Does it matter if Muslim representatives are elected in India?

by on July 27, 2013 at 4:34 pm in Education, Medicine, Political Science | Permalink

There is a new paper by Sonia R. Bhalotra, Guilhem Cassan, Irma Clots-Figueras,  and Lakshmi Iyer which says yes it does matter:

This paper investigates whether the religious identity of state legislators in India influences development outcomes, both for citizens of their religious group and for the population as a whole. To allow for politician identity to be correlated with constituency level voter preferences or characteristics that make religion salient, we use quasi-random variation in legislator identity generated by close elections between Muslim and non-Muslim candidates. We find that increasing the political representation of Muslims improves health and education outcomes in the district from which the legislator is elected. We find no evidence of religious favoritism: Muslim children do not benefit more from Muslim political representation than children from other religious groups.

The NBER version is here, there is an ungated pdf here.

Jacob A. Geller July 27, 2013 at 4:47 pm

I was disappointed to see that the results aren’t significant for races where the margin of victory was 1% or less. They’re only significant when the margin of victory was 2% or less (and also 3% or less). I also wonder how representative of other districts these ones are (which districts do and don’t have close races between Muslim and non-Muslim candidates? what sort of districts are those?). There is a sort of Lucas critique here…

…interesting nonetheless.

Jacob A. Geller July 27, 2013 at 4:50 pm

PS — How close is a race where the winner won by 1.9 percentage points? Is that close enough for us to say that the religion of the winning candidate is “random”?

Arun July 27, 2013 at 7:21 pm

Confusing symptom and cause maybe? Constitutencies that elect Muslim candidates are probably intrinsically more progressive in the first place.

Dylan July 27, 2013 at 8:09 pm

That was my thought – constituencies willing to look past religious preferences are more likely to have characteristics that lead to better social outcomes.

Cimon Alexander July 28, 2013 at 1:46 pm

This shows me how far I have drifted from the liberal mainstream nowadays. My first thought was “these districts must have a lot of Muslims”. After alll, there is a lot of historical conflict between Hindus and Muslims that extends into modern times. But a liberal leaps to the conclusion that liberal Hindus are voting for Muslim candidates. I wonder who is correct?

Rahul July 28, 2013 at 1:57 pm

The pesky problem is reality is multi-dimensional: religion is not the only (perhaps not even the main) thing that matters.

A Hindu voter may be faced with a corrupt Hindu candidate versus a clean Muslim. Or a Muslim businessman may vote for a pro-business Hindu. Or an agnostic (plenty among the young) may vote for the candidate that promises change. Or feeds him most moonshine on election eve. Caste politics is another twist.

It’s a folly to analyse Indian politics on religion alone.

Rahul July 28, 2013 at 1:51 am

The paper is filled with shoddy writing like this snippet:

“[Muslims] are, on average, poorer than Hindus: 31% of Muslims were below the poverty line in 2004-05, much higher than the figure of 21% for upper-caste Hindus and comparable to the figure of 35% for lower castes.”

So what they really mean is Muslims are, on average, poorer than upper-caste Hindus. It’s as inane as writing “Protestants are, on average, poorer than Catholics of royal lineage”

You cannot compare the bulk of a cohort with a sub-set of another and make any useful comparison. It’d seem that Muslims are, on average, richer than lower-caste Hindus.

Either compare all Muslims with all Hindus, or lower-caste (or class, education whatever) Muslims with lower-caste Hindus. What’s worse, don’t use a skewed comparison with a sub-class to make a sweeping statement about the larger category.

Jacob A. Geller July 28, 2013 at 1:56 am

Or, explain why the proper comparison is between Muslims and upper-caste Hindus, if there is indeed a reason.

shrikanthk July 28, 2013 at 3:08 am

Yes. Also how do they define “upper caste” Hindus? If the term strictly includes only the priestly castes / trading castes and old warrior castes, I would expect the “below poverty line” % to be much much lower than 21% (10% is my hunch).

Also I’m pretty sure atleast 50-60% of “Hindus” don’t fall in the “upper caste” bracket the way I’ve defined it above.

Rahul July 28, 2013 at 3:20 am

This was their primary source (interesting read):

“Social, Economic and Educational Status of the Muslim Community of India”

Though, I couldn’t find anything on upper-caste Hindu poverty in there. You might be luckier (I’m sincerely hoping they didn’t pull those numbers out of a hat……).

Ashok Rao July 28, 2013 at 4:39 am

That’s all great. Except for the fact, of course, that every claim in the statement – “[Muslims] are, on average, poorer than Hindus: 31% of Muslims were below the poverty line in 2004-05, much higher than the figure of 21% for upper-caste Hindus and comparable to the figure of 35% for lower castes – is true. Aside from your vastly hyperbolic analogy.

