The TELEPHONE variable and the TELNONCEL variable

I’m finding the variable on the micro-data level to analyze the wireless-only (cell-phone-only) household as the NHIS early release reports depicted about “wireless substitution”( NHIS is providing this report starting from 2003.

In NHIS documentation, the variable CURWRKN from 2006 to 2018 provides this information by gathering responses to the question “Is there at least one telephone INSIDE your home that is currently working and is not a cell phone?” and variable CURWRK from 2003 to 2006 provides this information by gathering responses to the almost the same question.

In IPUMS NHIS dataset, I find the harmonized variable TELNONCEL is provided combining CURWRK (2003-2006) and CURWRKN (2007-2010), which provide a easier way to study “wireless-only household” across years. However, this variable is not available for data after 2010.

Additionally, I find the variable CURWRKN with data available after 2010 (2007-2018) is actually combined with INSIDE (before 2007) together as TELEPHONE variable. Apart from not including “is not a cell phone”, the INSIDE question in NHIS questionnaire asks the similar question about telephone inside home.

I find the fact that the variable TELNONCEL is not available after 2010 is quite confusing. At the beginning, I thought it was NHIS that stopped to ask this questions after 2010. It took me quite a while to find that information about this question is still available after 2010 but in another variable called TELEPHONE. I think a few sentences explaining this situation added in the variable description page, if it would not cause too much trouble to IPUMS team, will make it more convenient for other users.

What’s more, the description for the variable TELEPHONE is “Household has telephone”, which is not accurate. After 2006, TELEPHONE is using CURWRKN and I think that survey question is asking about the family characteristics instead of household characteristics. I also find out from the data downloaded from IPUMS NHIS that people in the same household (cases same NHISHID) but from different family (different FMX) can have different value for TELEPHONE.

Hope it makes sense. Thanks!

Hello! Hope my post doesn’t confuse anybody.

Actually I have another question about the TELNONCEL variable. I find that the NIU consists of the majority of TELNONCEL in the years 2003-2006.

From what I learned from the NHIS early release report about “wireless subscription”, it seems they didn’t get so many missing values for this information. Like this report containing data from 2003 to 2006: Their data point are large enough to calculate the proportion of landline-only and this metric are always larger than 20%.

It would be of great help if anyone can explain why there are so many NIUs in this variable or alternative way to get the micro-data level information about “landline-only” and “wireless-only” of NHIS respondents in the years 2003-2006.


Thanks for the note. I think your explanation of the construction of these variables over time is correct. The TELNONCEL variable is only available from 2003 through 2010. The TELEPHONE variable is available from 1963 through 2018 (or the most recent integrated survey). The questions informing the TELEPHONE variable do change in 2006. As is stated on the description tab: “For all persons, TELEPHONE indicates if there is at least one telephone that is currently working inside the house. Beginning in 2006, the question explicitly excludes cell phones.” I’ll discuss with the IPUMS NHIS Team about any sort of clarification that can be made to the documentation.

Regarding the relatively large share of NIU cases in the TELNONCEL variable from 2003 through 2006, this is due to a change in the universe of the questions informing this variable over these years. As is stated on the universe tab: From 2003-2006 the universe includes “All persons who report their home phone number as a cellular phone number, or for whom it is unknown whether their home phone number is a cellular phone number.” From 2007-2010 the universe expands to “All persons in families who have a phone, or who didn’t know or refused to report their phone number.”

You are right. I missed the “universe text” information in the documentation. Thanks for the clarification! Then it seems we can only know the “cell-phone-only” households and households that have both cell phone and landline phone from the data.
Then I’m curious how they compute the “landline households without a wireless phone” percentage in their early release report. Do you by chance know how they make that?


Although I’m not entirely certain how the CDC calculated these figures, it looks like the information in the TELEPHONE and TELCEL variables could be helpful. In particular, TELEPHONE indicates (since 2006) if the household has access to a landline phone and TELCEL indicates if a member of the family (within the household) has access to a cellphone.

Thanks. The TELEPHONE, TELCEL and TELNONCEL variables are all very useful and from them I knew the complete household telephone status after 2006.

Sorry I didn’t make myself clear. It is in the 2003-2006 period that I cannot know the household telephone status regarding their landline telephone ownership. And from the CDC early release report, I find that they know this information for 2003-2006. The report is here: In the table 1 from their report, they know the proportion of “landline only” households, which means they knew the household landline ownership no matter whether they have a cell phone. This makes me very curious.

Hope it makes sense. Thanks!

After looking into the questionnaire text a bit more closely, it seems there was an error in our documentation. It actually looks like the phrase “and not a cellular phone” is added to the questionnaire starting in 2003. So, the detail in the documentation that says this phrase was added in 2006 is misleading and explains the seeming discrepancy in the available information. That is, starting in 2003 is when TELEPHONE only refers to landline phone access. This error has been corrected and it will show up online with the next data release.

Yes. Glad you find that. It looks like the change in 2007 is only the universe that extend to all respondents, which makes it possible to find the “landline telephone only” households.