Issue Brief

Introduction
A cornerstone of the Afordable Care Act (ACA) is the reorganization of the individual group
market. Te ACA institutes guaranteed issue, community rating, and subsidies in order to
increase access to coverage in the individual market.
Survey data play an important role in evaluating the efect of the ACA on individual market
coverage. However, estimates of the size of the individual market, often referred to as direct
purchase coverage, derived from surveys typically exceed counts from administrative records
(Abraham et al., 2013). Estimates from the American Community Survey (ACS) are particu-
larly high (Mach and OHara, 2011; Abraham et al., 2013). Previous authors have suggested
that the excess number of cases identifed as having direct purchase coverage in the ACS is
partially driven by the large number of direct purchase reports that are accompanied by reports
of other plan types (Mach and OHara, 2011). Mach and OHara suggest that the potential
false-positive reports could be caused by confusion among people that only have employ-
er-sponsored insurance or by respondents that are referring to single service plans.
Previous work on the Current Population Survey (CPS) suggested that the over-counting
of direct purchase could be tied to the Medicaid undercount. Cantor and colleagues (2006)
suggest that Medicaid and CHIP benefciaries enrolled in managed care plans may perceive
their coverage as directly purchased from an insurance company because of the extent of their
interaction with the managed care company and their familiarity with its name. However,
other authors conducting a record-check study of the CPS found that managed care partici-
pants tend to be better, not worse, reporters of their Medicaid coverage compared to partic-
ipants enrolled in fee-for-service plans (Plotzke et al., 2010). Validation studies also indicate
that Medicaid recipients enrolled in managed care are more accurate reporters of their coverage
than those in fee-for-service Medicaid (Call et al., 2008/2009).
Te purpose of this brief is to present preliminary analysis of the contribution of Medicaid
misreporting in the ACS to estimates of direct purchase coverage.
Methods
Data comes from the 2009 ACS which we linked to the enrollment records from the Medicaid
Statistical Information System (MSIS). Tese data are available under contract with the
Census Bureau at the Minnesota Census Research Data Center. A description of our linking
methodology and weighting approach is described in detail in a previous report (Boudreaux et
al., 2013). We include only non-elderly people in the civilian non-institutional population.
We focus on the non-elderly because the ACAs individual market provisions are targeted on
that segment of the market.
Estimates of Direct Purchase from the ACS and
Medicaid Misreporting: Is there a link?

|
STATE HEALTH ACCESS DATA ASSISTANCE CENTER 1
BRIEF 38 MAY 2014
The University of Minnesotas State
Health Access Data Assistance
Center (SHADAC) is funded by the
Robert Wood Johnson Foundation
to collect and analyze data to inform
state health policy decisions relating
to health insurance coverage and
access to care. For information on how
SHADAC can assist your state with
small area estimation or other data
issues relevant to state health policy,
please contact us at [email protected]
or call 612-624-4802
Authors
Michel Boudreaux
Kathleen Call
Joanna Turner
Brett Fried
State Health Access Data Assistance
Center (SHADAC)
Summary
Are the relatively high estimates
of direct purchase coverage in the
American Community Survey (ACS)
caused by people misreporting their
Medicaid coverage? We examine
this question using a unique version
of the ACS that has been linked to
Medicaid enrollment data. We fnd that
a relatively small number of people
with direct purchase in the ACS are
enrolled in Medicaid on their interview
date. We conclude that misclassifcation
of Medicaid coverage is not the main
driver of high direct purchase estimates
in the ACS.
The ACS estimates that
25 million non-elderly
people have direct
purchase coverage.
We primarily rely on two measures to investigate
the contribution of Medicaid misreporting to levels
of direct purchase. First, we report the percent of
the non-elderly that report direct purchase coverage
that are found to be enrolled in Medicaid or expan-
sion CHIP (Medicaid Plus) according to the MSIS.
Second, we report the size of the population reporting
direct purchase before and after logically editing the
data such that anyone identifed on MSIS as enrolled
in Medicaid Plus is not coded with direct purchase
coverage in the ACS. While it is impossible to be
certain that Medicaid enrollees are not also enrolled
in an individual market plan, we agree with Mach
and OHara (2011) that the likelihood is very small
that a person having low enough income to qualify
for Medicaid would choose to (and be able to given
eligibility rules) supplement their Medicaid coverage
with direct purchase coverage. We examine these
measures by age, poverty and state.
Finally, we examine whether the state-level percent
of those reporting direct purchase that are enrolled
in Medicaid per MSIS varies as a function of the
proportion of Medicaid enrollees participating in
managed care plans. Tis is an indirect test of the
hypothesis that managed care participants misclassify
themselves into direct purchase.
