GHMC results analysis: part 3

Update from Warangal & Khammam

As everyone knows by know TRS swept Warangal winning 44 wards with TRS rebels coming in second with 8 seats. Congress bagged 4 wards while BJP & left won one seat each.

Rather unexpectedly TRS did extremely well in Khammam too winning 34 wards with Congress taking a creditable 10 seats. Left with 4 & YCP with 2 bring up the rear.

The official results from the Warangal & Khammam elections are not available. In any case I did not plan to analyze these. However information from secondary sources may be of interest:

·         Warangal: TRS 51.7%, Congress 13.5%, BJP 10.4%, TDP 2.4%

·         Khammam: TRS 42.4%, Congress 20.5% Left 12.3% YCP 10.7% TDP 8.4% BJP 1.6%

TDP not only drew a blank in both the cities but suffered a humiliating defeat in Warangal with just one of its nominees retaining his deposit. One of its candidates won a grand total of 4 votes J

YCP polling 10%+ in Khammam is an interesting takeaway. BJP did reasonably well in Warangal reinforcing our earlier finding that it may be better off fighting on its own.

Congress can find solace from the fact it appears to be the second popular choice outside of Hyderabad. They need to bridge the gap as well come back in the capital in order to have a serious fling in 2019.

A quick recap

In the first part of this report, we came to the following conclusions:

·         TRS scored a historic win dominating the non-Majlis areas

·         TRS did quite well even in the Majlis strongholds

·         Majlis performed predictably well retaining its traditional base

·         This election was a virtual nightmare for both NDA & Congress

Continuing the analysis, in the second part we found:

·         If assembly elections are held today TRS would win 18 seats in GHMC, Majlis retaining 4 constituencies and the other two being too close to call

·         If Loksabha elections are held today TRS would win 3 of the 4 seats with Majlis retaining Hyderabad

·         Vote swing in favor of TRS is around 24% in the period 2014-16 while Majlis broadly stood its ground

·         NDA and, to a lesser extent, Congress both lost ground in the last two years

·         BJP appears to have been dragged down by TDP

I will now move on to find the reasons behind this historic win. By necessity, this exercise though data driven is somewhat subjective. In the present post I will look at some outlandish "reasons" offered by the losing parties or their sympathizers. I will examine more serious matters in the next (concluding) part.

Vote tampering accusations

Immediately after the results were out, former Congress mayor Banda Karthika Reddy accused TRS of tampering with Electronic Voting Machines (EVM). Nalamada Uttam Kumar Reddy, state Congress president, made the same allegation earlier about the 2015 Warangal Loksabha bye-elections.

Such allegations have been levied quite a few times in the recent years ever since one Hari K. Prasad "demonstrated" a simulated attack. Prasad has since removed the "demonstration" videos from his site retaining only a transcript. I do not consider the paper serious scientific research in view of several defects: that is a subject for a different post though! For the present let us assume the attacks claimed by Prasad are possible at a laboratory level to see if these can be repeated in a real world situation.

The authors list several tampering possibilities in their "vulnerability analysis". I prefer to restrict myself to what they claim to have actually done to an EVM they managed to obtain instead of chasing speculation:

·         Dishonest display

·         Inserting a chip-on-memory manipulator

Both these methods require physically tampering the control unit. The first method is not only cumbersome but also useless as the machines can output data to a computer. Tampering display is meaningless when the results are based on machine read output. The authors' claim "officials manually record the totals from each machine and add them together to determine the election result" does not sound correct to me as the extensive statistical reports are not easy to generate from manual compilation. Even assuming the contention is correct, the existence of an audit trail that can be used in case of a dispute rules this method out for any serious mischief maker.

The second method is more realistic in attempting the stated objective of stealing votes. This involves inserting a chip-on device consisting of a microcontroller, three light emitting diodes (LED) and a ten position rotary switch. The rotary switch is set to any position between 1-9 corresponding to the ballot position of the "stealing party" candidate.

Physically tampering with the control unit is not as easy as it sounds. Even though security may be lax at times some kind of break-in or trespass would still be required. The attack can be executed only between the time the machines are transported to the storage site and the votes are counted, a window that can be just a couple of days as in the GHMC case. The organizational element of reconnaissance, planning, staffing, communications & execution can be daunting. In addition unusually high criminal movements can be easily detected by intelligence & law enforcement.

