Analyzing India’s OIympic Performance

With the curtains having come down on the 2008 Summer Olympics, it is worthwhile pondering over the reasons for our traditionally dismal performance in the world’s greatest sporting event. While it is true that our medal tally has gone up this time to three with the first ever Gold being won in an individual event (congratulations to Abhinav Bindra for that), the number is still far too small for comfort and discomfiting particularly when compared to other large countries across the world including, most notably, our northern neighbor China.

Anirudh Krishna and Eric Haglund published an article in EPW a few weeks ago analyzing this very question (a brief item on their findings is here). Previous investigators have shown that per capita GDP and population are key predictors of a nation’s medal tally. They point out that according to a model taking into account both these factors, India was expected to win 19 medals at Athens (2004) but actually managed only one. Besides, GDP has grown consistently over the last six decades since independence but our olympic victories have not (see here for a summary of our successes since 1900) – on the contrary, we almost regularly won a single Gold in pre-independence days in events post-1928, a success we have been unable to repeat afterwards. Other factors such as climate and whether the nation happens to be hosting the Olympics also have a significant association with the medal tally but these too cannot account for the extreme paucity in our case. Likewise is the case regarding our political system: single party and communist ruled nations do have a consistently higher medal tally (Johnson, 2004) but most democratic nations fare better than we do making it an unlikely explanation for our failure (read this and this post on the Becker-Posner blog for a discussion on the reasons for this).

The authors argue that it is not the actual population of a country but the effectively participating population that matters, i.e., the proportion of the public that is healthy, has access to education and most importantly, is well connected with the outside world. Good health is essential to compete in sports; schools provide the avenue where talent finds recognition and information via the mass media raises the aspiration levels of society or so goes the reasoning. As a proxy to measure the last of them, they estimate the number of radios per 1000 residents. They show that their model which incorporates these factors does a better job of explaining inter-country variation than models using population and GDP alone. They also give the example of Jamaica, a low-income country with 430 radios/1000 residents which is well above the world median figure of 258, winning many more medals than a simple analysis using these two factors would have predicted.

The trouble is that their predictions are limited to a single cross-sectional event, namely, the total medal tally in the 1996 Summer Olympics. While they do acknowledge this fact, how likely is it that their assertions would survive in a longitudinal case study? Connectivity and access to information have undoubtedly improved in India over the last six decades and so have education and health care to some extent. At the very least, our medal tally ought to have seen a modest improvement over the years but none is evident (the current tally of 3 is no doubt an all time high but unless maintained in the future, it would have to be considered an outlier). As for the example cited, it may well be that Jamaica has a higher density of radios than India but even then, is the overall number of connected people (which is what really matters to determine the effectively participating population) lower than in the Caribbean country? The same question needs to be asked about education and healthcare as well: even assuming that only a fraction of our roughly 300 million strong middle class enjoys access to these amenities, is that number not sufficient to match that of any small sized country across the world? It is therefore difficult to believe that our own inability to match the 11 medals that Jamaica won – nearly four times our own tally – is due to our lack of connectivity.

Finally, all of this would only explain our failure if it is the outcome of poor talent at the recruiting stage. The study presents no evidence of that. If our sportsmen/women match those of other countries in their performance before undertaking professional training but subsequently fail to excel, the system that prepares them for such events is the likely culprit. One of the reasons for the phenomenal success of communist countries has been attributed to their rigorous training methods. A previous study also found that in both labor-intensive and physical capital intensive sports, higher income countries performed better than their poorer counterparts though the difference was much more pronounced for the latter kind indicating that financial investment matters for excellence in every sport albeit in different forms – perhaps more for equipment in some and better training in others. K.P.S.Gill, not surprisingly, blames the government for failing to help build turf stadiums and says that this contributed to the decline of Indian hockey in recent years.

The authors may well be right that people in India do not aspire to take up sports in a big way. That however is not something that can be quantified by such broad social indicators. This is part of a larger criticism I have with this study: they approach this issue much like any other developmental question and seek to argue that economic advancement of society will also inevitably lead to Olympic success and vice versa. As pointed out previously, the inverse temporal relationship between the two renders that claim questionable in India’s case. Furthermore, it is eminently possible to have development even absent a focus on sports in which case, higher income and connectivity would still translate only to mediocrity of the many, not the excellence of the few. The exalted performance vital to success in such competitive events involves identifying the talented outliers and subjecting them to intensive training, neither of which are considerations in formulating policies for social development. Future studies would do well to take these factors into account.

Join the discussion

This site uses Akismet to reduce spam. Learn how your comment data is processed.

4 comments
  • It seems to me that you are mixing up what can be called "positive" analysis and "normative" analysis.

    A positive analysis would simply seek to explain the olympics medal tallies of countries in terms of some determining factors. To this end, typically, a model would be put forward and estimated via econometric analysis. I am guessing this is what Krishna & Haglund do since I don't have access to their paper. This is fine, even if I don't find the research question – what factors explain the olympics medal tally of a country? – particularly interesting.

    We move from the positive to the normative when start making judgments regarding the number of medals we as a country should win in the Olympics. When we say that our olympics medal tally is "too low" and hence, something is "wrong" which has to be "corrected", we are making judgments. Now, notwithstanding the fact that a sizable section of our middle and upper classes also feels this way, this is a subjective judgment with which I happen to disagree. I don't really see why we should care about doing well in the Olympics. I feel that there are more important things we as a society should care about. It seems to me you also feel the same way.

    Let's now take up the issue of what Krishna & Haglund do. Are they really saying that there is something "wrong" with our "low" medal tally which has therefore, to be "corrected"? Are they really saying that "economic advancement of society will lead inevitably to Olympic success and vice versa"? From what I can see, the answer seems to be no.

    What the authors seem to be doing is critiquing the hitherto available (positive) econometric models, saying that they do not do a good job of explaining variations in inter-country performance. In place, they propose what they think is a better econometric model. You can agree or disagree that their model is "better" than the earlier ones and I am sure the argument will continue but, as I said, I don't find this research question particularly interesting.

    However, what Krishna & Haglund seem to be doing is positive, not normative analysis and that is why I think your criticism of their work is wrong.

  • Anon,

    You are correct that they are offering a model they believe to be superior to the ones proposed earlier because it explains variation better. A part of their analysis involves predicting how many medals a nation would have been expected to win based on population alone (model I), population and GDP (model II) and their proposed model (model III) and comparing these figures to the number of medals actually won. Based on the average deviation of several countries, they argue that their model is a better fit than the previous ones. My contention is that as far as India is concerned, their model is not significantly better than the others in explaining (or rather, failing to explain) the deviation from predicted values. As per models I, II and III, India would have been expected to win 157, 16 and 14 medals respectively in the 1996 games; the number actually won was only 1. The authors agree that I and II cannot explain our performance. My question is whether III does any better (yes, 16 > 14 but even with that marginal improvement of two, we have a difference of 13 between expected and actual value). I am making no judgment of how many medals India ought to win ideally, only indicating what it would have been predicted to win as per the model.

  • Dilip,

    I’m not sure one can construct a model that accurately explains the performance of each of the participating 190 or so countries. Even if one can do so, the resulting model may be so complicated as to render it useless. The complication, in part, would result from the fact that each country has some peculiarities that it is particular to itself. A model incorporating all the peculiarities of all countries would lead to a model involving 100 or more variables (assuming each country contributes one variable denoting its own peculiarity). As you can see, a model involving so many variables is not only difficult -if not impossible – to estimate econometrically but also quite unwieldy.

    Any model of this type therefore has to involve some simplification to keep the analysis tractable. The cost of this simplification is that the model will typically not explain the performance of all countries accurately. So, sure, none of the models come close to predicting India’s actual performance in 1996 but that is not how one would judge the Krishna-Haglund model; a better judge would be whether it does better on the whole than the existing models and this seems to be the argument that Krishna & Haglund make in their paper.

    If one is interested in predicting India's actual performance in the Olympics, then one would do better to construct a model involving all the factors peculiar to India and estimating that model. Such a model may be useful for predicting India's performance but may not be useful for predicting inter-country performance. A model is like a map: its usefulness depends on the purpose. If you are travelling from Delhi to Chennai (say), then a detailed road map of Delhi or Chennai is useless. On the other hand, if you are going from Moti Bagh to Delhi University, then a highway map of India is not going to help you much either.

  • You are right of course that their model is a general one but you miss the point about the focus of their inquiry. Here are the questions they set forth in the first few paragraphs:

    “Compared to its share in the world’s population, India’s share of Olympic medals is abysmally low. In the 2004 Olympic Games, for example, India won only one medal. Turkey, which has less than one-tenth of India’s population, won 10 times as many medals, and Thailand, which has roughly 6 per cent of India’s population, won eight times as many medals. India’s one-sixth share in the world’s population translated into a 1/929 share in 2004 Olympic medals. While Australia won 2.46 medals per one-million population and Cuba won 2.39 medals per one-million population, India brought up the bottom of this international chart, winning a mere 0.0009 medals per one-million population. Nigeria, next lowest, had 18 times this number, winning 0.015 medals per one-million population. Why does the average Indian count for so little? [Emphasis added]

    What prevents the translation of India’s huge number of people into a proportionate – or even near-proportionate – number of Olympic medals? The gross domestic product certainly matters, as previous analyses have indicated [Bernard and Busse 2004], but something else also seems to be making a difference, given that Cuba, Ethiopia, Kazakhstan, Kenya and Uzbekistan – countries not known for having high average incomes – have won many more medals than India, despite having a far smaller national population. Why do 10 million Indians win less than one-hundredth of one Olympic medal, while 10 million Uzbeks won 4.7 Olympic medals? [Emphasis added]

    In this article, we explore the concept of effectively participating population, arguing that not everyone in a country has equal access to competitive sports – or for that matter, to other arenas, including the political and economic ones. Many are not effective participants on account of ignorance or disinterest, disability or deterrence.“

    They also look at opportunities and achievements in two villages of Karnataka and Rajasthan. There can therefore be no doubt that their focus is on India. The model therefore needs to be judged based on whether it does a reasonably good job of explaining India’s case. That begs the question: if you started off by proposing that India is actually in the mainstream because its effectively participating population is low even if its overall population is not but your model fails to back up that view, does it not point to its inadequacy? (for some inexplicable reason, they ignore an important paper by Daniel Johnson where a more complex model was proposed including climate, host country and political system; if these factors were all added, the difference between their model and the previous one would probably be even less).