Results
Te universe in Table 1 is comprised of ACS
records that report direct purchase coverage in any
combination with other coverage types. Overall, the
TABLE 1. ANY DIRECT PURCHASE BY AGE &
POVERTY, 2009 ACS-MSIS LINKED FILE
Count
%
Linked SE
Corrected
Count
Age*
0-18 6,399,022 10.9 0.18 5,702,811
19-64 18,601,214 3.1 0.05 18,026,243
Total Non-Elderly 25,000,236 5.1 0.06 23,729,054
Income (% FPL)*
0-138 3,981,709 16.0 0.27 3,343,854
139-249 4,454,547 6.9 0.16 4,145,345
250-399 5,548,737 3.2 0.09 5,371,474
400+ 10,338,846 1.1 0.04 10,222,470
Source: 2009 ACS-MSIS Linked File, Non-Elderly Civilian Non-Institutional Population.
Notes: The % Linked column refers to the percent of cases reporting direct purchase
coverage that are enrolled in Medicaid or expansion CHIP on the date of interview based
on MSIS data. The corrected count is obtained by logically editing direct purchase
responses to No if the case is found as enrolled on date of interview.
* p<0.001. The Wald test suggests that % Linked is dependent on the row variable.
TABLE 2. ANY DIRECT PURCHASE BY STATE,
NON-ELDERLY, 2009 ACS-MSIS LINKED FILE
State* Count
%
Linked SE
Corrected
Count
Alabama 400,575 4.0 0.36 384,474
Alaska 41,071 1.8 0.54 40,316
Arizona 502,790 4.1 0.52 481,999
Arkansas 221,755 4.4 0.54 212,009
California 3,464,643 4.6 0.16 3,303,639
Colorado 516,877 1.5 0.20 509,301
Connecticut 289,221 4.9 0.52 274,937
Delaware 52,658 4.9 1.35 50,077
DC 76,796 10.5 1.88 68,761
Florida 1,495,186 4.2 0.24 1,432,137
Georgia 781,420 4.4 0.40 747,086
Hawaii 117,654 6.7 0.97 109,760
Idaho 168,750 2.5 0.38 164,477
Illinois 938,657 4.8 0.30 893,337
Indiana 456,634 3.5 0.37 440,707
Iowa 310,940 3.7 0.36 299,385
Kansas 247,871 3.2 0.41 240,043
Kentucky 311,476 5.5 0.61 294,460
Louisiana 370,546 6.5 0.53 346,527
Maine 94,002 5.0 0.80 89,256
Maryland 479,541 6.6 0.46 448,042
Massachusetts 562,094 5.8 0.43 529,289
Michigan 714,933 3.4 0.28 690,930
Minnesota 496,699 5.4 0.44 469,655
Mississippi 230,818 5.5 0.66 218,108
Missouri 473,889 4.2 0.36 453,943
Montana 105,874 1.4 0.35 104,430
Nebraska 186,150 2.1 0.39 182,327
Nevada 243,443 3.3 0.67 235,456
New Hampshire 86,384 2.6 0.49 84,116
New Jersey 620,405 4.5 0.40 592,238
New Mexico 146,558 5.3 0.83 138,829
New York 1,635,927 12.2 0.49 1,436,392
North Carolina 812,251 3.5 0.25 783,613
North Dakota 78,731 2.8 0.54 76,544
Ohio 757,849 5.8 0.44 713,583
Oklahoma 277,427 5.9 0.47 261,098
Oregon 349,725 2.1 0.32 342,455
Pennsylvania 1,025,954 9.9 0.51 924,374
Rhode Island 88,999 15.7 2.20 75,001
South Carolina 368,979 4.5 0.47 352,551
South Dakota 101,688 4.1 0.79 97,538
Tennessee 510,369 7.2 0.45 473,607
Texas 1,659,303 2.8 0.15 1,612,811
Utah 276,168 2.2 0.38 270,159
Vermont 44,830 6.3 1.14 42,019
Virginia 684,194 3.4 0.31 661,201
Washington 560,965 3.2 0.32 542,754
West Virginia 114,241 5.3 0.97 108,175
Wisconsin 391,060 5.0 0.41 371,409
Wyoming 55,267 2.8 0.99 53,720
Source: 2009 ACS-MSIS Linked File, Non-Elderly Civilian Non-Institutional
Population.
Notes: The % Linked column refers to the percent of cases reporting direct
purchase coverage that are enrolled in Medicaid or expansion CHIP on the
date of interview based on MSIS data. The corrected count is obtained by
logically editing direct purchase responses to No if the case is found as
enrolled on date of interview.
