Paulitics: Paul’s Socialist Investigations

….

….

Paulitics: “#1 most accurate election forecaster” (2007 Ontario General Election)
-Freddy Hutter, TrendLines Research

~

Paulitics 2008 Canadian Election Seat Projections
(Meta-analysis)

Last updated: 27 September, 2008

							Other
27-Sep	Hill & Knowlton (5-poll ave.)†	154	60	42	51	0	1
25-Sep	DemocraticSpace††	146	80	32	48	0	2
27-Sep	Paulitics Arithmetic Projection*	149	71	36	51	0	1
27-Sep	Paulitics Geometric Projection**	150	65	41	51	0	1
27-Sep	Trendlines.ca‡	150	81	32	44	0	1
20-Sep	electionprediction.com‡‡	146.7	96.5	23.8	38.3	0	2.6
26-Sep	LISPOP projection¤	153	86	27	41	0	1
21-Sep	Nick Boragina of ‘niXtuff’¤¤	151	90	28	37	0	2
							Other
	Average Seat Projection:	150.0	78.7	32.7	45.2	0.0	1.5
	Party Standings at Dissolution:	127	95	30	48	1	7
	+ / -	+23.0	-16.3	+2.7	-2.8	-1.0	-5.6
	Government Type Predicted:	MIN.	no	no	no	no	no
	Maximum Seat Projection:	154	96.5	42	51	0	2.6
	Minimum Seat Projection:	146	60	23.8	37	0	1

NOTES:

† The Hill & Knowlton seat projection data was obtained by entering the most recent Paulitics rolling five-poll average into the Hill & Knowlton election forecaster ’split’ function (as opposed to their ’swing’ function).  The rolling-five poll average was entered as a ‘national’ split as opposed to entering the  provincial/regional data from the Paulitics provincial/regional polling resource.  This was done because Hill & Knowlton’s predictor for certain provinces has highly irregular properties which in my opinion reduces its predictive value.

†† Greg Morrow of DemocraticSpace does a riding-by-riding prediction to obtain his seat projections.  Thus, while both the Paulitics arithmetic and geometric measures by definition do the same thing, for the purposes of this meta-analysis, Greg Morrow’s final seat projection is the important detail.

* This projection data was obtained by taking increases/decreases in support for each political party and applying them to every riding’s 2006 election results as an “arithmetic” function (as opposed to a geometric ratio).  An “arithmetic” function takes increases or decreases in a given party’s support as more or less evenly distributed throughout the province or region in question.  What this means is that, if Party “x” increases its support from, say, 5% to 10%, using an “arithmetic” function, we will add 5% to the vote total of Party “x” to every riding to determine the final seat projection.  Because of this, this way of predicting seat totals is sometimes called the “universal swing” measure and is most commonly used in the U.K..

The application of arithmetic increases/decreases in each party’s support was not done at the national level, but rather was done individually in every region — British Columbia, Alberta, Prairies [MB & SK], Ontario, Québec and Atlantic Canada — to increase accuracy.

** This projection data was obtained by taking increases/decreases in support for each political party and applying them to every riding’s 2006 election results as a “geometric” ratio (as opposed to an “arithmetic” function).  A “geometric” ratio takes increases in a given party’s support as more or less concentrated in areas where the party in question is already strong.  What this means is that, if Party “x” increases its support from, say, 5% to 10%, using a “gemetric” ratio, we will multiply the vote total of Party “x” in every riding by 2 to determine the final seat projection.

The application of geometric increases/decreases in each party’s support was not done at the national level, but rather was done individually in every region — British Columbia, Alberta, Prairies [MB & SK], Ontario, Québec and Atlantic Canada — to increase accuracy.

‡ Trendlines puts out both a current seat projection (located on the graph) as well as something of a seat prediction which assumes that the current trends in party support will continue on unabated until the election day. For the purposes of this meta-analysis, Paulitics uses the current seat projections for trendlines.ca and NOT the seat predictions.

‡‡ Electionprediction.com’s current seat projections still list many seats as ‘too close to call’.  For the purposes of this meta-analysis, these seats have been redistributed amongst the various parties in proportion to each party’s current seat projections on electionprediction.com.  It is important to note here that the “other” column for electionprediction.com may contain Green Party seats OR simple independents.

¤ Unknown methodology

¤¤ Unknown methodology

~

If you would like to make a prediction yourself (please: no gambling for money here, just for bragging rights) or if you have any questions about any of the methodology used in this meta-analysis, please feel free to leave a comment below.