Consistency of variable "lineno"

To whom this may concern–

I have been linking the responses of those who participated in both the Jan/Feb and March supplement in year x (specifically in 1995, 1997, 1999, 2001, 2003, 2005, 2007). To do this, with the guidance of IPUMS staff member Tim Moreland, I merged the IPUMS CPS data with the NBER data to gain unique person identifiers, and believe to have done this successfully.

My remaining question, however, is in regards to the usefulness of the variable “lineno.” I have been using lineno as one of the several variables to uniquely identify respondents and link them throughout months in year x. However, it got me thinking that even if a person participated in both Jan/Feb and March of year x, isn’t it possible that they have a different “lineno” in Jan/Feb than they do in March of year x?

Thanks in advance for any help you can provide–


The line number variable (LINENO) records the order in which members of a household are listed by the interviewer. Typically, the head of household is listed first, followed by the spouse, their children, members of subfamilies and unrelated persons. The respondent’s LINENO should then remain constant over time, barring a recording error.

In addition, it is possible to identify any LINENO changes by verifying that a respondent’s age, race, and/or sex remain constant from month to month, with the exception of age being allowed to increase by one. These few mismatches can then either be dropped or an effort can be made to update a respondent’s LINENO by utilizing demographic variables. Thus, for the small potential subset of respondent’s whose line number does change, it should not cause an issue for your analysis.

See this paper for a more in-depth discussion on the accuracy of matching CPS respondents across time and potential errors.

Hope this helps.


Since my question regarding lineno is similar to the one above, I thought I would continue the thread.

Based on the definition provided above, lineno seems to be a relational identifier rather than a unique identifier. If this is the case, then I can think of a few examples in which the lineno will be the same, but the individual within the household it refers to will differ. For example, if the head of the household divorces his/her spouse and remarries a new spouse before a subsequent wave of the survey, would the new and old spouse have the same lineno?



Yes, in the case you describe, the new and old spouse would likely have the same line number. This is why it is always important to check for the validity of links using CPS identification variables. Back in the summer of 2018 we held a workshop on linking the CPS samples. This included a session specifically on validating links over time. These resources should be helpful to you.