.W., 1918, Food price policy and income distribution in low income countries, Economic Development and Cultural Change 21, l-26. Minhas, B.S. and A. Vaidyanathan, 1965, Growth of crop output in India, 1951-54 to 1958-61: An analysis by component elements, Journal of the Indian Society of Agricultural Statistics 11, 23G-252. Myrdal, G., 1968, Asian drama (Twentieth Century Fund, New York). Parthasarathy, G., 1984, Growth rates and fluctuations of agricultural production: A districtwise analysis in Andhra Pradesh, Economic and Political Weekly 19, Al4A84. Pray, C.E., 1980, An assessment of the accuracy of the official agricultural statistics of Bangladesh, Bangladesh Development Studies 8, l-38. Rao, C.H.H., 1915, Technological change and distribution of gains in Indian agriculture (Macmillan, Delhi). Ray, S.K., 1983, An empirical investigation into the nature and causes for growth and instability in Indian agriculture: 195c-1980, Indian Journal of Agricultural Economics 38, 459-414. Rudra, A., 1910, The rate of growth of the Indian economy, in: E.A.G. Robinson and M. Kidron, eds., Economic development in South Asia (Macmillan, London) 35-53. Sawant, S.D., 1983, Investigation of the hypothesis of deceleration in Indian agriculture, Indian Journal of Agricultural Economics 38, 415496. Sen, S.R., 1961, Growth and instability in Indian agriculture, Journal of the Indian Society of Agricultural Statistics 19, l-30. Tisdell, C.A., 1983, The optimal choice of a variety of a species for variable environmental conditions, Journal of Agricultural Economics 34, 115-185. World Bank, 1982, Bangladesh: Foodgrain self-sufficiency and crop diversitication (Annexes and statistical appendix), Report no. 3953-BD (World Bank, Washington, DC).