It was a photo-finish between Tony Tan and Tan Cheng Bock, with the former pipping the latter by only 7,269 votes, or about 0.3 percent. However, this margin was a wee bit more than the total number of overseas electors (5,504) and so the local count was considered conclusive. Tony Tan will be the seventh President of Singapore.
Immediately, recriminations began. Supporters of Tan Cheng Bock accused supporters of Tan Jee Say of “throwing” the contest to Tony Tan. This suggested that supporters of Tan Jee Say ought to have engaged in tactical voting to deny Tony Tan the presidency.
But maybe they did, except that it was not enough to overcome Tony Tan’s support.
I had long suspected that such recrimination would happen and that was why I ran a survey Thursday through Saturday. I did not announce the aim of the survey earlier and I think very few readers of Yawning Bread managed to read my intention from the survey questions, but the main objective of the study was to see how many Singaporean voters engaged in tactical voting in the presidential election. Between the possible answers “not enough tactical voting” and “no tactical voting”, the data confirmed my hunch that it would be closer to “no”. The results show that there was little of it. Only about 2 or 3 percent of voters switched to voting for Tan Cheng Bock when they would really have preferred Tan Jee Say.
The cookie that is Singapore politics often crumbles into a pro-People’s Action Party (PAP) and an anti-PAP camp. In general elections, the pro-PAP camp has long been the larger one, thus giving the party a huge majority in Parliament, but the presidential contest was not one that would determine the government. A by-election effect was expected where voters would use the occasion to register their unhappiness with the PAP by voting against it (or its proxy) when the government was not at stake.
As the actual results showed, only about 35 percent percent voted for Tony Tan, the candidate that everybody knew the government preferred. Nearly two-thirds voted against him.
The problem in this contest was that there were three alternatives to Tony Tan, and from the beginning, there were fears that the anti-government vote would be badly split.
Usually, when such a situation occurs, voters will consider tactical voting — i.e. where they vote for the candidate most able to unseat the one they most dislike, even if that candidate they end up voting for is not their first choice.
At least two Straits Times reporters asked me, fairly early in the campaign period, if I foresaw Singaporeans voting tactically. I said while they might like to, it would be hard to actually carry it out. The reason is that a key piece of information was necessary — which candidate stood the best chance of unseating the most disliked candidate? So long as opinion polls remain illegal during an election period, it was difficult for Singaporeans to know who was the leading alternative.
Tactical voting is largely an issue only for the 40 percent of the electorate who chose opposition parties in the general election of May 2011. They would probably refuse to vote for any candidate the government is seen to prefer, and that was Tony Tan. However, as alternatives, they had a choice of Tan Cheng Bock, Tan Jee Say and even Tan Kin Lian, though in the latter’s case, it was widely believed that he stood little chance. For the 60 percent who voted for the PAP in GE2011, tactical voting would not be an issue, because Jee Say would have been so unacceptable, the choice would only be between Tony Tan and Tan Cheng Bock.
It so happens that Yawning Bread largely reaches the 40 percent who voted for the opposition; that made this site a good vehicle for carrying out a survey on tactical voting.
Why do I say that this site largely reaches this 40-percent pro-opposition block?
Results from an earlier survey by Yawning Bread around the time of the May 2011 general election showed a good correlation between Yawning Bread readers’ preferences and those who voted for opposition parties. The party-percentage split from the May 2011 survey was very close to the actual party vote-shares among opposition parties that eventuated, as can be seen from this table:
Thus, one could roughly say that respondents reached by Yawning Bread mostly came from the 40-percent block that voted for opposition parties and in a fairly representative way — except for age and gender skewing, which is commonly seen in internet surveys on politics.
Since this site reaches the same segment of the electorate that was faced with the issue of tactical voting, a Yawning Bread survey would be a good vehicle for studying this phenomenon.
In fact, from this presidential election survey alone, I was able to predict the final vote shares of Tan Jee Say and Tan Kin Lian. This is because Jee Say and, to a lesser degree, Kin Lian would be drawing their support from the 40-percent pro-opposition block. Since my presidential election survey gave me data on how this block was voting, I could extrapolate it to a nationwide level. I told three friends at around 3 p.m. on Saturday, about midway through Polling Day, that Tan Jee Say would get about 25 percent nationwide and Tan Kin Lian would get less than 5 percent. It would turn out correct.
