Probability based telephone surveys must utilize a dual frame approach in order to capture the ever increasing cell phone only population.   Until the day comes where it’ll be a single frame approach of only cellular numbers, researchers need to ensure they get the appropriate blend of cell only vs. dual phone users in their sampling allocations.

Marketing Systems Group (MSG) produces quarterly estimates of Cell-Phone Only (CPO) rates for every county in the country^[1]. These rates are based on telephone households (as opposed to occupied housing units) and lag one quarter. A Cell-Phone only household is defined as one in which there is no operational landline telephone and cell phones are used exclusively to make and receive calls.  The CPO estimates are derived by using a combination of survey-based information along with large commercial and administrative data sources.  This triangulation approach provides us the ability to create CPO estimates at the state and county levels.  Moreover, the primary sources used to create these estimates are updated on a continuous basis.  This enables us to create updated CPO estimates each quarter.

MSG is currently the only source of county level CPO estimates available.  Having this granular level of information can benefit survey planning and weight computation/calibration at the sub-state level.  We analyzed our CPO estimates at the county level and observed that many states have big differences in the CPO rate from county to county (see Figure 1). This could have a big impact on studies that infer a given county level CPO rate using the state level estimate.

Figure 1:  CPO variations among counties (minimum vs. maximum) within state compared to the state level estimates.

One limitation of utilizing administrative data in our triangulation methodology is that it does not take into account behavior patterns of dual phone households.  Dual phone households are those that have both land line and cellular numbers but only make and take calls on their cell phones.  With that being said, these estimates are the closest approximations at the county level available.

Whether it be for statistical or productivity reasons, rely on MSG’s CPO estimates as a criterion to obtain the appropriate mix of land line vs. cellular numbers for sub-state dual frame RDD surveys and for weighting calculations.

[1] A Total of 32 counties were combined with other counties or removed due to lack of information.