The swing in Malay votes from BN to PH, together with solid support for PH from non-Malays, can overcome the negative effect of Pas’ withdrawal from the Pakatan coalition, says Subramaniam Pillay.
In two recent public presentations (9 and 30 March), Invoke Malaysia, a non-governmental organisation established by PKR’s Rafizi Ramli, the MP for Pandan, made a couple of bold predictions for the coming general election:
- Pakatan Harapan (PH) will win at least 89 seats in Semenanjung Malaysia versus 76 seats for Barisan Nasional (BN)
- Pas will not win a single seat in the next Parliament.
Reactions have been negative as expected. BN politicians and others have labelled Rafizi as delusional, biased and “syok-sendiri”. These predictions are in complete contrast to other forecasts made by political pundits who claim that the BN will do much better than in the 2013 general election.
According to them, this better performance would be due to the recent redelineation and more importantly due to Pas splitting anti-BN votes by engaging in three-way contests. Rafizi, however, has turned this claim on its head by arguing that three-way contests favour PH more than BN.
In this article, we will discuss the plausibility of Invoke’s forecast by examining the methodology and underlying assumptions used.
How reliable is the methodology?
Invoke has collected a lot of data on voters across Semenanjung by i) phone banking and canvassing by volunteers, ii) digital tracking over social media, iii) traditional phone polling, and iv) digital polling over Facebook.
Based on this data, a logistic regression model is used to predict the likelihood of a voter choosing BN, PH or Pas depending on age, location, ethnicity, gender and other variables.
To ensure representativeness of the sample, a stratified randomised sampling technique using 24 strata of voters is being used. After going through all the slides and explanations provided during the presentation and the question-and-answer session that followed, I have no good reason to doubt the reliability of the methodology used.
But this does not mean that one cannot draw slightly different conclusions based on the assumptions one makes using the same polling results. Let us now look at the polling data.
Popular support for BN, PH and Pas survey results
Invoke runs monthly surveys on voter preferences in three-way contests by getting complete responses from about 3,500 to 4,000 respondents.
Figure 1 below extracted from Rafizi’s presentation on 9 March shows the support for the different parties among Malay voters
There are some interesting observations we can draw from the data:
- Support for Umno/BN has declined over the past three months from 41% to 29%. This decline is mirrored by the increase in the proportion of “Undecided” voters.
- Support for PH among Malay voters is increasing but is still at a relatively low level of about 14%.
- Support for Pas is stable around 18% of Malay voters across Semenanjung. However, this overall figure may mask a large variation between Pas’ support in states like Kelantan and Terengganu where it would be much higher compared to the southern states of Melaka and Johor where support would be much lower.
- The most significant finding in the poll is the exceptionally large number of respondents in the “Undecided” and “Refused to disclose” categories. In the 17 February poll, they formed a combined 39% of the voters!
Invoke then probed the non-PH respondents with more questions. Based on those answers, Invoke is concluding that “PH Malay support is very close to Umno’s”. According to Rafizi, many respondents may be afraid to openly support PH because they are civil servants or receiving BR1M money.
Figure 2 shows the support for the different parties among Chinese voters
From this poll, we can conclude that:
- Chinese support for PH is high at about 45% compared to less than 20 % for BN.
- As expected, there is zero support for Pas.
- “Undecided” and “Refused” voters together form about 36% of the Chinese sample.
As before Invoke, using more probing questions on the “Undecided/Refused” respondents, concludes that “Chinese votes have come back to GE13 level ie split between 80% to PH and 20% to BN”.
The chart below shows the support for the different parties among Indian voters:
The following observations can be made from this:
- It is quite evident that Indian voters express the strongest support for BN at about 45% while PH has only 25% support.
- Just like the Chinese, Pas has virtually zero support from Indians.
- The “Undecided”/”Refused” category is smaller than those for the Malays and Chinese at less than 30% of the sample.
According to Rafizi, “while Indian voters are the most angry group of voters on economic issues, PH has not been able to capitalise on the anger to get higher share of Indian votes.” Based on further probing, his estimation of the vote split of Indians is 55% for BN and 45% for PH.
