Evaluating the Consequences of SCHIP Expansions for Household Well-Being

advertisement
Evaluating the Consequences of
SCHIP Expansions for Household
Well-Being
Lindsey Leininger, University of Wisconsin
Helen Levy, University of Michigan
Diane Whitmore Schanzenbach, University of Chicago
Children's health insurance coverage, 1988 through 2008
Source: Current Population Survey
80.0%
75.0%
73.3%
70.0%
64.8%
65.0%
60.0%
55.0%
50.0%
45.0%
40.0%
35.0%
30.0%
24.4%
25.0%
20.0%
15.0%
13.1%
10.0%
5.0%
1988
1989
1990
1991
1992
1993
1994
1995
1996
Private
1997
Public
1998
1999
2000
Uninsured
2001
2002
2003
2004
2005
2006
2007
SCHIP crowd-out
• Estimate of crowd-out associated with SCHIP expansion
is about 60% (Gruber and Simon 2008).
• If households switch from private to public coverage,
they save money:
– Premium contributions
– Out-of-pocket medical spending
• Schaefer et al. (2009) estimate savings from switching
to be about $2,500 annually.
• Total household consumption should increase.
• How do households spend this “windfall”?
Our approach
• Data: Consumer Expenditure Survey
• Dependent variable = consumption
• Consumption as measure of well-being
– Mayer and Jencks (1989); Meyer and Sullivan
(2003, etc.)
• Empirical approach taken from crowd-out
literature: “simulated eligibility” as IV
Empirical approach (1)
• Gruber and Simon (2008):
INSijt = α + β· ELIGijt + φXijt + ν j + ρt + εijt
where i=individual, j=state, t=year
• “Simulated” instrument for ELIGijt uses
variation across states & over time in
generosity of public coverage.
Empirical approach (2)
• Our approach is similar except:
– Dependent variable is household consumption
instead of private coverage.
Yijt = α + β· ELIGijt + φXijt + ν j + ρt + εijt
– i indexes households, not individuals.
– Eligibility is fraction of family members who are
eligible.
– Cf. Gruber and Simon “family” regressions.
Data (1)
• Consumer Expenditure Study, Interview
Component, 1996 through 2002.
• Total spending plus detailed categories (food at
home, food away from home, etc.).
• We impute consumption of housing and
transportation.
• Our sample is households with children only.
• N = 50,488 households for 1996 through 2002
• Sample n in a state-year ranges from 0 to 1,334;
median is 296.
Data (2)
• Instrument constructed using SIPP, 1996 and
2001 panels
• National sample of households with children
run through public insurance eligibility rules in
each state and year
– > an IV that reflects the generosity of public
insurance that varies at the state-year level
Average simulated eligibility for Medicaid/SCHIP, 1996 - 2002
Source: SIPP, 1996 and 2001 Panels
0.35
Average fraction eligible for public insurance
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1996
1997
1998
1999
2000
2001
2002
Average budget shares for families with children
Consumer Expenditure Survey, 1996 - 2002
Alcohol/tobacco, 2%
Personal care, 1%
Childcare, 2%
Education,
Maintenance, 1%
1%
Medical care, 2%
Health insurance, 2%
Furniture, 3%
Food away, 4%
Housing, 23%
Clothing, 5%
Entertainment, 5%
Transportation, 16%
Utilities, 9%
Misc, 11%
Total real (2002$)
quarterly spending
during this period
was $11,540.
Food at home, 15%
Average real (2002$) quarterly spending on medical care and health insurance,
by income decile and insurance status
Source: CE 1996 - 2002
$800
$700
$600
$490
$500
$400
Difference is about $430,
or almost 4% of total
spending
$300
$200
$100
$57
$0
0
50% FPL
100% FPL 150% FPL 200% FPL 250% FPL 300% FPL 400% FPL 500% FPL >500% FPL
Private only
Public only
Results (1)
• Recall that our specification is:
Yijt = α + β·ELIGijt + φXijt + ν j + ρt + εijt
• Y is consumption, overall and in each of 16 different categories.
• Xijt includes race, education, family composition, age, employment,
MSA, and homeownership dummies.
• νj are state dummies; ρt are year dummies
• I will show you βOLS and also βIV.
