2020 India assembly elections - Delhi, Bihar

[jaichind]

India Today-Axis My India exit poll results by community

                  NDA       UPA        LJP        GDSF (RLSP-BSP-AIMIM)
Muslims         9          79           1            7
Mahadalit      43         37           8            6
Paswan         31         31         30            2
EBC              59         22           8            5 
Yadav           10         83           1            1
Kurmi           66         20           4            4
Koeri            52         26           7            9
Other OBC    57         26           7            3
Tribal           53          27           8            2
Baniya          64         19           6            2
Bhumihar      54         19         17            2
Rajput          54          17        13            3
Brahmin       62          17          8            2
Other Upper  62          21          8            2

Overall         39          44           7            4

India Today-Axis My India pointed out that

a) The BJP Upper caste base was not enthusiastic about it alliance with JD(U).   Bhumihar and Rajput defected in large number to LJP
b)  Tejashwi Yadav's strategy seems to be Muslim+Yadav+Youth of all castes which worked out well
c) RJD's alliance with the Left was effective in bring in Mahadalit and EBC votes
d) LJP clearly did well with Paswans as expected and denied that vote to NDA
e) JD(U)'s attempt to corner the Mahadalit vote failed in face of Tejashwi Yadav's youth appeal and Left's effective consolidation of its Mahadalit base behind UPA
f) RLSP failed to corner the Koeri vote

[jaichind]

My exit poll based algorithm for predicting results

Again my algorithm (adjusted somewhat) of converting exit polls to seat projection

Again the algorithm is
1) First un-skew Todays Chanakya's exit polls to adjust for its pro-BJP house affect
2) Look at pollsters which did pre-election surveys and exit polls to see if all/most of them point to a particular party is doing better in the exit polls relative to pre-poll survey.  Then
  a) If Yes, then the most optimistic of the exit polls for said party and use that as the projection
  b) If No, then just use the average of exit polls as the projection
3) Is the ruling party/front being defeated using the projection from 2)
 a) If Yes, add some extra seats to the winning opposition party/front
 b) If No, leave result from 2) alone
4) If the winning party is BJP then repeat starting at step 2) WITHOUT un-skewing Todays Chanakya's exit polls
 

                                   NDA       UPA        Others
Today's Chanakya           50         180           8
India Today-Axis             80         150           4
Democracy Times           99         135           9
Republic-Jan Ki Baat      104        128          12
TV9 Bharatvarsh            115        120           8
ETG                              114        120           9
Times Now-C Voter        116        120           7
PollDiary                       112         98          33
Dainik Bhaskar              125         77          41
-------------------------------------------------------------
Average                        102       125          16

The only exit poll that did a pre-election poll was Times Now-C Voter

                                    NDA        UPA        Others
Pre-election                   147          89             7
Exit                               116         120            7

So the momentum is significantly behind UPA.  There is no need to throw out Today's Chanakya in this case since they do not show signs of a pro-BJP bias.  This means we should NOT go with the average but go with the more pro-UPA polls.  I will go with the second most pro-UPA poll which is India Today-Axis

                                   NDA         UPA        Others
India Today-Axis             80         150           4

But since this is a defeat of an incumbent you have to shift more seats toward the winning opposition bloc which gives us

                                   NDA         UPA        Others
India Today-Axis             80         150           4
Shift                              -8          +8
-------------------------------------------------------------
Projection                      72          158           4

So my exit poll based projection algorithm projects a massive UPA victory with around a 2/3 majority.  Totally unexpected at the beginning of the campaign when I thought it would be NDA by a tiny margin or UPA by a tiny margin and that was one of the more pro-UPA projections out there.

[jaichind]

Its interesting that the two exit polls that has NDA with an edge, PollDiary and Dainik Bhaskar, has Others at 33 and 41 respectively.  So their exit poll results must have LJP and GDSF overperforming and eating into both the NDA and UPA vote shares and in the resulting 4 way chaos NDA ends up with more seats as GDSF hurts UPA more than LJP hurts NDA. 

