Today in AgBioView Special (March 5, 2003):
'Science' Paper on Bt Cotton Performance in India - Further Criticisms
plus Response by the Author
Here is the final posting of the protracted debate on the performance of
Bt cotton in India. You may recall that earlier on Feb 21, 2003 I had
posted comments by Shantharam on the paper 'Yield Effects of Genetically
Modified Crops in Developing Countries' by Matin Qaim and David Zilberman
published in Science 299, 900- 902 (2003) along with response by the
author Matin Qaim. This elicited much feedback from readers (Ammann,
Kameshwar Rao, Jain and others) which I posted on Feb 25.
Now, in this final round of discussion on this topic, I post below
criticism of the paper by Suman Sahai and Abjhjit Sen with a response from
Qaim, followed by a round of arguments from both. Also posted below is
Qaim's response to Devinder Sharma from Bio-scope.org site and Dr.
Kameshwar Rao's comments on Sahai and Sen critique.
This discussion thread would close now.
- C. S. Prakash
Sahai and Sen Comments
Comments on the Paper 'YIELD EFFECTS OF GENETICALLY MODIFIED CROPS IN
DEVELOPING COUNTRIES' By Matin Qaim And David Zilberman [Science 299, 900-
- Dr. Suman Sahai, Gene Campaign, New Delhi, India, firstname.lastname@example.org; and
Dr. Abhijit Sen, Jawaharlal Nehru University, New Delhi, India;
The paper by Qaim & Zilberman showing a dramatic increase in the
performance of Bt cotton in India [ Science 299, 900- 902 (2003) ] is far
too premature and has several methodological problems. The authors draw
rather sweeping conclusions about the "Yield Effects of Genetically
Modified Crops in Developing Countries" based on meagre cotton trial data
derived from a single location, Maharashtra, in just one country, India.
The title is misleading, claiming far more than the scope of the study
The data presented are not derived from commercial plantings that were
authorised in 2002 but only from selected field trials. They appear to be
based on only the first picking. Cotton in this region is harvested in
four to five pickings. Protection of the Bt gene tapers off with age and
the final harvest would show this effect. In this case yield has been
recorded selectively from the most protected phase, so any conclusions
based on these data are untenable. In any case these data are decisively
contradicted by the real results coming in from farmers fields at the end
of the cotton season. Bt cotton performance is not clearly superior and
farmers in Maharashtra are reporting mixed results ( V. Jawandhia,
Shetkari Sangathana Pyne, Maharashtra, personal communication ).
The authenticity of the data in this paper is questionable since they are
derived from the field trial plots of Mahyco-Monsanto, the company that
owns the Bt cotton varieties under study. No data from farmers' fields or
from the All India Coordinated Varietal trials conducted by ICAR (Indian
Council of Agricultural Research )have been included in the study. It is
astonishing that the authors who never visited even the trial fields to
collect the data first hand, have the confidence to make such
Given the history of the misleading 'scientific' data put out ( e.g.
tobacco companies denying the ill effects of nicotine) , leading medical
journals have taken a decision to prevent the concerned corporate sector
from influencing the interpretation and reporting of research results (
see Krimsky & Rothenberg, 2001; Deangelis et al, 2001) . It would be
advisable for journals like Science to take a similar policy stand and
maintain a strict rigor in the editorial process so that biotechnology
companies promoting their GM crops can not misuse a peer reviewed forum.
The paper reports unprecedented yield increase, (upto 87%) which can
influence policy makers and farmers but strains credibility. Such
spectacular performance has not been reported from anywhere else in the
world where Bt cotton is cultivated. Yield in plants is known to be
multifactorial and polygenic and there is no known evidence for a single
gene being largely responsible for yield. Attributing such large effects
on yield to just one copy of the Bt gene is inherently faulty.
The authors do not provide a credible cause and effect relationship
between the Bt gene and the phenomenal yield effect. Their explanation is
that Indian farmers do not have adequate access to chemical pesticides so
they have very poor yields and when the Bt gene prevents crop loss,
staggering yield jumps are achieved. This analysis is not based on fact.
The relatively minor cotton crop is the single largest user of chemical
pesticides in India, with expenditure exceeding US$ 83 million annually.