• βIV results will be shown with and without additional controls at the
state-year level:
•
•
•
•
Unemployment rate (BLS)
% of private sector workers offered HI (MEPS-ICNET)
Average family insurance premium (MEPS-ICNET)
Average employee share of family premium (MEPS-ICNET)
1996 avg.
Total consumption
Food at home
Food away from home
Alcohol & tobacco
Housing
Maintenance
Utilities
Childcare
Furniture
Clothing
Transportation
Health insurance
Medical care
Entertainment
Personal care
Education
Miscellany
State-year controls?
$11,142
$1,321
$409
$142
$2,656
$85
$806
$190
$367
$527
$1,967
$226
$229
$657
$104
$184
$1,272
OLS
β
SE
-3,429
149
-65
11
-121
12
-45
5
-354
43
-32
6
-86
8
-113
12
-155
18
-131
13
-567
32
-51
9
-82
11
-221
27
-19
2
-26
17
-1,359
35
N
IV (1)
β
SE
2,542 2,403
-88
207
189
190
-114
103
59
828
184
152
49
122
228
229
209
301
11
313
888
515
-169
155
-260
151
-215
357
-4
36
276
347
1,300
638
N
IV (2)
β
SE
3,742 2,477
-8
227
283
195
-54
110
-84
788
189
151
179
132
142
239
160
289
26
307
1,225
546
-200
162
-155
166
-60
388
-17
39
255
347
1,862
772
Y
OLS
β
SE
$11,142
-3,429
149
$1,321
-65
11
1.$409
IV estimate-121
is big and
12
imprecise.
$142
-45
5
$2,656
-354
43
$85
-32
6
$806
-86
8
$190
-113
12
$367
-155
18
$527
-131
13
$1,967
-567
32
$226
-51
9
$229
-82
11
$657
-221
27
$104
-19
2
$184
-26
17
$1,272
-1,359
35
N
1996 avg.
Total consumption
Food at home
Food away from home
Alcohol & tobacco
Housing
Maintenance
Utilities
Childcare
Furniture
Clothing
Transportation
Health insurance
Medical care
Entertainment
Personal care
Education
Miscellany
State-year controls?
IV (1)
β
SE
2,542 2,403
-88
207
189
190
-114
103
59
828
184
152
49
122
228
229
209
301
11
313
888
515
-169
155
-260
151
-215
357
-4
36
276
347
1,300
638
N
IV (2)
β
SE
3,742 2,477
-8
227
283
195
-54
110
-84
788
189
151
179
132
142
239
160
289
26
307
1,225
546
-200
162
-155
166
-60
388
-17
39
255
347
1,862
772
Y
OLS
β
SE
$11,142
-3,429
149
Total consumption
Food at home
$1,321
-65
11
Food away from home
$409
-121
12
2. State-year controls make
Alcohol & tobacco
$142 slightly
-45
estimate
larger. 5
Housing
$2,656
-354
43
Maintenance
$85
-32
6
Utilities
$806
-86
8
Childcare
$190
-113
12
Furniture
$367
-155
18
Clothing
$527
-131
13
Transportation
$1,967
-567
32
Health insurance
$226
-51
9
Medical care
$229
-82
11
Entertainment
$657
-221
27
Personal care
$104
-19
2
Education
$184
-26
17
Miscellany
$1,272
-1,359
35
State-year controls?
N
1996 avg.
IV (1)
β
SE
2,542 2,403
-88
207
189
190
-114
103
59
828
184
152
49
122
228
229
209
301
11
313
888
515
-169
155
-260
151
-215
357
-4
36
276
347
1,300
638
N
IV (2)
β
SE
3,742 2,477
-8
227
283
195
-54
110
-84
788
189
151
179
132
142
239
160
289
26
307
1,225
546
-200
162
-155
166
-60
388
-17
39
255
347
1,862
772
Y
OLS
β
SE
$11,142
-3,429
149
Total consumption
Food at home
$1,321
-65
11
Food away from home
$409
-121
12
Alcohol & tobacco
$142
-45
5
Housing
$2,656
-354
43
Maintenance
$85
-32
6
Utilities
$806
-86
8
Childcare
$190
-113
12
3. Health insurance/medical
Furniture
-155
18
care results$367
are
Clothing
-131
13
insignificant.$527
Transportation
$1,967
-567
32
Health insurance
$226
-51
9
Medical care
$229
-82
11
Entertainment
$657
-221
27
Personal care
$104
-19
2
Education
$184
-26
17
Miscellany
$1,272
-1,359
35
State-year controls?
