Social Statistics, Counterfactuals and the Green Revolution

Features

  • Author: Chris Smaje
  • Date: 25 Oct 2013
  • Copyright: Image appears courtesy of iStock Photo

The Green Revolution has been a spectacular success. Without it, billions would have died from hunger and precious wilderness would have fallen to the chainsaw or the plough. It has also been a spectacular failure. Without it, the billion people in the world who still today go hungry may have been better fed, and the scourges of agro-chemical pollution, pest resistance, yield decline and overuse of water would have been averted.

thumbnail image: Social Statistics, Counterfactuals and the Green Revolution

Each of these radically different judgments has received a wide airing from academics and journalists, and each of them involves that most difficult of analytical beasts - the historical counterfactual. Since it seems impossible to know what would have happened if what did happen hadn’t happened, it’s surely plausible to claim that there’s no place for counterfactuals in serious social science – and certainly in serious social statistics. The very basis of statistics is correlation. Therefore counterfactuals, generating no data, resist statistical analysis. That point can be generalised at a more philosophical level: a counterfactual is implicitly a theory of causation, and as the statistician Terry Speed famously pronounced “Considerations of causality should be treated as they have always been treated in statistics: preferably not at all” (1).

More recently, Judea Pearl has reconstructed a logic of causality in statistics which admits counterfactual analysis (2). Pearl ingeniously formalises how to address counterfactuals in algebraic terms on the basis of the kind of bounded system analysis common in engineering (‘what would [have] happen[ed] to X given Y in system Z?’). This may allow causal counterfactual inferences in policy analysis of the type ‘what would have happened if the government had raised income tax in the last budget?’, but possibly not ones defined more nebulously around concepts such as ‘the Green Revolution’. Nevertheless, discussions of the latter sort are common both in everyday political commentary and in academic treatments because resort to counterfactuals reflects something at the heart of social science, both its joy and its misery, namely that it cannot avoid addressing simultaneously both what is and what ought to be. In this article, I use the case of the Green Revolution to illustrate some of the implications.

The ‘Green Revolution’ refers to coordinated international efforts between the 1940s and the 1970s to develop new crop varieties – most importantly, wheat and rice – that produced higher unit area yields than traditional varieties when combined with appropriate inputs of fertiliser, pesticides and water. It’s often seen as a successful humanitarian effort to tackle hunger at a time of rapidly growing population and unprecedented pressure on agroecosystems. However, this line of argument has also been identified as what Raj Patel calls ‘a dominant narrative of the Green Revolution’, a story actively constructed for more questionable political ends (3). For critics like Patel, this dominant narrative serves to prepare the ground for approaches to global hunger which prefer to seek technocratic solutions for its consequences rather than political solutions to its causes, and in doing so divert attention from the underwhelming legacy of the original Green Revolution in the persistence of hunger, the loss of rural employment and accelerating ecological problems.

I have to admit that my sympathies are on Patel’s side of the debate, but the point of this article isn’t so much to argue a particular case as to look at the way these larger policy arguments are oriented by social statistics and data analysis more generally. We’ll encounter some familiar methodological problems in the social sciences along the way, but ones that are particularly acute when the debate turns on counterfactual analysis. Here, I divide them into four themes.

1. Problem definition: proxies and artefacts

To answer the question ‘Has the green revolution been a success?’ one obviously has to define both ‘the green revolution’ and the criteria of ‘success’. Regarding the former, this is where the system boundaries mentioned earlier in relation to Pearl’s work are important. Specific, positive questions (e.g. what effect did the introduction of ‘improved’ grain varieties have on surpluses, yields, social mobility etc. in a particular place and time) are easier to answer than generic, normative ones such as whether the Green Revolution was successful. This is also the case because criteria of ‘success’ are debatable. As Green Revolution critic John Perkins puts it,

    "If success means an increase in the aggregate physical supply of grain, the green revolution was a success. If success     means an end to hunger, then the green revolution was a failure. People without access to adequate land or income,     regardless of their country of residence, remain ill fed" (4).