* p<0.001. The Wald test suggests that % Linked is dependent on the row
variable.
2
|
STATE HEALTH ACCESS DATA ASSISTANCE CENTER
MAY 2014
|
www.shadac.org 3
ESTIMATES OF DIRECT PURCHASE FROM THE ACS AND MEDICAID MISREPORTING
ACS estimates that 25 million non-elderly people
have direct purchase coverage. Tis far exceeds the
administrative count of 6.7 million reported in
Abraham et al. (2013). Among the full non-elderly
population, 5.1% of direct purchase reporters are
found to be enrolled in Medicaid or expansion CHIP
based on MSIS data. Te linkage rate is higher for
children than non-elderly adults (10.9%, p<0.001)
and higher for people at lower-levels of poverty.
Te fnal column presents the size of the direct
purchase population after logically editing the data
such that no MSIS-identifed Medicaid enrollee is
coded to direct purchase in the ACS. Overall, 23.7
million people are estimated to have corrected
direct purchase. Tis table suggests that while there
is some misclassifcation from Medicaid enrollees,
Medicaid misreporting is not the driving factor in the
misreporting of direct purchase.
Table 2 reports the same measure across the states.
Results from the Wald test suggest that the linkage
rate (i.e., the percent of those reporting direct
purchase in the ACS who are enrolled in Medicaid
according to MSIS) varies by state, ranging from
15.7% of those reporting direct purchase coverage in
Rhode Island to 1.4% in Montana.
Figure 1 plots the state level linkage rate (y-axis) as
a function of the percent of Medicaid participants
enrolled in managed care on the x-axis (managed
care penetration rates obtained from the Centers
TABLE 3. LINKAGE RATE AMONG DIRECT
PURCHASE CASES AND PERCENT IN MAN-
AGED CARE, BY STATE
State*
% Managed
Care % Linked
Alabama 60.5 4.0
Alaska 0.0 1.8
Arizona 90.4 4.1
Arkansas 78.2 4.4
California 55.7 4.6
Colorado 94.9 1.5
Connecticut 63.3 4.9
Delaware 78.3 4.9
DC 63.7 10.5
Florida 83.3 4.2
Georgia 83.3 4.4
Hawaii 98.7 6.7
Idaho 80.7 2.5
Illinois 58.1 4.8
Indiana 72.4 3.5
Iowa 81.7 3.7
Kansas 97.5 3.2
Kentucky 89.1 5.5
Louisiana 62.9 6.5
Maine 65.0 5.0
Maryland 77.2 6.6
Massachusetts 53.8 5.8
Michigan 89.7 3.4
Minnesota 63.6 5.4
Mississippi 75.9 5.5
Missouri 97.2 4.2
Montana 64.2 1.4
Nebraska 86.3 2.1
Nevada 86.1 3.3
New Hampshire 0.0 2.6
New Jersey 72.8 4.5
New Mexico 73.3 5.3
New York 69.0 12.2
North Carolina 84.1 3.5
North Dakota 62.4 2.8
Ohio 74.9 5.8
Oklahoma 84.5 5.9
Oregon 96.9 2.1
Pennsylvania 82.3 9.9
Rhode Island 68.5 15.7
South Carolina 100.0 4.5
South Dakota 76.8 4.1
Tennessee 100.0 7.2
Texas 66.1 2.8
Utah 84.1 2.2
Vermont 54.9 6.3
Virginia 59.8 3.4
Washington 90.0 3.2
West Virginia 47.6 5.3
Wisconsin 61.3 5.0
Wyoming 0.0 2.8
Source: 2009 ACS-MSIS Linked File, Non-Elderly Civilian Non-Institution-
al Population and Centers for Medicare and Medicaid Services (CMS).
Notes: % Linked is the linkage rate among cases reporting direct pur-
chase coverage (in any combination). % Managed care pertains to data
as of December 2010 obtained from CMS available at http://www.cms.
gov/Research-Statistics-Data-and-Systems/Computer-Data-and-Systems/
MedicaidDataSourcesGenInfo/Downloads/2010December31f.pdf. Rates
describe the percent of all Medicaid benefciaries that are enrolled in
managed care.
FIGURE 1. STATE LINKAGE RATE AS A FUNCTION OF PERCENT
MEDICAID PARTICIPANTS IN MANAGED CARE
Source: 2009 ACS-MSIS Linked File, Non-Elderly Civilian Non-Institutional Population and Centers for
Medicare and Medicaid Services (CMS). Managed care penetration as of December 2010 obtained from
CMS available at http://www.cms.gov/Research-Statistics-Data-and-Systems/Computer-Data-and-Systems/
MedicaidDataSourcesGenInfo/Downloads/2010December31f.pdf.
4
|
STATE HEALTH ACCESS DATA ASSISTANCE CENTER
for Medicare and Medicaid Services (CMS)). Te
graph demonstrates a very slight positive trend line
that could easily be driven by the handful of outliers.
Table 3 presents the tabular data behind the Figure 1
graphic.
Table 4 is similar to Table 1, except the universe is the
population that reports only direct purchase and no
other coverage type. Tis measure of direct purchase
that includes only those reporting direct purchase is
often suggested as a simple way to reduce the error in
the ACS direct purchase estimate (Mach and OHara,
2011; Abraham et al., 2013). Overall, the size of this
population is 16.3 million, considerably closer to the
6.7 million in the administrative count, but still over
twice as large. Among the 16.3 million identifed as
having direct purchase alone, 3.0% are found to be
linked to the MSIS. Te gradient by age and poverty
is the same in Table 4 as Table 1. After logically
editing the data, 15.8 million people are estimated to
have direct purchase alone.
Table 5 presents the same statistics as Table 4
for direct purchase alone, but at the state level.
Again the Wald test rejected the null hypothesis of
independence suggesting that the linkage rate among
the population with direct purchase varies by state.
Conclusion
Tis preliminary analysis of the connection between
survey error in direct purchase and Medicaid suggests
that while there is a plausible connection, it is small
and does not substantially contribute to the
TABLE 5. DIRECT PURCHASE ALONE BY STATE,
NON-ELDERLY, 2009 ACS-MSIS LINKED FILE
State* Count
%
Linked SE
Corrected
Count
Alabama 237,542 2.1 0.45 232,537
Alaska 21,920 NA NA 21,920
Arizona 362,673 3.1 0.65 351,513
Arkansas 141,597 2.6 0.51 137,900
California 2,526,355 2.7 0.17 2,457,672
Colorado 397,340 0.6 0.13 395,065
Connecticut 184,987 4.2 0.65 177,275
Delaware 28,674 2.3 1.05 28,000
DC 38,386 6.5 2.99 35,884
Florida 1,080,334 2.6 0.27 1,051,924
Georgia 486,363 2.9 0.41 472,089
Hawaii 60,370 5.6 1.12 56,980
Idaho 124,950 1.7 0.40 122,855
Illinois 634,266 3.2 0.31 614,137
Indiana 291,426 2.4 0.43 284,312
Iowa 210,248 1.9 0.26 206,219
Kansas 172,014 1.9 0.40 168,733
Kentucky 197,850 4.5 0.80 188,928
Louisiana 232,070 3.6 0.53 223,723
Maine 58,163 2.2 0.65 56,867
Maryland 282,231 5.3 0.51 267,373
Massachusetts 330,089 3.0 0.40 320,252
Michigan 448,426 1.5 0.23 441,507
Minnesota 344,484 3.6 0.43 332,039
Mississippi 137,607 3.0 0.62 133,413
Missouri 318,938 2.5 0.33 310,807
Montana 73,422 0.9 0.40 72,765
Nebraska 133,570 0.9 0.24 132,367
Nevada 121,827 2.3 0.86 119,003
New Hampshire 61,322 1.4 0.48 60,444
New Jersey 374,359 2.5 0.36 364,853
New Mexico 90,138 3.7 0.89 86,838
New York 861,225 7.3 0.50 797,958
North Carolina 548,891 2.0 0.29 537,721
North Dakota 57,397 1.3 0.50 56,645
Ohio 467,981 4.0 0.47 449,276
Oklahoma 178,492 3.7 0.50 171,944
Oregon 247,313 1.0 0.30 244,792
Pennsylvania 616,127 6.1 0.53 578,392
Rhode Island 57,930 15.1 2.75 49,187
South Carolina 222,372 1.9 0.33 218,154
South Dakota 71,224 2.2 0.67 69,655
Tennessee 338,555 4.3 0.43 324,152
Texas 1,070,516 1.5 0.14 1,053,955
Utah 188,476 0.9 0.25 186,797
Vermont 31,527 3.5 1.04 30,417
Virginia 407,135 2.2 0.39 397,996
Washington 387,971 1.8 0.29 380,883
West Virginia 49,532 2.3 0.84 48,399
Wisconsin 255,462 3.0 0.44 247,771
Wyoming 38,027 0.4 0.34 37,867
Source: 2009 ACS-MSIS Linked File, Non-Elderly Civilian Non-Institutional
Population.