Indians and the Orang Asli communities have been the most neglected ones under the Alliance/BN rule yet they are the most loyal to BN! Prior to the Hindraf movement in 2007, Indian support for BN was in the range of 75-80%. Now it is more evenly split with slight variations based on the rural-urban divide.
We can summarise the discussion so far as follows:
- Pas has zero support among non-Malays.
- Chinese support for PH has returned to the 2013 level of about 80%.
- Indian support is split roughly equally between PH and BN.
- The relatively high Malay support for BN in the 2013 general election has suffered a somewhat significant decline; the anti-BN Malay vote is divided between PH and Pas differently in different regions.
- The presence of three-way fights will play a critical role in the outcome of the coming general election.
Dynamics of three-way contests
Given the breakaway of Pas from the old Pakatan Rakyat and the emergence of new Malay/Muslim-based parties in the new Pakatan Harapan coalition, Malay votes will be split three ways between BN, Pas and PH in Semenanjung Malaysia.
To the best of my recollection, there has never been a genuine nation-wide three-way fight for the Malay vote in any previous election. As such, extrapolation based on past voting patterns would not be reliable for predicting the outcome of GE14.
In three-way contests using a first-past-the-post system, the winning candidate does not have to get more than 50% of the voters. All he or she needs is to do better than the second candidate. Because of this, the number of seats won by a party can change dramatically with a small change in the percentage of votes.
In the UK and Canada, which both follow the first-past-the-post system, the winning party rarely gets more than 50% of the votes. An interesting example of this is the 2015 Canadian elections. Table 1 presents some polling results and seat projections during the more than two-month campaign period.
In early August 2015, when the election was called, Justin Trudeau’s Liberal party was languishing in third place.
By the time the election was held on 19 October, their popularity had increased from 25.8% to 39.5% while the seats won increased from a projected 84 seats (25% of the seats) to 184 seats (54% of the seats). In other words a 13.7% increase in popular vote more than doubled the number of seats won.
In many constituencies, where the Liberals were trailing just behind the leading party (be it NDP or the Conservative party), a small increase in popular vote brought them just a bit ahead to win the seat.
A more interesting observation is what happened to the incumbent Conservatives. Although their share of the popular vote was about the same at the beginning and at the end of the campaign, their seats fell from 130 to 99.
This is because of strategic voting by many left-of-centre voters who switched from the left-leaning NDP to the more centrist Liberals; this in turn led to the Conservatives becoming second in the 30-plus seats which they would have won if the elections had been held in early August.
Many NDP supporters disliked Stephen Harper’s Conservative government so much that they decided to vote to prevent a Conservative victory – which meant voting for the Liberals in constituencies where the Liberals were trailing the Conservatives narrowly. In other words they did not want to waste their vote on their own NDP candidates, who had no chance of winning.
Different impacts of three-way contests in the coming general election
Before we go further, it is important to note that Invoke makes a very critical assumption that compared with 2013, Malay support for BN has declined noticeably. Two reasons were advanced.
First, due to the natural process of citizens who have turned 21 in the last five years registering to vote, support for PH has increased more than for BN. The premise here is that younger voters are disproportionally more anti-establishment, ie anti-BN.
Secondly, the entry of Tun Mahathir into the electoral fray by forming Pribumi, a Malay based party is drawing some support away from Umno-BN. The impact has been especially noticeable after the selection of Mahathir as prime ministerial candidate for the PH coalition a few months ago.
Whether one likes him or not, one cannot deny that he has shown a remarkable level of energy and tactical smarts in recent months by getting PH component parties focused on:
- finalising the allocation of seats among the four parties
- bringing out a common manifesto
- going around the country campaigning jointly
- and now, pushing through the adoption of a common symbol for the PH coalition.
He doesn’t appear to have lost his political instincts despite his age.
Let us now look at Semenanjung Malaysia’s three-way contest which has a twist: there is a major difference in the contest for votes from the Malay electorate and the non-Malay electorate. Among the non-Malays, it is a two-way contest between BN and PH while among the Malays, it is a three-way contest between BN, PH and Pas.