The extravagant claims that Bt cotton technology is more beneficial in
India, does not pass elementary statistical tests. The paper makes the
following claims: (a) That the yield of the Bt hybrid is significantly
higher than of non-Bt varieties (b) That the yield of non-Bt counterpart
is no different from those of popular varieties (c) That this implies
significant effectiveness of the Bt gene in controlling bollworm
infestation and through this to the increased yields (d) That the
resulting yield difference in India is much higher than found elsewhere:
50 to 80 per cent more with Bt than non-Bt, as compared to only 12 and 15
per cent found in similar situations in the US and China (e) That, since
crop losses from bollworm are much higher in tropical climates and
insecticide use is relatively low, much greater benefits can be expected
from Bt in India and other tropical countries in South and Southeast Asia
and in sub-Saharan Africa than elsewhere.
It should be noted that the results from field trials analysed in the
paper relate only to claims (a), (b) and (d). Claims (c) and (e) are
derived conclusions rather than hypotheses, which can be tested with the
data, presented. The really important claim made, however, is claim (e).
But the paper does not present any direct evidence that crop losses from
bollworm are inherently greater in tropical climates.
On the other hand, it is clear that claim (e) cannot be established
without establishing claim (d), and this requires much more than the
statistical validity of claim (a). But:
* The paper reports a mean Bt yield of 1501 kg/ha with standard deviation
857 kg/ha while both the non-Bt counterpart and popular checks have almost
identical mean yields averaging 818 kg/ha with standard deviation of
around 572 kg/ha. Thus, while the results do show that the mean Bt yield
is 80 per cent higher than of the controls, this is associated with
extremely large standard errors, for both Bt and non-Bt.
* Based on these same summary statistics, Bt yields would be lower than
non-Bt yields in about 30 per cent of cases and, more importantly,
although the coefficient of variation is slightly lower with Bt than
non-Bt, Bt yields do not appear to have significantly lower variability
than non-Bt yields. From the point of view of an hypothesis that claims
that yield variability is largely a result of bollworm infestation and
that the Bt-gene effectively controls this, these statistical results are
* In view of this large yield variability, an attempt is made in the paper
to control for other factors through production function analysis which
explicitly embeds a "damage control function" to assess the relative
efficacy of Bt and insecticides. Although the various factors considered
in this analysis still leave unexplained about 60 per cent of yield
variation, the results do establish claim (a). The basis for this is that
the coefficient of 2.199 estimated for the Bt dummy is found to be
significantly different from zero at the 95 per cent statistical
confidence level. This value of the coefficient implies that Bt yields are
46 per cent higher than non-Bt yields at current insecticide use levels
and that insecticide use would need to be tripled on non-Bt to achieve Bt
* However, since the production function leaves most of the yield
variability unexplained, the coefficient estimated on the Bt dummy has a
large standard error. In particular, the 90 per cent lower confidence
level for the coefficient works out to 0.738 and the 95 per cent lower
confidence level only 0.325. At the lower end of its 90 per cent
confidence interval, the coefficient on the Bt dummy implies Bt yields
only 5 per cent higher than non-Bt at current insecticide use levels, and
a requirement of only 24 per cent increase in insecticide use on non-Bt to
achieve Bt yields. At the lower end of its 95 per cent confidence
interval, implied Bt yields would in fact be lower than non-Bt even at
current levels of insecticide use.
* This means that although there is statistical support for claim (a),
claim (d) is not established at conventional levels of statistical
confidence. It is not possible statistically to reject the alternative
hypothesis that the Bt yield differential may actually be lower in the
limited population from which they have drawn their Indian sample than
elsewhere. Since claim (d) is not established statistically, the results
presented in the paper do not provide an acceptable basis for its claim
Even if the nature of the meagre data are not questioned, and the heroic
generalisations about developing countries are ignored, the results
presented in the paper do not demonstrate that Bt cotton technology has or
will prove more beneficial in India and other tropical regions of South
and Southeast Asia and sub-Saharan Africa, than it has proved elsewhere.