Assuming that this election is an anti-incumbency election the anti-incumbent vote tends to get consolidated so this scenario is not likely.  That was the gamble RJD made when they refused to accept HAM and RLSP seat requests and let them leave UPA on the premise that the anti-JD(U) vote will consolidate behind UPA.   As long as these two exit poll results does not take place it seems Tejashwi Yadav and RJD most likely made the right gamble.

[jaichind]

If Nitish Kumar does lose the election as the exit polls predicts then the start and end of his reign would all rotate around the LJP factor.  Nitish Kumar would have rode into power based on the LJP and now lose power based on LJP.

In the 2004 LS elections a grand RJD-INC-LJP alliance was formed and defeated JD(U)-BJP.  LJP was a JD(U) splinter that stayed in NDA but bolted from NDA and joined RJD-INC to protest the  2002 Gujarat riots under then CM Modi.  The 2004 RJD-INC-LJP victory in the 2004 LS election was a shock since it was assumed that the JD(U)-BJP would repeat their 1999 LS election and on anti-incumbency against the Bihar RJD state government.

As the 2005 Bihar assembly election approached it seems that as long as RJD-INC-LJP alliance held together they were certain to win re-election.  But it was at this stage that LJP had other ideas as it wanted to carve out a space for itself and not play second fiddle to RJD.  So LJP, while staying a part of the UPA government at the center, ran separately from RJD.  The Bihar INC did nothing to stop this as it also wanted to cut RJD down to size and formed tactical alliances with both RJD and LJP on the premise that there is no way JD(U)-BJP can win so a post election government with a weakened RJD is in the best interest of the INC.

The Feb 2005 Bihar assembly election threw up a fractured mandate with RJD INC and LJP a majority between them but with LJP holding the balance of power between RJD-INC and JD(U)-BJP. LJP negotiated  with both sides  for so long that a good bloc of the LJP MLAs defected to JD(U) to try to break the deadlock.  The pro-INC governor of Bihar refused to accept this defection and called for another Oct 20015 Bihar assembly election.  INC this time around saw the danger of a JD(U)-BJP victory and allied with RJD while LJP ran separately hoping to play kingmakers again.  The result was a resounding victory for JD(U)-BJP and the start of a 15 year Nitish Kumar reign in Bihar.  

Now 15 years later it seems the BJP was trying to play the same game INC did back in 2005 with the LJP but with the goal of cutting JD(U) down to size by tactically accepting LJP running to split the JD(U) vote while LJP stays with the NDA government at the federal level.  The BJP was certain that after the 2019 JD(U)-BJP-LJP landslide victory in the LS elections there is no way the RJD-INC could win the 2020 assembly elections.  Just like 2005 it seems it is about to blow up in the BJP's and JD(U)'s face.  So Nitish Kumar's rise to power starts with a LJP defection and is likely to end with a LJP defection.

[jaichind]

Is there normally that much difference in exit polls?

This is par for the course, especially for Bihar.  In 2015, for example, the exit polls were 



Worse, there was an exit poll that had JD(U)-RJD-INC 170 NDA 64 which ended up being accurate that was not published because CNN-IBN did not believe the result and pulled the poll from its TV program.

The number of parties and alliances involved, many races being multi-polar, and different swings bases on caste makes Bihar very hard to poll and project.

[jaichind]

Most exit polls that has detailed breakdown by party has RJD-INC Maoist ally CPI(ML) at around an amazing 12 !! seats with CPI and CPM with 1-2 seats each.  This would be the best electoral performance of a Maoist party in an Indian election if this ends up being true.

If next year's WB assembly goes the way one would expect, then in 2021 you would have the Left having more seats in Bihar than WB.  If anyone had predicted this in the mid 2000s that person would have been put in an insane asylum.

