Downloadable Data

Many people have asked: "Which polls do you use?" To be as transparent as possible, all the raw polling data used is available for viewing and downloading as described below.

A different, but relevant question is: "Do you use every poll by every pollster?" The answer is no. By far, the vast majority of pollsters are consultants who are hired by candidates and parties to help them win elections. Many of them are quite good at keeping their eye on the ball—winning. To them, being honest is a tool to be used selectively, when it helps their overall goal. In general, these companies will release numbers when they make their candidate look good but if the numbers don't look good, they won't say anything. In some cases, they have been known to fudge the numbers or even make them up out of thin air. Furthermore, almost none of them publish their questions, methodology, or crosstabs making independent analysis impossible, whereas legitimate pollsters frequently publish this information (sometimes only to subscribers though). None of their polls are used on this site. Only polls from neutral pollsters, who are generally hired by the media or organizations not working for specific candidates, are used.

There are two borderline cases: Rasmussen Reports and Public Policy Polling. Rasmussen Reports is owned by Scott Rasmussen, who makes no bones about being a staunch Republican and who works for Fox News. He pioneered robopolling (automated calls made by a computer) and can thus run polls very quickly. While no one doubts that he actually runs the polls he says he does, the demographic models he uses to weight his data have been criticized as being skewed towards the Republicans. Public Policy Polling frequently works for left-of-center organizations like some unions but is not a campaign-strategy consultant-type company trying to win an election. It also uses robopolling. Nate Silver, a professional statistician now writing a blog at fivethirtyeight.com, has analyzed the polling data from past elections and concluded that Rasmussen's data was biased towards the Republicans by about 4%. Interestingly enough, PPP's data is also skewed toward the Republicans, but only by a tiny amount, probably due to random fluctuations. Scott Rasmussen has left the company he founded, and it has gone downhill since then. We don't use any of their polls. In contrast, PPP is a pretty respectable firm, so we will use its polls. In general we don't use pollsters that Silver rates below C-.

Viewing the Polling Data

If you just want to view nicely formatted and color-coded polling data (rather than downloading it for use in Excel or Open Office), click on one of these links.

Downloading the Polling Data

The raw (unformatted) polling data is available for the presidential and Senate races here. Depending on your browser settings, these files may be automatically downloaded when you click on them.

Every poll occupies one line.

The flat ASCII text fields are:
State  Dem-%    GOP-%    Ind-%    Starting-month    Starting-day    Ending-month    Ending-day    Pollster

The .csv files are formatted with the following comma-separated fields:
Day,  Poll-length,  State,  (empty),  Dem-%,  GOP-%,  Ind-%,  Ending-date,  (7 empty fields),  Pollster-poll-length

The Day field is the day of the year at the midpoint of the poll, so a poll from Feb 02 to Feb 04 would have Feb 03 as the midpoint and this is day 34.

Polling Data in Graphical Form

If you prefer your data in graphical form, here it is.