N
1996 avg.
IV (1)
β
SE
2,542 2,403
-88
207
189
190
-114
103
59
828
184
152
49
122
228
229
209
301
11
313
888
515
-169
155
-260
151
-215
357
-4
36
276
347
1,300
638
N
IV (2)
β
SE
3,742 2,477
-8
227
283
195
-54
110
-84
788
189
151
179
132
142
239
160
289
26
307
1,225
546
-200
162
-155
166
-60
388
-17
39
255
347
1,862
772
Y
Robustness checks
βIV becomes zero (still imprecise) if either:
• dep. var. = ln(Y) instead of Y
• Sample = below-median education only.
Conclusions/next steps
• Effect of SCHIP expansions on total consumption
is big – implausibly so.
• Effect of SCHIP expansions on health/medical
care spending is fairly small but with big standard
errors.
• What can we do to make estimates more precise?
– Add additional state-year controls?
– Use MEPS to estimate effects on medical care/health
insurance spending more precisely.
Backup slides start here.
The Basic Idea
• If I sign up my kids up for SCHIP and drop them from
my private policy, my premium contribution goes
down.
• I can spend that money on other things.
• For example, at the University of Michigan…
“self + kids” coverage is $186.38/month
“self only” is $90.52/month
• If I’m in the 25% marginal tax bracket my monthly takehome pay goes up by about $864/year.
• A “typical” SCHIP-eligible family is around 200% FPL, or
about $36,000/month for a family of three.
– The increase is about 3% of my after-tax income.
Variations on this theme…
• Wage offset: My consumption will increase by
even more if you think my wages will increase
because my employer’s contribution has gone
down by about $180.
• Lower out-of-pocket spending: If SCHIP is more
generous than the coverage I had before my outof-pocket medical spending will decrease.
• Precautionary motive: Spending might also
increase because better insurance may mean less
precautionary saving.
The consumer’s budget constraint
AOG
Without SCHIP, the consumer buys a
health insurance policy worth πp, giving
up πp(1-t) of other goods.
Y(1-t)
A = original optimum
πp
Y
HI
The consumer’s budget constraint
AOG
Y(1-t)
B
With SCHIP, the family can choose point
B; health insurance is free if they are
willing to accept exactly πg of it.
A = original optimum
πg= value of government HI
Y
HI
The consumer’s budget constraint
AOG
Y(1-t)
If B is the new optimum, the family can
spend πp(1-t) more on other goods.
B
πp(1-t)
Total consumption will be Y(1-t) instead
of (Y-πp)(1-t).
A
πg= value of government HI
Y
HI
Fraction eligible for public insurance in families with children
Source: CE, 1996 - 2002
1.000
0.900
0.800
0.700
1996
Axis Title
0.600
1997
1998
0.500
1999
2000
0.400
2001
2002
0.300
0.200
0.100
0.000
0
50
100
150
200
250
300
400
500
>500
What do we expect?
• Increase in total consumption of ~$500.
• Simulate increase in spending on non-medical
goods as a result, based on estimated
marginal propensities to consume.
– Estimate MPC for each good around 175% FPL.
– Multiply by $500.
Hypothetical effects of a $500 windfall with no medical care and
insurance spending
$160
$140
$138
$120
$100
$100
$83
$80
$60
$40
$20
$0
$29
$25
$22
$22
$19
$18
$15
$12
$7
$4
$4
Hypothetical effects of a $500 windfall with no medical care and
insurance spending
$300
$249
$250
$200
$164
$150
$138
$100
$100
$83
$50
$38
$29
$25 $21
$22
$19 $24
$22
$34
$18
$15
$3
$12
$25
$19
$7
$4
$4
$0
-$11
-$1
-$8
-$50
Hypothetical
Observed
-$7
-$2
Download