Partial Interviews in the Consumer Expenditure Survey Laura Erhard, - - PowerPoint PPT Presentation

1 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Exploring the Characteristics of Partial Interviews in the Consumer Expenditure Survey

Laura Erhard, Supervisory Economist Division of Consumer Expenditure Surveys DC-AAPOR/WSS Review-Preview Summer Conference July 10, 2019

2 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Outline

 Brief overview of the CE  Response Rates and Partial Interviews  Research Questions  Findings

3 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Consumer Expenditure Survey

 Collects spending data on the U.S.

Population

 Sponsored by the Bureau of Labor

Statistics, collected by the Census Bureau

 Survey participants report dollar amounts

for all non-investment purchases. Business expenses and reimbursements are excluded.

 Provide expenditure weights for the U.S.

Consumer Price Index (CPI)

 Quarterly Interview and Diary Survey

4 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Data

 2017 CE Quarterly Interview (CEQ)

Personal interview Rotating panel, collected quarterly, each

household interviewed 4 times.

3-month recall Designed to collect larger, recurring

expenses that are easy to recall

5 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Response Rates and Partials

2017 CEQ n Rate (% eligible) Eligible 40,193 100 Complete Interviews 24,479 60.9 Type A Non-interview 15,714 39.1 Insufficient Partials ?? ??

 In CE, interviews are counted as complete if they cover all expenditure

sections (through section 20).

 “Sufficient Partials” are interviews that complete everything, but the

income, assets, and liability questions (section 21 and 22).

 “Insufficient Partials” are classified as Type A non-interviews and are

indistinguishable based on our final interview classification codes.

6 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Research Questions

 Who are those that drop out?  Can they help provide insight on other non-responders?  Can their break-off point inform us about survey design?  Could their data somehow be used in our processing?

7 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Defining a Partial: Data

 Audit trail data: paradata created during a Blaise programmed

Computer Assisted Personal Interview (CAPI) instrument that records a detailed history of the sequence, timing, and flow of an interview

Summarized Audit Trail timing data in survey files Detailed, accessible audit trail tables

8 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Defining a Partial

 Type A (Noninterview) cases with time > 0 for expenditure

sections

 Eliminate cases where cumulative expenditure time < 65

seconds or where demographic questions were never asked.

2017 CEQ n Rate (% eligible) Eligible 40,193 100 Complete Interviews 24,479 60.9 Type A Non-interview 15,714 39.1 Insufficient Partials 294 0.7

9 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Who are these partials?

10 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

10.7% 22.6% 31.5% 35.2% 12.0% 26.1% 27.1% 34.7% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0%

Less than h.s. High school Some college Bachelors or higher

COMPLETE (N=24,477) TYPEA_PARTIAL (n=291)

Respondent Education Level

c2(3) = 3.8, p = 0.28 Cramer's V = 0.01

11 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

80.7% 11.5% 5.2% 2.6% 74.9% 16.2% 5.8% 3.1% 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0%

White Black Asian Other

COMPLETE (N=24,479) TYPEA_PARTIAL (n=291)

Respondent Race

c2(3) = 7.07, p=0.07 Cramer's V = 0.02

12 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

6.5% 15.9% 15.8% 18.0% 19.6% 14.7% 9.4% 10.3% 19.6% 17.2% 18.9% 16.8% 10.3% 6.9% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0%

25 and under 26-35 36-45 46-55 56-65 66-75 75 and older

COMPLETE (N=24,479) TYPEA_PARTIAL (n=291)

Respondent Age

c2(6) = 16.19, p = 0.01 Cramer's V = 0.03

13 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

63.6% 36.4% 59.9% 40.1% 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0%

Owner Renter

COMPLETE (N=24,479) TYPEA_PARTIAL (n=294)

Household Tenure

c2(1) = 1.76, p = 0.18 Cramer's V = 0.01

14 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

29.7% 34.4% 14.4% 21.5% 29.6% 26.9% 18.0% 25.5% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0%

1 CU member 2 CU members 3 or 4 CU members 5 or more CU members

COMPLETE (N=24,479) TYPEA_PARTIAL (n=294)

Household Size

c2(3) = 9.60, p = 0.02 Cramer's V = 0.02

15 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Region

17.8% 21.9% 35.8% 24.5% 23.8% 21.1% 29.6% 25.5%

0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0%

Northeast Midwest South West

COMPLETE (N=24,479) TYPEA_PARTIAL (n=294)

c2(3) = 9.27, p = 0.03 Cramer's V = 0.02

16 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Participation - Wave

 Wave of data

Count of complete interviews Percent of Completes (n=24,773) Percent of Partials (n=294) 1 25.8 28.6 2 24.7 25.5 3 24.2 28.2 4 25.4 17.7

17 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Participation – other waves

 With 4 waves of data collections, what happens in the other

waves?

Count of complete interviews Percent of Completes (n=24,773) Percent of Partials (n=294)

• 36.1

1 8.25 30.3 2 10.7 20.1 3 17.25 13.6 4 63.81

18 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Where do respondents break-off?

19 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Where Respondents Break-Off

10 20 30 40 50 60 70 80 Count of respondents

20 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

14 14 16 34 23 31 17 15 48 40 21 47 61 37 45 45 2 16 16 34 41 10 20 30 40 50 60 70 10 20 30 40 50 60 70 80 Median time to break-off (minutes) Count of Respondents

Where Respondents Break-Off

44.9 %

21 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

How do the expenditure data compare?

22 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

How do their expenses compare? (Rented Property)

N Mean Average monthly rent (1st property) Completes 8,225 $ 900 Partials 62 $ 1,100

23 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

How do their expenses compare? (Owned Property)

N Mean Annual property tax (1st property) Completes 13,169 $ 3,200 Partials 98 $ 5,700 Original mortgage amount (1st loan) Completes 8,315 $ 186,000 Partials 73 $ 349,000 Amount of last monthly payment (Fixed rate mortgage) (1st loan) Completes 8,343 $ 1,500 Partials 75 $ 1,600

24 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

How do their expenses compare? (Utilities)

N Mean Monthly telephone/internet/cable bills Completes 22,571 $ 204 Partials 120 $ 216 Monthly electric/gas/water/etc Completes 22,406 $ 203 Partials 98 214

25 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Research Questions

 Who are those that drop out?

Slightly younger, larger, non-white households, in the Northeast

 Can they help provide insight on other non-responders?

 Reported expenditures are higher for those that break-off

 Can their break-off point inform us about survey design?

Around 15-30 minutes seems to be where the majority of partials lose

interest…or is it content of the sections? More research is needed.

 Could their data somehow be used in our processing?

Data for partials were very sparse, so not likely.

26 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Next Steps

 Closer look at the definition of a partial (presence of

respondent)

 Investigate respondents that do stick around for >1 hour and

then drop out

 Look at data quality of sections (e.g. item nonresponse) for

partials

 Investigate interviewer reported reasons that the respondents

didn’t complete the interview (“doorstep concerns”)

Contact Information

27 — U.S. BUREAU OF LABOR STATISTICS • bls.gov

Laura Erhard Branch Chief, BRPD Division of Consumer Expenditure Surveys www.bls.gov/cex 202-691-5115 Erhard.Laura@bls.gov