The legend of 'votekataua'

I heard this term for the first time in Benaras a month ago, at a lassi shop, in the context of Arvind Kejriwal. Almost a month has passed since, and from what I hear Arvind and AAP have since risen much in Varanasi's esteem - he is considered a serious contender now. Though used in a negative connotation, in reality, the phenomenon of 'votekataua' is a very natural consequence of the voting system that we used. In the first past the post system (or simple plurality system the 'winner takes all'.  The candidate with the maximum votes (not the majority of votes) wins the seat. Let us consider a hypothetical constituency with party A and party B engaged in a tight two cornered contest. In a FPPS system getting more vote for oneself is not the only way for party A to win an election. The other way obviously is to cause a division in the votes of party B such that both parts are now less than party A. The new entity that wins enough votes to severely damage the chances of the challenger but not quite enough for itself to win, advertently or inadvertently plays the role of a vote-cutter or 'votekataua'. Come to think of it, if we followed a proportionate representation system, this word would probably not exist.

The word is new but the phenomenon is not. I have talked about this in detail in the context of the Karnataka assembly election (Will Yeddyurappa do a Bangarappa?).  We went on to see how Yeddyurappa spoiled the prospects of the BJP in that election (Karnataka 2013). The Congress romped home with a simple majority without increasing its vote share by much.

In this post I want to look at 2 more parties who earned this title in the last election. Specifically we will look at the case of estranged nephew Raj Thakray's Maharashtra Navanirman Sena (MNS) in urban Maharashtra and Matinee idol Chiranjeevi's Praja Rajyam Party (PRP) in Andhra.

The Maharashtra Navanirman Sena was born out of a fratricidal war of succession within the Shiv Sena. When Bal Thackray retired from active politics in 2006 he appointed his son Uddhav as leader of the party resulting in Raj Thackray breaking away to form a party of his own. The MNS in many ways was an ideological clone of the Shiv Sena, with similar 'marathi manus' DNA and hindu nationalist leanings. In the initial years of its formation the MNS adopted a particularly virulent  xenophobic stance against north Indians in Mumbai, almost as a way to differentiate itself from its parent, leading to several bouts of violence in the city. All the while amidst clamor for action against the MNS to contain the violence the Congress/NCP state government stayed away, letting MNS build its own constituency by cutting into Shiv Senas base. It turned out to be a masterstroke. As we can see in the figure below, even though the MNS was not able to win a single seat in Maharashtra, it still managed to put up a reasonable contest. The result was that what was expected to be primarily a 2 cornered fight across much of the state did not quiet pan out that way. Much of Maharashtra saw three cornered contest, with the MNS coming up as a strong third force in urban centers and the BSP consolidating its position in much of Vidharba to emerge as a contender there (Graphic below)

Learn About Tableau

Between - 2004 and 2009 - the Congress +NCP increased its number of seats from 22 to 25 inspite of its vote share falling by almost 4%, primarily because the BJP-Shiv Sena combine lost even more (7.5%). 4% of that drop is a straight shift of Sena voters to the MNS.

Learn About Tableau

 The effect of the MNS was particularly severe in the Urban centers. A little quick math shows how in all the 10 of the 11 constituencies that the MNS  contested from  it severely dented the prospects of the BJP-SHS alliance while not managing a single seat for itself. 

Had it not been for the MNS the results from Maharashtra would have read 30-15 in favor of the BJP-SHS instead of the final 25-20 in favor of the Congress-NCP. As far as 'Votekatauas' go, the MNS in 2009 was a textbook example.

Moving on to Chiranjeevi and his Praja Rajyam Party. The PRAP was the other impressive debut of 2009. In their first election The Praja Rajyam Party won a whooping 15% of the vote share in Andhra, but did not have a single Lok Sabha seat to show for their efforts. The PRAP however did manage to turn to contest in much of Andhra into a 3 cornered affair. 

Learn About Tableau

With the breakup of the NDA and the emergence of the PRAP the the divided opposition was no match to the strong Congress. As per a lot of observers - because of the emergence of the PRAP the anti incumbency vote against the Congress got split in half, making it possible for YSR Reddy to deliver Andhra for the Congress for the second time in a row, with a higher number of seats in spite of a fall in vote share.