For example:

“This study finds on the basis of NSS data that there are striking variations in the level of poverty among different religious groups, both within and across states in India. The paper shows that religious diversity in India also has an economic dimension. A comparison of the average consumption level of religious groups for all India shows that the average monthly per capita expenditure of Muslims is the lowest in both rural and urban India. The average MPCE of Sikhs and Christians is the highest in rural and urban India, respectively. The prevalence, depth ,and severity of poverty are found to be highest among the “others” in rural India and among Muslims in urban India.”

There’s shoddy writing on this thread, but it isn’t from Bhalotra and colleagues.

prasad July 28, 2013 at 7:43 am

If upper and lower caste hindus have poverty rates of 21 and 35% then for the weighed average to be 31% you need to define about 30% of Hindus as upper caste. Things are complicated, because for example Mandal says 25% of Hindus are upper caste, while other reports would place the upper caste ratio as high as 38%.

Using the 25% figure the Hindu poverty rate is 31.5% while with the 38% number it’s more like 30%. Regardless, I’m seeing rather less evidence than I would have expected for the common wisdom that Muslims in India are much poorer than Hindus.

Ashok Rao July 28, 2013 at 8:08 am

What more proof than average MPCE per capita of Muslims vs. Hindus – by the most comprehensive survey across the country – would we need?

prasad July 28, 2013 at 9:06 am

If nothing else, the statement that Muslim and Hindu poverty rates are the same is new to me. Did you know this already?

Rahul July 28, 2013 at 8:27 am


You make a good point. My point though, was even more basic: To demonstrate mean(M) < mean(H) you cannot use as evidence mean(M) mean( another-subset(H) ) does not bolster their claim.

I wasn’t appealing to external data or prior knowledge. Merely the fatally flawed logical structure of that argument they make. As to the underlying reality of who’s poorer (when averaged correctly), I don’t claim to know.

prasad July 28, 2013 at 9:12 am

Rahul, I was just replying to Ashok.

I completely agree with your point; the para comparing Muslims just to upper caste Hindus is plain weird. You can compare Muslims to Hindus. Or if you want caste in the story you should break up both populations by caste. (In fact that seems like something worth doing period – there could easily be some lurking Simpsons paradox type effect). You can’t compare Muslims to upper caste Hindus and expect to be making sense.

“To demonstrate mean(M) < mean(H) you cannot use as evidence mean(M) mean( another-subset(H) ) "

It's not just "subset of H" – it's a subset you know a priori is not representative of H!

Ashok Rao July 28, 2013 at 9:32 am

Actually not so simple.

Quick question – are Dalits Hindus? Or, more importantly, should they be disaggregated separately?

prasad July 28, 2013 at 9:38 am

ISTM Dalits should be treated about the same as Arzals…

Incidentally, I am not as sure incomes and expenditures vary with religion between Hindus and Muslims either:

They’re looking at household income/expenditures, maybe the picture at the individual level looks different if household sizes are very different. (But last I checked children per household was converging pretty fast. Plus anyway fertility is the right handle on *that* sort of effect.)

Jacob A. Geller July 28, 2013 at 9:23 am

“There’s shoddy writing on this thread, but it isn’t from Bhalotra and colleagues.”

Be nice…

Ashok Rao July 28, 2013 at 4:49 am

This really is an interesting claim. Of course not of policy significance (subsidizing Muslim legislators doesn’t seem too secular).

By the way, note that this supports (weak, but not negligible) evidence in favor of the Sen/Dreze hypothesis where provision of basic public services is a crucial goal. In such districts we see that because Muslims are on average poorer (contrary to what others on this thread have said this is quite established fact) and are also the minority in many areas, they are far more eager to institute inclusive public programs. Clearly this improves the outcomes towards not only Muslims, but all groups – perhaps because Muslims are more able to represent the poor.

Salem July 28, 2013 at 8:51 am

Yet if you look at the paper, this result only holds when you control for Congress/BJP support. This makes the result almost backwards, given that Congress is the party of Indian Muslims. Compare the claim “richer people have more children” – true, if you control for education. Not at all clear that you should control for this.

Contrary to Ashoka Rao, this flies in the face of the Sen/Dreze hypothesis. If the hypothesis were true, we wouldn’t need the control. I think the most straightforward account of this result is that we are looking at a proxy for a secular, non-clientalist approach to politics.

Ashok Rao July 28, 2013 at 9:08 am

I hope you know that richer people are far more clearly associated with education than Congress is with the “Indian Party of Muslims”. Not the least because you can define one association and the other is just rhetoric.

