Bush is tanking. Here are the market reports, as of a few minutes ago. Traffic on the sites is heavy, and it's hard to get through.

Tradesports:

                    Bid    Ask    Last    Change

PRESIDENT.GWBUSH2004    36.0    37.0    37.0    -16.0

PRESIDENT.KERRY2004    61.0    62.4    62.0    +15.0

By the time I finished formatting this table, the spread had increased by eight cents to

Tradesports members have Bush below 50 -- and falling -- to win in Florida, Iowa, New Mexico, Ohio, Pennsylvania, and Wisconsin. Colorado's "bid" price for Bush to win is 64 cents; it closed yesterday at 77.

Iowa market at 3:45 Central Time (by party, and by margins of greater/lesser than 52% of the pop.vote):

Symbol        Bid    Ask    Last    Low    High    Average 

DEM04_G52    0.264    0.288    0.288    0.130    0.299    0.191 

DEM04_L52    0.371    0.398    0.395    0.301    0.400    0.332 

REP04_L52    0.180    0.230    0.219    0.186    0.398    0.337 

REP04_G52    0.079    0.099    0.100    0.050    0.163    0.119 

Iowa Market: Vote Share. These figures are a few minutes older than the winner table:

KERR        0.511    0.519    0.515    0.495    0.520    0.509

BU|KERR    0.480    0.501    0.482    0.478    0.521    0.484

In the June 2004 federal election in Canada, ALL the posters got it wrong. There was a last-minute slump in opposition party support, and a last-minute surge in incumbent Liberal party support, that resulted in a Liberal parliamentary minority government, rather than the pundit-expected Conservative minority, or even majority. What was especially interesting, though, was the introduction into Canada of two relatively new types of polls. SES Research, on behalf of CPAC (the Canadian equivalent of C-SPAN) conducted a three-day-rolling tracking poll. JMCK, an Alberta outfit, conducted several publicly-released automated polls. ("Push 1 for Liberal, 2 for NDP"..., etc.) Both boasted of having large sample sizes relative to the typical telephone poll. The last public polls by the various major firms predicted:

		Lib	Con	NDP	BQ	Oth
JMCK		32.2	33	17.2	9.4	8.1
Compas		36	31	17	11	6
Ekos		32.6	31.8	19	11.2	4.9
SES		34	30	20	12	4
Ipsos-Reid	32	31	17	12	8
Environics	33	33	18	11	5

The final election-night result was:

	Lib	Con	NDP	BQ	Oth
ACTUAL	36.7	29.6	15.7	12.4	5.6

That was an average per-party absolute error of:

		Lib	Con	NDP	BQ	Oth
Compas		0.7	1.4	1.3	1.4	0.4
SES		2.7	0.4	4.3	0.4	1.6
Ipsos-Reid	4.7	1.4	1.3	0.4	2.4
Ekos		4.1	2.2	3.3	1.2	0.7
Environics	3.7	3.4	2.3	1.4	0.6
JMCK		4.5	3.4	1.5	3.0	2.5

The average per-party "off" score, by pollster, was:

		Ave
Compas		1.0
SES		1.9
Ipsos-Reid	2.0
Ekos		2.3
Environics	2.3
JMCK		3.0

This was admittedly the first time out for JMCK, but its automated push-button polls did the worst on the national figures of any of the public polls. (Its regional breakdowns were even more abyssmal, especially in Quebec.) SES was good, but not noticeably better than its competitors over similar polling periods. Lesson learned: take polls with a grain of salt. Take automated polls with an even bigger grain of salt.