Wednesday, July 18, 2007

Cell Phones & Political Polling

As we enter the presidential primary polling season articles are already surfacing questioning if cell-phone only users are distorting polling data. The common response to these allegations is that cell-phone only households account for 12.8% of US households, and their responses are not likely to be much different from their associated demographic group. Initial research has shown cell-phone only voters not to be married, younger, more mobile, less affluent, more likely to be a minority, and more liberal. Pollsters have summarized up to now that the cell-phone only demographic are those less likely to vote and likely are not all that different from other voters in their (18-34) group. We've consistently heard that cell-phone only voters have a minimal impact on results (one to two percentage points) and fall within the margin of error on most standard surveys.

Younger voters, the go-to demographic to describe cellular-only voters, do indeed vote and are a significant part of the electorate (We've explained this before on this blog.) Recent research from the Pew Research Center has dispelled another myth about cell-phone only voters. They are not the same as their land-line counterparts. The Pew Research Center conducted a dual frame survey between Land Line and Cell-Only voters. Across the survey's 46 different questions, there was an average of 7.8% difference between the two groups, with a range of differences from 0% to 29%. The Pew Research Center estimated that the maximum change in the final survey with cell-phone responses added would only account for 2%. The mean change accounted for less than a percent (0.7%). Those results are within the margin or error for most surveys.

While today Cell-phone only voters are concentrated among the demographic groups outlined above, there is evidence to suggest that the proportion of cell-phone only households is only likely to increase and by a result of growth in “wireless” services and aging demographics, the profile of a cell-phone only voter is evolving. Wireless substitution will grow exponentially as the cost of services continues to decrease and their quality and convenience increases. Wireless substitution is also complicated by the ever-evolving technology available. For example, T-Mobile recently launched a Wi-Fi and Mobile Calling integrated phone (dual mode GSM/WiFi) that allows their cellular phone to become a “landline” phone while in range of their wireless router. Other service providers are currently developing similar services.

Some pollsters suggest weighting samples to account for the uncovered population (such as weighting up certain demographics in a college town based on known demographics and voting behavior). While this tactic in general may work in larger, national surveys, a political survey for a state legislative district that captures less than 5% of the 18-34 demographic cannot be reliably weighted within that age group for several reasons. Weighting may serve well as a stop-gap attempt to keep polling data accurate if we believe that as more Americans become cellular-only voters that their opinions will normalize to the population as a whole.

Other “soft sciences” have explored the “digital divide” and its effect on communities. As communication technologies change and integrate into internet hybrid services, such as VOIP (voice over internet protocol) and WiFi based UMA (Unlicensed Mobile Access), voters will continue to be further segmented along various demographic lines associated with access to digital communication technology. We do not believe there will be a dramatic "normalizing" effect between the various groups. No one can be certain yet where demographic lines will be drawn along still emerging technology channels.

Calling cellular phones is complicated for several reasons. Federal law prohibits the use of automated dialing devices when calling cell phones. Calls to cellular phones have to be done by hand, greatly reducing the efficiency of call centers. On the other end of the cost spectrum, cellular users typically have a cost per minute rate plan. Cellular-based voters would incur a cost to participate in a survey, which would likely drive down participation rates. While compensation / reimbursements could be offered for the use of a participant's cellular minutes, this would add additional cost to conduct to surveying. There are also issues of liability. What legal responsibilities does the pollster have if a respondent is driving while participating in a survey and gets into an accident? Cellular phones are also not tied to a geographic location and create difficulties in screening eligible participants. A respondent may live in one area but maintain previous address's area code and phone number. As we've discussed, cell-only voters of today tend to be younger and create some challenges in accurately developing a balanced sample for a political survey.

Moving beyond the problems of contacting cellular voters, some pollsters have increased their reliance on internet polling, such as pollster John Zobgy of Zogby International. While internet polling has its own drawbacks, a conversation for another day, the process of mixed-use polling (to borrow a term from urban planners) seems like the most reliable and cost-effective method of reducing sampling and non-response problems associated with cellular-only households. Mixed-use polling is a blending of surveys such as traditional telephone surveys and accompanying internet polls. The methodology needs to be refined and adapted, but the future of traditional telephone polling is apparent. Adapt or Bust.


No comments: