Chris Carroll Johns Hopkins University on Improving Consumption Measurement Chris Carroll Johns Hopkins University Remarks At CRIW-NBER Conference On …

Comments on Improving Consumption Measurement

Chris CarrollJohns Hopkins University

Remarks At CRIW-NBER Conference On Improving Consumption MeasurementCambridge, MA

July 13, 2009

It’s A Global ProblemNew Technologies

Checks and BalancesOpenness, Transparency, Accountability

Conclusion

Why Are Some Surveys/Countries Better Than Others?My Best Guess

We Have A Serious Problem

1980 1985 1990 1995 2000 20057.5

7.6

7.7

7.8

7.9

8

8.1

8.2

8.3

Food Consumption Per−capita

Year

Log

− 2

003

$

1980 1985 1990 1995 2000 2005

8.8

8.9

9

9.1

9.2

9.3

9.4

9.5

9.6

Nondurable Consumption Per−capita

Year

Log

− 2

003

$

1980 1985 1990 1995 2000 20057.1

7.2

7.3

7.4

7.5

7.6

7.7

7.8

7.9

Durable Consumption Per−capita

Year

Log

− 2

003

$

1980 1985 1990 1995 2000 2005

7.4

7.5

7.6

7.7

7.8

7.9

8

8.1

8.2Housing Services Per−capita

Year

Log

− 2

003

$

NIPACEXPSID

NIPACEX

NIPACEX

NIPACEX

Figure 3: Comparison between per-capita averages in CEX and NIPA: various definitions ofconsumption

Carroll Improving the CEX



Conclusion


CE vs NIPA for CanadaFigure 2: Mean Nominal Non-Durable Consumption

050

0010

000

1500

0C

onsu

mpt

ion

1975 1980 1985 1990 1995 2000 2005Year

SNA FAMEX/SHS

the SCF had unusually small samples in 1977, 1979, 1981 and 1984, which explains

the anomalies prior to 1990.11 The post 1990 discrepancies can be explained by the

2003 historical revision of the weights discussed in the previous section. As a result

of this revision, which was applied retroactively to 1990, higher weights were assigned

to low/no earnings individuals as well as high earnings individuals and had the effect

of increasing substantially the fraction of the population not in the labor force. This

is reflected in Figure 3 as a low employment to population ratio.12

11Survey years 1977 and 1979, which contain data pertaining to 1976 and 1978, have otherproblems—in particular hours worked is missing—so we dropped these years for the rest of theanalysis.

12This is confirmed by a similar figure (not shown) constructed using weights available prior tothe revision, which shows essentially no discrepancy between the SCF and LFS numbers over the1990’s.

12




Conclusion


CE vs NIPA for Germany

1985 1990 1995 2000 20051

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8 x 104 Per Capita Disposable Income from NIPA, EVS & GSOEP

Year

Per

Cap

ita In

com

e in

Eur

os, P

rice

s of

200

0

EVSGSOEPNIPA




Conclusion


CE vs NIPA for Italy

11.

21.

41.

61.

8

1980 1985 1990 1995 2000 2005

National Accounts SHIW

Figure 1: Average per-capita total consumption in the SHIW

and in the National Accounts

.91

1.1

1.2

1.3

1.4

1980 1985 1990 1995 2000 2005

National Accounts SHIW

Figure 2: Average per-capita non-durable consumption in the SHIW

and in the National Accounts

23




Conclusion


CE vs NIPA for Russia

40

Figure 2: Comparison of RLMS with Official Statistics

1000

2000

3000

4000

5000

1994 1996 1998 2000 2002 2004

RLMS NIPA

A.Income per capita

1000

2000

3000

4000

5000

1994 1996 1998 2000 2002 2004

RLMS NIPA

B.Consumption per capita

500

1000

1500

2000

1994 1996 1998 2000 2002 2004

RLMS HBS

C.Food consumption per household member

500

1000

1500

2000

1994 1996 1998 2000 2002 2004

RLMS HBS

D.Non-food consumption per household member

Notes: For comparability purposes, the following RLMS measures are selected: yD in panel A, cD in panel B, cF + consumption of home-grown food in panel C, cD - cF in panel D. The RLMS sample is unrestricted. All RLMS measures are deflated using monthly CPI and the date of interview. All NIPA and HBS measures are deflated using annual average CPI. RLMS income and consumption for 1997 are imputed using the lagged RLMS value multiplied by the 1997 growth rate from NIPA.