References: 1. KRIMSKY, S., & ROTHENBERG, L.S. (2001) Conflict of interest
policies in science and medical journals: editorial practices and author
disclosures, Science and Engineering Ethics, 7, 205-218. 2. DEANGELIS,
C.D., FONTANAROSA, P.B., & FLANAGIN, A. (2001) Reporting financial
conflicts of interest and relationships between investigators and research
sponsors, JAMA, 286,89-91.
First Response by Matin Qaim (Senior Author of the Science Paper)
>> The paper by Qaim & Zilberman ......far too premature and has several
>methodological problems ...
>> authors draw rather sweeping conclusions about the "Yield Effects of
>...." based on meagre .. cotton trial data derived from a single
>location, Maharashtra, in just one country, India. The title is
*** If Suman Sahai had read our paper more carefully, she would know that
we are reporting about 157 different locations in three states of India
(Maharashtra, Madhya Pradesh, and Tamil Nadu). The results from the
2001/02 trials are generally consistent with trials in earlier years.
Based on crop protection theory, we argue that it is likely that also
other countries in the tropics and sub-tropics will see higher yield
effects of pest-resistant crops than what examples of China and the US
would suggest. This is not a premature statement, and also not only pure
theory, because it is backed by empirical evidence from South Africa and
>> The data presented are not derived from commercial plantings that ....
>They appear to be based on only the first picking. ......
*** We have waited until farmers had cleared their fields for collection
of yield data. For some farmers this took until March 2002. That is, all
pickings are included. The performance of the first commercial crop in
2002/03 has not yet been evaluated scientifically.
>> The authenticity ... is questionable since they are derived from the
>field ..plots of Mahyco-Monsanto..
*** Again, had Ms. Sahai read our paper more carefully she would have
known that the trials were conducted under close supervision by the Indian
Government authorities. Also she would have known that ICAR trials in
2001/02 even showed higher yield effects than the Mahyco-Monsanto trials.
In our paper, we make reference to the respective Annual Report of the All
India Coordinated Cotton Improvement Project (references 15 and 29). Also,
it is quite surprising that Ms. Sahai knows so many apparently wrong
details about our research. I have visited India twice during this
research and have personally carried out numerous interviews with small
cotton producers and trial participants.
>> Yield in plants is known to be multifactorial and polygenic and there
>is no known evidence for a single gene being largely responsible for
*** Ms. Sahai is correct when she says that the genetic yield potential is
a polygenic trait. But we are not at all talking about an improvement of
this genetic yield potential. The yield effects that we report are
entirely due to avoided crop losses. Thus, Bt can narrow the wide gap
between yield potentials and actual yields. We clearly state this in our
paper. It is not the first time that resistant plants have shown such big
yield effects. But, given the experience with Bt cotton in other
countries, the Indian results were quite surprising, and this is the
reason why it was worth reporting about them.
>> The relatively minor cotton crop is the single largest user of chemical
>pesticides in India, with expenditure exceeding US$ 83 million annually.
*** We are not arguing that no pesticides are used in India. Otherwise,
there could not have been such big savings in pesticides that we also
mention in our paper (Table 1). We know very well that cotton is the
biggest pesticide consumer in India. But application rates in the US and
China are still much higher. Furthermore, especially the cheaper
pesticides in India are not very effective anymore, because of problems
with resistance. Our cause-effect relationships are based on econometric
estimation, which is explained in the supplementary online material of our
>> It should be noted that the results from field trials analysed in the
>paper relate only to claims (a), (b) and (d). Claims (c) and (e) are
>derived conclusions rather than hypotheses, which can be tested with..
*** Although there is a lot of variation also within regions, it is a fact
that pest-related crop losses are bigger in tropical climates. See, for
instance, Oerke et al. (our reference 20). By the way, we do not say that
benefits are bigger, because pesticide savings are also associated with
economic gains. What we say is that positive yield effects are bigger.
>> standard deviation of around 572 kg/ha. Thus, while the results doshow
>that the mean Bt yield is 80 per cent higher than of the >> controls,
>this is associated with extremely large standard errors, for both Bt and
***This is very typical for survey data, which do not come from controlled
experiments. Remember that farmers themselves managed the trials, and
decided about input amounts, according to their own knowledge and
tradition. We perform the necessary statistical tests in a very rigorous
way, and the yield differences are statistically significant even at a 1%
>>> lower variability than non-Bt yields...is largely a result of bollworm
infestation and that the Bt-gene effectively controls this, ... these
statistical results are not convincing.
