"The Cellphone Effect" in Political Polls

One of the top electoral poll tracking sites observes that polls that include cellphones in their samples have Obama ahead by 9.4 percentage points, while the landline-only polls show just a 5.1 point lead.

Nate Silver, who runs the site, noted:

Mark DeCamillo of California's vaunted Field Poll, which include[s] cellphones in their samples, suggested … that it was much easier to get the cooperation of cellphone users on the weekend than during the week. How come? Because most cellphone plans include free weekend minutes. Conversely, one might expect that young people are particularly difficult to reach on their landlines over the weekend, since they tend to be away from home more (especially on a weekend when some nontrivial number of them are out volunteering for Obama). So, while I haven't tried to verify this, it wouldn't surprise me if the "cellphone gap" expands over the weekend, and contracts during the week.

I suspect this effect is some combination of:

1. Higher mobile penetration among younger users who, according to polls, tend to track higher for Obama;
2. Wireline substitution among this same demographic, where many simply don’t have landline phones, so polls that don’t include cellphones will simply not reach this demographic;
3. Cellphone users are typically more tech-savvy, again which is more consistent with Obama voters;
4. Cellphone coverage is better in urban areas, which tend to vote more Democratic, while less densely-populated areas, where coverage is worse and people are less reliant on mobiles, are more conservative (notwithstanding that rural Vermont voters probably cancel out rural Idaho voters);
5. Obama has more effectively harnessed the power of mobile with text-messaging campaigns (e.g., announcement of his VP choice);
6. Day of week and time of day of poll relative to price of minutes, as suggested above; and
7. A lot more that will come out in post-election analysis.

Having recently spoken with several telephone number database providers, mobile telephone number databases are simply not as good as those for landlines, and it’s not clear how the numbers that are available skew demographically.

When the election results are in, and the winners and losers among pollsters are declared, I suspect much more will be written about this fascinating topic.  

Comments

[Brad Schorer]

If database providers can get access to the name and address associated with the cell #’s, then we can get some demographic data applied to those statistics. However, as you know, the mobile carriers do not release that information so the only way to get the data is through opt-in sources on the Internet which are not accurate or current. It's also important to remember cellphones reach an “individual” where a landline reaches a “household”. The one-on-one conversations that can occur through that personalization (i.e., asking for Jim specifically versus the Smith household) goes a long way to an engaging conversation regardless of the plan or weekend minutes available.

[Jay Young]

Good post and really on point. The final results of the election were much closer to the cell phone numbers than the landlines. At some point wireless information is going to have to catch up if further applications and customers are going to materialize. This could really be a valuable information channel.