Data to Examine Consumption Poverty and
I
Inequality
lit in
i the
th U.S.:
U S 1960-2008
1960 2008
CE Survey
y Data Users’ Needs Forum
June 22, 2010
Bruce D. Meyer
University of Chicago and NBER
Based on work with
J
James
X
X. S
Sullivan
lli
University of Notre Dame
I. Introduction

Question: How have poverty and inequality
changed over the past five decades?



We look at both income and consumption based
measures of well-being
well being
We emphasize the importance of measurement
issues for understanding poverty and inequality
patterns
We refine the methods to convert expenditure
d t iinto
data
t consumption
ti data
d t
I. Income v. Consumption: Conceptual


Meyer and Sullivan (2003, 2007)
C
Conceptual
t l iissues ffavor consumption.
ti



Permanent income (Cutler and Katz 1991;
Poterba 1991)
Public and private insurance
Access to credit
II. Income v. Consumption: Data Quality

Reporting issues are split between income and
consumption




Ease of reporting v. sensitive topics
Nonresponse
Under-reporting
Low percentiles of expenditures greatly exceed
low percentiles of income
Nonresponse Rates
Table
T
bl 5
Survey Nonresponse and Imputations Rates, CPS and CE Interview Survey,
1993-2007
Survey Nonresponse
CPSASEC/ADF
CE Survey
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
(1)
0.154
0.154
0 154
0.154
0.157
0.144
0.161
0.144
0.159
0.162
0.150
0.160
0.174
0.167
0.171
0.156
(2)
0.156
0.167
0 194
0.194
0.211
0.199
0.201
0.202
0.200
0.220
0.220
0.214
0.240
0.255
0.234
0.262
Imputation Rates
CPS-ASEC/ADF
Pre-tax
Money
Income
(3)
0.153
0.156
0 180
0.180
0.190
0.204
0.219
0.217
0.248
0.255
0.262
0.254
0.256
0.239
0.252
0.251
After-tax
a
Income
(4)
0.252
0.259
0 295
0.295
0.316
0.344
0.375
0.382
0.428
0.434
0.422
0.389
0.401
0.373
0.403
0.398
CE Survey
After-tax
b
Income
(5)
0.444
0.456
0 496
0.496
0.518
0.548
0.574
0.589
0.626
0.628
0.604
0.565
0.583
0.557
0.592
0.591
Total
E pendit res
Expenditures
(6)
0.104
0.104
0 104
0.104
0.125
0.128
0.129
0.149
0.154
0.163
0.179
0.184
0.167
0.194
0.228
0.130
Share of Dollars Imputed, CPS
04
0.4
0.3
0.2
0.1
AFDC/TANF
Food Stamps
OASDI
SSI
UI
WC
Meyer, Mok, and Sullivan (2009)
2008
8
2007
7
2006
6
2005
5
2004
4
2003
3
2002
2
2001
1
2000
0
1999
9
1998
8
1997
7
1996
6
1995
5
1994
4
1993
3
1992
2
1991
1
0
Reporting Rates for Dollar Amounts of Transfer Programs, CPS
1
0.8
06
0.6
04
0.4
AFDC/TANF
Food Stamps
SSI
0.2
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
1983
1981
1979
1977
1975
1973
0
Ratios of CE Expenditure Measures to National Aggregates,
Aggregates 1980-2008
1980 2008
Food at home
Rent plus Utilities
Alcoholic beverages
Tobacco
1.2
1.1
Food away from home
Gasoline and motor oil
Transportation
Clothing
1
CE/NIPA Ratio
C
0.9
0.8
0.7
0.6
0.5
0.4
03
0.3
0.2
0.1
1980
1984
1987
1992
1994
1997
2002
2004
2007
2008
II. Income v. Consumption: Data Quality



Refinement to income tend to move it toward
consumption (alternative poverty,
poverty housing
housing,
MOOP, etc.)
CE provides much info to approximate
consumption that is missing in CPS (housing
and vehicle characteristics
characteristics, MOOP)
MOOP).
Consumption is more strongly associated with
other measures of well-being
II. CPS Income Data

Current Population Survey – ASEC/ADF




1963-2008
1963
2008
Taxes calculated using TAXSIM
Census has imputed noncash benefits since 1980
1980.
These imputed benefits have some drawbacks
III. CE Consumption and Income Data