[jaichind]

There were 7 by-elections in UP assembly.  In 2017 UP assembly elections BJP won 6 of the 7 and SP won the last one.  India Today-Axis My India had an exit poll for these 7 seats which was

            seats    vote share
BJP         5-6      37%
SP          1-2      27%
BSP        0-1      20%
INC           0        8%

India Today mostly presented it as the BJP retaining their vote base.  Not sure if I agree.  If you look at these 7 seats in 2017 and mapped out their vote shares it was

BJP        44.8%
SP-INC   22.6%
  SP          19.9%
  INC          2.7%
BSP        24.2%
RLD          2.4%
NISHAD    3.5%

NISHAD has since supported BJP.   In 2017 SP and INC ran as allies but this time are running separately.   BSP is breaking with tradition and taking part in by-elections.  If the exit poll vote shares are correct both BJP and BSP have lost ground to SP and INC.  The erosion of the BJP vote since 2017 is quite significant and I would not read this as good news at all if I were the BJP.

[jaichind]

NDTV analysis of exit polls

Gender gap with Men voting for UPA and Women for NDA (mostly has to do with Nitish Kumar's policy of prohibition).  Good news for NDA is women turnout is higher.  Note that men register a a higher rate then women so more men votes were cast than women




BJP is expected to lose ground relative to 2019 LS elections



A 17% swing against BJP which is the average so far in 2019-2020 assembly elections leads to UPA 150 NDA 80 which looks suspiciously to my projection



When Modi is on the "ballot" BJP does better, even in assembly elections



Key voting pattern by community



Where Muslims are more numerous (RJD and INC base)



Where Yadav are more numerous (RJD base)



Where Paswans are more numerous (LJP base)



Where Kurmis are more numerous (JD(U) base)



Where Upper Castes are more numerous (BJP base)

[jaichind]

If RJD does win it seems it has to do with a whirlwind of rallies by Tejaswi Yadav.  In a 20-day campaign,  Tejaswi Yadav completed 247 rallies which averages to over 10 rallies in one day.  The size of the rallies were huge (and mostly youth) and only got bigger as the campaign went on

Sample of what the rallies looked like
https://www.youtube.com/watch?v=xrKef_XI0RA

No masks or social distancing here.  Doctor Fauci  would not approve

[jaichind]

Any chance we see a "red mirage" type effect in today's elections?

See my old writeup on how vote counts work

Jaichind, can you explain to us exactly why the "winning party's seat count keeps growing as the counting goes on" phenomenon occurs? That was clearly the case tonight, since early seat counts had UPA flipping 50 seats or so from the NDA, and almost all those flips evaporated as counting went on. I have no idea why that would happen, though. I could understand if the counting bias always favored the UPA early on and the NDA always grew as the night went on, but you seem to suggest that whatever party is winning will be underestimated early on, which I can't think of a mechanism for.

The way to explain is that the way the count works is they count a block of precincts, report results then count another bloc of precincts and then report the combined result.  Because of that the count is NOT random and will have clear biases.  The bias is random but large.  So in the earlier round the vote share could be off as much by, say 10%, or even greater from the real final vote share.

Now, lets imagine an election where Party A defeated Party B 55% to 45% in overall vote share.  Lets assume the vote share per seat is a bell distribution around the 55/45 A vs B point.  So the real result will have party A winning 75% of the seats.  Now in the early round each result will be randomly add 10% or subtract 10% from A's vote share.  Say in half the seats A's vote share is artificially reported as 10% higher than it is and half of the seats A's vote share is artistically reported as 10% lower than it is .  Now, since A is beating B by 10% on average in the seats that we artificially add 10% to A's vote share does nothing to the reported leader of the seat since most likely A was already ahead in the real result anyway. But in the half the seats where we artificially reduce the reported vote share of A then it could lead us to report that B is ahead.  Now as the real results do come in as more and more precincts are counted the true vote share of that seat is revealed reverting a lot of B reported leads as really A leads.

Now if the election is 50/50 or neck-to-neck between A and B then this random shift does nothing to the seat count on average.  That is why I said

Counting will start 8am Delhi time (10:30PM EST).  The way Indian count works there is always a bump for the landslide winner at the very end.  So if early in the count we see NDA below a majority then the result will most likely be NDA without a majority or NDA with a narrow majority.  If early in the count we see NDA with a comfortable majority then it will be a mega NDA landslide in the end. Rarely do the side that is losing suddenly pull ahead or reduce the winner's margin of victory at the end of the count.

