SNP on course for majority, Alba goes to Holyrood


When it comes to trying to predict election results with modelling you need to take the bad with the good.

The results of your modelling may not show the outcome you desire.

But you must resist the temptation to alter your modelling to reflect your own personal world view.

However, at the same time, you can theorise why the results your model produce do not match your expectations.

This brings us to the well know phase in the computing world "Garbage in = garbage out", in other words, the results you get are only as reliable as the data you give the model.

So in this context that is the polling results, is the polling reliable?

The truth is we won't know until the actual election results are in.

But why might the polls be inaccurate, well historically new political parties have a hard time being picked up by the polling companies.

We are also getting some very different results from different pollsters.

This could be in-house bias, sampling variation.

There are two pollsters in particular who are are the extremes when it comes to Alba.

Comres have always been bad for Alba, Panelbase has been the best for Alba.

Comres initially put Alba on 1% but have since seen it climb to 2%

Panelbase started at 6% and have since fallen to 4%.

It is expected that different pollsters results will start to converge the closer we get to the actual election day and this analysis assumes no more polls come out tonight or tomorrow.

I do have to say I have concerns about Comres given the recent difficulties the firm has faced with having to change results after publication and having to issues statements to correct reporting on another poll.

Both of these are highly unusual things to happen in the polling industry.

Despite this you cannot discount them, because even if their polling looks off, it could be that they alone are actually correct - we'll know soon enough.

But this does leave me less confident about the results my model has produced, given that it uses data from all the polling companies including Comres. As I said before garbage in, but we just don't know yet if we are feeding the algorithm garbage.

Anyway here's my results.

For the Constituency vote

The headline here is obviously the SNP gaining 6 seats at the expense of both Labour and the Tories.

The next headline is Labour not being able to win any constituency seats.

The other fact to take away is the Green are at best wasting their own time and voters time standing in the constituencies and at their worst damaging the Independence cause.

For the List vote


The headline here is Alba picking up 3 seats in Central, North East, and the South.

It really disproves the common myth that it is a waste of time voting for Alba in the South Scotland region.

In fact, the reason they pick up a seat here is because the SNP take 2 seats from unionists in this region on the constituency vote, 1 from Labour and one from the Tories.

That could be due to SNP popularity, but it is also just as likely due to Alba voters giving their constituency vote to the SNP proving that Alba voters are not doing damage and are in fact helping both parties.

I know Alex Salmond is predicting the SNP will not gain any list seats, my model says they will pick up 1 in the Highlands & Islands region.

You also might notice the Lib-Dems doing surprisingly well in my results, and the Greens not making the expected gains, it certainly surprised me.

At first, I thought I'd made a mistake and swapped the Greens and Lib-Dem figures around, but then again I knew from past runs that my model has always been quite unfavourable to the Greens.

So looking through the results in detail it became clear that it is a result of the unique way my modelling algorithm works, so it'll be interesting to see if this is reflected in the final results. It could prove my model to be far more accurate than anyone else's or reveal a flaw in my logic, assuming the data from the pollsters turns out to be good and not garbage.

Overall results


So SNP gets the overall majority they want, Alba get represented at Holyrood and Tories & Labour lose 11 seats between them. The Lib-Dems double their seats but remain third.

I'd call that a win.

Comments

Post a Comment

Popular posts from this blog

Update on DC Thomson poll

Image
So we have more info on the DC Thomson poll. There are some reports that the questions asked were pretty standard and did indeed ask specifically about Alba, although the popularity ratings result still look very suspicious to me, that could be an in-house bias with the polling company or evidence that the concerted effort from all political sides, including the SNP, has had the desired effect. Someone on ScotGoesPop suggested that the fieldwork was from the 18th of March, but that doesn't make sense and the official fieldwork dates are given as 29-30 March. That's probably as quick as you could possibly put together a poll after the launch of Alba so we're still left with the impression that this was a rush job to try and kill off Alba before they had any chance to make any headway in the polls. And indeed the unionists and SNP loyalists have jumped on it as proof that Alba is dead in the water. However, others seem quite reassured and positive that Alba managed 3% just 3 ...
Read more

My thoughts on Independent candidates

Image
I have always found party politics to be largely off-putting. It inevitably attracts careerists, especially to the most popular party, who want the party money, machinery, and status to add legitimacy to their campaign and careers. Independent candidates, on the other hand, choose to reject all those advantages. Some will be doing this because they have fallen out with, kicked out, or been rejected by the established parties. Others will be standing on single issues, probably not expecting to win, but hoping to raise awareness of whatever issue it is they feel is not being addressed. Then of course you'll get the people doing it for a laugh or the attention, the joke candidates who are not stealing any votes from serious candidates and can be safely ignored. It doesn't need to be said that the chance of an independent candidate beating the favourite party candidate in any constituency is low, but the chances are better standing as a list candidate. It can only really be done by...
Read more

Alba soar in latest Poll

Image
As predicted in my previous post the latest poll from Panelbase as reported by James Kelly HERE  makes the Courier/DC Thomson poll look like the rush job designed purely to try a kill the threat of Alba at birth. With the dataset finally released for the Courier poll, we can finally say that support for Alba varied by quite a margin, with some regions as high as 5% and some as low as 1%. Of course, this was to be expected as nobody would expect the traditionally conservative South of Scotland to be attracted by the prospect of another independence party. So this will of course mean that the 6% Panelbase shows in this latest poll is equally varied by region and will result in more seats than others in those regions where support is higher. The other point I made was the exclusion of Alex Salmond from the leaders' debate and his complaint to OFCOM in regard to this exclusion. I said we needed another poll showing increased support before OFCOM made any ruling and here it is showing A...
Read more