This document summarizes research comparing Medicaid reporting in the American Community Survey (ACS) to administrative enrollment data. Key findings include: - The ACS appears to undercount Medicaid enrollment, though not as much as some other surveys. The undercount increases with age, income, and varies by state. - Misreporting Medicaid enrollment in the ACS translates to an overestimate of the uninsured population of around 1.2 percentage points or 3.2 million people. - The undercount is larger for those with more limited Medicaid benefits or shorter enrollment tenure, though the ACS still represents a valuable data source for policy analysis.

1. MEDICAID UNDERCOUNT IN THE AMERICAN
COMMUNITY SURVEY
Joanna Turner
State Health Access Data Assistance Center (SHADAC)
University of Minnesota
JSM meetings, Boston, MA
August 5, 2014

2. Click Acknowledgments
to edit Master title style
Click • Funding to edit for Master this work text styles
is supported by the U.S.
Census Bureau
Second level
Third level
Fourth level
• Collaboration between Census Bureau and SHADAC
Fifth level
to extend previous Medicaid undercount research to
the American Community Survey (ACS)
• Collaborators:
• Brett O’Hara (Census Bureau)
• Kathleen Call, Michel Boudreaux, Brett Fried (SHADAC)
2

3. Click Background
to edit Master title style
• Administrative data on public assistance programs are not
Click to edit Master text styles
sufficient Second level
for policy making
• Not timely
Third level
• No population Fourth level
denominator
• Incomplete or lower quality covariates
Fifth level
• Population surveys fill these gaps
• Yet they universally undercount public program enrollment
described in administrative data
• Food stamps, public housing, TANF (Lewis, Elwood, and Czajka 1998;
Meyer, 2003)
• Medicaid (Call et al 2008, 2012)
3

4. Research Click to edit focus
Master title style
• Click Describe to edit Master the concordance text styles
of Medicaid reporting in
the Second ACS level
and enrollment data in administrative
records
Third level
Fourth level
• Compare “misreports” across characteristics of:
Fifth level
• Individuals: age, income, state
• Medicaid enrollment: scope of benefit, tenure
• Bias to uninsurance estimates
4

5. Click Data source: to edit Master American title Community
style
Survey (ACS)
Click to edit Master text styles
• Large, continuous, multi-mode survey of the US
Second level
population residing in housing units and group
Third level
quarters
Fourth level
• Added health Fifth level
insurance question in 2008
• One simple multi-part question on health insurance
type
• Unique data source due to its size
• Subgroup analysis (small demographic groups and low
levels of geography)
5

6. ACS health insurance question part “d”
Click to edit Master title style
Click to edit Master text styles
Second level
“Is this person CURRENTLY covered by any of the
following Third level
types of health insurance or health
Fourth level
coverage Fifth plans?
level
d. Medicaid, Medical Assistance, or any kind of government-assistance
plan for those with low incomes or a
disability?”
6

7. Data Click source: to edit Master Medicaid title Statistical
style
Information System (MSIS)
Click to edit Master text styles
• Medicaid enrollment records
Second level
• Longitudinal database of enrollment
Third level
• Records Fourth level
originate in the states and are reported to the
Fifth level
federal government
• Includes regular Medicaid and Expansion CHIP
• Tracks all levels of enrollment (e.g., emergency & family
planning)
• Not a perfect gold standard
7

8. Definition Click to edit differences
Master title style
Click • MSIS to edit includes Master comprehensive text styles
and partial coverage
Second • ACS level
comprehensive coverage is a subset
Third level
Fourth level
• ACS includes Medicaid, CHIP, and state-specific
Fifth level
public programs (will refer to coverage as “Medicaid
Plus”)
• MSIS Medicaid and Expansion CHIP coverage is a
subset
8

9. Investigating Click to edit Master survey title response style
errors
• Click Discordance to edit Master between text styles
MSIS and ACS can come from
definitional differences and survey response error
Second level
Third level
Fourth level
• Our focus here is on survey response errors which
Fifth level
we investigate by merging the ACS and the MSIS
9

10. Investigating survey response errors (2)
Click to edit Master title style
• Click Use to linking edit Master methodology text styles
developed by the Census
Bureau’s Second level
Center for Administrative Records
Research Third level
and Applications
• Personal Fourth level
Identification Key (PIK)
Fifth level
• Research conducted at the MN Census Research
Data Center located at the University of Minnesota
• http://mnrdc.umn.edu/
10

11. Investigating survey response errors (3)
Click to edit Master title style
Click • Consider to edit Master a case text to have styles
Medicaid enrollment if they
are Second covered level
on the day of ACS interview by full
benefit Third coverage level
from Medicaid or Expansion CHIP
Fourth level
• Adjust ACS person weights to account for unlinkable
Fifth level
records
• Although all persons were linked estimates reported
here are for the civilian non-institutionalized
population
• Explicit reports of coverage (not edited)
11

12. Linked and unlinked data and misreports
Click to edit Master title style
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
12

13. Click Coverage to edit by Master age
title style
10% 8%
Click to edit Master text styles
13%
“Misreports”
Second level
Third level
Fourth level
Fifth level
Any Medicaid Plus
13
12%
77% 81%
14.9%*
13%*
1%*
27%*
72%* 71%*
Total 0-18 19-64 65+
Uninsured
Other coverage
Source: 2009 MSIS and ACS civilian non-institutionalized population as analyzed by SHADAC.
Explicit reports (not edited)
* Indicates that estimate is statistically significantly different (at p<.01) from the estimate
for the 0-18 age group.