Learn About Tableau

And so it happens every time, it is something like an adoption barrier in marketing parlance. New parties have to always face the challenge of being seen as nothing but a vote cutter. Its kind of a self fulfilling prophesy, this thing about being a 'votekataua', if people think that's what a party is, that is what it is likely to end up being being (tactical voting in FPPS).  But then again there are parties that are able to cut through this, communicate their message to the voters and convince them of a possibility of victory, the AAP campaign in Delhi is an excellent example of this. New parties are the needed in democracy, they keep it fresh and bring in new ideas and leaders. While it is important to understand the phenomenon of the votekataua as a tactical weapon in democracy, it is also important to get rid of some of this negative connotation with wasted votes.

No country for cubism.

The brief study of the types of democracy and voting systems was interesting as well as threw up a couple of insights for me. Still, mulching Wikipedia is really not my favorite pastime, this is a datablog and yours truly is a data-wrangler, its time to go back to the roots. The plan in the next couple of posts is to go deeper into the consequences of the 'first past the post' voting system that India uses for its elections with the lens of data.  In this post I also want to make an attempt to understand the twisted relationship between vote shares and seats won by different parties. In the process I also want to take a closer look at polygonal contests in this voting system.

There is a tendency in large parts of the media to simplify the narrative of this election. Straight contests lend easily to tv debates and montages, Congress vs. BJP,  Rahul Gandhi on one side, Narendra Modi on the other and the parliament house in between. A rather simplistic picture, with undertones of a wave of voters moving from one camp to another that needs to be exploded further. The graphic below shows the vote shares and seats of the primary Parties/agglomerations over the years as also combined vote share of the two main parties- the Congress and the BJP in the last few elections.

Learn About Tableau

Two observations are notable here. Firstly, from (b) and (d) the strength of the two remaining national parties in terms of the votes they poll seems grossly overrated. Even if we were to add up the vote shares secured by the BJP and Congress that number often doesnt even cross 50% and if anything seems to be going down in the last couple of elections. More than half of the electorate votes for neither the BJP nor the Congress. This remaining 50% is highly fragmented among hundreds of parties (mostly regional) and this makes elections across large parts of India multi-cornered (polygonal) affairs.

Secondly from (a) and (c) the correlation between the vote shares and seats won is not straightforward.  However It is well known that the first past the post voting system accentuates differences in vote share and tends to reward parties with higher vote shares disproportionately- case in point being the 1984 election when the congress with 49% vote share swept the election winning almost 80% of the seats. Again this is not all that straightforward. In the three elections between 1996 and 1999 the BJP won more seats than the congress even though it had lesser vote share nationally.   A more nuanced region by region view, with an understanding of the 'polygonality' (how many corners) of the contest is needed to get to better answers and that's where I intend to go to in this post. Also there are empirical observations to correlate the the vote share and seats won - for example the the "cube law" which  states that in a two cornered contest the ratio of seats won is proportional to the cube of the ratio of  vote shares. I want to round up this post with a quick empirical test of the law of cubes in the Indian context.

To understand 'polygonaility' by region i.e which seats/regions in the country have straight, triangular, quadrangular and even pentagonal contest, some definitions are needed first. Given that most seats have 10 or more contestants, how do we define whether a contest is straight/triangular/quadrangular? The graphic below shows the frequency distribution of the vote share of the top 2 candidates in all seats in the 2009 election  



In about 22 constituencies, the top two candidates together secured between 95% to 100% of the vote share. In the context of Indian democracy this is as direct a contest as it gets. A quick look at the cumulative numbers reveals that the top two contestants together constitute more that 80% of the vote share in 272 constituencies out of the 543 (roughly half). 80% is the number I chose as the threshold for whether a contest is classified as 2 cornered, 3 cornered or 4/5 cornered. Essentially, if the top 2 make up more that 80% of the vote share it is classified as  straight or  2 cornered contest. If that is not the case but the top 3 constitute more than 80% of the vote share then it is classified as a triangular or 3 cornered contest and similarly for 4 cornered (quadrilateral) and 5 cornered (pentagonal) contests.