This ability to predict from survey results is consistent with the assumption I am making that Yawning Bread surveys reach a roughly representative section of the pro-opposition camp. This is how I arrived at my prediction:
From the survey, Tan Jee Say was getting about 66 percent of responses. This meant that he’d be getting about 26.5 percent of the nationwide vote (rightmost column of grey-green table), provided:
(a) my responses were roughly representative of the 40-percent opposition block, and
(b) virtually none of the 60-percent pro-PAP block would be voting for him.
I think (b) is a fair assumption. If even supporters of some opposition parties buy the notion that Tan Jee Say was too “confrontational” — a comment I have personally heard — then what more of PAP-sympathetic voters?
That Tan Jee Say’s final vote tally amounted to 25 percent nationwide kind of confirms that assumption (a) is valid too, since the prediction I made from my survey came very close to that.
Given the above experience with Yawning Bread surveys, for the rest of this article, you can assume that what we are analysing here is the behaviour of the 40-percent pro-opposition block, with a slight skewing to younger and male voters.
Method and responses
There was a vague sense that Tan Cheng Bock could be the one to back if one wished to vote tactically, simply from the fact that he could pull support from both the 60-percent block who voted for PAP candidates in the May 2011 general election, and from the 40-percent block who voted for opposition candidates. However, as the campaign progressed, Tan Jee Say gained traction. I heard, for example, that neighbourhood bookies were steadily shortening the odds against Tan Jee Say as Polling Day neared.
Anti-government voters could thus have been unsure whether to back Tan Cheng Bock or Tan Jee Say, even if they wanted to vote tactically.
The way to look for tactical voting is to compare how people actually voted against how they would really want to vote if the need for tactical voting is eliminated. The survey did so by setting up a series of hypothetical straight fights to see how voter preferences changed.
The survey form was available on Google Docs slightly after midnight in the early hours of Thursday 25 August, and remained active till 10 p.m. on Saturday 27 August (i.e. until about 2 hours after polling stations closed, but well before the results were announced). You can see the questions by clicking the thumbnail at left.
Other than this site, myself and a few readers advertised the survey on Facebook, but I believe the reach via Facebook was not great.
Clearly, as an internet survey, no claim can be made that it is a controlled representative sample of the voting population as a whole. However, as explained above, I am confident that my survey is reaching a rough representation of pro-opposition voters, the same segment that was faced with the possibility of tactical voting.
How well they think they know the candidates
Questions 2 and 3 of the survey made a digression. They were questions asking how well voters thought they understood the candidate they had chosen to vote for, on a scale of 1 to 5:
Tan Cheng Bock had the lowest mean scores on both measures. Yet, he nearly won the race. What would explain that? Some possibilities:
(a) These were pro-opposition voters; The pro-PAP voters (not captured in my survey) might have thought much more highly of him;
(b) Although he didn’t have strong positives, he had the least negatives;
(c) Although values-affinity and performance-predictability were low, he had relatively high personal likeability.
Tan Jee Say had the highest mean score for values-affinity. He communicated best what he stood for. He also had the second-highest mean score for performance-predictability, suggesting that he also communicated well what he was going to do with the office.
Tony Tan tied with Tan Cheng Bock in having the lowest score for values-affinity. However, readers must take care to note that the sample size was small (only 60 respondents) and anyway respondents were generally opposition-sympathetic, when the vast bulk of his support came from the pro-PAP block not captured in my survey. Tony Tan had the highest score in terms of performance-predictability; people felt they knew how he was going to perform in office.
Tan Kin Lian had middling scores for both values-affinity and performance-predictability. Again, take care that the sample size was relatively small (79 respondents).
Hypothetical straight fights
In this section, I am restricting the analysis to only two groups of respondents: those who voted for Tan Cheng Bock and those who voted for Tan Jee Say. The sample sizes of those who voted for Tony Tan and Tan Kin Lian are too small to be meaningful.
The aim is to see how many of those who voted for Tan Cheng Bock would really have preferred another candidate. Likewise for those who voted for Tan Jee Say. In other words, how many engaged in tactical voting.
Starting with those who actually voted for Tan Cheng Bock:
The interesting one is the bar chart on the left. If Tony Tan and Tan Kin Lian were not in the contest, 5.6 percent of those who voted for Tan Cheng Bock would have preferred to vote for Tan Jee Say.
