“Thousands of Indian farmers are committing suicide after growing GM crops.” It is no minor claim. Genetically modified crops revolutionise agriculture – but are controversial. They will feed the world, reduce the need for pesticides and fertilisers, and add health-protecting nutrients to those who consume them, say some. They are an ecological disaster in the making, say others, and impoverish the Third World farmers who grow them. Often quoted is the example of suicides among Indian farmers who grow GM crops. Ian Plewis examines the data and the conclusion.
Arguments continue to rage about the safety and the costs of genetically engineered or genetically modified (GM) crops, and about their relevance to problems of world food production and the public health in developing countries. Those who promote them claim increased yields, fewer inputs, and economic and ecological benefits. The scientific consensus1 emphasises their advantages, but does not regard GM as a panacea for all the agricultural problems of the world. Anti-GM campaigners2 emphasise possible risks to health and to ecosystems, over-reliance on pesticides (so-called “Roundup ready” GM crops rely on intensive spraying with glyphosate, to which they are resistant), and the dominant position of multinational companies in the GM seed market – which, they say, badly affects small farmers in poorer countries. Previously self-reliant, such farmers now have to buy expensive seed each year; a crop failure can ruin them. Many of the arguments, for and against, rely on statistical data and are therefore outwardly convincing. But these data are very often not subject to critical assessment and analysis. Closer analysis can tell very different stories.
Genetically modified cotton is a boom crop in India. But are its financial risks driving farmers to kill themselves?
One example is the assertion that the introduction of GM cotton in India has led to a surge in farmer suicides. In 2008, HRH Prince Charles pointed to “the truly appalling and tragic rate of small farmer suicides in India, stemming in part from the failure of many GM crop varieties”3. His speech was reported with the headline in our standfirst. It is, sadly, true that thousands of Indian farmers commit suicide each year – but then there are many millions of farmers. And, overall, male and female suicide rates in India are not notably high – higher than in the UK but lower than in France, according to the World Health Organization. But if suicide rates have indeed risen among farmers who have begun growing GM crops it would be important evidence in the GM debate. Here, I try to throw some statistical light on trends in farmer suicide rates over the last 15 years.
For many, the introduction of GM cotton to India is a resounding success story. Indeed, the proportion of farmers growing GM varieties since the official launch in 2002 has risen spectacularly: over 90% of cotton growing land is now planted with one of over 800 varieties of what is usually known as Bt cotton. This is seed that has been genetically modified by adding genes from the soil bacterium Bacillus thuringiensis to provide resistance to cotton bollworm. Yields have increased. More sceptical commentators argue that the observed increase in yields is part of a longer-term trend brought about by better farming practices and irrigation, that Bt crops are prone to failure, and that the expense of the seeds, allied to the need to buy new seed every year, has contributed to rising farmer debt that eventually drives farmers to suicide. The Indian sceptics have gained ground recently and the introduction of GM food crops, notably Bt brinjal (aubergine or eggplant), has been postponed.
There is evidence4,5 of a positive effect on profit for farmers growing Bt cotton arising not only from higher yields but also from reduced pesticide costs. One might therefore suppose that this would reduce the endemic problem of debt and lower the risk of suicide. But it is also apparent that this profit-enhancing effect varies across the nine states (see Figure 1) responsible for nearly all India's cotton production.
It is not easy to estimate how many farmers there are, nor how many suicides
Are the data fit for purpose?
We clearly have to examine the data of suicides and of farmers. Are the data up to the job? The short answer is that they are not ideal. We cannot draw on data from a very large prospective study that follows individual cotton farmers before and after their adoption (or non-adoption) of Bt cotton. No such study exists. Instead, we must rely on data from official sources at the state and national levels over time, with all their strengths and weaknesses, in order to get a better picture of the trends in the rates of Indian farmer suicides, of any association of trends in suicides with trends in cotton yields, and whether there is any evidence of a discontinuity after Bt cotton was introduced.
The National Crime Records Bureau (NCRB) collects extensive data on suicides, broken down (since 1996) by state, occupation and gender, and these data have been widely used by researchers6. One of the occupations coded is “self-employed farming/agriculture” and these are assumed to be farmers or, as they are often referred to, cultivators (landless agricultural labourers are coded separately). It is, however, suspected that not all suicides appear in the NCRB counts because of social stigma and fear of prosecution. (Suicide is, in principle at least, a criminal offence in India.) In addition, occupational classifications are subject to error. This is a potentially important problem for women who farm land that is in a male relative's name as they might not be coded as farmers in those circumstances. Fortunately, there is a good source of cross-sectional data on suicides at the national level in the form of a 2012 survey published in the Lancet7, from which we can estimate that, for all occupations, male and female suicides are, respectively, 35% and 47% higher overall for the nine states in question than they are in the NCRB data. I applied these correction factors to the NCRB data and assumed that they also apply to farmers, and that they do not vary over time.
We do not know the number of farmers who have joint holdings, nor the number who have more than one farm
Previous analyses have, perforce, analysed numbers of farmer suicides rather than suicide rates6. This is potentially misleading if the denominator – the number of farmers at risk – is changing over time. It is not easy, however, to establish just how many farmers there are, state by state over time. There are two sources of data. The first is the Census of Population, conducted every 10 years, with the most recent in 2011. The Census asks for main job, but this can be unpaid work by females and children working on the family farm where decisions are made by men, and is therefore likely to overestimate the number of cultivators, especially female cultivators. The second source is the Agricultural Census, conducted every 5 years. This collects data on the number and size of agricultural holdings, broken down by state and gender. The data separate individual from joint holdings, but we do not know the mean number of farmers in joint holdings; joint holdings are common in Gujarat, Haryana and Rajasthan, but rare in the other six states. It is, at least in principle, possible for a farmer to have more than one holding in different villages. I considered a range of between two and four farmers per joint holding. In addition, I assume that those who cultivate very small holdings (less than 0.5 hectares) do so as a supplement to their main income, and they are therefore excluded from the denominator.
The estimates from the two sources for the number of male cultivators when the nine states of interest are combined are relatively close: between 41.3 and 43.1 million in 2011. There are, however, marked discrepancies at the state level, where the ratios of the two estimates of the numbers of cultivators vary from 0.6 in Punjab (many more cultivators estimated in the Census of Population) to 1.27 in Maharashtra (more cultivators estimated by the Agricultural Census). Moreover, for some states, the time trends are different. For example, in Andhra Pradesh, the number of cultivators is decreasing according to the Census of Population but increasing according to my calculations from the Agricultural Census. This points to the need to see whether conclusions about trends in farmer suicides are robust to assumptions about the numbers of farmers.
Are suicide rates for Indian farmers high?
Are suicide rates for Indian farmers high? For males, they are not when we compare them to the rates for non-farmers in 2011 – see Figure 2. The figure considers the nine states individually, and only in three of them are farm rates the greater. Nor are they high when we consider all nine states together (and these states account for 44% of the Indian rural population in 2011). If anything, farmer suicide rates, at about 29 per 100 000, are a little lower than for non-farmers (35 per 100 000). But there are some marked differences between states, as we can also see in Figure 2. Male farmers were much less likely than non-farmers to commit suicide in Rajasthan and Tamil Nadu, whereas the reverse is true for Maharashtra. The states are grouped by region (north to south) in Figure 2, and this brings out the fact that suicide is generally more prevalent in the south of India.
The position is, however, different for female farmers. There is a lot of noise at the state level but, for the nine states as a whole, the rate is about 32 suicides per 100 000 female farmers, compared with 19 per 100 000 for women who do not farm. So, for the cotton-farming states of India, male and female farmer suicide rates are similar, the female suicide rate for farmers is much higher than it is for non-farmers, and the suicide rate for men who do not farm is higher than the corresponding rate for women.
Are suicide rates for Indian farmers increasing?
If the introduction of Bt cotton has improved the livelihoods of Indian cotton farmers then we would predict a change in any time trend after 2002. But we would not expect this change to be the same across all cotton-growing states, because the proportion of farmers who grow cotton varies from 26% in Punjab and Gujarat to just 5% in Karnataka; the adoption of Bt cotton also varies across states, averaging from 33% in Gujarat to 57% in Andhra Pradesh since 2002; and, as we see below, the effects on yields are not uniform.
I modelled the trends for each state and for all the states combined between 1996 and 2011. (Technically, the method was to fit linear and non-linear terms for year to the logit of the suicide rate8.) Conclusions about trends rest on consistency of estimates across models according to the assumptions made about the numbers of farmers. Figure 3 shows the estimated suicide rates from the best-fitting models for male and female farmers for all the nine states. We see an increase for men up to 2005 and a decline thereafter. But for women we observe a consistent decline such that, from 2009 onwards, the male and female rates are essentially the same.
We find that there is no consistent evidence for a change in the estimated suicide rate for male farmers for the northern state of Rajasthan, but for Punjab the rate declines up to around 2004 but then increases (Figure 4), and for Haryana it increases steadily. In the central region the rate is declining in Gujarat, essentially constant in Madhya Pradesh, but in Maharashtra it increases up to 2005 and declines thereafter (Figure 4). And in the south, the rate is constant for Karnataka and Tamil Nadu but increases steadily in Andhra Pradesh. So the pattern varies considerably between the states.
Our conclusion must be that the data do not support the view that farmer suicides have increased following the introduction of Bt cotton. Indeed, taking all states together, there is evidence to support the hypothesis that the reverse is true: male suicide rates have actually declined since 2005, having been increasing before then. The picture at the state level is less clear-cut, especially the contrast between Maharashtra and Punjab. In one, farmer suicides have gone down, in the other they have gone up. Can we bring any more evidence to bear to understand this contrast better?
Since growing GM cotton farmer suicides have increased – but only in Punjab. In other states, farmer suicide rates have gone down since the introduction of GM crops
Both Punjab and Maharashtra have high proportions of farmers growing cotton (26% and 20%), and Bt adoption rates are the same (56%). However, when we model the effect of the introduction of Bt cotton on cotton yields, we find that yields have risen in Maharashtra but have fallen in Punjab8. We cannot, of course, say that this is a causal effect, but the results for these two states are in line with our hypothesis that there is an economic component in the explanation of suicides.
To that extent, then, the claim that farmer suicides have increased since growing GM can hold – but only in Punjab. And those who would base their opposition to GM upon it must accept also that the opposite holds in Maharashtra – that farmer suicide rates there have decreased in the GM era, as they have for India as a whole.
The Indian farmer suicide story has become received wisdom for some anti-GM campaigners. In fact, we find that the suicide rate for male Indian farmers is slightly lower than for non-farmers. And Indian suicide rates as a whole are not notably high in a world context. The pattern of changes in suicide rates over the last 15 years is consistent with a beneficial effect of Bt cotton, albeit not in every cotton-growing state. The widespread adoption of Bt cotton means that we will never have the opportunity to study this question in the depth that it merits, but it might still be possible to do so in a different context if permission to grow GM vegetable crops is granted by the Indian government. as it has been in Bangladesh.