## HOW TO PREDICT AN ELECTION OUTCOME

Predicting the outcome of an election is not easy, as many political pundits have found to their chagrin.
Published February 4, 2024

In our discourse up till now, we have explored individual motivations to vote, how these motivations translate into votes, vote banks, political parties and party clusters. As a continuation to this, we are reminded of an important question: can we predict electoral results? If so, how?

Building up on our earlier discussion, we seek to address this question by introducing our simple model called the Gallup and Gilani Model of Electoral Calculus. The calculus comprises an interplay of four factors which together determine the direction of the electoral outcome.

The first of the four factors is the size of the vote bank of a given party and the nature of its distribution across Pakistan’s 12 electoral territories. The second is the size of the electoral territories where its vote bank is concentrated. The third is the nature of the electoral contest in the electoral territories in which the given party is an important player; we define this phenomenon as the polarity level of an electoral contest. The last of the four factors is the set of rules which convert votes into seats in parliament. All four factors have been explained in detail in the rest of this chapter.

The Gallup and Gilani Model of Electoral Calculus enables us to execute a complex interplay of the four factors which can determine the direction of the electoral outcome. Once a model has done its work, an expert with domain knowledge of the history of elections in different parts of Pakistan is greatly facilitated to interpret the numerical findings.

In an odd sense, data and the model serve as the reports produced by a clinical laboratory whose results require to be interpreted by a specialist doctor in the given specialisation.

Predicting the outcome of an election is not easy, as many political pundits have found to their chagrin. Predicting it in a country as diverse and complex as Pakistan is even more difficult. Dr Ijaz Shafi Gilani is one of the country’s leading polling experts, founded Gallup Pakistan, and has been grappling with election-related data for decades. Eos is presenting here, with permission, an excerpt from his book The Ritual of Elections in Pakistan (1970-2018): A Process Without A Product, published less than two weeks ago by Lightstone Publishers…

FACTOR 1: SIZE OF PARTY VOTE BANKS

In an earlier chapter, we identified five prominent voting clusters in Pakistan: the PPP Cluster, the Muslim League(s) Cluster, the Religious Parties Cluster, the Regional Parties Cluster and the PTI Cluster.

While comprehending these clusters is crucial, it is equally imperative to examine individual parties in isolation and assess their historical performance and significance within Pakistan’s electoral landscape.

It is noteworthy that there are approximately 17 parties with significant vote banks, each having secured roughly 1 percent or more of the total votes cast in all elections since 1970 – with PPP at the top with 27 percent of votes received. In total, these parties have accumulated around 300 million votes.

The Gallup and Gilani snapshot of Pakistan’s electoral history outlines the historical size of the vote banks of all political parties of Pakistan, which have contested any election since 1970. The list is ranked in terms of the total votes scored by the party during the entire period. Those interested may access further details through a Snapshot on the Electoral History of Pakistan.

FACTOR 2: SIZE OF THE 12 ELECTORAL TERRITORIES OF PAKISTAN

The size of the gain through concentration as opposed to dispersion of vote banks depends on the size of the electoral territory.

Predicting the outcomes of an election before the actual voting day is always a central focus for political analysts and pundits and national opinion polls are often used as the primary measurement tool. However, relying solely on national polling data can be misleading unless it is analysed at a more granular geographical level. This is especially true for parliamentary systems like Pakistan’s where the electoral outcome is determined by approximately 300 contests.

The Gallup and Gilani Model of Electoral Calculus has organised the 272 contests into 12 regional electoral territories. This categorisation is based on a thorough analysis of historical voting patterns. It is a necessary step in Pakistan due to its diverse population where different regions exhibit unique political leanings and voting behaviours. Thus, to achieve accurate predictions regarding the electoral outcome, it is essential to analyse polling data at a subnational level.

Let’s delve into the specifics of how we have divided the 272 contests into electoral territories. Balochistan is represented by two territories, Khyber Pakhtunkhwa (KP) by four, Sindh by two and Punjab by four. This division takes into account various factors such as population differences, ethnic composition, voter preferences and the concentration of support for specific voting clusters within each region.

Notably, the number of seats per region varies, with Southern Punjab having the highest number of seats at 49. It is crucial to note, however, that these divisions are time-bound. Demographic changes, population variations, migration patterns and fluctuations in the number of constituencies between elections can influence the allocation of seats. However, for the purpose of our discourse, we focus on the broader perspective over a period of five decades since 1970, deliberately excluding recent developments.

