Is it religion which makes Hindus defecate in the open more than Muslims?

A mosque and a temple side by side

Vivek Dehejia and Rupa Subramanya

A new salvo has been fired in the "culture wars" -- and this one, so to speak, hits bottom: the bottom of the latrine that is.

In a recent paper, Michael Geruso and Dean Spears claim to have "solved" an important "puzzle" in the development economics literature. While the data show that Indian Muslims fare worse than Hindus, on a wide range of socio-economic indicators, one area where they fare far better is in infant and child mortality.

The numbers show that, by the age of 5, mortality among Muslims is about 18% lower than it is among Hindus. This is a puzzle, because conventional economics suggests that a group faring worse on most socio-economic indicators would also fare worse on this one, whereas in fact they are doing better.

Geruso and Spears' novel solution to the puzzle is to point to the large gap, as measured in the data, in the propensity to engage in open defecation (OD) between Hindus and Muslims.

Their data, drawn from publicly available sources, show that Hindus are 40% more likely than Muslims to engage in OD.

The startling part of their paper is to claim that this single difference, taking into account both the "direct" impact (a household not using a toilet) and the "indirect" effect (having neighbours who similarly don’t), can fully explain the difference in mortality rates.

Key to their explanation is that Hindus and Muslims tend to cluster in their own communities -- sometimes a result of discrimination, at other times of self-selection -- so that a Muslim household using a toilet is more likely to be surrounded by other Muslims households who also use toilets, and vice versa for Hindus.

In fact, Geruso and Spears suggest it’s this indirect effect which is the more important of the two, creating what economists call a "negative externality" that might, perhaps, call forth a policy response (unless, of course, government intervention makes things worse rather than better -- but that's a debate for another time).

So far, so good.

Geruso and Spears' interesting work has been picked up by major publications including the New York Times and The Economist. 

So what caused it to hit the fan for us and a few others, coming out of what should have been an interesting but hardly explosive scientific paper?

The reason is that, as careful as their primary statistical and economic analysis is, they’re extremely careless in offering what can best be described as an Orientalizing (it's not just leftists who can draw on Said), armchair cultural/religious explanation for the difference in OD rates between Hindus and Muslims.

Section 2 of their paper begins portentously, quoting first from an Islamic sacred text and then from the Manusmriti, a Hindu text. 

The snippet from the Islamic text basically tells you not to defecate in the open. The snippet from Manu, however -- extracted out of context from a much larger manuscript -- tells you to defecate away from home -- presumably in the open. 

Coupled with these two quotes and a potted sociology of the caste system, pointing to notions of ritual impurity and the role of the "manual scavenger" class, with an inevitable little detour via a pearl of wisdom from Mahatma Gandhi thrown in, the authors appear to provide us with a simple (in reality, simplistic) cultural explanation for this difference. 

While they concede that it’s beyond the scope of their paper to evaluate these cultural claims, they’re dismayingly content to trot out these cultural "theories", which, to say the least, are contentious and, in some respects, downright problematic.

Nor -- and forgive the expression  -- do they pass a basic smell test. Most contemporary Hindus, when making a decision whether to defecate in the open or not, are not so far as we're aware whipping out their pocket Manu for advice. (Nor are Muslims, for that matter, whipping out their sacred texts.) Like most normal people, we just go when nature calls, whether it’s in a latrine at home or in the field.

Conceivably, Geruso and Spears could have made a more sophisticated and nuanced sociological or cultural argument in favour of the theory that Hindus, for some deep-seated reason, have a preference to defecate in the open, but they fail to do this.

It’s a shame, really, because the scientific part of their paper -- which tests statistically for a relationship between differences in defecation rates and mortality outcomes -- is pretty well done. 

Not surprisingly, of course, in full fledged Orientalist mode, the New York Times and The Economist unerringly homed in on this pseudo-cultural explanation, drawn it would appear more or less directly from the paper. without, it would seem, doing any additional homework.

The more egregiously crude of the two is Gardiner Harris in the Times, who, without flinching, can pen a paragraph like this:

"Open defecation has long been an issue in India. Some ancient Hindu texts advised people to relieve themselves far from home, a practice that Gandhi sought to curb."

The Economist is just a little less bald -- but, as you would expect, snarkier and more sanctimonious -- and puts it this way:

"Leaders need also to confront the cultural reasons for bad sanitation. Hindu tradition, seen for example in the 'Laws of Manu', a Hindu text some 2,000 years old, encourages defecation in the open, far from home, to avoid ritual impurity."

Ironically, in presenting data in support of their basic theory -- the relationship between OD and mortality -- they undercut the other bit of their theory, which relates to purported cultural differences between Hindus and Muslims. 

What’s relevant here is Figure 4 from the paper (reproduced below), which has two panels, and uses state level data across three different surveys, to capture the geographical variation in OD rates and infant and child mortality, respectively.

Each panel shows, on the vertical axis, the difference in the relevant mortality rates between Hindus and Muslims, and on the horizontal axis, the difference in open defecation rates. Each circle (the larger the circle, the bigger the state) represents a particular data point, and the authors insert a regression line -- a line of "best fit" --  with a confidence interval around it.

What the authors draw from this figure is that the line is upward sloping. In other words, where there’s a higher gap in open defecation rates, they’ll be a higher gap in mortality rates. They take this -- correctly -- as validation of their main hypothesis, which links open defecation to mortality. 

As they say themselves, oblivious, it seems, to how damaging it is to their cultural theory: “We exploit the fact that the size and even sign of differences in latrine use between Hindus and Muslims varies across the vast geography of India.” (Section 5.2, page 22) 

What Geruso and Spears entirely miss is that these figures -- and their own statement quoted above -- utterly undercut their first hypothesis, which claims there’s some sort of cultural reason why Hindus are more likely to defecate in the open than Muslims. 

If that theory were really correct, all of the points would be clustered in the upper right hand corner of the figure (the "nonnegative orthant", in mathspeak), that is, above the (0,0) point. 

In fact, the points aren't all clustered in the upper right, as the authors themselves tell us, and as you can see from the pictures. Further, they even tell us there are some states where the Hindu open defecation rate is lower than the Muslim! (Those are the points at the bottom left of the figures.)

So while this is all good for their big hypothesis relating OD and mortality, their armchair cultural theory is a crock of you know what.

In fact, if you look at these pictures without ideological blinkers and without the helpful regression lines to guide the eye, we challenge you to say decisively that you see a tight relationship. What we see is a cloud of points in all four quadrants and some really big outliers far away from the line of best fit.

What these figures also tell us is that there’s enormous regional variation both in the difference in open defecation rates and in mortality rates. Indeed, as the authors themselves note, again apparently without realising its double significance, the regression line passes through the (0,0) point. This is good for the OD/mortality hypothesis but sucks big time for the cultural theory.

Why?

If you look closely there are a whole bunch of circles around those two (0,0) points. What that means is that there are a bunch of states in different time periods in which there’s a negligible or no difference in the rates of open defecation between communities and correspondingly a negligible or no difference in mortality rates. How does Manu help explain this?

What's more, if you look closely at the figures, the overall negative relationship that the authors find -- their big scientific result -- is driven by the large circles in the upper right hand part of the figure: in other words, the big states drive the overall result. But that aggregate result masks enormous variation at the state level.

Unfortunately, rather than exploring what is one of the most interesting facets of their data, the authors bury this figure as just one more piece of evidence in favour of their big scientific hypothesis. What a shame. 

We would have wanted to know: what are those states near the (0,0) point where there isn’t much of a difference between the two communities, either in OD or mortality rates? What are the states in the lower left quadrants where Hindus are doing better on both fronts?

The authors don’t tell us and, of course, our friends at the New York Times and The Economist didn't bother to ask.

The deeper question that Figure 4 raises is this: with such vast heterogeneity in the data, how meaningful is it in the end to boil the discussion down to the labels "Hindu" and "Muslim"?

By our reading of the paper, and related literature, there are very likely a large and possibly unobservable set of regional and local differences, interacting in complex ways, which, despite the best econometric techniques currently available, cannot be distilled out when comparing Hindu versus Muslim.

Could the "Hindu" versus "Muslim" differences at the aggregate level across a range of indicators, which were picked up in the original "paradox" to be explained, themselves represent some sort of artefact of statistical aggregation?

Could they be hiding a whole slew of deeper, local, regional, even household-specific -- and difficult to measure -- differences which all get swept under the rug if everything is aggregated up into these big categories, Hindu and Muslim?

This, of course, is a heretical suggestion, and only a conjecture on our part.

Yet its very possibility suggests that researchers ought to dig deeper before reducing a complex pattern of socioeconomic -- and, yes, also cultural and religious -- differences into these big aggregated boxes, "Hindu" and "Muslim", and create "puzzles" for others to solve. 

Geruso and Spears' principal scientific finding -- that there's an important relationship between open defecation and higher mortality -- seems to us to be sound and unobjectionable, and after all matches what both science and common sense tell us.

If that had been their paper, there would have been little if anything to object to. Indeed, Spears has co-written several other papers which focus specifically on the effects of open defecation on various health and mortality outcomes.

The problem is when sophisticated statistical analysis is grafted on to dubious and facile cultural theorising. It gets worse when the hyped-up results get picked up by journalists and editors who find it convenient to slot India and Indians into convenient, stereotypical categories, all the better to reinforce conventional prejudices.

Sometimes, to get the real story, you have to dig deeper -- into that pile of you know what.

Vivek Dehejia (@vdehejia) is an economics professor at Carleton University in Ottawa, Canada. 

He and Rupa Subramanya (@rupasubramanya) co-authored "Indianomix: Making Sense of Modern India" (Random House India, 2012).