The interaction between these two types of contests within each constituency would have different impacts on each of the 165 constituencies depending on its ethnic composition.
To simplify our discussion, I have taken the 2013 general election voter registration data and grouped the 165 constituencies in various categories by proportion of the two major ethnic groups in Table 2 below.
We can now try to make some broad generalisations on the impact of three-way contests on the different categories as follows:
Category SM: Malay super-majority seats
Almost all of the 46 Malay super-majority seats are easy pickings for BN. They are mostly in rural areas where PH support is not as strong. The drop in support for BN among rural Malays is not as severe as among urban Malays. In the northern states, where Pas is strong, PH will draw away the anti-BN vote, thus hurting Pas and ironically helping BN!
For example, let us take the east coast state of Terengganu. None of the eight seats have more than 10% non-Malays. In fact most of them have less than 5% non-Malay voters. Traditionally these seats have seen straight fights between Pas and BN.
Table 3 presents the outcome in the past four elections:
Both parties have a core support of not less than 40%. It is the 10-15% of swing voters who determine the outcome. In the coming general election, with the entry of PH in the fray, the anti-BN vote will be split leading to BN victories in all these seats. Even if BN obtains a little more than 40% of the votes, it can win all the seats.
This is the logic that led to Invoke’s prediction that in Terengganu, Kelantan, and Perlis as well as in other Malay super-majority rural and semi-rural constituencies, Umno will be the winner.
Pas may still get around 30-40% of the votes in its traditional areas of Kelantan, Terengganu, Perlis and northern Kedah – and yet not win a single seat.
PH will not do well in this kind of seats either because there are not enough non-Malay votes to counter the split in anti-BN Malay votes by Pas and BH.
There might be some exceptions in seats where a top leader like Mahathir contests (eg Langkawi or Kubang Pasu) where the individual candidate’s uniqueness negates this logic. This happened in the 1986 election when Pas lost all its seats except in Pengkalan Chepa, where the late Tok Guru Nik Aziz won. Let us not forget that even Haji Hadi Awang, the current Pas president, has lost in straight fights in his Marang constituency in 1986 and 2004.
We can thus conclude that in Malay super-majority seats (>80% Malay voters), there is an extremely high probability that Pas will lose all the parliamentary seats if it faces three-way contests.
Category CM: Chinese-majority seats
Taking the Semenanjung as a whole, as shown in Table 2 above, in the 2013 general election, there were 22 seats in this category. All of these seats were contested by DAP, and it won all of them.
PH component parties will again win all such seats in the coming election. With the Chinese support predicted to be close to 80%, even if the constituency has only 50% Chinese voters, that is already 40% (50% x 80%) of the popular vote in the bag. All the PH candidate needs is a fifth of the non-Chinese voters for him to her to get a simple majority and win the seat.
If Pas decides to contest against DAP as they claimed a few weeks ago, it will end up drawing more votes away from BN and making it easier for DAP to win the seats.
Unfortunately for PH, there are not too many seats with a Chinese majority. The latest redrawing of boundaries has made the situation even worse by concentrating Chinese voters in fewer constituencies.
Category MX: Mixed seats
The 29 mixed seats, where no single ethnic group forms a majority, are also favourable to PH. Most of them are located in either urban or semi-urban areas, where Malay support for PH is much better than rural areas.
For example if a constituency has 45% Malay, 40% Chinese and 15% Indian, then assuming PH can get a minimum of one-third of the Malay votes (15%), 80% of Chinese votes (32%) and half of Indian votes (7.5%), it will win by getting a total of 54.5% of the votes.
Category MM: Malay-majority (50% to 80%) seats
These 68 seats are in urban, semi-urban and rural areas. This will be the main battleground in the coming general election. If the seats are in urban areas (eg Bandar Tun Razak with a 58% Malay vote), it will be easier for PH to win. On the other hand, if a MM seat is in a rural area it will be more difficult for PH.
Prediction – PH to trump BN in Semenanjung?
