Bihar Election: What Opinion Polls Can Tell Us & What They Can’t

Renowned psephologist Sanjay Kumar writes about unreasonable expectations people have from opinion polls.

Bihar Election 2020: It is absolutely unreasonable to expect pre-poll survey to be a tool for correctly forecasting the voting day outcome.

(Photo: The Quint)

No beating around the bush: yes, conducting a pre-poll survey is challenging, whether it is in Bihar or any other state. It is, however, far more challenging to conduct a survey in Bihar than in many other states.

It is difficult to conduct pre-polls surveys due to multiplicity of political parties, shifting of alliance partners, and multi cornered contests. What makes the process more complicated is the presence of large number of castes and a very strong caste-based voting.

A much bigger challenge for any pre-poll survey is meeting people’s expectation, what people expect a pre-poll to tell you.

What has added to the challenge of conducting an opinion poll in present times is the COVID-19 pandemic, which makes it mandatory for the field investigator and respondent to follow various precautions, maintain distance, wear mask – all of which in some manner dampens the exercise. Travelling long distance has been very difficult, so surveyors needed to arrange for local investigators for conducting the field work.

• It is far more challenging to conduct a survey in Bihar than in many other states.
  
• Whatever CSDS estimates are, they can’t help us in predicting the final result when votes are counted on 10 November.
  
• Bihar has diverse population and a large number of political parties.
  
• Difficult to calculate “swing” because of changing alliances.
  
• No pre-poll survey can account for the impact of rallies and other campaign elements on voting choices.
  
• Pre-poll surveys only capture political preferences at the time of data collection.
  
• The media disseminate pre-poll findings as gospel truth and the final result.
  
• It is, therefore, absolutely unreasonable to expect pre-poll survey to be a tool for correctly forecasting the voting day outcome.
  

What CSDS Bihar Survey Tells US

Our most recent pre-poll survey in Bihar estimates:

• 38% votes for the NDA,
• 32% for the RJD+,
• 7% for the RLSP+
• 6% for the LJP, and
• 17% for other smaller parties and independents.

Based on these vote-share estimates, we calculate the seats for

• NDA to be in the range of 133-143,
• RJD+ to be in the range of 88-98,
• LJP to be in the range of 2-6 and
• Others between 6-10.

Remember, these are estimates of vote share based on the prevailing voting preference of the voters of Bihar when the pre-poll survey was being conducted, between 10-17 October. It can’t help us in predicting the final result when votes are counted on 10 November.

Voting is still almost two weeks away; people will cast their vote on 28 October, 3 November, and 7 November. Prime Minister Narendra Modi is still to campaign in Bihar for NDA and his campaign is likely to make an impact on people’s voting choices, as has been witnessed in other states during last few years.

Challenges in Assessing Voters’ Mood in Bihar

Larger the number of political parties, bigger the challenge for estimating votes from a diversity of voters. Bihar has both these traits: the population is diverse and voters belonging to different castes have different voting preferences. In fact, some parties are even identified as caste parties in Bihar.

There is a large number of regional political parties with limited support base in specific districts and regions, making it extremely difficult to estimate the vote share.

The seat estimate for a party is worked out based on the vote estimate, so if the vote estimate is inaccurate, there are enormous chances of inaccurate estimate of seats for parties and alliances.

The estimate of seats is calculated by especially devised program using what is referred to as “swing model”. It estimates/projects seats based on the swing of votes in favour of or against any political party or alliance. It uses the votes polled by the party or the alliance during the last election as the base, and estimate of votes based on current survey is applied to this base. A positive swing in favour of an alliance or party is expected to indicate larger number of seats, while a negative swing is expected to suggest reduction in seats.

Fast Changing Alliances in Bihar Perplex Psephologists

Assessing vote share in a multi-party contest is hard, but what becomes even more challenging is working out the base vote of alliance if alliance changes between two elections as in the case of 2020 Bihar Election.

During the 2015 assembly elections JD(U) and RJD were in alliance with the Congress, while the BJP allied with the LJP. In 2020, JD(U) is in alliance with the BJP and the LJP is contesting separately. The BJP has formed a sub-alliance with Vikassheel Insaan Party (VIP) and JD(U) has done the same with Hindustani Awam Morcha (HAM).

This is too complicated to assess the current alliance based on the vote share of these parties in 2015 when some of them contested against each other.

Patna: Bihar Chief Minister and Janta Dal United (JDU) national President Nitish Kumar (Photo: IANS)

Furthermore, if there is split in a party it poses challenge for estimating votes for the two factions of the same party based on their vote share during last elections. What proportion of votes should be assigned to which faction? The merger of two or more parties also poses a similar challenge as there may be some voters who may not want to vote for the new parties formed.

Campaign Is a Strong Factor for Final Outcome

Campaign has its own influencing role, political parties and candidates do try to mobilise voters through public rallies, road shows, et. al. Some candidates and parties are able to mobilise voters very successfully, some not so much. Voters are known to change their voting decision based on political campaigns, some on the last day and few just at the time of casting their vote.

Prime Minister Narendra Modi (Photo: PTI)

No pre-poll survey can account for the impact of rallies and other campaign elements on voting choices. If that is taken into account and added to the estimates from the actual survey, it would not be doing justice to the opinion poll process.

While these challenges could be overcome by using statistical tools, there is hardly a tool to handle the burden of people’s expectations. What people expect from pre-poll surveys is only forecast: who will win election, which party will get how many seats etc. Most people don’t even bother about the vote estimates, which is the main element that any pre-poll survey captures.

Why People Expect Opinion Polls to the Predict Final Outcome

Pre-poll surveys are expected to tell the counting day results assessed from people’s political preference obtained weeks before election, without discounting for the fact that opinions, voting preference may change in the meantime. Whatever is indicated by the pre-poll survey is taken as the ultimate truth, which it is not. Pre-poll surveys only capture political preferences at the time of data collection.

It should be possible for any pre-poll survey to predict election correctly – or very close to that – if things remain the same as they were at the time of collecting data. This will mean minimising the effect of campaign.

The problem does not lie only with people who watch these opinion polls. The agencies or the media which disseminate pre-poll findings as gospel truth and the final result. This packaging is responsible for creating unreasonable expectations.

It is, therefore, absolutely unreasonable to expect pre-poll survey to be a tool for correctly forecasting the voting day outcome.

(Sanjay Kumar is a Professor at Centre for the Study of Developing Societies (CSDS). He is also a well known psephologist and a political Commentator. This is an opinion piece, and the views expressed are the author’s own. The Quint neither endorses nor is responsible for them.)