State-level Wireless-only Estimates

March 12, 2009. Most of us who work in the health survey field have been aware of the increasing number of households that rely on cell phones and do not have traditional landline phones, with the latest estimates showing over 17% of U.S. households as wireless-only. This creates challenges for telephone surveys as the cell phones samples are more difficult and costly to reach but important to have in a representative sample. While we have known the national estimate of wireless-only households based on the work of Stephen Blumberg and his colleagues at the National Center for Health Statistics (NCHS), very little is known about these estimates at the state level.

Sharon Long, senior researcher at the Urban Institute and one of our SHARE grantees suggested that SHADAC researchers work with the NCHS to develop state-specific prevalence estimates of wireless-only phone coverage at the state level. Stephen Blumberg and Julian Luke at the National Center for Health Statistics collaborated with SHADAC’s research team of Gestur Davidson, Amanda Yu, Karen Soderberg and Mike Davern to develop a research plan to develop such estimates using non-public data from NCHS. These new estimates, released yesterday by the NCHS, will of great use to not just Sharon but other folks in the SHADAC network who conduct state surveys.As you will see in the brief, the modeled estimates show quite a lot of variation between states and demonstrate that a single national estimate is not necessarily effective to apply to state surveys. Wireless-only estimates are low for California and New York, states that have high-density populations, yet high for the rural states of Iowa, Nebraska and Oklahoma. Understanding why these variations exist will be interesting future research.

The state-level estimates are important for state health surveys because they provide a population estimate for the ever-increasing prevalence of wireless-only households. These households are typically excluded from telephone surveys – the mode commonly used to conduct state health surveys. Understanding the size of this population within each state will help researchers understand the implications of excluding them from the target population. This will also help refine post-stratification adjustments to account for the resulting bias.

SHADAC has also released a new issue brief that describes the issues around wireless substitution as it relates to state surveys, and also presents the NCHS estimates. The NCHS release is also available for download, and offers a technical appendix to explain the model used to develop these estimates. SHADAC researchers will continue to stay abreast of this topic and present information as it becomes available.

This collaborative work represents another example of the fine collaboration between SHARE grantee technical assistance requests, SHADAC researchers along with the experts from federal government agencies.