Monday, April 15, 2019

Alberta election projections

I wanted to go through the process of making an election projection with Alberta as the target. I began by doing a poll average of all 10 polls since the election began.

This is what I end up with. Next I need to add my adjustments for Edmonton and Calgary, as, Calgary is less pro-blue and Edmonton more pro-blue when compared to last election.

As you can see, not much change in sum, but a few ridings do flip.

An important step is to take account for who is running and where. This changes further ridings. Note a counting error on the map (which I've decided not to change to explain how these happen) with the Alberta Party. Since there are more UCP ridings than any other, I simply counted all the non-UCP seats, tapping my finger, or a pen, on the screen to count each one.

After I got that total, I added this up, and subtracted 87, which gave me 56. As such, both the Alberta Party and the UCP have the wrong number of seats shown. However, it is not the number of seats, but the seats themselves which I've calculated in my spreadsheet. As such, that is correct, and the numbers are simply a bonus I've included so you don't have to count them yourself; however in the event the two do not match, its the number of splotches on the map which is correct.

When I colour in the ridings, I paint by number. If my spreadsheet tells me riding #04 is NDP, I find it, and colour it in orange. This can lead to errors when the number on my spreadsheet is wrong, or, when my spreadsheet does not have numbers, only names. In these cases I can misread things, or, in the worst case, mix up a Moncton Centre for a Saint John Centre, as I've done before.

Regardless, on to the next map, which takes trends, expectations, and my gut into account!

This is where things really change. Its harder to explain "why" a lot of this happens, as its heavily based on my gut and instinct. In order to make sure I'm still grounded, this is the next step:

When I make my projections, this is the step I adjust the most. During any random given election, I will tend to intentionally do at least one projection with this dialed up to 11, and one with this set to minimum; and this is how you can sometimes see wild swings in my numbers when the polls are stable.

This is because "how biased am I" is not an easy question to answer. One question in particular that I do not get nearly as often as I should, is "what do (I) want to happen"

This is the answer to that question. you'll note some perhaps strange riding swaps. I like it when all parties are competitive in all areas. I like it when smaller parties do well. In particular, I like the fact that Alberta changed their government from the one who governed it since the 70s. Am I thus always biased towards the NDP? Heck no, I want the BC NDP to lose. I also want the NB NDP to do poorly, but I want the Manitoba NDP to do well. I want Tories to win in Nova Scotia, but lose in New Brunswick. I want the Greens to win in PEI but to fail to win any seats in Manitoba.

It is very rare that who I want to win and who I want to lose is based on policy. Its frequently based on novelty, change, and making pretty maps. It is very rare that I actually have real position on a big issue.

Despite that, I do still want the NDP to continue in government in Alberta. No government in Alberta has ever "returned" to power. No government has lasted less than three terms. I want these two things to continue.

Despite that

I can't see how the UCP does not fails to win a majority tomorrow.

No comments:

Post a Comment