BJP-ruled states still more communally violent

In an article posted on the website Newslaundry on 14 October, Rupa Subramanya argues that government statistics do not support the view that “there’s been some sort of upsurge in communal violence since the election of Narendra Modi”, contending that figures that show an increase of 25% in communal incidents in January-May 2015 vs. January-May 2014 are unlikely to be statistically significant.

She also states the following:

This is a reference to my 15 February 2014 blog post titled BJP ruled states more communally violent. In her article, Subramanya asserts that:

  1. These findings are questionable because “one can get just about any result [by] using different start and end dates”; and
  2. The persistence of communal violence among a variety of different states makes it “impossible for a fair minded person to assert that there’s a greater prevalence of communal violence in either BJP or Congress ruled states.”

Note that the original one-and-a-half year old blog post used 2010-13 data because that’s what was available at the time. So let’s be fair-minded and run the analysis with the data used in the Newslaundry piece (2010-January 2015).

Screen Shot 2015-10-16 at 11.29.13 am

Lo and behold, there’s no change in the ranking of states. Zip, zero. BJP-ruled states have an intensity of communal violence (measured by casualty rates) that is 61% higher than that of INC-ruled states. Note that both INC- and BJP-ruled states are above the national average, which means that states ruled by other parties are, on average, more peaceful. I’d say that debunks the debunker.

Let’s include communal violence data from 2007-09 to include as much information as we easily can. And here is what we find:

Screen Shot 2015-10-16 at 11.46.23 am

There is some reshuffling in the ranking of the states, but the basic pattern holds: BJP-ruled states have 64% higher intensity of communal violence than INC-ruled states, and 78% higher than the national average. The only states where neither the BJP or INC was dominant thoughout this period are Jharkhand, Kerala and Rajasthan. The BJP was in power for most of the time in Karnataka and all of the time in Gujarat and Madhya Pradesh. The INC was in office in Assam throughout and in Andhra Pradesh (including Telangana) for most of the period.

The Newslaundry piece does note that several states changed government in 2013 and 2014 (and earlier), so let’s focus only on those that have experienced lengthy periods of government by one party.

Screen Shot 2015-10-16 at 11.39.28 am

(No) surprise! BJP-ruled states still have a casualty rate 73% higher than INC-ruled states, and 60% greater than supposedly polarised Uttar Pradesh (UP) (remember, the UP figures include the Muzaffarnagar riots).

These findings, though robust, need not comprise the whole story. A thoughtful critique would note that the political party running a state isn’t the sole determinant of communal violence, and that factors such as the nature of party competition (Wilkinson 2004), the presence of institutionalised riot systems (Brass 1997), the density of civic ties among communal groups (Varshney 2002) and other contending explanations could also shape levels of communal violence.

A considered critique might also seek to distinguish between low-level communal friction at the level of locality, town or village, and outbreaks of communal violence that go beyond these. Subramanya hints at this in her observations about Uttar Pradesh but appears too focused on trying to fix the responsibility for the 2013 riots on Chief Minister Akhilesh Yadav to look at this in a considered way.

Instead, we get baseless generalisations about cut-off points, and a digression involving temperatures and climate change. Sorry, Newslaundry, this just doesn’t cut it.

7 thoughts on “BJP-ruled states still more communally violent

  1. shanks

    what is this intensity you speak of? Is that a metric that is used worldwide or you made that up? Let’s see, did you scale it proportionally for the religious mix of the people in each state? Have you factored in, that for larger populations, the likelihood of any kind of clash is higher? What happens if you remove outliers, e.g. the biggest riot in each state?

    As a statistician, I find your definition of ‘intensity’ weird; I’ve seen (incidents/accidents/fatalities etc)/million/year but not intensity. Your data just showed that rioting and violence is a fairly steady affair across all states….remove BJP/Congress subheading and the numbers simply tell that story.

    1. I use intensity as shorthand for casualties per million. The Home Ministry dataset presented in parliament does not disaggregate down to the level of individual riot.

      The numbers could indeed tell their own story, and political parties need not be the only driver (as I acknowledge above) and could even be endogenous to the process.

      This particular discussion has its genesis in the following post, which evolved into a more detailed look at communal violence by state and party:

      1. shanks

        >I use intensity as shorthand for casualties per million

        hmmm….fine. But what’s wrong or ambiguous about casaulties/million? Intensity, in this context only tends inflame the discussion as if the riots were fought harder in BJP states when you yourself say you don’t have info on individual riots

        >The numbers could indeed tell their own story,

        which they do. And while your attempt at a causality is a good start(whatever your political leanings), it might be a good idea to be a little more rigorous about the methodology

        >and political parties need not be the only driver (as I acknowledge above)

        then why the clickbaity title?

        >and could even be endogenous to the process.
        Sorry, didn’t understand this. What is the ‘process’? If you’re saying that the rioting itself might be something trivial/small time issue in nondescript small town not triggered by exogenous incitement, then the core edifice of your statement comes crashing down? Why would you then correlate that to the political party in power in the tables above?

        Net result

        1. Political parties *may* be a driver to violence
        2. No info on cause of rioting makes it doubly difficult to ascribe motive(endogenous or exogenous)
        3. Year after year, we had riots in the low hundreds in each state irrespective of the party in power like a law & order background noise.

        All I see is, when I open the papers, there’s at least 1 riot going on somewhere in the country.

  2. Disagreement over variable names and title aside, you’re debating causality here, and the blog post quite clearly states that causality is not clear cut.

    “Endogenous” here refers to the possibility that communal violence could affect who is elected to power as much as the party in power influences the level of violence. Core edifices crashing? I think not.

    This blog is a response to Newslaundry’s questioning of data. It demonstrates that the basic pattern holds even when you expand the time frame, exclude ambiguous cases etc. It establishes that BJP-ruled states have higher levels of violence than INC-ruled states, and states run by other parties have lower levels still. That’s it.

    As for your conclusion:

    1. Agree.
    2. Irrelevant, since I am not ascribing motive.
    3. Non-sequitur. Year after year, we’ve had communal incidents and casualties in most states, but the background noise is louder in some states than others.

  3. fopgames

    These are the states which have always had communal violence. This only shows BJP is in power in most of those states. does not show relation between the two.

    What would be a better measure is to compare the stats of the same states with when Congress or other party was ruled vs when BJP ruled.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s