Cell phones and Nonsampling Error in the American Time Use Survey Brian Meekins Stephanie Denton AAPOR 2012
American Time Use Survey (ATUS) A Bureau of Labor Statistics survey, conducted by the U.S. Census Bureau Continuous survey about how, where, and with whom Americans use their time Nationally-representative survey of persons age 15 and over 20-minute computer-assisted interview conducted by telephone 2
ATUS Estimates Number and percent of individuals engaging in activities on an average day Average hours spent doing activities Time of day when individuals do the activities 3
ATUS Sampling Frame People are selected from households that recently completed the monthly labor force survey the Current Population Survey (CPS) CPS Survey 2 to 5 months after the end of the CPS survey, selected individuals are interviewed for the ATUS survey ATUS Sample ATUS Survey 4
ATUS Sample During the 8 th wave of the CPS, the respondent may provide a phone number for recontact If selected for ATUS and no phone number is on file, sampled persons are mailed information to call in and complete the survey Do not know if phone number given is cell phone or landline phone 5
Potential Problems with Cell Phone Numbers May result in significant nonresponse bias More difficult to complete Lower levels of cooperation (Link, Battaglia, Frankel, Osborn, Mokdad, 2007; Scott, 2006; Brick, Brick, Dipko, Presser, Tucker, Yuan, 2007) Cell phone households different from landline households Younger, less likely to have children, Hispanic, unmarried (Blumberg, Luke, 2010) 6
Potential Problems with Cell Phone Numbers Cell phone numbers typically linked to individuals versus households ~40% of ATUS sample members are not same reference person for CPS Increases amount of effort needed to interview sampled member Greater nonresponse 7
Potential Problems with Cell Phone Numbers Underrepresented groups in ATUS similar to those that rely primarily on cell phones ATUS has fewer completed interviews than the CPS with single, young renters (Meekins, Downey, Fricker, 2010) Housholds that do not provide number in CPS may be cell only households May contribute to nonresponse bias from lower response among no number households 8
Study Questions What is the impact of calling cell phone numbers on nonresponse? Do cell phone interviews differentially impact the measurement error associated with ATUS estimates? 9
Methods ATUS sample from October 2009 to October 2010 matched to existing telephone number database (Marketing Systems) to identify cell phone numbers Two versions of the Telcordia database Multiple versions of Neustar database 35,298 cases total 36% of these were identified as cell phones 10
Disposition of Cases CPS ref person Disposition Cell Landline Cell Landline Completion 71.5 77.8 71.4 79.1 Overall completion rate is 78.5 for landline and 71.5 for cell phone Not CPS ref person Refusal 10.9 10.9 10.0 9.1 Noncontact 5.9 5.3 7.6 6.6 Not eligible 3.6 1.3 4.1 1.9 Unknown eligibility 8.0 4.7 7.0 3.2 11
Completion Rate Little difference in rate of refusal Contact is more difficult with cell Cell phones only slightly more likely to complete interview during afternoon Telephone type Hard to Reach Reluctant Landline 31.9 17.5 Cell 36.7 17.3 12
Results: Nonresponse Effort by telephone type and CPS reference person Telephone type (CPS ref person) Mean Attempts Mean NC Attempts Landline 9.5 7.1 Cell 11.5 8.9 Telephone type (Not CPS ref person) Mean Attempts Mean NC Attempts Landline 9.9 7.5 Cell 12.0 9.4 13
Demographic Profile Demographic variables from CPS Variable Name Category Cell Landline Housing Tenure Owns 57.2 80.8 Rents 42.8 19.2 Marital Status Married 49.1 60.4 Sep, Div, Wid 18.0 17.5 Never Married 32.9 22.1 HH income Lowest 25 th 20.2 13.7 Middle 50 th 53.8 51.6 Highest 25 th 26.0 34.7 14
Demographic Profile Variable Name Category Cell Landline Age Under 18 7.6 8.2 19 to 30 28.7 11.3 31 to 45 39.4 34.3 46 to 65 21.2 31.5 66 + 3.1 14.7 Race White 77.9 80.8 Black/AA 14.8 13.0 Other 7.3 6.2 15