That said, I do take your point that non-clientalism may be the key driver. But in poor, close districts there are good reasons to believe that this result isn’t a red herring.

prasad July 28, 2013 at 9:15 am

You’re being needlessly tendentious here. Do you really mean to say that the claim about Muslims voting Congress is “just rhetoric”? You must think it’s impossible to break down opinion and exit polls by religion if so.

Ashok Rao July 28, 2013 at 9:30 am

I would be wary of listening to someone who thinks all (or much) of Indian politics – !!!! – can be boiled down to “Congress is the Muslim party”. I’m not trying to be difficult, I just don’t even know how to quantify that.

And I know – for a fact – the relationship (assoication) between education and income is far, far more clear and demonstrable. I was wrong to say that it is “just rhetoric”, but that is a large part of it.

I know poor Muslims living in Mumbai slums who’ve voted for Shiv Sena backed guys. But Shiv Sena is the Hindu party, yes?

prasad July 28, 2013 at 9:42 am

I would have thought – regardless of wording – that all he needs for his statement is something about correlations between party support and religious identification, that you need to control for. As such, the obvious way to operationalize “X is the party of Ys” is that the frequency of voting X is higher among Y’s than among non Y’s.

Rather than “each Y votes X” which is the sort of statement that is refuted by the fact that some Muslims vote Shiv Sena etc.

Salem July 28, 2013 at 1:08 pm

I would be wary of listening to someone who thinks all (or much) of Indian politics – !!!! – can be boiled down to “Congress is the Muslim party”. I’m not trying to be difficult, I just don’t even know how to quantify that.

Well, considering that I didn’t say that, that’s handy! And it’s pretty easy to quantify. For example, “What % of Muslims vote for Congress?” Indeed, the authors quantify it for their purposes (table A4) which is how come they’re able to use a control.

As Prasad says, you are being needlessly tendentious. It’s pretty obvious that I didn’t mean that each and every Muslim votes Congress, any more than “richer people have [more/less] children” is intended to be true for each and every couple. I strongly suspect that you are throwing around FUD because you want your claims to go unchallenged, e,g,:

But in poor, close districts there are good reasons to believe that this result isn’t a red herring.

If you are capable of a civil discussion about this, I would be genuinely interested to know what you think those reasons are.

Rahul July 28, 2013 at 11:02 am

Had this study shown, for arguments sake, that representation of Muslims adversely impacts health outcomes, would Bhalotra and colleagues then conclude it was reasonable to synthetically suppress Muslim representation in politics? I’m just wondering. (Apologies if I’m wrong in gauging the subliminal message of this article as a case for increasing Muslim quotas )

I don’t want to single out the Muslim issue, but studies of this sort universally (be it Jews, blacks, Hispanics, women etc.).

Once one accepts the premise of secularism, it ought to be considered sacred and axiomatic (or close to). Once you start positing arguments for preferential representation of a group on performance grounds, you encourage the other, far less palatable variant of disenfranchisement arguments on sundry similar excuses.

Jacob A. Geller July 28, 2013 at 3:52 pm

I think the short answer is “no.”

The longer answer is that they’re comparing Muslim candidates vs. non-Muslim candidates — NOT Muslim candidates vs. some kind of “control candidate,” or baseline of zero. So you can think of this as effectively saying that *non*-Muslims DO worsen health and education outcomes, AND you can think of it as “Muslims improve health & education outcomes” — both interpretations are technically correct, and interchangeable. But for obvious reasons they decided to frame it as “Muslim candidates improve health outcomes,” NOT “Hindu candidates kill infants and keep kids out of school.” So if the results were instead that non-Muslims improved health & education outcomes, I think they would simply have reported the results in the same way as these, only with “Muslim” and “non-Muslim switched, i.e. “non-Muslim candidates improve health & education outcomes,” NOT “Muslim candidates kill infants and keep kids out of school, so they should be resisted.” Note that the authors are also pretty keen to show that Muslims do not seem to favor their own people over others, and that if anything Muslim candidates support a lot of their non-Muslim constituents — so I don’t see much in the way of anti-Muslim animus here, nor any anti-Hindu animus either for that matter.

They’re just a couple of social scientists who went p-value fishing (as many social scientists do), came up with some convenient p-values which might or might not reflect some actual causal relationships (see my skepticism above, or the paper itself), and so they wrote a paper about those p-values and threw in some speculation about what the cause of those relationships might be. They’re not out to suppress anybody, just doing what social scientists do.

Rahul July 29, 2013 at 2:36 am

Fair enough. I’m happy if people ignore the work as publish-or-perish crap. :)

Nathan W July 28, 2013 at 10:42 pm

I’d be tempted to say “yes it matters” because there are so many Muslims in India. But you’d think more practical considerations than which dude told us to try to be decent would be more relevant to electoral consideration and democratic representation …

Comments on this entry are closed.

Previous post:

Next post: