This may well be the final 'beta test' post for the model, as it's starting to shape up as being stable and making sense. Quite a lot of work put into the model fixing bugs that were introducing nonsense into the prediction, in particular correcting normalisation errors, and generating a more suitable standard deviation for the monte-carlo simulation.
The model will now tighten up the standard deviation the more the
'revert to mean' diminishes. In general this means that it'll have a
tighter range of variation in the simulation the closer we get to the
Some 'fudge factors' have been removed, to be replaced with real data.
In particular Scottish regional polling will be taken into account, but
only on a monthly basis.
As you can see, I've also increased the amount of data you see in the
graph, now showing the probabilities for coalition-government. A future
fix might need to add UKIP to this, but that can wait till they start
winning seats in a significant portion of predictions. The seat count is
now in actual seats, not a percentage, as I'm using a suitable
normalisation method to give a realistic seat-count using whole numbers,
no more fractions of a seat given to a party.
Looking at the simulations, it seems that the Conservative party is at
even more of a disadvantage than they were in 2010. Due to the
continuing constituency-concentration effect the party suffers from,
they have significant issues contesting marginals as their votes are all
stacked up in safe seats. The Conservatives need a significant margin
in any lead over Labour to be the largest party, let alone a majority.
However, Labour don't appear to have a comfortable benefit-margin from
this. Their chances of a being in government as a majority or leaders of
a coalition do not go above 50% until they have a significant lead over
This leads to there being a wide margin in the simulation where neither
party is able to put together a coalition with the LibDems. This doesn't
quite mean that there's a constitutional crisis on the table, as most
results also leave the SNP able to sit in coalition if they wished,
particularly accounting for Northern Ireland party SDLP's tactic
acceptance of the Labour whip. That might lead to a different kind of
constitutional crisis, but not one of there being no possible coalition.
originally posted here