The following graphic shows the geographical distribution constituencies based on the polygonality of the contest in 2009.

Learn About Tableau

It is interesting to see that the contest is essentially 2 cornered (by the given definition) in large parts of Punjab, Himachal, Rajasthan, Madhya Pradesh, Chhattisgarh, Gujarat, Northern Karnataka, Kerala and West Bengal. Meanwhile  large parts of  Tamil Nadu, Andhra Pradesh, Orissa, Assam and Maharashtra  witnessed three cornered contests where as other parts of UP, Bihar and Jharkhand witnessed triangular/quadrangular and in some rare cases even pentagonal contests. In Nawada, Bihar the top 4 condidates secured just 62% votes between them, with the winner getting away with merely 22% vote.

The following graphic shows different states characterized by the polygonality of the contest in its seats.

Learn About Tableau

Looking at the major states one by one, in UP we find the usual suspects(BJP, BSP, SP, Congress) engaged in a 3 or 4 cornered contest in most seats. The notable exceptions to this are the usual pocket boroughs of various parties where one predominant contestant gives the appearance of a straight contest. For example Mainpuri, Etawah (SP strongholds) and Amethi, Raebareily (for INC).

 In the case of Maharashtra one would be expecting it to be a straight contest between the NDA (BJP+SHS) and UPA (INC+NCP) but it is not quiet. In a large part of Vidharba the BSP among others offers a third alternative making it a triangular contest and similarly in the Urban areas of Mumbai, Pune and Nasik, the MNS was able to eat into significant vote share of the Shiv Sena to make it a 3 cornered contest.

AP sees quiet a number of 3-4 corner contests with Congress, Telegu Desam, Telangana Rashtra Samiti, Chiranjeevi's Praja Rajyam party and in a few case the BJP fighting for the spoils.

Rajasthan, Gujarat, Madhya Pradesh, Chhattisgarh, Punjab, West Bengal and Kerala witness primarily straight contests. The only nuance being that while in the case of the first 5 it is primarily between the Congress and the BJP, the latter 2 were primary between Congress vs Left/others.

In both Bihar and Jharkhand , multiple regional parties (some of whom fight elections in both the contiguous states) lead to 4 cornered and 5 cornered contests.

To wrap up I want to take a look at the applicability of the law of cubes in smaller subsets of seats, where there is a direct contest. Of the 272 seats where there seams to be a direct contest , The NDA and the UPA were locked in direct contest in only about 150 seats. The states of Gujarat, Madhya Pradesh, Maharashtra,  Rajasthan, Chhattisgarh,  Delhi, Punjab constituted a bulk of these seats.  Of the remaining 122 direct contest seats - the UPA was competing with a regional party/Third front in 107 seats (primarily West Bengal, Kerala and Tamil Nadu), the NDA was competing with regional parties/3rd front in 8 seats and the contest was between 2 regional parties in the remaining 7. Must be said here that in West Bengal and Tamil Nadu, the UPA had significant allies in 2009 which it has since lost. The contour of elections in West Bengal will be more and more 3/4 cornered and the same would go for Tamil Nadu where the old Congress alliance with DMK no longer exists.


 The graphic on the side shows a quick analysis shows that the law of cubes is only roughly applicable to even this set of constituencies where we have as close to a direct contest as possible in the Indian context. What the law can do for us is that it can give us a rough idea of what a change in vote shares would do to the number of seats won (on the 150 seats where there was a direct contest). The above calculation also illustrates the effect of a 10% change in vote share. As for trying to predict the number of seats at a national level based on a shift in vote shares (from poorly opinion polls), well thats a job better left to the professional psephologist.