One problem here is that social scientists often lack direct data on the issue in question, and use proxy data instead. For example, Gordon Conway writes “there has been a significant reduction in poverty and hunger in those countries directly affected by the Green Revolution, apparently because of cheaper food” (5). A heavy methodological weight hangs on the ‘apparently because of’ in that sentence, which I cannot examine in detail here. But just to touch on a couple of issues, Conway refers to FAO data showing declining prices for four key food staples from the 1950s to the 2000s; however, this tells us nothing directly about actual levels of absolute or even relative poverty and hunger over this time period. Conway does concede that in some parts of the world the absolute number of undernourished people has increased as a result of population increase, and the relative number of undernourished hasn’t decreased by much either: in this analysis, only East and Southeast Asia emerge strongly as beneficiaries of the Green Revolution, subject to the limitations of the proxy data.

...a big picture analysis claiming that declining global food prices are benefitting the poor in low income countries may be an artefact of implicitly biased data construction. Only careful statistical under-labouring in the belly of the data can reveal a more complex reality.

Conway also cites an interesting analysis by SOAS economist Andrew Dorward (6) which suggests that the food prices borne by poor people in Green Revolution countries may be higher than global data suggest, because these data are built on consumer price index assumptions from wealthy countries. More locally relevant comparators suggest a much shallower price decline. In other words, a big picture analysis claiming that declining global food prices are benefitting the poor in low income countries may be an artefact of implicitly biased data construction. Only careful statistical under-labouring in the belly of the data can reveal a more complex reality.

2. Cherry-picking

Choosing the data that best supports the argument one wishes to make is the hoariest of subterfuges in social statistics, and perhaps is inevitable to some extent inasmuch as data and its analysis are rarely free of the interpretive intentions of the investigator. Critics of the Green Revolution allege cherry-picking in the analyses of its proponents in the mediums of both space and time as well as context: space in relation to using exemplarily successful places such as the Punjab’s Sangrur district to represent the success of the Green Revolution overall, time in relation to limiting the analysis to the initial yield increase, with the introduction of the new varieties without addressing the slowly diminishing returns over a longer timeframe as pest problems multiplied; and context in relation to emphasising the Green Revolution as primarily a biotechnological breakthrough, with its socioeconomic elements of subsidies and farm extension underplayed.

Regarding the first point, Raj Patel writes ‘To allow the Sangrur district to represent Punjab, and Punjab to stand as a metonymic case study of the Green Revolution in Asia, is problematic, and not merely because the idea of the case study is flawed....Sangrur may not have been representative of Punjab, and Punjab certainly wasn’t representative of India’ (7). Of course, statisticians have gone to great lengths to eliminate such ‘metonymy’ in their quantitative understanding of sampling bias, but this generally depends on the notion of a random sample taken from a unified population – rarely the case in the world of public policy case studies.

3. Comparators

Data analysis is almost always comparative – rigorously so in the case of experimental designs employing a control group, but usually only implicitly in the case of historical counterfactuals. As already mentioned, a difficulty of counterfactuals is that we can scarcely know what would have happened had history run a different course.

Let me generalise the issue thus. Historical situation A is transformed by intervention B. A common counterfactual strategy is then to project the consequences of A forward into the present in the absence of B (an ‘A→A scenario’), but of course it’s possible that in the absence of B any number of other alternative interventions B’ could have occurred (‘A→B’ scenarios’). Often, these alternatives are not politically neutral – and this is clear in the case of the Green Revolution, whose architects promoted it explicitly as an alternative to the ‘Red Revolution’ they feared if the rural poor went hungry. Peasant political revolution is one B’ scenario that may have occurred in the absence of the Green Revolution, but others are possible, including the widespread adoption of poverty relief, land reform and pro-peasant policies that might have tackled hunger without the biotech boost of the new Green Revolution crop varieties. For this reason, the counterfactual claim that the Green Revolution saved billions of lives seems open to question. It’s a claim based on A→A thinking, which implicitly rules out A→B’ possibilities.

The fact that analysts can’t agree whether the Green Revolution saved billions of lives or was a diversion from the real issues of hunger relief may not be the greatest advert for the acuity of the social sciences. But, whether or not one accepts Patel’s argument that a ‘dominant narrative’ of the Green Revolution has been explicitly constructed for political ends, at least the lack of agreement provides further evidence that there is never an objective or ‘ideology-free’ point of judgment in the social sciences.

4. Implicit theory: the problem of developmentalism

As previously mentioned, it’s impossible to provide a theory-neutral statistical analysis of the Green Revolution (or of anything much else) to determine its relative success. But it’s a good idea to be as clear as possible about one’s underlying theoretical suppositions, which can often be unstated or even unconscious. There’s an implicit ‘developmentalism’ in much of the scholarship on rural poverty. Beyond the fairly uncontroversial assumption that poor farmers would prefer to be richer, this assumes that poor farmers would prefer not to farm at all but instead pursue more modern and ‘developed’ livelihoods. Such developmentalism draws from the persisting legacy of social evolutionary thinking from nineteenth century social science, emphasising human ‘progress’ – from the supposed backwardness of hunter-gatherer societies through various stages of agricultural intensification to contemporary industrial society. It’s implicit in influential studies of agricultural change, such as Ester Boserup’s The Conditions of Agricultural Growth (8), and in descriptions of peasant villages as resembling ant hills ‘built by creatures motivated largely by inherited animal instincts and devoid of any inclination to depart from a fixed hereditary pattern’ described in the Green Revolution literature (9).

This thinking finds its way into popular accounts of the Green Revolution based on anecdotal reports of improved farmer welfare and generalised evidence of increased access to the cash economy (10). But whether this constitutes adequate evidence to make counterfactual judgments concerning the Green Revolution’s success is debatable. One B’ alternative is to reconstruct the development paths that poor small-scale farmers might choose for themselves, given the opportunity, in the place of top-down packages like the Green Revolution that were chosen for them. The anthropologist Paul Richards has described such processes of ‘indigenous agricultural revolution’ on the basis of mostly qualitative evidence for certain West African societies (11). How to infer from such work any overall judgment concerning the success of the Green Revolution and alternative possible development paths is, of course, equally challenging.

Conclusion

Historical counterfactuals are methodologically troublesome, but they exert a continued appeal in the social and policy sciences because we can rarely do controlled experiments yet need to make policy decisions based on the best available evidence. The fact that analysts can’t agree whether the Green Revolution saved billions of lives or was a diversion from the real issues of hunger relief may not be the greatest advert for the acuity of the social sciences. But, whether or not one accepts Patel’s argument that a ‘dominant narrative’ of the Green Revolution has been explicitly constructed for political ends, at least the lack of agreement provides further evidence that there is never an objective or ‘ideology-free’ point of judgment in the social sciences.

The lesson to be drawn from the debate from the viewpoint of the social statistician is perhaps the old adage that what really matters is not so much the destination as the journey to it. Policy analysts can debate the rights and wrongs of the Green Revolution. In doing so, they need access to good social statistics, and hopefully what I’ve illustrated here is the importance of attending to some basic rules of thumb in providing them: define your question clearly and carefully, be clear about your comparators, assess the weaknesses of proxy indicators, attend to artefactual issues in your data construction, avoid cherry-picking results, consider whether your analysis is conditioned by an implicit social theory, and if so try to make it explicit.

References

(1) Speed, T. (1990). ‘Complexity calibration and causality in influence diagrams’ in Oliver, R. and Smith, J. (Eds.) Influence Diagrams, Belief Nets and Decision Analysis, New York: John Wiley, p.58.

(2) Pearl, J. (2009). Causality: Models, Reasoning and Inference, Cambridge: Cambridge University Press.

(3) Patel, R. (2013). ‘The long Green Revolution’, Journal of Peasant Studies, 40, 1: 1-63.

(4) Perkins, J. (1997). Geopolitics and the Green Revolution, Oxford: Oxford University Press.

(5) Conway, G. (2012). One Billion Hungry: Can We Feed The World? Ithaca: Cornell University Press.

(6) Dorward, A. (2011). ‘Getting real about food prices’ http://www.soas.ac.uk/cdpr/publications/dv/file66348.pdf

(7) Patel, op cit

(8) Boserup, E. (1965). The Conditions of Agricultural Growth, London: George Allen & Unwin.

(9) Perkins, J. (1990). ‘The Rockefeller Foundation and the Green Revolution’, Agriculture and Human Values, 7, 3/4: 6-18.

(10) Pearce, F. (2010). Peoplequake, London: Transworld.

(11) Richards, P. (1985). Indigenous Agricultural Revolution, London: Hutchinson.

Related Topics

Related Publications

Related Content

Site Footer

Address:

This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on StatisticsViews.com are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and StatisticsViews.com express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.