With an understanding of the dynamics of three-way contests, let us now look at Invoke’s prediction. Based on the polling and on the demographic characteristics of each constituency, Invoke has made the following state by state prediction for the coming general election:
To facilitate our discussion, I have grouped the states into four groups:
Kelantan, Terengganu and Perlis
All the 25 constituencies in Kelantan, Terengganu and Perlis fall in the Malay super-majority (SM) category. It is predicted that BN will win all of the seats in three-way contests. This prediction makes sense based on our discussion on Malay super-majority (SM) seats in the previous section. Next let us look at the urban belt.
Penang, Selangor and Wilayah Perseketuan – the urban states
There are 47 seats in these three states. Invoke predicts that PH will win 42 of them.
Only two seats (Putrajaya and Sabak Bernam) had more than 80% Malay voters; 25 seats were either mixed or Chinese-majority seats where PH should win easily. The remaining 20 were Malay-majority (MM) seats.
The recent boundary changes had the biggest impact in Selangor and the Federal Territory by reducing mixed seats and increasing Malay-majority seats.
According to Invoke in spite of this blatant manipulation by the Election Commission, PH is expected to do very well in these three states because a) surveys show that Penang and Selangor residents are generally supportive of the Pakatan governments and b) PH support among urban Malays is relatively strong. As such, this prediction seems plausible.
Perak, Pahang, Negeri Sembilan and Melaka
There are 52 seats in these four states, and Invoke predicts that the outcome of the coming general election will not differ very much with BN expected to win 31 seats, the same number as before. In these four states, the urban seats are already in the hands of PH. The contest will be in semi-urban areas.
Although some PH leaders believe that they can defeat BN in Perak and Negeri Sembilan, apparently Invoke’s polling shows that BN’s support among Malay voters is strong in Perak. In contrast to Invoke’s prediction, a very recent poll by Institut Darul Ehsan predicted that PH could capture the states of Negeri Sembilan and Melaka. Overall, Invoke’s prediction is plausible.
Kedah and Johor – PH’s trump card?
Kedah has 15 parliamentary and currently it is 10-5 in BN’s favour.
Invoke is predicting that BN will lose five of those 10 seats. The main explanation is the impact of the Mahathir factor. Mahathir wields enormous influence in Kedah and as such, there appears to be a groundswell of support for PH. On top of it, there is disunity within Kedah Umno.
So even in seats with large Malay majorities, PH is predicted to do well and win 10 out of the 15 seats. Such an explanation appears plausible.
Invoke’s prediction for Johor’s 26 parliamentary seats is the most controversial. Currently, BN holds 21 out of the 26 seats. Invoke is predicting that it will lose 11 of those seats to PH and end up with only 10 seats while PH is expected to win 16 seats.
I find it a bit too optimistic. In the 2013 general election, there were only six mixed seats and three Chinese-majority seats. The impact of Pas among Malays is also relatively weak in Johor. So basically, in Johor the contests can be viewed as de facto straight fights. Is it possible that Pribumi has been able to draw significant support away from Umno?
The recent Institut Darul Ehsan (IDE) study claims that “Johor will likely remain under BN, but projected PH would win Kedah, Melaka and Negeri Sembilan”. IDE’s deputy director “Redzuan said PPBM president and ex-deputy Prime Minister Tan Sri Muhyiddin Yassin, who is from Johor, did not possess ‘charismatic leadership’ qualities to rival that of Dr Mahathir” (themalaymailonline.com, 5 April 2018).
I think that Invoke’s remarkably optimistic prediction will materialise only if there is an extraordinary turnout. So I am a little sceptical of the Johor forecast.
Sabah and Sarawak – the kingmakers?
In this article, there has been no discussion of Sabah and Sarawak for two reasons.
Firstly, Invoke does not have much data on these two important Borneo states due to a lack of resources and as such could not make predictions.
Secondly, if PH can get 100 or more seats in Semenanjung, it is not inconceivable that PH and its allies in Sabah and Sarawak can easily get 15 seats to exceed the 112 seats to form a government. With the recent announcement of an electoral pact between PH and Warisan in Sabah, the prospect for these parties have improved considerably.
