Introduction

The American Community Survey (ACS) is designed as a monthly mail-out survey with followup by Computer Assisted Telephone Interviewing (CATI) and Computer Assisted Personal Interviewing (CAPI) operations during a three month interview cycle. The ACS is an annual survey of three million addresses with approximately one-twelfth of the sample mailed out each month. All households with a mailable address are sent a mail questionnaire during the first month of the interviewing cycle. The mailable cases are sent an initial letter and reminder card to return the survey. Also a second questionnaire is delivered if the housing unit does not return the first questionnaire within a few weeks time. If a mail form is incomplete or has more household members than allowed on the form then a telephone failed edit follow-up (TFEFU) operation is conducted to obtain the missing information. During the second month, all households which did not return a mail form and for which we can obtain a telephone number are sent to CATI. During the third month, all households which did not return a mail form or for which we did not obtain a CATI interview are sent to CAPI. Those eligible for CAPI are subsampled at two different rates: 2-in-3 for units without a mailable address and 1-in-3 for all other units.

Is the current assumption of CAPI subsampling at 1-in-3, the correct rate? This paper will look at the assumptions behind that rate and see if they are still valid at the current time. The 1-in-3 rate was calculated in Alexander (1993) as 0.38 and changed to 1-in-3 for operational simplicity.