Stealing votes would require opening each EVM control unit and inserting the manipulation device. The manipulator executes a program in two parts: reading the actual votes polled and rewriting votes to the programmed level. After this is complete, the "done" LED lights up. The authors went to the extent of providing a rudimentary but clever error recovery mechanism with an "error" LED. Looks brilliant at a first glance but how feasible is this?

The sheer volumes are a challenge to any wannabe thief. Assuming an average of 1,000 electors to a booth, we are looking at around at around 7,500 units in GHMC. Considering just the wards won by TRS, this figure is around 5,000.

As Indian EVM's are totally different from those used elsewhere, the manipulators have to be built specially for India. The microcontroller, like most sophisticated components these days, is a surface mount device (SMD). Assembling the chip-on manipulator therefore requires access to an anti-static production facility with pick-and-place (P&P) system together with as reflow/wave soldering equipment.

Only the rotary switch & LED's are available off-the-shelf. The custom PCB (printed circuit board) and microcontrollers can be traced back to the buyer: something that no thief would like J

Assuming for the moment that thousands of devices are somehow assembled and the transactions are buried deep, we are still left with the programming issues. While burning an EEPROM (electrically erasable programmable Read-Only Memory) is quite easy, programming is not so easy.

The ballot positions of the major party candidates vary widely. The 150 GHMC wards resulted in 100 unique ballot positions. This means the party intending to steal the elections must plan several program combinations.

The ballot positions are the same inside all booths in a ward thus posing a different & difficult problem. An EEPROM typically store only a small amount of data (e.g. EVM EEPROM is 64 kb). This limited memory does not provide much room for complex/smart pieces of code. Unless a different program is used in every booth, the mischief will result in clearly identifiable trends that can result in invalidation of the result.

Programming the EEPROM is not a mere technical job. The trick is not to steal too few votes and lose the election or steal too many leading to raised eyebrows! In addition to this, the "stealing party" needs to balance booth wise variations. This calls for an accurate understanding of the mood not only at a ward level but also at booth level. A tall order given the poor performance of surveys and even exit polls J

The authors glibly claim "the device fits in a shirt pocket" omitting to mention the thief needs to carry a screw-driver too. Any thief with any knowledge of electronics would realize the device needs to be carried in a suitable anti-static bag. Otherwise he could end up damaging or inducing latent damage in the EVM thereby risking serious scandal. This leaves him with the problem of disposal of the bags.

I see no way the control unit can be used after the manipulator is inserted. The entire mischief must be conducted with the unit open. Reading the transcript of the alleged "demonstration" video, it appears that Prasad demonstrated the eyeball catching but useless "dishonest display" method.

What is the "stealing party" left with now?  The political crook needs to predict with a great deal of accuracy the vote stealing plan at thousands of booths. The technical accomplice would need to create thousands of programs, label each burnt EEPROM to avoid unintended results, pack it in an anti-static bag and create a "kit" for the heavy.

A semi-skilled thug needs to carry the "kit", surreptitiously break into the premises, identify the appropriate control unit, break the seal, open & switch the unit on, conduct the crime, close, shut down the unit and reseal taking care to leave no incriminating evidence like screws or the anti-static bag behind. Unless the miscreants take additional risks deploying several individuals, this needs to be repeated one after the other for each unit. All this calls for a great deal of organization: this is no lone wolf operation J

An audit trail is available even after this rigmarole. The memory manipulator can only work like a postscript leaving the original vote records intact.

Even ignoring the standing of the authors, slanted presentation and serious defects in the so called technical paper, it is clear that this is at best an amateurish attempt. There is no conceivable way this "experiment" can be scaled up to any reasonable degree.

The authors let the cat out the box by observing "in close races an attacker might be able to change the election outcome by tampering with only a few machines. A small number of tightly contested seats often determine which party holds a majority in the parliament, so a national-level attacker could focus on tampering with machines in these districts".

The present case can in no way be described as "tightly contested". As we saw earlier TRS won by 5,000+ majority in two-thirds of its wards. The party outscored all the other major parties put together in three-quarters of the wards it won.

Gerrymandering theories

A channel believed to support TDP came up with a gem in the immediate aftermath of the results. They attributed the TRS victory to gerrymandering claiming that boundaries were manipulated to transfer slums and areas with weaker sections to wards where TRS was perceived to be weak. We may note that the media & the cyberspace were abuzz with stories about officials (including Somesh Kumar, the then GHMC Special Officer) in the run-up towards the elections. The insinuation of misuse of official powers was clear though carefully couched in vague terms.

