This is a bummer for the Right, because we watched the Democratic Convention and saw a plethora of presentations that we thought would be weak tea against the cold hard facts of 44 straight months of unemployment above 8 percent: an unaccomplished mayor playing attack dog with a smile; a fluffy depiction of the president’s life from his wife; a Bill Clinton speech that started effective and then just seemed to go on forever; the vice president alternating between shouting and whispering and the president reheating the leftovers in the back of his speechwriter’s refrigerator.
If you think that there’s a bit of a Democrat tilt to weekends, you may dismiss these results a bit. Does weekend polling give you different results? Some pollsters insist it doesn’t. Here’s a comment attributed to Tom Jensen of Public Policy Polling:
I like weekend polling because I can attempt to reach everyone at 10 AM Saturday, 3 PM Saturday, 8 PM Saturday, 4 PM Sunday, and 8 PM Sunday. . . . The vast majority of people are at home at least one of those times. I find that we get a higher response with that than if we call for instance at 7 PM Monday, 2:30 PM Tuesday, 7 PM Tuesday, 10 AM Wednesday, and 7 PM Wednesday.
(Of course, if the respondent household is on a weekend trip, PPP is out of luck.)
But common sense would tell you that people’s schedules are pretty different on the weekend than they are on the weekday or weeknight: high-school football on Friday nights, college football on Saturdays, the NFL on Sundays, religious services. And while some pollsters may call back the same randomly selected number in order to try to reach that household, some may just go through their list. Thus, if you poll Sunday mornings, you’re going to lose a lot of Christians. If you poll small towns in Texas on a Friday night, you’re probably not going to get many respondents.
This article from Slate in 2006 argued that polling on non-school nights would make the samples likely worse for Democrat . . .
It’s harder to get people on the phone on Friday and Saturday nights. Pollsters try to collect a fair sample of potential voters by dialing phone numbers selected at random. But the people who are most likely to answer their phones may not be representative of all Americans. For example, the poll samples could have a bias toward the older people who are more likely to be at home on any given night.
Pollsters don’t want to skip anyone with an active social life, so they try to call back when people don’t pick up the first time. A typical opinion poll might be conducted over five school nights, from Sunday to Thursday, with up to 10 calls made to each number. (They keep calling — during the day and at night — until they get through.)
As an election approaches, pollsters try to gather data more quickly. They might run a two-day poll instead of a five-day poll, and they might conduct it over weekend nights instead of during the week. (Some magazines, like Newsweek, time their polls to coincide with a weekend publication deadline.) In theory, younger people are more likely to be out on Friday and Saturday nights, which would make them less likely to be included in the sample.
What would that mean for the results of a given study? Weekend polling would skew the sample away from the young and active types and toward the oldsters who sit at home. That doesn’t mean the weekend poll gives more credence to the elderly vote. It might mean just the opposite: Pollsters can correct for having too many old people by giving extra weight to everyone else. In that case, the opinions of the few young people who are in the sample would count extra.
That means one possible source of bias in a weekend poll comes from the young voters who happen to be staying in on Friday and Saturday nights. If they happened to vote more Democratic than their more socially active cohorts, the poll will skew toward the Democrats.
Of course, who’s out more on Friday and Saturday nights if we’re playing the stereotype game? Those hip, young Democrats, or those wealthier Republicans who have the disposable income to be out?
Either way, one of the fascinating facets of the polling during Obama’s presidency has been the fact that the monthly unemployment report has little if any discernible effect on Obama’s approval rating or head-to-head numbers against Romney in the tracking polls. This may meant that nearly half of Americans don’t blame Obama at all for the high level of unemployment for the past four years. But I’m starting to wonder if the traditional release date of the numbers — 8:30 a.m. on the first Friday of the month — is actually good news for the Obama administration. The numbers are released late enough in the morning to miss the morning commuters. Americans who don’t work in the financial sector may or may not hear about the latest numbers during the day; by Friday night, they’re already thinking about the weekend. The newspaper articles about the new numbers appear on Saturday, the least-read day of the week. Maybe there is additional coverage in the Sunday paper (the most-read day of the week), maybe not. But by Monday, the news cycle has moved on — anyone who’s talking about the jobs numbers is talking about something that’s roughly 72 hours old, ancient history in the worlds of the blogosphere, cable news, etc.
A future president who felt really good about the monthly jobs numbers would want them released on a Monday morning, so that they set the tone for the week . . .
As usual, Allahpundit reminds us to keep an eye on which sample is showing what size a bump: “A few readers have e-mailed to note that Gallup’s job approval number is based on a sample of adults, not registered voters. Right; it’s the head-to-head with Romney, where O now leads by three, that’s based on a sample of registereds. Apologies if that wasn’t clear. The point is, if we’re seeing his job approval rise among adults, that’ll probably have some similar but lesser effect among registereds and then a similar but still lesser effect among likelies. Not a huge bounce, but a little something.”