It is also possible that current BN component parties in Sabah and Sarawak might switch to become component parties of PH if PH wins big in the peninsula.
PH victory not inconceivable
In overall terms, Invoke’s prediction is not that dramatic as many have made it out to be. It is predicting that BN will lose a mere nine seats from its previous 85 seats. Given the unpopularity of Prime Minister Najib and the split in Umno leading to the formation of Parti Pribumi, it is entirely plausible for BN to lose that many or even more seats.
Many pundits claim that Pas taking part in three-way fights will affect PH. However, the split is only among Malay voters.
Moreover, BN has suffered an erosion of support among Malay voters. That swing in Malay votes from BN to PH together with the solid support for PH from non-Malays can overcome the negative effect of Pas’ withdrawal from the Pakatan coalition.
What can go wrong with Invoke’s prediction?
- The logit model used to compute the probability of a voter choosing BN, PH or Pas has not been tested in any real election as it is a new model. What happens if the margin of error in the computed probability is wider in practice due to incomplete information and omitted variables? I do not want to get into a very technical discussion here but until the model is tested out in the coming general election, one cannot be super confident about its predictive ability.
- During the election campaign period, any inadvertent blunder by one or more PH senior politicians could have a dramatic impact. A loss of 1-2% in popular votes at the tipping point could have a big impact on the number of seats won or lost.
- An insufficient turnout especially among younger voters will have a disproportionally negative effect on PH.
- The recent redelineation may have had some impact on PH’s overall chances although Rafizi in his 30 March forum argued that it would have only a marginal effect in a couple of seats. In Selangor, he mentioned Shah Alam. However, as Mahathir said, the redelineation effect can be overcome if there is a larger turnout in the marginal seats.
- The most serious threat is the potential for abuse and voter fraud in the postal voting process. Throwing open postal voting to a larger pool of public servants makes it easier for the regime to cheat. Unlike early voting, where the candidates’ polling agents are present, in postal voting we would not know who actually voted. Again, a higher turnout can negate the effect of any potential postal voting abuse.
What can be done to help bring about change?
- The most important agenda should be to ensure a very high turnout especially at the marginal seats, which are the mixed seats and Malay simple-majority seats. Political parties and concerned citizens should work on this. This effort should be particularly targeted at the younger voters. Individually, we can help by providing transport on the election day, for example.
- PH should concentrate their time, energy and resources on marginal seats. For example, much of the resources that are being used to campaign in safe urban seats should be reallocated to nearby marginal seats. For example, DAP and PKR should focus on marginal seats on the Penang mainland and help Amanah and Bersatu to make inroads into parliamentary seats like Kepala Batas, Tasik Gelugor and Balik Pulau. This requires closer cooperation among the component parties. Don’t waste time speaking to the converted. Holding mammoth rallies in your strongholds is an act of syok-sendiri. There is a larger than usual number of undecided voters this time around. So please focus on them.
- PH also needs to pay more attention to Indian voters, especially in constituencies where they make up a substantial number of the voters. The swing in their support can tip the balance in marginal seats in Selangor, Perak and Negeri Sembilan.
- Do not make foolish statements that hurt the feelings of any ethnic or religious groups. Always pay attention to the B40 group and emphasise inclusivity.
- Do not waste too much time attacking Pas. Unlike their president, many Pas supporters are hard core anti-Umno who may not be very happy with the current bromance between Pas and Umno. By attacking Pas, you may drive Pas supporters away. Instead encourage them to vote strategically like what many NDP supporters did by voting for Justin Trudeau’s Liberal party to ensure the defeat of the much-disliked Conservatives in the 2015 Canadian elections. The message to them should be “Vote PH to get rid of Umno-BN”.
In conclusion, a PH victory is not inconceivable. The large number of undecided voters is an indication of the fluidity of the situation. If PH’s momentum increases throughout the election period, in three-way contests, it could result in a big increase in the number of seats. Let us be more positive and work towards bringing about a change.