For instance, we have intentionally excluded the seats from the Federally Administered Tribal Areas (Fata) since they have recently been merged into Khyber Pakhtunkhwa. Similarly, as we write, the preliminary census report of 2023 suggests a significant population increase of 56.1 percent in Balochistan since 2017, which could potentially lead to changes in seat distribution such as a decrease in Punjab’s seats projected by some analysts.

By intentionally excluding recent developments, which undoubtedly have the potential to impact the future electoral landscape, we aim to achieve a comprehensive understanding of Pakistan’s electoral dynamics. This approach is crucial for enhancing the accuracy of our predictions and painting a more holistic picture of the country’s electoral landscape.

The allocation of seats in Pakistan’s National Assembly is significantly influenced by the unequal population sizes of its four provinces. Punjab, with the largest population, holds approximately 55 percent of the seats followed by Sindh with 25 percent, Khyber Pakhtunkhwa (KP) with 15 percent and Balochistan with 5 percent.

The distribution of the 272 National Assembly general seats is as follows: Balochistan has 14 seats, Khyber Pakhtunkhwa has 35 seats, Sindh has 61 seats, Punjab has 148 seats, the Federal Capital has two seats and erstwhile Fata has 12 seats. [Editor Magazines Note: As per the latest allocations, Balochistan seats have been increased to 16, KP to 45 (including the merged Fata seats), Federal Capital to three, Sindh remains at 61 seats, while Punjab’s seats have been reduced to 141. These numbers do not include the 60 reserved seats for women allocated to the provinces or the 10 seats for non-Muslims allocated at a national level.]

Elections in each of Pakistan’s four provinces follow distinct patterns, which have become increasingly apparent over the last three decades since 1988. As such, understanding provincial-level politics is critical to comprehending the broader national electoral picture. Each province has its own political landscape, with unique contenders and varying levels of political strength.

FACTOR 3: TYPE OF COMPETITION, UNI-POLAR, BI-POLAR OR MULTI-POLAR

Elections are held to represent the voter in Parliament. Thus, in the end, what counts for political parties is not necessarily the size of the individual votes scored by them but the number of parliamentary seats which they won. This is the process which we call the conversion of votes into parliamentary seats.

Our experience in electoral research convinced us that the forecast of the seats which any party will win in Parliament is much more complicated than measuring the percentage of the parties’ appeal in terms of finding out the percentage of population which has an intention to vote for it.

The process of converting votes into seats is heavily influenced by factor 3 and factor 4. This part of the chapter is devoted to explaining the two factors and the method to compute their impact.

Polarity level of the electoral contests of each of Pakistan’s four provinces

Based on the electoral history of each of Pakistan’s four provinces, from 1970 to 2018, we have classified each of the four provinces into one of three types, uni-polar, bi-polar or multi-polar.

It appears that the type represents the standard behaviour of the province, except the situations in which the electoral lab of the province is undergoing a transformational change. The data provided below will help the reader arrive at this or any other interpretation.

The Province of Punjab: The typical contest in Punjab is bi-polar

In Punjab, the most populous province, the political arena has predominantly been marked by a bi-polar contest between the Pakistan Peoples Party (PPP) and the Muslim League (PML(s)) for a significant period. These two parties have consistently secured an average of over 75 percent of the vote in recent elections, leaving all other parties to compete for the remaining 25 percent or less. However, the bi-polar nature of Punjab’s politics has experienced destabilisation since 1997.

In 1997, the historical bi-polar contest in Punjab underwent a seismic shift, transitioning into a uni-polar one, as the Muslim Leagues gained a significant vote bank of 59 percent, which was more than double that of Pakistan Peoples Party (PPP) at 22 percent.

However, this uni-polar contest transformed into a tri-polar one during the 2002 elections, when the Muslim League split into two factions: the Pakistan Muslim League-Nawaz (PML-N) and Pakistan Muslim League-Quaid-i-Azam (PML-Q). It is important to note that this split was not a natural outcome of the political process but was influenced by external coercive factors exerted by the establishment.

The tri-polar contest persisted in the 2008 election, where all three parties secured the same percentage of votes (29 percent) while the combined Muslim League obtained twice as many votes as PPP. This indicated that the expected turbulent sympathy wave following the assassination of PPP leader Benazir Bhutto was not as significant as anticipated.