Consumer Expenditure (CE) Interview
Component



1960/61, 1972/73, 1980-1981, 1984-2008
1982-1983
1982
1983 only urban consumers; also because
summary measures of aggregated expenditures
are not provided it makes it difficult to use
Recent improved timeliness of data releases is
welcome; talk of speeding up releases further?
III. CE Consumption and Income Data

Consumer Expenditure (CE) Interview
Component raw data



Mostly use family files: summary expenditures
Also use detailed expenditure files: vehicles,
vehicles debt,
debt
Medicaid enrollment, HI coverage
And member files: exact age composition of CU
III. CE Consumption Data

We modify expenditures to approach consumption


We make many improvements in the measurement of
consumption at the bottom
Rental equivalent for owner-occupied housing



We checked
W
h k d th
thatt it relates
l t sensibly
ibl tto reported
t dh
home values
l
Others have related reported home values to sales prices in
other datasets.
Impute value of public/subsidized housing using detailed
housing characteristics


Make adjustment
j
based on PSID info on rental equivalent
q
Adding rental equivalent to survey would be helpful
III. CE Consumption Data

Flow value of vehicles (based on more than
350,000 purchase prices)





Use equations to predict purchase price for those
without it
Complicated set of regressions to determine implicit
prices of vehicle characteristics depending on what
information is missing.
Use data to determine depreciation that goes into
flow value
Validated using NADA data
Unfortunate that make but not model available
beginning in 2006
2006.
III. CE Consumption Data

Medical care, health insurance


Subtract out MOOP
MOOP. Imputed in CPS in proposed
Supplemental Poverty Measure.
Use information on Medicaid
Medicaid, Medicare
Medicare, and Private
HI coverage.
III. General Issues in CE




We use annualized quarterly data
W have
We
h
compared
d one quarter
t to
t four,
f
and
d
there is some understatement of dispersion
inherent in relying on one quarter
quarter.
This problem is likely to be much more severe
if one relies on two weeks of expenditures to
infer consumption in the diary data.
C
Comparisons
i
off one week
k off di
diary d
data
t tto ttwo
weeks shows differences in the distribution of
expenditures.
expenditures
III. CE Income Data

We use TAXSIM as reported income tax
payments are very different from estimated
taxes.




NBER willing to supply code to implement in CE
State IDs missing for 16 percent of the sample
we used from the 1990s in our AER paper
paper.
Imputation of income began in 2005.
C ld reconciliation
Could
ili ti off iincome and
d
consumption be brought back?
Figure 1: Real After-tax Income Plus Food Stamps at Various Percentiles, 1980-2008, CPS &
CE Survey
5th Percentile CPS
10th Percentile CPS
25th Percentile CPS
5th Percentile CE Survey
10th Percentile CE Survey
25th Percentile CE Survey
16000
14000
12000
8000
6000
4000
2000
200
08
200
06
200
04
200
02
200
00
199
98
199
96
199
94
199
92
199
90
198
88
198
86
198
84
198
82
0
198
80
2005 $
10000
Figrue 2: Real After-tax Income Plus Food Stamps at Various Percentiles, 1980-2008, CPS &
CE Survey
50th Percentile CPS
75th Percentile CPS
70000
90th Percentile CPS
50th Percentile
P
til CE S
Survey
60000
75th Percentile CE Survey
90th Percentile CE Survey
40000
30000
20000
10000
200
08
200
06
200
04
200
02
200
00
199
98
199
96
199
94
199
92
199
90
198
88
198
86
198
84
198
82
0
198
80
2005 $
50000
VI. Core Consumption

We look at a subset of total consumption that
p
spending
p
g categories
g
that tend to
includes important
be well reported:





Housing
Food at home
Transportation
For those near the poverty line, Core is 80% of nonmedical
di l consumption
i iin early
l 1980
1980s
Information is needed to compare CE totals to NIPA
aggregates Knowing which categories of
aggregates.
expenditures line up well with NIPA would be
helpful. An earlier Garner et al. paper did this, but
NIPA categories have changed
changed.
VII. Predicted Consumption




We regress total consumption measures on a
p
a cubic in the age
g
cubic in core consumption,
of the head, education of the head dummies,
family type dummies, and race dummies.
We use data from 1980
1980-81,
81 because total
expenditures in the CE Survey compare more
favorably to NIPA in the early 1980s than in
recent years.
Coefficients from this regression are then
used to predict a value of the consumption
measures for each consumer unit in all years.
R-squared = 0.72
VIII. Results

Do income and consumption poverty and
inequality differ?
Summary of Changes in Income and Consumption Inequality
0.3
0.25
Change
e in 90/10 R
Ratio
0.2
Log After-Tax Income
Log Consumption Excluding HI
0.15
0.1
0.05
0
1963-1972
-0.05
-0.1
1972-1980
1980-1990
1990-2000
2000-2008
Figure 4: Consumption Inequality 1961-2008
65
6.5
6.0
5.5
4.5
4.0
3.5
After-tax Money Income (90/10)
Expenditures (90/10)
Consumption (90/10)
Consumption Excluding HI (90/10)
Core Consumption (90/10)
30
3.0
2.5
47
45
43
41
39
37
35
33
31
29
27
25
23
21
19
17
15
13
11
9
7
5
3
2.0
1
90/10 Ratio
9
5.0
XI. Conclusions


Consumption data extremely useful to look at
measures of well
well-being.
being
Consumption poverty and inequality are quite
different from their income cousins
cousins.
XI. Conclusions: Data Suggestions

Recent improvements helpful



Imputation of income
Improved timeliness of data release
Opportunities for improvement






Information on categories compatible with NIPA or
more regular comparisons to NIPA totals
Suppression of vehicle model starting in 2006
High fraction of units with suppressed or recoded
state id
Make data available at RDC?
Use TAXSIM?
Reconcile Y and Expenditures?
IX. Comparisons Across Data Sets

How does consumption inequality in the CE
Survey compare to that in the PSID?


Food and housing
How does income inequality in the CPS
compare to that in the PSID?

Pre tax money income
Pre-tax
Consumption Inequality, CPS and PSID 1980-2007
0 14
0.14
0.12
0 10
0.10
0.06
0.04
0.02
Food at home & housing consumption (CE Survey)
-0.04
Food at home & housing consumption (PSID)
-0.06
-0.08
2008
2006
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
-0.02
1982
0.00
1980
n 90-10 Since 1980
Change in
0.08
Pre-Tax Money Income Inequality, CPS and PSID 1967-2006
04
0.4
Pre-tax Money Income (CPS)
Pre-tax Moneyy Income (PSID)
(
)
0.2
0.1
-0.1
-0.2
20
007
20
005
20
003
20
001
19
999
19
997
19
995
19
993
19
991
19
989
19
987
19
985
19
983
19
981
19
979
19
977
19
975
19
973
19
971
19
969
0.0
19
967
Change in
n 90-10 Since 1980
0.3
September 6, 2007 | Issue 43•37
In The Know: Are America's Rich Falling
Behind The Super-Rich?
Panelists discuss a new study showing
the gap between the wealthy and the absurdly
wealthy is widening, and how we can help the
merely rich catch up.
up
I. Introduction

Some previous work has examined income
and/or consumption poverty or inequality







P60 Series Reports (annual)
Gottschalk and Danziger (2005)
Burkhauser et al. (various)
Johnson Smeeding
Johnson,
Smeeding, and Torrey (2005)
Krueger and Perri (2006)
Heathcote Perri
Heathcote,
Perri, and Violante (2010)
Attanasio, Battistin and Ichimura (2004)
Figure 2: Real Changes in Consumption at Various Percentiles, 1972-2008
0 25
0.25
5th Percentile Consumption Excluding HI
10th Percentile Consumption Excluding HI
90th Percentile Consumption Excluding HI
25th Percentile Consumption Excluding HI
50th Percentile Consumption Excluding HI
75th Percentile Consumption Excluding HI
0.15
0.10
0.05
-0.05
-0.10
2008
8
2006
6
2004
4
2002
2
2000
0
1998
8
1996
6
1994
4
1992
2
1990
0
1988
8
1986
6
1984
4
1982
2
1980
0
1978
8
1976
6
1974
4
0.00
1972
2
Log Differen
nce Relative tto 1980
0.20
Figure 5: Consumption Inequality 1961-2008
65
6.5
6.0
Consumption Excluding HI (90/10)
5.5
Predicted Consumption Excluding HI (90/10)
4.5
4.0
3.5
3.0
2.5
2007
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
1983
1981
1979
1977
1975
1973
1971
1969
1967
1965
1963
20
2.0
1961
90/10 Ratio
5.0