Bellwether Polling Divisions
Predicting the 2019 Sri Lankan Presidential Election
The term bellewether referred to the ancient English practice of placing a metallic bell around the neck of a castrated ram (known as a wether) that led a flock of sheep. A shepherd could identify the flock movements by the tinkling of the bell. Even when the flock out of sight.
In modern usage, the term bellwether refers to an entity or individual that indicates trends in a larger system or collective.
For example, whichever candidate won the state of Ohio has won the presidency in all but two of 58 US Presidential Elections. Hence, states like Ohio, Florida (also two misses), Nevada, Missouri, New Mexico, and Tennessee (all with three misses) are known as “Bellwether states”.
I was curious if Sri Lanka’s Presidential Election had any Bellwether trends. For example, were there any polling divisions which always backed the winning candidate?
Sri Lanka did indeed have eight bellwether polling divisions: Moratuwa and Ratmalana (Colombo Electoral District), Ja-Ela and Katana (Gampaha Electoral District), Mawathagama (Kurunegala Electoral District), Medirigiriya, Minneriya and Polonnaruwa (Polonnaruwa Electoral District). Whichever candidate won these polling divisions, went on to win the presidency.
For example, these are the results for Moratuwa.
I could see no discernible pattern in the distribution of bellwether polling divisions across the country:
We defined a perfect bellwether as a polling division which was always won by the winning candidate.What about the percentage of votes? Are there bellwethers that also match the winning percentage?
For each of the 160 polling divisions in Sri Lanka, I looked at the absolute error between the percentage of votes a winning candidate won island-wide, and the percentage of votes that winning candidate won in the polling division.
The polling division with the lowest error was Moratuwa. The percentage of votes a winning candidate received island-wide differed by an average of 1.2% in Moratuwa.
Interestingly, the polling division with the second-lowest error (1.6%), Yatinuwara, in the Kandy Electoral District, is not a Perfect Bellwether, backing the losing candidate in 1988. However, its percentages have matched the island-wide result almost precisely in the last two presidential elections.
The top percentage bellwethers are as follows:
Bellwether Electoral Districts
We defined and analyzed “bellwether-ness” at a polling division level. What about at the electoral district level? Have any of the 22 electoral district consistently managed to pick the winner?
Observent readers would have noticed that all three polling divisions in the Polonnaruwa electoral district were bellwethers. Hence, the Polonnaruwa electoral district as a whole is itself a bellwether.
No other electoral district has backed the winner in all seven presidential elections, though Colombo, Gampaha, Kurunegala, Puttalam, Kandy, Kegalle, Ratnapura, Monaragala and Hambantota have backed six.
Interestingly, the best percentage bellwether electoral district (i.e. that which matches the winning presidential candidate’s islandwide percentage vote best) is not Polonnaruwa, but Puttalam. In fact, Polonnaruwa is only ninth best.
“Perfect Bellwethers will always be Perfect Bellwethers.”
Suppose you are watching the results as they come in, after the November 16th election. Let’s suppose you notice that candidate Sajith Premadasa (say) wins the Moratuwa polling division.
Can you conclude that the UNP candidate has won the election?
You cannot. The “perfectness” of a bellwether polling division is how well it matches the results of the country as a whole. Bellwether polling divisions support political parties in the same proportions that the country as a whole supports these parties. This support can change, and the polling division could diverge from the country.
In the very first presidential election (1982), the UNP candidate J. R. Jayawardena won 137 out of 160 polling divisions. Hence, after this first election, 137 polling divisions were (trivially) “perfect’ bellwethers. Given the UNPs big victory, many electorates that might have voted for its opponents, ended-up voting for the UNP. Hence, the high number of bellwether polling divisions. Also, with a single election, we have only one sample. Thus, the statistical significance of our bellwether statistic is low.
In subsequent elections, the Bellwether count drops, while the significance of the metric increases:
On the eve of the 2015 election, there were 44 perfect bellwether polling divisions. Avissawella was one such polling divisions that had backed the winner in all six presidential elections.
Suppose you were watching the results as they come in, after the 2015 election. Let’s suppose you noticed that candidate Mahinda Rajapaksa won the Avissawella polling division.
Could you conclude that the UPFA candidate had won the election? Not quite.
In 2015, Avissawella backed the losing candidate, lost its bellwether status, and Maitripala Sirisena ended-up winning the election.
“Candidates should try to win all the Perfect Bellwethers.”
Bellwethers are bellwethers because they correlate strongly with the voting behaviour of the country as a whole. However, there is no causal relationship between the result of the bellwether and that of the country.
Thus, trying to win a bellwether (for example, by campaigning more in such polling divisions), would not lead to more votes in the country as a whole. The candidate might win the bellwether, and lose the country, leading to the polling division losing its bellwether status (as Avissawella did).
On the other hand, bellwethers might be good places to learn about voter preferences and needs. They could act as “samples” for the country as a whole. But after such insights are gained, they must be applied to the whole country. Not just to the bellwether.
“Polonnaruwa is special.”
As we learnt, not only is the Polonnaruwa electoral district a bellwether, so are all three of its polling divisions. So is there something special about the ancient, former-capital of Sri Lanka? When it comes to presidential elections?
Probably not. And here’s why. We can divide our “history of presidential elections” into three groups:
- 1982 and 1988: The Green Era. In this Era, much of the Sinhala majority heartland supported the UNP. Including Polonnaruwa. So it’s not surprising that Polonnaruwa voted for the UNP.
- 1994, 1999, 2005, 2010: The Blue Era. Following Chandrika Kumaratunga’s 1994 victories in both general and presidential elections, the heartland turned Blue. Again, it’s not surprising that Polonnaruwa voted for the Blue Party.
- 2015: The Polonnaruwa Era. In 2015, much of the Sinhala heartland remained Blue, voting for Mahinda Rajapaksa. One exception was Polonnaruwa, which supported their very own Maitripala Sirisena, who went on to win the election.
Hence, it is not surprising that Polonnaruwa supported the winning candidate in every election.
So who is going to win?
As usual, I’m going to utter my favourite phrase: “I don’t know”.
I also don’t know what the voters in our current perfect bellwethers (i.e. Moratuwa, Ratmalana, Ja-Ela, Katana, Mawathagama, Medirigiriya, Minneriya and Polonnaruwa) are thinking. But maybe you do.
If you go to Ratmalana, talk to a bunch of Ratmalaneans and find out that they are all supporting candidate X, you could make one of three conclusions:
- Either candidate X will win, and Ratmalana will remain a Perfect Bellwether.
- Or else candidate X will not win, and Ratmalana will loose its Perfect Bellwether status.
- Or else, you didn’t walk to enough Ratmalaneans.
It’s that simple.