*** These are mere statistical speculations. In more than 95% of all
observations were the yields of Bt hybrids higher than those of
>>> statistical confidence. ... the results presented in the paper do not
>>> provide an acceptable basis for its claim (e) either.
*** That around 40% of the variation is explained (R2=0.4) is a pretty
good result for cross-section survey data. This cannot be compared to
results from controlled experiments. We chose the damage-control
framework, because it incorporates biological relationships between crop
loss and pest control agents into the production function. For our
purpose, namely to show where exactly the yield effects are coming from,
this is the appropriate approach. Yet this approach requires non-linear
estimation procedures, which produce less robust results than OLS
estimates. Against this background, our results are very satisfactory.
If you just want to establish that there is a positive yield effect of Bt,
you would take a normal production function and include Bt as a dummy on
the right-hand side. We have done this, too, controlling for all other
inputs and farm and farmer characteristics, and result in a net yield
effect of 94%, when using the common Cobb-Douglas specification. The
coefficient for the Bt dummy is 0.937 with a t- value of 7.11. This
unquestionably proves that there is a statistically significant yield
Counter Response from Sahai and Sen:
Here are our responses to Mr. Qaim's comments.
1. The author defends his use of selective field trial data saying they
were done under 'close supervision 'of the Indian government authorities.
No amount of 'close' supervision makes field trials anything other than
planned trials in the field, results from which are almost never
replicated in commercial cultivation. These data are from field trials,
not commercial cultivation and therefore premature. Extrapolating this to
farmers' fields for the future is speculative.
2. Qaim & Zilberman's data are now being increasingly challenged by the
results of the first commercial crop of the Mahyco-Monsanto cotton that
have come in. The indifferent performance is there for all to see and has
been reported at length in the media . The Bt crop in Maharshtra and
Andhra Pradesh appears to have failed. The state government in Andhra has
admitted, "Farmers have not experienced very positive and encouraging
results". Farmers echo this.
3. Mr Qaim attempts to deflect the evidence of poor performance by saying
'scientific evaluation' of the commercial planting has not yet been done.
It is unlikely that 'scientific evaluation' will reveal the crop to have
performed any differently than what farmers harvests are showing.
Actually, to get any real idea of the performance of Bt cotton, it will be
necessary to see the results of at least two more years of commercial
Qaim & Zilberman have done a great disservice to science by jumping the
gun in this fashion and so have the editors of Science, by letting this
paper through. These sensational data have led to a spate of media reports
about the 'superlative' performance of Bt cotton. Such misleading reports
can end up influencing policy makers in a direction that could be
ultimately detrimental to farmers.
4. Mr. Qaim bases his projection for high yields (from meager Indian
data), on crop protection theory. This is pure speculation. Crop yields
are multifactorial and have differing dynamics in different countries,
based on a multitude of local factors, which need to be studied in each
location. They cannot simply be constructed from a crop protection theory
and extravagantly applied to all developing countries.
5. The paper argues that the yield effect (nearly doubling ) is "entirely"
due to crop losses avoided by the presence of a single copy of the Bt
gene. This is an untenable, unscientific argument. The author's reasoning
also flies in the face of evidence from the ground.
If the yields in these trial plots were almost double simply because crop
loss due to bollworm was avoided, we would have to assume that nearly half
the cotton crop in India is destroyed due to the bollworm, because of
ineffective pesticides. If this were indeed the case, it would be
difficult to explain yields from local non-Bt hybrids, that
average1800-2000 kg/ha (as compared to the 'superior' 1500 kg/ha
performance of the Bt cotton) . The Bt cotton yields are inferior to many
local varieties, which are cultivated using the same 'ineffective'
6. Showing that much higher variability is usual in field trials than in
controlled experiments, as the authors do, does not prove that the yield
differential is larger in India than elsewhere. Mr. Qaim does not refute
this statistical observation and is unable to substantiate the claim of
spectacular performance in India.