Conclusion


CE vs NIPA for Sweden

1984 1986 1988 1990 1992 1994 1996 1998 200090

95

100

105

110

115

120

125

130

NIPA

HUT

SEK

1000




Conclusion


CE vs NIPA for UK

15

Figure 2.6: Income and Consumption Coverage

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

119

78

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

Year

Cov

erag

Income CoverageConsumption Coverage

In order to try to understand what may be behind the declining performance of

the expenditure data, it is worth looking at some of the components behind the

total. Figure 2.7 shows the percentage coverage of certain items included in our

consumption basket. Expenditure on food, clothing and catering matched the

national accounts extremely well, both in levels and in dynamics until the late

1980s (and before the beginning of our sample period). Coverage for these items

rarely fell below 90%. On the other hand, alcohol and tobacco have always had

low coverage, but this is common for items that carry a social stigma. 1988 saw a

sudden collapse in the coverage of catering, which suggests that there was a

sudden change in measurement for this category in one of the datasets.

However, for all other categories there has been no sudden shift, but a gradual

decline in coverage, approximately since 1993. Therefore the decline in coverage

of the aggregate since 1993 is not confined to certain items but seems to be a

broad trend across many expenditure categories. The case of food expenditure is

puzzling since the national accounts data for this item are formed mainly from

the FES data. It may therefore be sensible to use the FES food coverage as a basis

for comparison.




Conclusion


Commonalities

In Many Countries, Coverage Has Declined Seriously

In Some, Deterioration Is Accelerating




Conclusion


Commonalities

In Many Countries, Coverage Has Declined Seriously

In Some, Deterioration Is Accelerating




Conclusion


Differences:

In Person Vs Phone Interviews

Diary Vs Recall

Incentives To Participate

Quality of Interviewers

Quality of Survey (Focus Groups!)

Use Of Budget Constraint




Conclusion


Differences:


Diary Vs Recall








Conclusion


Differences:


Diary Vs Recall








Conclusion


Differences:


Diary Vs Recall








Conclusion


Differences:


Diary Vs Recall








Conclusion


Differences:


Diary Vs Recall








Conclusion


Differences:


Diary Vs Recall








Conclusion


Biggest Problem: Increased Reluctance to Participate

Differential Sample Selection: Richer ⇒ Less Willing

Second Biggest: Changes In Technology (Automated Payment, Etc)

If You Asked Me These Questions, I’d Have No Idea!

I’m Concerned You’re Drilling A Drying Up Hole




Conclusion










Conclusion










Conclusion










Conclusion










Conclusion

Nielsen Data

Credit Card Data

Personal Finance Online

QuickenOnlineiBankhttp://www.mint.com/product/videos/mint-demo/Big Advantage: People Get Something Valuable At the End

50 percent chance that current structure should be ditched altogether




Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Nielsen Data

Credit Card Data







Conclusion

Comparison of CEX Data With Other Sources Should Be Routinized

Every CE Report Should Show Comparisons With Other Sources, Say:

NIPANielsenCPS

Biannual Review Report Should Contain Broader Comparisons

SCF, Labor Market, SOI Tax Income, and “special topics”Explain What It Should/Should Not Be Used For

Occasional External Reviews, Not Just Internal




Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion



NIPANielsenCPS







Conclusion

And Collaboration

Openness, Transparency, Accountability

How Do These Buzzwords Apply?

Unfolding Brackets of Complexity Of Data Use

Very Easy To Get Top Level Data

Clear Path to “Drill Down” to Deeper and Deeper Levels

A Lot Of Researchers Have Been Defeated By Complexity

Experiment, e.g., With HRS or PSID Instrument On CE Sample




Conclusion

And Collaboration











Conclusion

And Collaboration











Conclusion

And Collaboration











Conclusion

And Collaboration











Conclusion

And Collaboration











Conclusion

And Collaboration

Encourage Researchers With Setup Like Census Data Centers




Conclusion

A Golden Age!

CE Should Be A Part Of It, Not A Drowning Polar Bear


Documents

Chris Carroll Johns Hopkins University on Improving Consumption Measurement Chris Carroll Johns Hopkins University Remarks At CRIW-NBER Conference On …