Since 2008, the tri-polar contest has, in some respects, evolved into a multi-polar one, primarily due to the entry of the Pakistan Tehreek-i-Insaf (PTI) political party. However, it is important to recognise that the bi-polar nature of Punjab’s politics has not completely disappeared, and PTI is yet to establish a dominant position.

Type of contest in Sindh: Uni-polar contests

Sindh has emerged as a uni-polar contest in both the Karachi Division and all other divisions of Sindh,

where the Muttahida Qaumi Movement (MQM) and Pakistan Peoples Party (PPP) respectively enjoy uni-polar domination.

Since the year 1988, MQM has never scored less than 40 percent of the vote in Karachi, with the exceptions of 2002 and 2018, and the runner-up is generally well below half of that.

In all other divisions of Sindh, PPP has similarly enjoyed a uni-polar status, with its vote share never dipping below 40 percent, while its closest competitors remain far behind. This uni-polarity has been consistent and the trend was evident even during the 2008 general elections, when PPP retained its majority in the rest of the divisions of Sindh and MQM continued its dominance in the Karachi Division.

Type of contest in Khyber Pakhtunkhwa: Multi-polar contests

The electoral landscape in Khyber Pakhtunkhwa (KP) is markedly different from other provinces. It may be characterised as a multi-polar contest, with four major players vying for political dominance.

To be more specific, KP can be deemed quad-polar, as it comprises Pakistan Peoples Party (PPP), Muslim League (PML(s)), religious parties such as the Jamiat Ulema-i-Islam (JUI) and Jamaat-i-Islami (JI), and regional parties including Awami National Party (ANP) among others.

Throughout the last two decades, each of these four parties has garnered around 20 percent of the vote, with fluctuations observed from one election to another. Yet, no party has ever crossed the 40 percent threshold on its own, and thus alliances have become a crucial aspect of KP’s electoral politics.

This was evidenced by the success of the religious parties’ coalition, the Muttahida Majlis-i-Amal (MMA), during the 2002 elections. This triumph can be attributed in part to the decline of the regional parties and independent candidates and in part to the consolidation of the rightist vote, which had previously supported the PML(s) in earlier elections.

The data from the early 1990s until 1997 indicates a steady rise in the Muslim League vote, gaining 10 percentage points since 1988, which impinged on PPP’s earlier vote bank. Conversely, PPP experienced a loss of 13 percentage points in the same period.

During the exceptional circumstances of the 2002 elections, the PML-N was under immense pressure from the government and was handicapped as a result. In response, a substantial number of its supporters opted for MMA. As per the Gallup Pakistan Exit Poll, 54 percent of MMA voters in 2002 had previously cast their vote in favour of the [Nawaz Sharif-led] PML in 1997, 14 percent had switched over from PPP, while the remaining 32 percent were from its own constituent parties, namely JUI and JI.

In 2008, after the dissolution of MMA and the return of PML-N, the quad-polar competition resumed in KP, returning the provincial picture to the pre-2002 scenario.

It is worth mentioning, however, that, since the 2013 elections, there has been an increasing trend towards uni-polar domination by the Pakistan Tehreek-i-Insaf (PTI), as the party won an overwhelming majority during both elections. Nevertheless, our analysis covers not only these two elections but is also relevant to the overall history of elections in the Khyber Pakhtunkhwa province.

Type of contest in Balochistan: Multi-polar contests

The contest in Balochistan is multi-polar in nature. There are a host of parties and factions, most of which score less than 10 percent of the vote and rarely does any one of them cross the 20 percent mark on its own. It is a good case of extreme multi-polarity.

Periodic shifts in polarity levels

The 2013 and 2018 elections brought significant changes to Pakistan’s electoral landscape, primarily due to the entry of PTI. In Punjab, the traditional bi-polar contest between the PML-N and the PPP transformed into a multi-polar contest involving PTI, PML-N, PPP, and smaller parties such as the Pakistan Muslim League Quaid-i-Azam (PML-Q) and Tehreek-i-Labbaik Pakistan (TLP).

In Khyber Pakhtunkhwa (KP), which had traditionally been multi-polar, PTI emerged as the front-runner, making it increasingly uni-polar in both elections.

Urban Sindh, which had been dominated by unipolarity through the Muttahida Qaumi Movement (MQM), witnessed a splintering of MQM into MQM-Pakistan, MQM-London, and other factions. Meanwhile, PTI gained more seats, indicating a potential shift towards a bi-polar contest. Balochistan also experienced a change, with the emergence of the Balochistan Awami Party (BAP).

