Share this post
FaceBook  Twitter  

Will the Republicans retain their majority in the U.S. House of Representatives this year?


Polly

POLLY WAS RIGHT! Who would win in November - George W. Bush or John F. Kerry? This question consumed Polly since March, when her page was launched. She heard from many sources, including 268 polls, 10 forecasting models, three surveys over as many months of a select panel of American politics experts, and the Iowa Electronic Markets (IEM). All of it she dutifully reported in her tables and graphs. As well as passing along what she found, Polly performed a simple computation, something she calls “pollynizing.” Averaging across methods, which Armstrong (2001) shows reduces forecasting error, Polly calculated the Pollyvote, the share of the two-party vote (that is, omitting third-parties) that Bush was predicted to win. Once or twice a week our parrot would post the latest value of the Pollyvote on this page and, invariably, even as Bush’s standing in the polls sank in July (see Pollygraph 2) or plummeted in the IEM in August (see Pollygraph 3), this variable showed Bush would win on election day. Not once during the past eight months did the value of the Pollyvote dip below 50 percent (see Pollygraph 4). (In fact, the mean was 51.7 and the standard deviation 0.66.) Then, on the morning of November 2, Polly posted her final forecast: President Bush would take 51.5 percent of the two-party vote. This hit bull's eye! As far as Polly knows, no other forecast, by pollster, pundit, or scholar, got it exactly right.

Polly's Tables

Table 1. Polly's Forecasts, by month. This table displays monthly data on polls, Iowa Electronic Market quotes and quantitative models forecasts going into the very latest Pollyvote. The Pollyvote is a forecast of Bush's share of the two-party vote. It is computed by taking the mean of the following: the most recent three-poll rolling average of Bush's share of the two-party vote in the polls, the seven-day rolling average of the Bush|Kerry quotes in the week prior to the latest poll reported, and the average of the quantitative model forecasts. (PDF)

Table 2. This Table includes all poll data, by date and source, going into the Pollyvote, March 8 to November 2. (PDF)

Pollygraphs

Pollygraph 1. Pollyvote, Bush|Kerry Quotes in the Iowa Electronic Markets, and Bush's Share of Two-Party Vote in the Polls, March 8-present. This pollygraph displays the Pollyvote, the Bush|Kerry quotes in the Iowa Electronic Market, and Bush's share of the two-party vote in the polls across time. The Pollyvote is a forecast of Bush's share of the two-party vote. It is computed by taking the mean of the following: the three-poll rolling average of Bush's share of the two-party vote in the polls, the seven-day rolling average of the Bush|Kerry quotes in the week prior to the latest poll reported, and the average of the quantitative model forecasts.

Pollygraph 2. Bush's Share of the Two-Party Vote, March 8 - present. This pollygraph displays Bush's share of the two-party vote across time, along with a three-poll rolling average trend line.

Pollygraph 3. Bush|Kerry Quotes in Iowa Electronic Markets, March 8 - present. This pollygraph displays the quotes across time, along with a seven-day rolling average trend line. Bush|Kerry quotes are contracts specifying that Bush will beat Kerry by a certain percent of the two-party vote and are paid off according to the actual vote cast in November. Thus, the quotes reflect what traders in the IEM expect Bush's share of the two-party vote will be.

Pollygraph 4. Pollyvote, March 8 - present. This pollygraph displays Polly's forecast of Bush's share of the two-party vote across time.

Pollygraph 5. Polls and the IEM, March 8 - present. This pollygraph displays Bush's share of the two-party vote in the polls and the IEM across time.

Pollygraph 6. Pollyvote and the IEM, March 8 - present. This pollygraph displays Bush's share of the two-party vote in the Pollyvote and the IEM across time.