If you look at the bar chart to the right, you’d see that if Tony Tan and Tan Jee Say were not in the contest, 2.7 percent of those who voted for Tan Cheng Bock would have preferred to vote for Tan Kin Lian.
These are relatively small numbers, which is why I say that there was little tactical voting. Tan Cheng Bock did not enjoy much of a boost from that.
Now let’s look at whether Tan Jee Say enjoyed any boost from tactical voting:
Even less. In hypothetical straight fights, only 3.9 percent and 4.3 percent would really have preferred Tan Kin Lian and Tan Cheng Bock respectively. In other words, the vast majority of those who voted for Tan Jee Say had him as their first choice.
The survey does not capture what pro-PAP voters did. However, pro-PAP voters had little room for tactical voting since for most of them, the only relevant choices were between Tony Tan and Tan Cheng Bock. Tactical voting would not have been on their minds.
The data confirms what I had initially thought — while in theory, tactical voting was possible, in practice, I couldn’t see it happening unless either Tan Cheng Bock or Tan Jee Say gained an obvious advantage over the other. That didn’t happen and so tactical voting didn’t much take place.
The next concept is even more abstract. However, it is important to check this out.
Perhaps tactical voting didn’t take place simply because those who supported Tan Cheng Bock or Tan Jee Say did not think the other was a worthwhile substitute? Maybe their respective supporters had so little affinity with the alternative that switching their votes was out of the question?
To determine this, the survey posed three more hypothetical straight fights. For those who voted for Tan Cheng Bock, the survey asked them what they would have done if Tan Cheng Bock wasn’t even in the contest. Who else would they have voted for?
As you can see, if Tan Jee Say had been available (in the absence of Tan Cheng Bock), about 70 percent would have switched to Tan Jee Say. It is quite obvious that of the Tan Cheng Bock-supporters from the 40-percent opposition block, many would consider Tan Jee Say as a reasonable substitute.
Between Tony Tan and Tan Kin Lian, more of them would prefer Tan Kin Lian than Tony Tan.
I can express this state of affairs by deriving a “substitutability index” — where the higher the figure, the higher the possibility of substitution (or alternate attractiveness). The index is simply the proportion of voters who would switch to the alternative.
For example, if we wanted to know how substitutable Tan Kin Lian was for Tan Cheng Bock, then
1. We use for analysis those respondents in my survey who actually voted for Tan Cheng Bock (n=445);
2. Look at the results in hypothetical scenarios where Tan Cheng Bock was not an option, AND where Tan Kin Lian was — there are only two such straight fight scenarios, one pitching Tony Tan versus Tan Kin Lian, and the other pitching Tan Jee Say versus Tan Kin Lian.
3. In the former hypothetical scenario, 57.1 percent would choose Tan Kin Lian; in the latter, 29.4 percent would choose Tan Kin Lian; average 43.25 percent.
4. Thus the substitutability index for Tan Kin Lian in the minds of Tan Cheng Bock’s voters was 0.4325.
Using the indices so obtained, I can draw a diagram to represent them, where the further apart two men are, the less substitutable they are. In the minds of Tan Cheng Bock’s voters:
Now, let’s look at the group that voted for Tan Jee Say. What would they have done if Tan Jee Say had not been in the contest?
As you can see, they would have been very adverse to voting for Tony Tan. They would vote for Tan Cheng Bock or Tan Kin Lian, whoever was available. If both were available (the middle bar chart), there was a slight preference for Tan Cheng Bock.
The diagram for the substitutability indices looks like this, with Tony Tan held at arm’s length:
The conclusion therefore is that Tan Cheng Bock and Tan Jee Say were relatively inter-substitutable in the minds of voters who belong to the 40-percent block that voted for opposition parties in GE2011.
The potential for tactical voting was there, as we could see from the substitutability analysis. However, in this election, very little tactical voting actually took place. Therefore the most likely chief reason why it didn’t take place was because there were no signals as to how close Tan Cheng Bock was to winning.
If an opinion poll had been done and published a few days before Polling Day (e.g. by an overseas organisation, thus arguably outside the ambit of Singapore law that bans publication of such polls) which showed how close the result might be, quite likely Singaporeans would have altered their voting behaviour considerably. And we would have had a very different result from Presidential Election 2011.