The Overhang Oddity

In my polling and projection posts, I’m finding myself quite regularly referring to a concept called “overhang“. This is a quirk of mixed electoral systems like the Additional Member System (AMS) we use for the Scottish Parliament and similar systems used in New Zealand and Germany.

Basically, an overhang is when a party wins so many of the single-member constituency seats that it has won more than its proportional share of seats based on the list vote. Overhang is interesting because it can distort the overall result, which is why New Zealand and Germany compensate for it. Scotland does not.

In the early days of the Scottish Parliament, Labour had a significant overhang, thanks in part to sweeping so many Central Belt constituencies. Since the SNP became dominant, overhangs have tended to be much smaller, as their vote is spread much more consistently around the whole country than Labour’s ever was.

Indeed, for all that 2011 was seen by many as the SNP “breaking” AMS, it actually worked almost perfectly, as the only overhang was one seat too many in the Lothian region, which came at the expense of the Lib Dems. In 2016 it was also only one seat, this time in Mid Scotland & Fife at Labour’s expense. However, recent polling has tended to show a more significant overhang.

Partly as a result of this tweet, which I’ll get to a bit later, and partly because I’m referring to it so often, I thought it might be useful to have a whole post on the topic – and interesting to see how the Scottish Parliament might look if we did compensate for overhang.

Firstly, let’s take the average result of the past 5 polls to get the vote shares we’re working with – those polls stretch from the 27th of June to 21st of October (constituency/list);

  • SNP – 41.2% / 33.6%
  • Con – 25.4% / 23.2%
  • Lab – 22.4% / 22.0%
  • Green – 1.2% / 8.6%
  • Lib Dem – 7.8% / 8.4%
  • UKIP – 0% / 2.6%

As with New Zealand and Germany, we’ll want to aim for proportionality to match the list vote, as that is the vote designed for proportionality. If we project those figures into seats, how close do they come? Let’s see (seats – % of seats);

  • SNP – 54 (41.9%)
  • Con – 30 (23.3%)
  • Lab – 27 (20.9%)
  • Green – 9 (7.0%)
  • Lib Dem – 9 (7.0%)
  • UKIP – 0 (0%)

It’s not actually a million miles off for most of the parties that win seats – it’s just the SNP coming miles ahead thanks to overhang. So let’s compensate for that.

In New Zealand, overhang is compensated for by allowing the party with too many seats to keep them, but not allowing that to detract from list seats other parties might fairly have won, so long as those parties meet a threshold of 5% of the vote or win at least one constituency. That means Parliament is not a fixed size, but in 20+ years, the most it’s ever had has been 2 extra seats.

Note that New Zealand uses a national list rather than regional lists, but for Scotland we’ll just do overhang per region to match how our system works. This comes out at (seats – % of seats);

  •  SNP – 54 (40.0%)
  • Con – 32 (23.7%)
  • Lab – 30 (22.2%)
  • Green – 10 (7.4%)
  • Lib Dem – 9 (6.7%)

We’d end up with 6 extra seats in the Scottish Parliament (so 135 rather than 129), coming two apiece from the Central, Glasgow and West regions. But it doesn’t do much to help the proportionality, and in fact the Lib Dems don’t get any of those extra seats so are even worse off.

Now let’s do it German style. This is where that tweet I linked to earlier comes in. What’s interesting about Germany is that it aims to fully compensate for overhangs, and will hand out as many extra list seats as necessary to restore full proportionality across the country.

They’ve made those provisions even stronger in recent years, then when you factor in historic lows for the two main parties, and the CDU/CSU nonetheless taking the lion’s share of constituencies, this can result in a hugely inflated Bundestag. Electing 598 seats at a minimum, last year’s election was notable for being by far the largest the Bundestag has ever been, with a whopping 709 seats. The projection for the next election contained in that tweet would result in an incredible 815 seats.

Doing something roughly similar for Scotland, where we would then bolt on more list seats per region to match the national voter share, we need to work out the percentage we’re working to by excluding parties below 5%, which would be;

  • SNP – 35.1%
  • Con – 24.2%
  • Lab  – 23.0%
  • Green – 9.0%
  • Lib Dem – 8.8%

As Scotland is much smaller than Germany, I recognise you’d need 500 seats in total to catch a difference of 0.2% as between Greens and Lib Dems, so I won’t go QUITE that far and say so long as it’s within 0.5% we’re golden. That means we’d end up with (seats – % of seats);

  • SNP – 54 (35.1%)
  • Con – 37 (24.0%)
  • Lab – 35 (22.7%)
  • Green – 14 (9.1%)
  • Lib Dem – 14 (9.1%)

That’d be a total of 154 seats, inflating the size of Holyrood by nearly a fifth. As the SNP are the party that is overhanging, they don’t end up with any more seats, all 25 of them being spread between the four other parties.

Now, I’m not suggesting for a second that Scotland should adopt either of these compensation mechanisms. I recognise that although we use PR for most elections in Scotland, we nonetheless are still very wedded to the UK’s traditional FPTP political culture. The idea of adding more politicians (and the cost that would, or would be perceived to, entail) just to achieve proportionality would definitely get ordinary voters backs up. As a separate issue though, it’s always worth remembering perhaps it is time to permanently increase the number of MSPs given how much their responsibilities have grown since 1999.

Instead, the aim of this post is to act as another reminder that in trying to balance “local” single-member constituencies with fair, proportional representation, you’re always at the risk of ending up in awkward situations. Whether it’s inflating one party’s representation or inflating the whole parliament to combat that, this is just another way we need to find a compromise when using mixed systems. No system is perfect… but at least mixed systems are miles better than pure FPTP.