2018 assembly election results analysis: part 2

In the first part of this report, I provided a brief summary of the people's mandate in the 2018 Telangana assembly elections together with a historical comparison of the verdict.

The present post is related to the methodology of this report covering both the data & analytical models. Please note this section is likely to be of a technical nature and probably boring to those looking for political content. On the other hand, data enthusiasts with little interest in politics may find it interesting.

Raw data

As per my usual practice all my data is from the public domain. Data of the previous elections (1978 onwards) is from the Election Commission of India (ECI) while the present election data is extracted from the details published by Chief Electoral Officer's (CEO) Telangana.

Subject to certain IPR terms, I can provide details of my calculations to any reader interested for any reason including verifying my numbers or further analysis.

Party affiliations

ECI & CEO recognize party affiliations of individual candidates as per the prescribed formal process including individual's declaration and party nominations (B-form). The official list shows 1,940 "candidates" contesting on behalf of 94 "parties".

This throws up the following challenges for analysts including too many database elements distracting from serious analysis. For example, a group calling themselves "Samaikyandhra Parirakshana Samithi" put up three candidates who secured a grand total of 1,824 votes between them!

To reduce this clutter, I reclassified the raw party affiliation data into sixteen primary database elements (TRS, INC, TDP, TJS, CPI, BJP, MIM, CPM, BLP, MBT, TRS-REB, INC-REB, SFB, IND, OTH & NOTA) that I am treating as "recognized parties".

A small group of individuals ran on an interesting concept of trying to obtain votes through "symbol subterfuge" e.g. on the symbol of a truck that can be confusing to voters intending to vote for the TRS's "car symbol". Some of these innovative folks enjoyed a reasonable success in an interesting quirk! I am grouping these candidates under SFB (for the hitherto unheard of party called Samajwadi Forward Bloc that most of these chose to represent).

As explained earlier, I am treating the NOTA button in each constituency as an individual candidate and the overall NOTA category as a "party" in my analysis.

These sixteen "parties" are further regrouped to six formations: TRS, UPA, BJP, MIM (Majlis), BLF & OTH (others). BLF did not trouble the statisticians beyond the initial cursory glance.

Please also note "OTH" is a residual category that changes with the context. For example, when analyzing victory margins I group the lone BJP winner among OTH.

Rebel trouble

Formal accreditation as used by the Election Commission may actually be misleading in several cases e.g. when a known party rebel candidate fights on the B-form issued by an entity with no local presence. my solution to this second problem is to rebrand individual candidates denied tickets with the "correct rebel status".

As everyone knows by now, two rebels made it to the current assembly. Korukanti Chander Patel, the # 23 Ramagundam champion, officially ran as the nominee of All India Forward Bloc (AIFB) while Lavudya Ramulu, triumphant from # 115 Wyra was labeled an independent. My methodology puts these winners down respectively as TRS & Congress rebels.

Though I tried to be fair to party rebels, I am forced to give a pass to two lone rebels (one each from TDP & BJP) as they secured under 1.5% votes. With due apologies to these worthy individuals, I prefer to keep my database clean J

The great Indian party hopping game

I tracked the peculiar Indian game of party hopping in the months leading up to the elections. My pre-poll Aya Ram Gaya Ram tracker shows 72 individuals who jumped ship in season with 26 of these fighting the elections (22 on behalf of the new party). Readers interested in the fate of these individuals are requested to be patient for now.

A couple of interesting sidelights (both treated as Congress rebels as per the methodology followed):

·         Former MLA Vooke Abbaiah who had the honor of fighting the past elections on three major party tickets (CPI, TDP & TRS) defected to Congress just before the elections. Failing to win its B-form, he filed his nomination as an independent

·         Syed Ibrahim, twice unsuccessful on TRS ticket earlier, ran as a rebel and ended up unlucky when he was denied the party nomination in 2014. He subsequently "returned home" but defected to Congress after being refused the ticket again. Undeterred by Congress too ignoring his claim, he managed to obtain the B-form of Bahujan Samaj Party (BSP) but lost a fourth time

Geographic classification

Telangana state as originally formed under article 3 consisted of ten districts. Subsequently certain areas of the Khammam district (often incorrectly called "Polavaram submergence mandals") were transferred to AP through an ordinance (ratified later). This geographic transfer affecting two constituencies (# 118 Aswaraopeta and # 119 Bhadrachalam) is however not of material significance to our analysis. It may be noted in passing there was a certain amount of litigation arising out of the related vote transfer due processes but this need not deter the present exercise.

The only Loksabha constituency impacted by this geography transfer is Mahabubabad. All other Loksabha constituencies remain intact.