14. Click to edit Master title style
9% 11%* 11%* 9%*
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
83%
71%*
64%* 61%*
8%
18%*
26%* 30%*
0-138 139-249 250-399 400+
Uninsured
Other coverage
Any Medicaid Plus
Coverage by FPL
14
Source: 2009 MSIS and ACS civilian non-institutionalized population as analyzed by SHADAC.
Explicit reports (not edited)
* Indicates that estimate is statistically significantly different (at p<.01) from the estimate
for the 0-138 FPL group.

15. Click to edit Master title style
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
15
Percent of linked cases reporting Medicaid
Plus enrollment on the ACS interview date
Source: 2008 MSIS and ACS civilian non-institutionalized
population as analyzed by SHADAC; Explicit reports only

16. Coverage by benefit type and scope
Click to edit Master title style
Click to edit Master text styles
Second level
Third level
Fourth level
77% 75F%ift*h level
46%
77%*
13% 16%*
35%
13%*
10% 9%*
19%
10%*
Medicaid Expansion CHIP Partial Full
Uninsured
Other coverage
Any Medicaid Plus
16
Benefit Type Benefit Scope
Source: 2009 MSIS and ACS civilian non-institutionalized population as analyzed by SHADAC.
Explicit reports (not edited)
* Indicates that estimate is statistically significantly different (at p<.01) from the estimate
for those with Medicaid, and for those with partial benefits, respectively.

17. Click to edit Master title style
Click to edit Master text styles
Second level
Third level
Fourth level
82% Fifth level
73%*
59%*
41%*
12%
14%*
18%*
24%*
6%
13%*
23%*
34%*
Continuously
Enrolled from
Oct 2008-Mar 2010
(18 months)
Enrolled For 50-99%
of Months
Enrolled For 25-49% Enrolled for 0-24%
Uninsured
Other coverage
Any Medicaid Plus
Coverage by enrollment tenure
17
Source: 2009 MSIS and ACS civilian non-institutionalized population as analyzed by SHADAC.
Explicit reports (not edited)
* Indicates that estimate is statistically significantly different (at p<.01) from the estimate
for those who were continuously enrolled.

18. Click Bias to to estimates edit Master of title uninsurance
style
• Click A key to edit policy Master metric text is styles
the share of the population that
lacks any type of coverage
Second level
Third level
• Uninsurance is a residual category, so undercounting
Fourth level
Medicaid partially contributes to bias in uninsurance
Fifth level
• We cannot estimate bias from other sources of coverage
• We cannot estimate bias from those that report Medicaid,
but are in fact uninsured
18

19. Upper Click to bound edit of Master bias to title uninsurance
style
attributable to Medicaid misreporting:
Explicit Click to edit reports Master text only
styles
Second level
Third level
Fourth level
Fifth level
Count in
millions
Percent
(SE)
Original uninsured estimate 40.9 15.4
(0.05)
Share of the uninsured that
are linked
3.2 7.9
(0.07)
Partially adjusted uninsured
estimate
37.7 14.2
(0.04)
19
Source: 2008 MSIS and ACS civilian non-institutionalized
population as analyzed by SHADAC.

20. Click Summary to edit of Master results
title style
• Although not perfectly comparable, the undercount in the
Click to edit Master text styles
ACS appears in line with other surveys
Second • Large level
(23%), but slightly better than some other surveys
Third level
(keep in mind ACS includes Medicaid and other means-tested
Fourth level
public coverage)
Fifth level
• As with other surveys the undercount increases with age
and family income and appears to vary by state
• The undercount translates into an overestimate of
uninsurance of 1.2 percentage points or 3.2 million but it
is likely that there are other offsetting influences
• ACS represents a valuable source of policy analysis
20

21. Click to edit Master title style
Contact Information
Click to edit Master text styles
Second level
Third level
Fourth level
turn0053@umn.edu
Fifth level
Joanna Turner
612.624.4802
www.shadac.org
@shadac

22. Click Linked to file: edit enrolled Master title in Medicaid style
on
ACS interview date: Explicit reports
Click to edit Master text styles
only
Second level
Third level
Fourth level
Fifth level
22
ACS
Yes No
MSIS
Yes
No

23. Click to edit Master title style
Previous research
• SNACC Phases I-VI (2007-2010)
Click to edit Master text styles
• CPS (CY 2005) implied undercount of about 41%
Second level
• NHIS (CY 2002) implied undercount of about 33%
Third level
• MEPS (CY 2002) implied undercount of about 18%
Fourth level
• Known enrollees coded as uninsured was 18%, 9%, and 8%
Fifth level
respectively for CPS, NHIS and MEPS
• O’Hara (2009)
• ACS Content Test (CY 2006) implied undercount of 34.4% for the
non-elderly
• Turner & Boudreaux (2010)
• 2008 ACS produces coverage estimates similar to other population
surveys (e.g. 2008 NHIS)
23

24. Coverage by type for linked cases enrolled
in Medicaid on ACS interview date:
Explicit reports only
Click to edit Master title style
Click to edit Master text styles
Second level
Any Medicaid Plus
NOT Any Third Medicaid level
Plus
Fourth level
Fifth level
77.1 (0.12)
22.9 (0.12)
Employer sponsored insurance 8.3 (0.08)
Direct purchase 2.2 (0.05)
Medicare 3.3 (0.04)
TRICARE 0.3 (0.01)
VA 0.1 (0.01)
Uninsured 9.9 (0.08)
24
Source: 2008 MSIS and ACS civilian non-institutionalized
population as analyzed by SHADAC.
Percent (Standard error)