Satvik Dev, Sonipat, 16 July 2024
The year 2024 has been regarded by many international observers as a tipping point that will profoundly influence the lives of more than two billion people. With roughly 50 nations scheduled to hold elections this year, debates have been happening regarding the significance of these votes (Shamim & Lodhi, 2024). Amid these rising concerns, attention naturally had to be focused on the elections in the world’s largest democracy – India. What made this year’s polls particularly defining was the widespread criticism of India’s perceived fragility as a democracy, which placed its democratic setup at stake. This was further exemplified by the conflicting visions professed by the two national parties.
At the outset, the verdict appears to clearly reflect the general discontent expressed by voters regarding the erosion of civil liberties under BJP rule. The purposefully crafted campaigns of opposition parties, led by the Congress, seem to have resonated with the intrinsic fears among marginalized groups and minorities. These fears concern the potential dismantling of certain constitutional principles that have guided India’s values since independence (Aiyar & Sircar, 2024). For others, their vote represented a protest against the exclusionary power system, and this sentiment seems to have reverberated effectively (Aiyar & Sircar, 2024).
However, while a multitude of factors can be articulated to explain the results, it is simply impossible to overlook the astonishing inaccuracy of the exit polls that transpired. While BJP couldn’t even cross the majority mark on its own, most agencies predicted a landslide victory for the BJP, with many suggesting a comfortable 350+ seat share for the NDA alliance (Frontline News Desk, 2024).
While exit polls in India have historically demonstrated a spotty track record—in 2004, for instance, they erroneously forecasted a comfortable majority for the BJP-led alliance that ultimately did not materialize—their accuracy improved markedly in subsequent years. Specifically, in 2014 and 2019, exit polling data proved to be relatively reliable, providing projections that closely mirrored the actual election outcomes (Frontline News Desk, 2024).
Given this record, it is imperative to direct efforts toward understanding what precisely went awry. At the surface level, some may attribute the discrepancies to bias within the polling agencies, even going so far as to allege deceitful motivations, such as the use of the same sample across different polls. While it is true that competition among these agencies can lead to a rush to broadcast predictions, occasionally catering to specific interests, undoubtedly, more complex factors are at play.
First and foremost, it is necessary to thoroughly familiarize ourselves with the entire methodology employed in the conduction of exit polls. The primary motivation is to estimate the vote shares for different parties and alliances in a constituency by conducting structured interviews with selected participants. The data thus collected is subjected to further rigorous analysis. In this context, the critical importance of a large and, more importantly, representative sample becomes evident – one which can reliably capture the overall mandate (Kumar, 2024).
Often, it is precisely at this critical juncture where the frameworks of many polling agencies tend to falter. In a vast parliamentary democracy like India, with its 543 constituencies, the task of accurately measuring voter intent is herculean and inherently fraught with potential pitfalls. Conducting exit polls across such a wide expanse is a monumental undertaking, with a high likelihood of encountering various shortcomings. One significant issue is the potential for interviewers to record false responses, either inadvertently or intentionally. Additionally, the willingness of respondents to participate and provide truthful answers is frequently called into question (Kumar, 2024). As Yashwant Deshmukh, the founder of CVoter, recently emphasized post-elections, sample sizes and budget constraints are major impediments to the accuracy and reliability of polling data (Frontline News Desk, 2024).
Furthermore, what significantly adds to this challenge is the herculean task of predicting seat distribution results for each constituency. This process relies on a swing model, which utilizes previous vote shares of the respective constituency to make projections. Instantly, one can identify a series of shortcomings that may pervade this process. Inconsistencies can arise during the calculation of vote share distribution, and additional factors such as changes in political alliances, splits or mergers of parties, and the interplay of multiple parties can severely complicate this labour-intensive and time-consuming endeavor (Kumar, 2024). It has been highlighted numerous times how precariously deceptive this entire process can be when the final calculation of seats is based on the comparison of previous and present-day vote shares of each party. It is essential to understand that vote share does not directly correlate with seat count.
