The small constituency effect and WP factor: What statistical analysis says about PAP performance at GE2020

What statistical analysis says about PAP performance at the polls

Now that the Elections Department has published the results of General Election 2020, what can voters and the political parties learn from the numbers?

Before the election, the Electoral Boundaries Review Committee had discarded the mega six-member GRCs (group representation constituencies) in favour of smaller GRCs with four or five members.

With the growth of population and downsizing of GRCs, the number of constituencies up for contest has risen steadily over the years from 16 in 2006 to 31 in 2020. For GE2020, Singapore was divided into 17 GRCs and 14 single-member constituencies (SMCs).

In GE2020, the popular vote shifted against the ruling People's Action Party - down to 61.2 per cent from 69.9 per cent in 2015, but still up from 60.1 per cent in 2011. The Workers' Party seized a second GRC - Sengkang - and retained Aljunied GRC with a margin of over 19 per cent.

Referring to the electoral map, another obvious change was the erosion of PAP support in the far west, thanks to the new Progress Singapore Party, led by Dr Tan Cheng Bock, a former PAP backbencher from the area.

Besides sentiment and satisfaction with policies, other factors, including quality of the opposition, geography, constituency size, and timing, affect election results. A scientific analysis of election results must consider all of these factors.

Multiple regression is a statistical method that does precisely that. Importantly, multiple regression allows the analyst to focus on the contribution of each individual factor, holding all other factors constant.

I ran multiple regression estimates of the PAP vote share in the last four general elections - 2006, 2011, 2015 and 2020. Three clear patterns (to be precise, statistically significant results) emerged in the last two elections.

First, the PAP did relatively better in single-member constituencies than GRCs. In GE2015, the PAP vote share in six-member GRCs was 8.8 percentage points lower than in SMCs. It certainly was a good move for the Government to disband the mega-GRCs.

But, now the focus is on five-member GRCs. In GE2020, the PAP vote share in five-member GRCs was 10.9 percentage points lower than in SMCs. From the PAP's standpoint, bigger does not appear to be better.

However, I should qualify that correlation is not causation. The PAP may have fielded stronger candidates in single-member constituencies or kept longstanding popular MPs (for instance, Minister Grace Fu won a fourth term as MP for Yuhua SMC, with over 70.5 per cent, and Mr Lim Biow Chuan, a third term as MP for Mountbatten SMC, with 73.8 per cent) and assigned weaker candidates to ride on the coat-tails of big-name ministers in the GRCs. So, the apparently weaker performance in the GRCs might just be the result of PAP strategy.

 
 
 

Second, among GRCs, the PAP tended to do better in those with larger populations.

Other things being equal (in particular, controlling for GRC member size), in a constituency with 1 per cent larger population, the PAP vote share was over 6 percentage points higher.

In GRCs with larger populations like Ang Mo Kio (71.9 per cent and with 178,039 votes cast) and Sembawang (67.3 per cent and 142,742 votes cast), the PAP beat its national vote share of 61.2 per cent.

Here, it is again important to stress that correlation is not causation. Multiple regression reveals correlation, not causation. The stronger PAP support in more populated constituencies might just be the result of constituency boundaries being drawn as such.

Third, the Workers' Party has institutionalised itself as the party of opposition. In GE2015, the PAP vote share was 7.9 percentage points lower in contests against the Workers' Party.

In GE2020, this deficit widened to 15 percentage points. However, the change was in large part due to the general shift against the PAP towards the opposition.

 
 

And I again qualify that correlation is not causation. The improved showing of the Workers' Party might be due to its focusing limited talent on a smaller number of contests (down from 10 in GE2015 to six in GE2020).

By coupling the GE2020 results with demographic data, researchers and political party strategists could do a lot more to appreciate the underlying patterns and trends. For instance, they could investigate whether GRCs are necessary to maintain minority representation in Parliament, and whether support for the PAP is stronger among older and wealthier people.

Ivan Png is a distinguished professor at the National University of Singapore Business School and Department of Economics. He served as a Nominated MP from 2005 to 2006.

A version of this article appeared in the print edition of The Straits Times on July 14, 2020, with the headline 'The small constituency effect and WP factor'. Subscribe