7. The curious Indian results presented by Qaim & Zilberman are far more
likely to be artifacts, a view strengthened by the absence of a convincing
explanation for this phenomenal performance.
- Suman Sahai and Abhijit Sen
Final Response from Matin Qaim:
The entomological literature confirms that average pest- related yield
losses in Indian cotton production are 50-60% (our references 19-20). This
is one of the reasons why average yields in India are only one-half of the
yields in the US, and one-third of those in China. To a large extent (but
not all), these yield losses are due to the American bollworm, to which Bt
technology provides a high degree of resistance. Therefore, the yield
effects of Bt cotton in our paper are well explained, and the results are
far from being curious.
This is not to argue that Bt increases the yield potential of cotton
hybrids, nor that the particular hybrids into which the technology is
currently incorporated are the best available for all locations. Also, we
do not argue that chemical pesticides could not achieve a similar effect.
Figure 2 of our paper shows very clearly that, if farmers were to use much
more pesticides (and some do so), the level of damage control could be the
same as with Bt. But is this environmentally desirable?
My statement that the commercial crop harvest of Bt cotton in India has
not been evaluated scientifically up till now relates to the fact that all
the statements that are presently discussed in the media refer to
anecdotal evidence from individual farmers in certain pockets. A
scientific evaluation needs a representative random sample over all or at
least many locations. We are currently implementing an independent farm
survey in three states, to narrow this information gap.
Sincerely, Matin Qaim. March 5, 2003
Response from Dr. Kameshwar Rao:
The comments Drs Suman Sahai and Abhijit Sen on the paper by Qaim and
Zilberman in Science, say much the same as the others, posted on
AgBioWorld on February 25, 2003, that the source and the kind of data used
do not justify the conclusions drawn. Notwithstanding, the commentators
agreed that 'the results do establish the claim that the yield of Bt
Hybrid is significantly higher than of non-Bt varieties'.
I refrain from adding on to the points Dr Qaim has answered, one way or
the other, but Drs Sahai and Sen's statement that 'protection of the Bt
gene tapers off with age and the final harvest would show this effect', is
not entirely true. I have discussed this issue with several, who are
experts on Bt cotton technology, from the lab to the field, and gather
that about three ppm level of the Bt toxin assures full protection against
In general, the toxin levels in the Indian Bt cotton varieties are higher
than in the American varieties. Only after 120 days in MECH 162, and after
140 days in MECH 12 and MECH 184, the Bt toxin titres come down to about
10 ppm. Nevertheless, this is more than adequate to offer the required
protection. Beyond 100 days, no new larvae develop on the cotton crop, as
the moths 'celibate'. However, the farmers usually give the season's
parting gift of one pesticide spray, about this time. In effect, there is
no risk to the crop, if the toxin levels fall after 100 days, even
In the absence of any other means, the only way to assess yield increase
due to protection against loss, is to compare with the yield of non-Bt and
Speculative Research - Another Opinion on Bt Cotton In India
- 'AgbioIndia' Sharma D v/s/ Zilberman D; Qaim M
Abstract: Bt cotton in India: Devinder Sharma replies to David Zilberman
and Matin Qaim The science publication "Yield Effects of Genetically
Modified Crops in Developing Countries" (1) by David Zilberman and Matin
Qaim provoked a highly interesting scientific debate about the potential
benefits and the potential failure of bt cotton in India. Devinder Sharma,
tending to be very critical to genetic engineering, published a comment
entitled "A Scientific Fairytale: Providing a Cover-Up to the Bt Cotton
Fiasco in India" (2) at the AgbioIndia website. Qaim and Zilberman
answered to Devinders' comment which again was commented by Devinder
Sharma who keep his opinion. The full text contains the "commented
comment" in a dialog form as well as a list of articles related to this
Mr Devinder Sharma, trained as a plant breeder, is now engaged in policy
research and analysis on agriculture, genetic engineering, international
trade and food security issues. Email: email@example.com
Matin Qaim and David Zilberman: Mr. Devinder Sharma has named our recent
article in Science (1) "A scientific fairytale on Bt cotton" (2). That Mr.