This allegation is based on the stereotype that rich & middle classes (presumably dominated by Hindu upper caste folks) tend to vote for NDA in great numbers while the weaker sections do not. If true this is a recipe for disaster as no party can hope to do even reasonably well if it depends only or primarily on the well-to-do people. There is no conceivable way NDA could have polled over a third of the votes in 2009 & 2014 if this was the case.

What is gerrymandering? One can define this as "manipulating electoral boundaries to favor one party or class". This practice is not uncommon in some countries with United States (US) topping the list. American courts have consistently stuck down racial gerrymandering but there is some consensus that partisan redistricting may not be illegal as long as racial exclusion is not the primary purpose.

Hyderabad has over 1,400 slums distributed across the city with the core city (old MCH area) accounting for the biggest chunk. The suburbs, conventionally held to be a TDP stronghold, have fewer slums.

Within the core city however the slums are fairly well spread out. Even areas like Banjara Hills where the super rich live have their share of slums. The reasons are not far too seek. Slums supply unskilled & semi-skilled labor to both the domestic & business sectors. Outside the old city, only a handful of wards are dominated by slum dwellers with the rest being hybrid in composition.

I found no perceptible difference between wards dominated by the weaker sections and their hybrid neighbors. For example, Adikmet (# 85) gave TRS a 6,350 vote majority while TRS won its better-off neighbor Musheerabad (# 86) by 4,121 votes. The case of # 103 Borabanda (4,511 votes) & # 102 Rahmathnagar (2,330 votes) is similar.

What will happen when ward boundaries are manipulated? The first thing that comes to mind is that it results in increasing the mal-apportionment i.e. certain "receiving wards" become bloated. Looking at the distribution of votes across constituencies, I find evidence contrary to this assumption. The skew has actually come down to a good extent in the last seven years. The table below summarizes the facts:

Parameter	2009	2014	2016
Average electors	38,004	309,320	49,491
Standard deviation	14,816	135,464	8,869
SD/Average	39.0%	43.8%	17.9%
Skew	123.0%	87.6%	76.9%

The only conceivable way of gerrymandering without increasing the skew is to continuously work on several permutations of transferring districts out & in: a foolhardy recipe if there ever was one!

Gerrymandering by definition requires both an accurate demographic profile and a predictable & consistent pattern of voting behavior. The fact that pollsters provide abundant & constant data on both these combined with the sharp differences in the major party platforms makes US an ideal candidate for gerrymandering. In India, none of these conditions are applicable. Our pollsters depend almost totally on statistical methods with very little emphasis on either profiling or voter positions. This is probably the main reason why their "forecasts" end up looking like crystal gazing rather than scientific predictions.

In India it is the Election Commission, an independent constitutional body, that determines constituency boundaries. This dramatically reduces the ability of the ruling party to influence the delimitation exercise. The Commission may depend on the bureaucratic machinery for information but even here the possibility for manipulation is not high.

Finally gerrymandering historically works in marginal fights. The present case is anything but one.

Incumbency factor

Quite a few "expert commentators" attributed TRS victory to the so called honeymoon. In my opinion this argument is not worth serious consideration. In general the "feel good" period runs a few months at a maximum. Twenty months is far too long for an electoral honeymoon. The fact that the very same worthies often castigated TRS on television about its alleged anti-people decisions lends very little credence to their conclusions.

A variation of the honeymoon theme is the claim "ruling parties often tend to win local polls". This is clearly not true as much contrary evidence shows.

A case in point is the TDP performance between 1985-87. The party won the assembly elections held in March 1985 in a landslide. However it lost considerable ground when the mandal elections were held almost exactly two years later. Within a few months it lost badly in the municipal polls.

Hyderabad has not been immune to this trend either. During the 1987 MCH elections, TDP resorted to widespread rigging, voter intimidation & other malpractices but struggled to achieve joint second largest party position. The ruling NDA won the 2002 MCH mayoral elections with a slim majority while Majlis emerged as the largest single party. Congress fresh from its 2009 assembly elections victory managed to win the maximum seats in the GHMC elections later that year but failed to cross the 30% vote mark.

In any case this theory assumes the ruling party retains the support it received earlier. As we saw TRS vote share saw a dramatic positive swing. Mere retention would not have fetched the party a victory, much less a landslide J