Is true. If it is landslide then  as the count goes on the landslide lead gets greater. if it is neck-to-neck then it stays neck-to-neck the whole time

I noticed this watching Indian election returns years ago and after some thinking figured out why.  This is why I focus on the early vote share reports on ECI by state as a way to figure out what the final seat count is going to be.

[jaichind]

Any chance we see a "red mirage" type effect in today's elections?

This took place in 2015 where the early count had a BJP-LJP-RLSP-HAM landslide victory over JD(U)-RJD-INC but then later in the day it went the other way around.  The networks actually started to talk about what the BJP did right to win only to change their tune later in the day.

[jaichind]

Very early count (not really that relevant and very VBM heavy which tend to lean BJP) has

Leads from 87 out of 243 seats

UPA       44 (+10)
NDA      38 (-13)
LJP         2 (+2)
Others    2 (--)

JD(U) taking a lot of losses so far

[jaichind]

Still very early (mostly postal VBM)

UPA       74 (+19)
NDA      56 (-22)
LJP         2 (+2)
Others    4 (--)

postal votes usually favor BJP.  This is not looking good for NDA.  BJP actually gaining seats while JD(U) taking massive losses.

[jaichind]

With about half the vote counted the NDA has a surprising narrow lead.  Looks like the exit polls had a miss again.  The result are actually what was expected w/o the exit polls

NDA         131(+6)
  BJP           73(+20)
  JD(U)        50(-21)
  VIP             6(+6)
  HAM           2(+1)

UPA          102(-8)
 RJD           66(-14)
 INC           18(-9)
 CPI(ML)    11(+8)
 CPM           4(+4)
 CPI            3(+3)

LJP            2(--)

GDSF        4(+2)
 RLSP         0(-2)
 BSP          2(+2)
 AIMIM      2(+2)

IND           4(--)

vote share wise which I have to compute seems to be neck-to-neck so I suspect it might get closer from here but we will have to see

[jaichind]

The vote share so far are (with NOTA not stripped out)

NDA         37.57%
  BJP          19.70%
  JD(U)       15.33%
  VIP            1.78%
  HAM          0.76%

UPA         36.54%
  RJD         22.93%
  INC           9.43%
  CPM(ML)    2.95%
  CPM          0.62%
  CPI           0.61%

LJP             5.77%

So neck-to-neck vote share wise.

[jaichind]

Seat wise it is now (a bit more than half the vote counted)

NDA         131(+6)
  BJP           75(+22)
  JD(U)        48(-23)
  VIP             5(+5)
  HAM           3(+2)

UPA          101(-9)
 RJD           63(-17)
 INC           21(-6)
 CPI(ML)    11(+8)
 CPM           4(+4)
 CPI            2(+2)

LJP            1(-1)

GDSF        6(+4)
 RLSP         0(-2)
 BSP          2(+2)
 AIMIM      4(+4)

IND           4(--)

Strong AIMIM performance seems to indicate that AIMIM did cut into the UPA Muslim vote

[jaichind]

Updated vote share so far are (with NOTA not stripped out) with around 55%-60% of the vote counted.  Gap getting closer between the two blocs

NDA         37.32%
  BJP          19.62%
  JD(U)       15.16%
  VIP            1.76%
  HAM          0.78%

UPA         36.70%
  RJD         23.01%
  INC           9.29%
  CPM(ML)    3.13%
  CPM          0.66%
  CPI           0.61%

LJP           5.72%

GDSF       4.82%
   RLSP        1.97%
   BSP          1.68%
   AIMIM      1.17%

[jaichind]

Seat wise it is now also getting closer

NDA         126(+6)
  BJP           76(+23)
  JD(U)        42(-29)
  VIP             5(+5)
  HAM           3(+2)

UPA          107(-9)
 RJD           70(-10)
 INC           18(-9)
 CPI(ML)    12(+9)
 CPM           4(+4)
 CPI            3(+3)

LJP            1(-1)

GDSF        6(+4)
 RLSP         0(-2)
 BSP          2(+2)
 AIMIM      4(+4)