An increase or decrease in vote percentage does not necessarily translate to a corresponding number of seats won or lost. Gaining votes in less competitive states might not result in additional seat wins, while losing votes in highly contested states could lead to significant seat losses. For instance, a decline of approximately 3% in vote share for the BJP in Bihar, from 23.58% in 2019 to 20.52% in 2024, resulted in the loss of five seats. In Maharashtra on the other hand, despite a decrease of approximately 1.4% in the BJP’s vote share compared to 2019, the party experienced a substantial loss, securing only about half the seats it had previously won (Verma, 2024). Deshmukh also highlighted that while his agency accurately predicted the share of votes won by parties at the state level, the projections for seat numbers were incorrect, particularly in states like Uttar Pradesh, Rajasthan, and West Bengal, where the BJP experienced significant losses in support. Conversely, in states where the BJP secured sweeping victories, such as Madhya Pradesh, Bihar, and Gujarat, the seat projections were accurate (Frontline News Desk, 2024). A comparison between Rajasthan and Uttar Pradesh reveals the complexities involved: a 9.83% decline in vote share in Rajasthan led to a loss of 10 seats, whereas an 8.61% drop in vote share in Uttar Pradesh resulted in a substantial loss of 29 seats. This is in stark contrast to Himachal Pradesh, where despite a massive 13.27% decline in vote share, the BJP managed to retain all its seats (Verma, 2024).
This variability underscores the significant fluctuations in seat distribution that can occur despite similar vote shares from previous elections. In light of this, new methodologies are constantly being developed and implemented to achieve more promising results. For instance, KCore Analytics, a research firm led by Hernán Makse, utilizes artificial intelligence (AI) derived from people’s social media interactions—what they are reading, writing, and responding to on the internet—to predict voter preferences. The analysis also incorporates crucial data such as inflation, which can influence election outcomes. The underlying premise is that AI provides a clearer understanding of people’s preferences by leveraging the anonymity of the internet (Frontline News Desk, 2024).
In this dynamic environment, people also have come to rely on exit polls increasingly for their predictions about seat distribution, which has become the primary measure of their success. Indian television channels also commission and sponsor these agencies to cater to their audience’s demands. Consequently, pollsters operate in a highly competitive market, much like forecasters in economics and weather, and face inherent uncertainties. Their business model relies not only on the accuracy of their forecasts but also on how they compare to those of other pollsters (Lahiri, 2024). As a result, pollsters may sometimes base their forecasts on those of their competitors, which may not necessarily reflect the underlying realities (Lahiri, 2024). It is crucial in this context therefore to emphasize the fundamental purpose of these comprehensive poll studies i.e., to understand the voting behavior of people from diverse socio-cultural and economic backgrounds by determining the vote shares of parties and comparing them to previous polls. Any poll that focuses solely on projecting seat distribution is more akin to an estimate poll that would just lack the data and research that goes into the spirit of a comprehensive exit poll (Kumar, 2024).
References
Shamim, S., & Lodhi, A. (2024, January 4). The year of elections – Is 2024 democracy’s biggest test ever? Al Jazeera.
https://www.aljazeera.com/news/2024/1/4/the-year-of-elections-is-2024-democracys-biggest-test-ever
Sircar, Y. A. (2024, June 5). Election results 2024: How India read the election, how the BJP reads the result. The Hindu. https://www.thehindu.com/elections/lok-sabha/how-india-read-the-election-how-the-bjp-reads-the-result/article68252169.ece
Frontline News Desk. (2024, June 11). How India’s exit polls got the 2024 Lok Sabha election horribly wrong. Frontline. https://frontline.thehindu.com/news/exit-polls-were-wrong-lok-sabha-election-2024-bjp-narendra-modi-amit-shah-coalition-bjp-led-alliance-rahul-gandhi/article68262575.ece
Kumar, S. (2024, June 1). An Expert Explains: How are exit polls conducted, and how should they be read? The Indian Express. https://indianexpress.com/article/explained/exit-polls-expert-explains-lok-sabha-2024-9361322/
Verma, L. (2024, June 6). In BJP’s slight fall in vote share, Rajasthan saw sharpest dip of 10%, UP close second. The Indian Express. https://indianexpress.com/article/political-pulse/bjp-lok-sabha-election-results-vote-share-9375065/
Lahiri, A. (2024, June 15). Why the exit polls couldn’t get the nation’s pulse in 2024. The Indian Express. https://indianexpress.com/article/opinion/columns/why-the-exit-polls-couldnt-get-the-nations-pulse-in-2024-9393392/