Democracy, Voting Systems and thoughts on opinion polls

The last couple of months has been the season of opinion polls. Psephologists, astrologers and sundry television experts have been trying to foresee the results of the election. There has been a lot of controversy as well about the veracity of these opinion polls, with questions being raised on how these polls are conducted and funded. It is rather intuitive that results of opinion polls can have a prejudicial impact on voting, but it seemed that the effect is even stronger in the 'first past the post' voting system that we use vis-a-vis proportional representation systems used in other democracies. Suffice to say that I started off with the rather simple intent of finding more about opinion polls, how they are conducted and the nature of their impact on voter behavior only to come to the conclusion that some basic understanding of the theory of democracy would be necessary before going any forward. It is evidently a very deep subject, and the idea is to skim it at a sufficiently high level.

The graphic below describes the different forms of government based on the source of power and the system of governance

A deeper understanding of the differences and the pros and cons of these forms of democracy is both interesting and important given the current debate in the country. When the 'aam aadmi'  proclaims the the demand for 'poorna swaraj', at the face of it, what they seem to be asking for is direct democracy. Similarly there have been strong currents which give this current election the look of a presidential race. Our founding fathers chose the current form of government after much deliberation - perhaps better understanding of these reasons is needed, before we even consider quick remedies.

Within representative democracy there are multiple voting systems which are essentially different methods that voters can use to make a choice between options.The figure below details out some of the most prevalent ones

Again since the first past the post system is what is followed in this country it perhaps makes sense to get deeper into some of its  pitfalls. Some of the strongest criticism of the FPTP system is that it encourages tactical voting - where voters have the incentive to vote for one who they believe is likely to win, though they might have preferred someone else so as not to 'waste' their vote. In essence in the first past the post system voters optimize between preference and winnability. Examples of tactical voting can be seen across multiple elections in India - So in UP across multiple state elections now the Brahmans have voted tactically for BSP even though their preferred party is the BJP, and in the recent Delhi state elections many people did not vote for the AAP because they did not want to 'waste' their vote on a party that has little chance of winning.

This is where the media and opinion polls have the potential to have a particularly insidious effect. By creating a halo effect and a buzz that a certain party is winning, a serious disincentive is created to vote for anyone other than the winner or the next serious contender. In a FPTP system the media and opinion polls can influence the way individuals think, by creating a perception of what the collective is thinking , and eventually change the way the collective will act by influencing the individuals that make it, kind of like a self fulfilling prophesy.

An example will perhaps come in handy. Lets consider a voter in UP whose first preference is the BSP but has voted for the BJP in the past. If the media and opinion polls convince him that there is a wave in favor of the BJP, this person is likely to switch allegiance so as to not let is vote get wasted. Entire communities can switch tactically in a similar way. The caveat here is that this can work both ways, for example now if we consider a typical SP voter who has a strong aversion for the BJP - the perception for such a wave may make him vote tactically for the party most likely to beat them and hence have the effect of consolidating the opposition.

Apart from the effect of the perception created by media and opinion polls on voter behavior which prima facie does not look very desirable, it perhaps makes sense to take a look at the success rate of these polls. "Not very good" should sum it up. Multiple times in the recent past these opinion polls have gone horribly wrong, most notably in 2004. India is in some sense, a psephologists dream and worst nightmare rolled into one. As a democracy it is incredibly large, diverse and complex. Just to get a large enough sample that can represent the 800 odd million voters is a challenge in the first place, and a sampling error is introduced right there. The other big challenge is to convert the vote shares to seats. Given the multi-party nature of the contest more than half the constituencies see multi-cornered contests where simple mathematical approximations like rule of cubes do not apply.  Other measures like index of opposition unity sound obsolete given that its premise is that of the Congress as the default party of choice, something that most people will agree is not true this time.  Then, the dynamic is very different in each of the states and even within some of the larger ones, with multiple political parties and myriad castes and sub castes. So even with the most honest intentions and the best tools and methodologies it is still something like gazing into a crystal ball.

What has been a bit disturbing in the last couple of months has been the finality and the confidence with which these have been presented. As if the results have been declared already on the news channels. Some humility is called for especially given the track record. Caveats need to be stated candidly, explicitly and frequently especially given the possibility of influencing outcomes in our electoral system. Then again, has this country not seen this movie before, maybe caveats are not needed.