IND           3(-1)

[jaichind]

Due to COVID-19 there were a lot more voting booths which means the count is much slower.  Usually by this time around 80%-90% of the vote would have been counted but right now it seems to be around 55%-60%

[jaichind]

BJP is sweeping various assembly by-elections

In MP mini-assembly election it is sofar
 
                  seats     vote share(w/o NOTA stripped out)
BJP              18             50.76%
INC               7             40.20%
BSP               1              4.86%


Similar story in Gujarat where just like MP a bunch of INC MLAs defected to BJP leading to by-elections where the INC MLAs run as BJP candidates

                  seats     vote share(w/o NOTA stripped out)
BJP                8            54.75%
INC                0            34.52%


In Bihar LS by-election it is neck-to-neck between INC and JD(U)

JD(U)    37.70%
INC       35.39%
BPP       11.20%  (AAP like youth party)
RLSP       5.06%


In UP the BJP wins most races but loses ground in terms of vote share

                  seats     vote share(w/o NOTA stripped out)
BJP                 5            35.79%
BJP rebel                         2.55%
NISHAND rebel                6.28%   
SP-RLD           2            24.84%
BSP                0             18.36%
INC                0               7.64%

[jaichind]

Seat wise it is now also getting even closer as foretold by close vote share count so far

NDA         124(-1)
  BJP           75(+22)
  JD(U)        41(-30)
  VIP             5(+5)
  HAM           3(+2)

UPA          110(-2)
 RJD           72(-8)
 INC           20(-7)
 CPI(ML)    11(+8)
 CPM           4(+4)
 CPI            3(+3)

LJP            0(-2)

GDSF        6(+4)
 RLSP         0(-2)
 BSP          2(+2)
 AIMIM      4(+4)

IND           4(--)

[jaichind]

Updated vote share so far are (with NOTA not stripped out) with around 60% of the vote counted.  Gap getting closer between the two blocs

NDA         37.32%
  BJP          19.58%
  JD(U)       15.15%
  VIP            1.75%
  HAM          0.84%

UPA         36.85%
  RJD         23.11%
  INC           9.30%
  CPM(ML)    3.18%
  CPM          0.64%
  CPI           0.62%

LJP           5.63%

GDSF       4.79%
   RLSP        1.97%
   BSP          1.64%
   AIMIM      1.18%

[jaichind]

Seat gap getting closer with RJD overtaking BJP as largest party (for now)

NDA         121(-4)
  BJP           71(+18)
  JD(U)        42(-29)
  VIP             5(+5)
  HAM           3(+2)

UPA          113(-2)
 RJD           74(-6)
 INC           21(-6)
 CPI(ML)    11(+8)
 CPM           4(+4)
 CPI            3(+3)

LJP            0(-2)

GDSF        6(+4)
 RLSP         0(-2)
 BSP          1(+1)
 AIMIM      5(+5)

IND           3(-1)

[jaichind]

Updated vote share so far are (with NOTA not stripped out) with around 65% of the vote counted.  Gap getting slightly closer between the two blocs

NDA         37.21%
  BJP          19.51%
  JD(U)       15.10%
  VIP            1.74%
  HAM          0.86%

UPA         36.85%
  RJD         23.25%
  INC           9.20%
  CPM(ML)    3.14%
  CPM          0.67%
  CPI           0.59%

LJP           5.61%

GDSF       4.80%
   RLSP        1.95%
   BSP          1.63%
   AIMIM      1.22%

[jaichind]

Seat gap getting closer with RJD overtaking BJP as largest party (for now).  NDA no longer has majority.

NDA         121(-4)
  BJP           73(+20)
  JD(U)        40(-31)
  VIP             5(+5)
  HAM           3(+2)

UPA          114(-1)
 RJD           77(-3)
 INC           20(-5)
 CPI(ML)    11(+8)
 CPM           4(+4)
 CPI            2(+2)

LJP            0(-2)

GDSF        6(+4)
 RLSP         0(-2)
 BSP          1(+1)
 AIMIM      5(+5)

IND           2(-2)