Sharma does not believe in peer-reviewed scientific findings is one thing.
But his essay reveals that he has not read our paper carefully, and we
would like to clarify some of his misinterpretations.
Indeed, our research builds on the large-scale field trials with Bt cotton
in India that were initiated by Mahyco and Monsanto in 2001 under close
supervision by the government authorities. The trial design with a Bt
hybrid, a non-Bt counterpart, and a conventional check on adjacent plots
was excellent to analyze the net effects of the technology in that
particular growing season. We have used the official field-trial records
about pest infestation levels, such as larval counts per plant. Yet, the
data about yields and input amounts that we use in our study were
collected based on our own questionnaire and personal interviews with
participating farmers. Our research was certainly not on behalf of Mahyco
and Monsanto, and it was entirely funded by the German Research Council
(DFG), the biggest public research funding agency in Germany. Thus, the
results are independent and free of any commercial interests.
Devinder Sharma: The authors have confirmed our worst fears -- the entire
research was speculative. No wonder they succeeded in painting a glorious
picture for Bt cotton -- claiming that the crop yields increase by 70 to
80 per cent -- and that too without conducting any experiments on actual
yields harvested. The authors say that the data about yield and inputs was
based on personal interviews with farmers. What the authors probably
didn't know is that personal interviews are NOT the way to determine crop
yields in agricultural research. Furthermore, the authors accept that they
used the official-trial record for pest infestation. The 'official-trial
record' in this case was provided by the company, Mahyco-Monsanto, since
the trials were conducted by them. How can that data be considered
reliable? Mahyco-Monsanto have a vested interest in promoting Bt cotton
and how can the results be therefore called 'independent and free of any
Matin Qaim and David Zilberman: The reason why we randomly selected 157
farms in three states, instead of including the total of 395 trial farms
in seven states was simply determined by our limited research budget to
collect comprehensive data. Summary statistics of all 395 farms, however,
reveal that our sample is indeed statistically representative, and we
mention this in our paper. Interesting to note is that positive yield
effects of Bt cotton in trials that were carried out in different public
experiment stations were even bigger than those that we report, a fact
that we also mention in the paper. The overall finding that Bt cotton
leads to significantly higher yields than conventional cotton under Indian
conditions is consistent with all the field trials carried out in previous
Devinder Sharma: This only goes on to show how scientific research was
manipulated throughout to show 'significant' higher yields. The authors
use the public sector research to justify the yield increases that they
found. It is however interesting to see how the public research
experiments yielded such 'positive' results. For record, let me bring out
the fraudulent manner in which the government committees were overseeing
the research trials. The two committees set up by the department of
biotechnology -- the Monitoring and Evaluation Committee (MEC) and the
Review Committee on Genetic Modification (RCGM) -- had not even objected
once to the fact that the crop was sown late in all the years of trials.
So much so, that in the year 2000-2001, the crop was sown as late as two
months and yet it recorded a spectacular increase of 50 per cent higher
yields. Such trials are not even scientifically valid, and should have
been outrightly rejected.
Matin Qaim and David Zilberman: Mr. Sharma points out correctly that yield
levels in field trials are often higher than in commercial agriculture.
But this holds true for both Bt and conventional cotton. Since we compare
Bt and non-Bt hybrids on adjacent field-trial plots, the relative yield
difference is not affected. A scientific evaluation of the commercial
plantings of Bt cotton in India in 2002 has not been completed so far, and
anecdotal evidence about individual farmers is not enough to make
conclusive statements. We are currently in the process of carrying out an
independent scientific assessment for the 2002-03 crop harvest.
Devinder Sharma: There is no need for yet another 'independent scientific
assessment' for the 2002-03 crop harvest. Please don't waste the resources
of German Research Council, if they are again funding the research. We
already know the result. With or without 'peer-review', we know the
authors will show significantly higher yield increases to justify the
flaws in the so called scientific analysis that was undertaken in 2001
field trials. The way the research analysis is being conducted, I agree
there is no need for anecdotes to explain the nuances. It itself is a
scientific anecdote !
But let me still tell you another 'anecdote' that may perhaps lay bare the
reason for such a research being conducted by foreign institutes. An
Indian film actress of yesteryears, Simi Garewal, had posed topless in the
English movie 'Sidhdhartha'. Her revealing pictures were published by an
Indian magazine - The Illustrated Weekly of India. Simi Garewal felt
offended and took the magazine to court. And I recall, the editor of the
magazine Khushwant Singh, a very well-known writer, had commented
something like this: "Simi Garewal doesn't mind the whites to see her nude
but feels offended if her own people were also to admire her beauty."
No wonder, the Indian government refuses to share the Bt cotton data with
the civil society and researchers in India but has no qualms about sharing
it with foreign researchers who have no idea about the ground realities.
Matin Qaim and David Zilberman: Mr. Sharma states that Bt technology would
not break the genetic yield barrier of cotton, but would only reduce
current crop losses. This is certainly right, because Bt technology is a
new pest control strategy. But pest-related losses in India are so high
that theaverage cotton yields in farmers' fields are among the lowest in
the world. For climatic reasons, pest pressure in India is higher than in
the US and China. This has been shown repeatedly in the entomological
literature. At the same time, the amounts of chemical pesticides used and
their effectiveness is lower in India. Crop losses can only be reduced
where they occur, and when farmers in India suddenly obtain 600 kg of
cotton per acre instead of 320 kg, then this is obviously an increase in
effective yields. The simple reason why significant yield effects have not
been shown previously in other countries is that Bt crops have hardly been
used by smallholder farmers in the tropics and sub-tropics.
Devinder Sharma: 'Yield increase' and' reducing crop losses' are two
different things. You cannot justify the wrong usage of the term 'yield
increase' by saying that since the yield increases are very high so 'this
is obviously an increase in effective yield'. Such quantum jumps in
production were earlier achieved by the use of pesticides, does it mean
that pesticides increase yields? The answer is a big No. So there is no
need to unnecessarily defend a wrong. A wrong will remain a wrong, and
scientists should have the courage to accept it.
If for climatic reasons, pest pressure in India is higher than in China
and US, why aren't the scientists trying a different approach than what is
being followed in the US/China? After all, India is in the tropics, and we
should have science look for different and more effective answers than to
translocate a methodology that is being followed in the temperate regions.
And let me tell you what probably is the most effective way to control
pests -- ban pesticides on cotton. This will allow the natural cycle to
reoperate where the beneficial insect species, which number 27, taking
care of the American bollworm. But scientists wouldn't look at that for
the simple reason that such an approach is devoid of any commercial
Matin Qaim and David Zilberman: Mr. Sharma's comment that Bt cotton might
not be representative of other crops is valid up to a certain point.
Different crop species are attacked by different pests, so that one should
not simply extrapolate the exact numerical findings. Yet, significant pest
damage also occurs in other crops, and genetic resistance mechanisms to
various pests are available at the research stage. The general
relationship that yield effects of pest control agents are higher the
higher the crop loss, is a theoretical fact, which holds for any crop
species. This is supported by our empirical evidence. In our Science
paper, we use the example of Bt cotton to explain these broader
theoretical linkages. Hence, the more general title is appropriate. We
maintain that the yield effects of genetically modified, pest-resistant
crops will be higher in the small farm sector of the tropics and
subtropics than what hitherto examples of temperate climatic zones would
Devinder Sharma: Not only 'valid upto a certain point', it is not at all
correct to assume that Bt cotton is representative of other crops. But
tell me, how can the example of Bt cotton be extrapolated to other crops?
There is no empirical evidence to say that Bt cotton is a representative
of other crops. If that be so, shouldn' we stop conducting similar
experiments on other GM crops?
'The general relationship that yield effects of pest control agents are
higher the higher the crop loss, is a theoretical fact,' which does not
hold true for all crop species. How long can we go on quoting 'theoretical
evidence' to promote Bt cotton? This 'theoretical evidence' also existed
prior to the country wide analysis that the authors claim to have
undertaken. What was the need for such a research if you already had the
'theoretical evidence'? Unless of course the entire effort was merely to
justify the 'theoretical evidence'.
