Our polls, like many others on the Web and in the real world, are unscientific because we don't identify the population of people we want survey, randomly survey a sufficiently large subset of that population, or use a polling method that's designed to eliminate bias. With last week's poll, however, we had an unusual chance to compare what people said when responding to the poll to how they acted. We asked a simple question: "Which Macintosh Web browsers do you use on a regular basis?"
Here's how the results broke down for the four browser versions in common use. Note that the poll enabled people to submit multiple responses if they regularly use more than one browser, so this table shows the percent of people who voted for each answer, not the percentage of votes for each answer.
Netscape 4.x: 67 percent (955 responses) Internet Explorer 5.0: 28 percent (400 responses) Internet Explorer 4.x: 25 percent (357 responses) iCab: 21 percent (304 responses)
Our poll received 2,196 responses from 1,419 people, which could mean roughly half the respondents indicated they use more than one browser. However, it could also mean 1,200-odd people only use one browser, while 200 or so use four or more. There's no way to know.
Let's compare those results with the actual browsers used to view the main TidBITS Web site last week - even though this is spurious from a statistical point of view. I ran our log file through Active Concepts' FunnelWeb log analysis program, which is one of the few Macintosh log analysis programs that let me easily separate out hits attributable to Macintosh Web browsers and provided the results by visitor, rather than page views or overall file hits.
According to FunnelWeb, we had 11,203 visitors to our main Web site last week whom we could positively identify as using a Macintosh Web browser. The percentages break down as follows:
Netscape 4.x: 41 percent (4,639 visitors) Internet Explorer 5.0: 26 percent (2,950 visitors) Internet Explorer 4.x: 22 percent (2,500 visitors) iCab: 2 percent ( 238 visitors)
At first glance, although the numbers for Internet Explorer match fairly closely between the poll and the log, it would seem that more people claimed they use Netscape and iCab than actually used those browsers when they visited our site. This is borne out in terms of the percentage of requests made on our Web site last week by Macintosh browsers: 53 percent came from versions of Internet Explorer, whereas Netscape accounted for only 39 percent.
These results imply iCab was more likely reported as an additional response rather than as a solitary answer to our poll - which is hardly surprising, given iCab's pre-release status. It's surprising that the numbers could also imply the same is true for Netscape, although to a much lesser degree. Poll respondents were fairly likely to say they used a Netscape browser regularly, but site visitors were less likely to be using a Netscape browser.
Another possibility is the "PBS answer" - when people are asked to report on their television viewing habits, they often claim to watch educational shows on PBS instead of network sitcoms. Simply put, it's reasonable to assume at least some people treated the poll as a chance to vote against Microsoft in a popularity contest - particularly with the release of Internet Explorer 5 and the pending anti-trust announcements.
No matter what the reason for these discrepancies, it's fascinating to have a chance to see how the results of a poll compare to numbers we can actually measure, even indirectly.