Post Telangana formation, the erstwhile ten districts were reorganized into 31 districts. All references to the term "district" in this report pertain to the old districts unless explicitly stated otherwise. I did track the polling variations across the "new districts" if anyone is interested.

The regional variations in Telangana are subject to varying interpretations as per the analyst's choice. The delineation followed in this report is outlined below:

·         North covering Adilabad, Nizamabad, Karimnagar, Warangal districts in full apart from Medak district (excluding # 40 Patancheru included in Hyderabad below) totaling 53 constituencies

·         Hyderabad covering 24 constituencies of GHMC, please note four of these are partially outside GHMC limits but are included in the "Hyderabad region" for the purpose of this analysis

·         South covering Mahabubnagar & Nalgonda districts in full apart from apart from 6 constituencies of the Ranga Reddy district (not counted in Hyderabad above) totaling 32 constituencies

·         Khammam covering 10 constituencies of the Khammam district. Certain analysts include this in "North" but I believe the uniqueness of this district deserves to be treated separately

The term "old city" used often with historical and/or political implications is defined even more vaguely. In this report I will instead refer to the Hyderabad Loksabha constituency. It may be noted that # 65 Goshamahal falls under this limit but # 63 Nampalli does not do so.

Population density classification

The 2001 census data lists 173 census towns each rated against a four class system in the then united AP. I used this as the basis for classifying each of the 119 Telangana assembly constituencies by population density.

·         Urban: all constituencies in Hyderabad region plus three constituencies with two census towns plus constituencies with a single class-1 census town

·         Semi-urban: all constituencies with a single class-2 census town

·         Semi-rural: all constituencies with a single class-3 or class-4 census town

·         Rural: all other constituencies i.e. without any census town

Other demographic indicators

Demographic information is difficult to come by in India and, where available, of suspect quality. The most reliable data is available only to commercial users (e.g. FMCG businesses) and often amenable only to custom analysis suitable to the sponsor's field of interest.

In the light of this drawback, an analyst has little choice but to resort to "guesstimates" based on his own knowledge of the area. I picked up the following four parameters and studied constituencies influenced by these factors:

·         Coal mining belt accounting for 10.4% of the 2018 votes

·         Acute agrarian distress symbolized by out migration, farmer/weaver suicides, bidi industry presence and/or fluorosis contributing 19.5% to the 2018 turnout

·         Areas (excluding the seats contested by Majlis) with significant Muslim population coming to 19.3% of the 2018 votes

·         Constituencies with significant settler population working out around 9.4% of the 2018 votes

This classification (and indeed even the choice of parameters) is purely subjective. In addition, the broad "swing" pertains to the constituency, not the target demography. The fact that TRS gained 6.4% votes in the "coal belt" does not show which way the miners voted.

The word "settler" refers to Andhras who migrated to the Hyderabad state before 1956 and gave up links to their former homes. Contrary to the belief of several media observers, the number of settlers in the Hyderabad region is quite low. More recent Andhra migrants living in the Hyderabad region are not settlers. I did not treat them as a separate electoral category as I am anyway reviewing Hyderabad patterns vis-à-vis 2014 and 2016.

Please also note that the classifications are not mutually exclusive. For example as # 12 Bodhan is home both to sizable Muslim and settler population, it has been included in both segments.

There are certain demographic indicators (e.g. tank irrigated areas & Lambada strongholds) that I could not deconstruct. I would be thankful to anyone who can provide information on such factors.

Linguistic minorities do contribute a good chunk of votes in Telangana but do not constitute a coherent voting block. This is also true of non-Lambada tribals. I did not therefore study the constituencies with significant populations of these demographic groups.

Vote preference & "swing" bands

Readers familiar with American politics will probably know Prof. Andrew S. Tanenbaum as a veteran observer-cum-analyst of the local political landscape. The Tanenbaum model uses seven color coded vote preference bands ranging from "strongly Democratic" through "strongly GOP".

As our present context differs a good deal from the strong bipolar two party system prevalent system I am using a somewhat different and simplified system:

·         Vote share of >= 50.0% is deemed to be unbeatable (green)

·         Vote advantage and/or gain ("swing") of >= 5.0% is considered excellent (green)

·         Vote share of <= 10.0% is treated as below threshold (red)

·         Vote disadvantage and/or loss ("negative swing") of >= 5.0% is deemed to be terrible (red)

My unpublished (but available on request subject to my own terms) spreadsheet thus shows the TRS vote share (50.4%) and vote advantage (19.7%) in Mahabubnagar district in green. On the other hand though TRS 2014 vote share in Khammam district is red at 9.5%, the vote gain (31.1%) is green. I did not use the color coding ("conditional formatting") in the blog posts to avoid jarring eyeballs.