Matin Qaim and David Zilberman: This is neither the first nor last time
that we discover that a new technology has varying impacts in different
locations. Students of technology have found long ago that gains from
technology, in this case a pest control technology, specific
circumstances. High levels of untreated pest damage provide a fertile
ground for yield increase, as we found in the case of Bt in India. A yield
increase of eighty percents is high but it happened before. Hybrid corn
and some green revolution technologies have had similar effects in some
places. We had a theoretical base to suspect that Bt will have higher
yield effects in India that say the U.S. or China. But even in the U; S.
there were cases with substantial yield effects. The same theory will
apply to other crops. If there is high crop losses because of pests that
are vulnerable to Bt alternative pest control; strategy are not effective
or not used then Bt can increase yield. We do not consider Bt a silver
bullet to solve all pest and food problems, but we have evidence that it
is an effective in many locations and should be part of an integrated pest
management strategies that assesses technologies according to their real
impacts (on productivity cost and the environment) rather that their
Devinder Sharma: When questioned about the validity of the tall
claims,scientists always take refuge saying that Bt cotton is no 'silver
bullet'and that it should be part of an 'integrated pest management'
package. But please tell us, where is the IPM package that should include
Bt cotton? If scientists had instead focused on IPM strategies (rather
than waste resources on defending Bt cotton), cotton farmers the world
over would have been a happy community. They would have been happily
engaged in sustainable farming practices. It is the scientific community
which first pushed the cotton farmers into a 'pesticides treadmill' and
are now forcing the farmers to face, in addition, a 'biological
To have a 'theoretical base' to justify the yield increases is a clever
ploy to hoodwink the farmers and policy makers to believe that the
conclusion of the paper is based on 'actual' jumps in yield. Furthermore,
the authors have used the 'Science' platform to seek credibility to their
theory. I am amazed that the editors of Science journal failed to see that
the paper was based on mere speculation. Amazed that the authors still
maintain that 'the same theory will apply to other crops'. No wonder, the
claims of genetic engineering in plants are based on a theoretical
assumption and therefore have rightly earned the dubious title:
Matin Qaim and David Zilberman: Other misconceptions of Mr. Sharma relate
to some agronomic details. It is not true that Bt technology changes the
water or fertilizer requirements of cotton hybrids. During the field
trials, water and nutrient supply for Bt and conventional plots did not
differ. Also incorrect is that farmers would use one kilogram of Bt seeds
per acre. Whether Bt or conventional seeds, private sector cotton hybrids
are provided in 450 gram packages, which is the amount which farmers use
for planting on acre of cotton. If seeds are provided by other sources,
farmers often use higher amounts, but this is related to lower seed
quality and germination capacity and has nothing to do with Bt technology.
Devinder Sharma: If you haven't looked at the water requirement of Bt
cotton vis a vis other cotton varieties, there is no reason to call it an
assumption. A former advisor to the Department of Biotechnology,
Government of India, had gone on record saying that Bt cotton requires
more water. ICAR research trials have also established that some Bt
varieties have failed to perform in areas of water stress whereas the
other hybrids have done relatively better.
Matin Qaim and David Zilberman: Finally, Mr. Sharma has mixed up our
institutional affiliations. Matin Qaim is with the University of Bonn in
Germany, while David Zilberman is with the University of California at
Berkeley. But this is excusable, given that he has read our paper only
Devinder Sharma: Does it make any difference? University of California,
Berkeley, is known to be a research center with the clear objective to
promote GM crops. We also know the desperate efforts Centre for
Development Research (ZEF), Bonn, is making to push golden rice and Bt
cotton. Both the institutes are therefore engaged in promoting a risky,
unhealthy and unsustainable technology and that too in the name of poor
and hungry. But still, I'll accept my mistake at mixing up the
institutional affiliation of the authors.
What worries me more is the way good science is being sacrificed at the
altar of commercial interests. Scientists have no one to blame but
 Matin Qaim and David Zilberman: Yield Effects of Genetically Modifed
Crops in Developing Countries. Science 299, 900-903 (2003).  Devinder
Sharma: A scientific fairytale on Bt cotton. AgBioIndia, February 14, 2003
>> A Scientific Fairy Tale - Devinder Sharma, India Together