Demographic Profile Demographic variables from CPS by telephone type and nonresponse Cell Landline Variable Category Response NR Response NR Housing Tenure Marital Status HH income Owns 62.7 42.9 83.0 72.7 Married 53.9 36.5 63.5 48.5 Lowest 17.7 25.1 11.5 21.1 25 th Middle 50 th 52.0 57.1 51.5 52.2 Highest 30.3 17.8 37.1 26.8 25 th 16
Demographic Profile Cell Landline Variable Category Response NR Response NR Age Under 18 7.8 7.1 8.4 7.2 19 to 30 26.5 34.5 10.7 13.9 31 to 45 41.3 34.6 35.7 28.8 46 to 65 21.5 20.4 31.5 31.5 66 + 2.9 3.4 13.7 18.6 Race White 80.1 72.2 82.6 74.0 Black/AA 12.4 21.2 11.3 19.8 Other 7.6 6.6 6.2 6.2 17
Models: Number of Attempts Model with telephone type only Coefficient (cell=ref) Stat sig -0.36 <.0001 Model with covariates Coefficient (cell=ref) Stat sig 2.04 <.0001 LS Means Cell Landline 9.75 9.60 Sig interactions: cps ref person, length of int 18
Nonresponse Model Proportinoal hazard model Outcome survey completion Time attempts (up to 80) Examine the effect of cell phone after inclusion of demographics and process variables (with interactions) 19
NR Model Results Model with telephone type only Hazard ratio Stat sig 1.894 <.0001 Model including other covariates Hazard ratio Stat sig 1.217 <.0001
Results: Measurement Error Variety of Indicators: Not related with income missing on CPS Just as likely to have common activities Less likely to have earnings allocated Not more likely to have bad activities No more likely to have DK or refusal, in general 21
Measurement Error Category Cell Landline Interview time (min) 18.4 19.6 Total number of activities 15.1 19.3 Percent rounding earnings 15.5 17.5 Slightly more likely to round earnings Number of activities is less Interview time only slightly less 22
Time Use Estimates (unweighted) Time Use Major Category Cell Landline Personal care* 410.9 448.0 Household activities* 77.3 98.9 Caring for other HH members 37.5 35.7 Work* 133.7 123.0 Education 14.7 16.4 Consumer purchases* 16.4 20.7 Household services 0.4 0.5 Eating and drinking* 47.0 54.1 Socializing* 179.6 209.6 Sports and exercise 15.8 18.2 Religious* 8.0 9.9 Volunteer* 6.5 9.6 Telephone calls 3.8 4.5 Traveling 54.4 58.0 * Statistically significant at.0001 23
Time Use Estimates (weighted) Time Use Major Category Cell Landline Personal care 564.9 563.2 Household activities* 99.6 113.3 Caring for other HH members* 41.6 33.1 Work* 223.1 188.0 Education* 30.1 36.0 Consumer purchases* 19.9 22.8 Household services 0.6 0.8 Eating and drinking 65.5 67.4 Socializing* 242.8 265.1 Sports and exercise 21.5 22.8 Religious 8.9 8.6 Volunteer* 7.1 10.3 Telephone calls 6.1 5.7 Traveling* 77.2 72.8 * Statistically significant at.0001 24
Models: HH Activities Model with telephone type only Coefficient (cell=ref) Stat sig 13.72 <.0001 Model with covariates Coefficient (cell=ref) Stat sig 1.05 0.907 LS Means Cell Landline 93.45 101.93 Sig interactions: cps ref person, time of day, length of int 25
Models: Work Model with telephone type only Coefficient (cell=ref) Stat sig -35.09 <.0001 Model with covariates Coefficient (cell=ref) Stat sig -19.82.1978 LS Means Cell Landline 160.97 143.07 Sig interactions: cps ref person, length of int 26
Models: Education Model with telephone type only Coefficient (cell=ref) Stat sig 5.83 <.0001 Model with covariates Coefficient (cell=ref) Stat sig -2.98 <.0001 LS Means Cell Landline 49.57 55.99 Sig interactions: cps ref person, time of day 27
Summary Sample members that volunteered cell phone were: Somewhat more difficult to contact Spent slightly less time on phone and provided fewer activities Provided different time use estimates (weighting did not alter this) Nonresponse could not be fully explained by process variables 28
Summary Direct effects greatly diminished in models, but interactions with process variables still important resulting in significantly different estimates of time use. Interaction with length of interview always strong enough controls? Interactions with process variables of concern, but not surprising 29
Contact Information Brian Meekins Meekins.Brian@bls.gov