However, it remains to be seen whether these changes are permanent or temporary features of Pakistan’s electoral landscape. The analysis presented here covers the elections from 1970 to 2018, and only time will reveal the long-term implications of these shifts.

Implications of polarity levels

Below, we discuss the types of implications of electoral polarity, that is, the uni-polar, bi-polar and multi-polar nature of electoral contests in the various provinces of Pakistan.

Implication for electoral outcomes

In the provinces or electoral territories where election outcomes are uni-polar, it becomes very difficult to dislodge the ruling party, unless all the remaining political parties collaborate with each other to form an alliance. Alliance-building, therefore, has been a characteristic of uni-polar electoral territories. In contrast, bi-polarity may result in a shift in ruling parties from one election to the next. The multi-polar territories generally require the formation of government through alliances.

Implication for political behaviour

The polarity levels are important because the nature of the polities in the province is severely affected if political power rests in one political player. The numerical domination depends on matters of negotiating a consensus or bargaining to share authority.

All of this changes if there are two players with equally matching political power and others have a minor status. Understandably, the nature of negotiations, bargaining and consensual possibilities are different if there are three or more players, all with a sizeable share in political power.

Furthermore, the polarity levels in a province are important for the nature of the relationships between the political arm of the government and the administrative arm or the social segment of the power of the communities, the strength of whose power is not reflected in the Assembly or Parliament.

FACTOR 4: LAWS WHICH CONVERT VOTES INTO SEATS IN THE ASSEMBLY

As the fourth and final step in the electoral calculus, the winner-take-all electoral system plays a pivotal role in determining the ratio between the popular vote and parliamentary seats. Under the winner-take-all system, the candidate who secures the most votes in a constituency wins the parliamentary seat, leaving the runners-up without any representation.

While this system has its own advantages, it is important to recognise that, in electoral territories which are bi-polar, it can lead to unpredictability and volatility. This has been evident in the past eleven electoral cycles.

Obviously, it is the kind of law which can be reviewed, as several political parties in their election manifestos have repeatedly demanded to switch over from the ‘winner-take-all’ to a ‘proportional representation’ law.

As a general rule, it can be stated that the law which one chooses determines one’s choice about the method to manage government. The winner-takes-all system facilitates a majoritarian method to run the government, while the proportional representational method facilitates a consensual method to run the government.

CONCLUSIONS

The discussion in this chapter makes it abundantly clear that predicting the outcome of a national election which is conducted in a large country and is based on the rules of a parliamentary system, managed through winner-take-all elections in nearly 300 parliamentary constituencies, is a highly complex matter.

Our experience suggests that a hybrid method, man-machine, is the most appropriate. A completely machine-driven computation is unlikely to provide an optimum benefit. Instead, a qualitative assessment by experts in this domain with the help of our four-dimensional qualitative model is likely to be of greater value for political analysts, political parties, individual contestants or individual political leaders.

FIVE KEY LEARNINGS ON THE VOTERS DECISION-MAKING PROCESS

1. Voting decisions are not the momentary outcome of any one behavioural, structural or contextual factor. Instead, voting decisions are the result of a process which may span over one or more than one electoral campaign.

2. Voting decisions may initially originate at a personal level influenced by a variety of factors. However, once the voter has voted for a party in one or more elections, the party becomes a marker of their personal identity. From thereon, party identification takes on a rather stable role.

3. Voters periodically switch to another party but, in general, they switch to a party that falls within the same identity spectrum as the one from which they switched.

4. Occasionally, voters migrate from one cluster of parties to another. This generally happens when the broader socioeconomic landscape of the country undergoes a transformation.

5. Over several generations, we observe that a totally fresh identity cluster emerges. Since the inception of Pakistan, only two new party clusters have surfaced: the PPP Cluster in 1970 which voiced the concerns of what we have described in the book as a constituency of the New-Poor and the PTI Cluster in 2013 which voiced the concerns of the ‘New- Educated.’ When a new cluster emerges successfully and disturbs the prevailing balances, the existing party clusters end up in realigning themselves and reinventing their identities.

The author holds a doctorate in political science from the Massachusetts Institute of Technology in the US and is a specialist in public opinion research. He established Gallup Pakistan in 1980 and is currently Chairman of Gallup and Gilani Pakistan.

Excerpted with permission from The Ritual of Elections in Pakistan (1970-2018): A Process Without A Product by Dr Ijaz Shafi Gilani, published by Lightstone Publishers 2024

Published in Dawn, EOS, February 4th, 2024