Notes: The % Linked column refers to the percent of cases reporting direct
purchase coverage that are enrolled in Medicaid or expansion CHIP on the
date of interview based on MSIS data. The corrected count is obtained by
logically editing direct purchase responses to No if the case is found as
enrolled on date of interview.
* p<0.001. The Wald test suggests that % Linked is dependent on the row
variable.
TABLE 4. DIRECT PURCHASE ALONE BY AGE &
POVERTY, 2009 ACS-MSIS LINKED FILE
Count
%
Linked SE
Corrected
Count
Age*
0-18 4,297,001 6.7 0.19 4,008,956
19-64 12,003,093 1.7 0.05 11,799,200
Total Non-Elderly 16,300,093 3.0 0.07 15,808,156
Income (% FPL)*
0-138 2,703,083 8.0 0.27 2,487,191
139-249 2,964,403 4.2 0.17 2,839,737
250-399 3,582,945 2.3 0.11 3,499,960
400+ 6,588,961 0.8 0.05 6,537,239
Source: 2009 ACS-MSIS Linked File, Non-Elderly Civilian Non-Institutional Population.
Notes: The % Linked column refers to the percent of cases reporting direct purchase coverage
that are enrolled in Medicaid or expansion CHIP on the date of interview based on MSIS data. The
corrected count is obtained by logically editing direct purchase responses to No if the case is found
as enrolled on date of interview.
* p<0.001. The Wald test suggests that % Linked is dependent on the row variable.
MAY 2014
|
www.shadac.org 5
ESTIMATES OF DIRECT PURCHASE FROM THE ACS AND MEDICAID MISREPORTING
high levels of direct purchase in the ACS. While it
is informative to rule out Medicaid misreporting
as a source for bias in estimates of direct purchase
coverage, the exact mechanism that leads the
ACS to over-estimate direct purchase, relative to
administrative counts, remains allusive. Previous
work has found that a substantial number (roughly
10 million) of those reporting direct purchase do
so in combination with other coverage types. Tese
apparent multiple covered cases could be people
that interpret the direct purchase item as referring
to a single service plan. Te 2009 National Health
Interview Survey suggests that there are 5.9 million
non-elderly people that have single service plans.
Nearly all single service plan enrollees are also covered
by employer sponsored insurance (5.8 million). Tis
suggests that if all ESI enrollees with a single service
plan reported that coverage as directly purchased
comprehensive insurance, it would only explain about
half of the multiple coverage cases. Furthermore,
even after removing all the double counted cases, the
ACS still estimates twice as many individual market
enrollees compared to administrative data. Further
work is needed to understand the sources of bias in
the ACS estimate of direct purchase coverage. Tis
will be an increasingly important problem to solve as
the ACS starts to be used for tracking the impact of
the Afordable Care Act.
Acknowledgements
Tis research is funded by a contract with the U.S.
Census Bureau. We appreciate the contributions and
guidance of staf at the U.S. Census Bureau: Brett
OHara and Jennifer Cheeseman Day (Social, Eco-
nomic, and Housing Statistics Division).
About SHADAC
Te State Health Access Data Assistance Center is an
independent health policy research center located at
the University of Minnesota School of Public Health.
SHADAC is a resource for helping states collect and
use data for health policy, with a particular focus on
monitoring rates of health insurance coverage and
understanding factors associated with uninsurance.
For more information, please contact us at
[email protected], or call 612-624-4802.
Suggested Citation
Boudreaux, M., Call, K.T., Turner, J., Fried, B. 2014.
Estimates of Direct Purchase from the ACS and
Medicaid Misreporting: Is there a link? SHADAC
Brief #38. Minneapolis, MN: State Health Access
Data Assistance Center.
REFERENCES
Abraham, J.M., Karaka-Mandic, P., & Boudreaux, M. (2013). Sizing up the individual market for health insurance: A
comparison of survey and administrative data sources. Medical Care Research and Review, 70(4), 418-433.
Boudreaux, M., Call, K.T., Turner, J., Fried, B., & OHara, B. (2013). Accuracy of Medicaid reporting in the ACS: Prelim-
inary results from linked data. State Health Access Data Center and U.S. Census Bureau.
Call, K.T., Davidson, G., Davern, M.E., Brown, E.R., Kincheloe, J., Nelson, J.G. Winter 2008/2009. Accuracy of self-re-
ported health insurance coverage among Medicaid enrollees. Inquiry, 45(4), 438-456.
Cantor, J.C., Monheit, A.C., Brownlee, S., & Schneider, C. (2006). The adequacy of household survey data for evaluat-
ing the nongroup health insurance market. Health Services Research, 42(4), 1739-1757.
Mach, A., & OHara, B. (2011). Do people really have multiple health insurance plans? Estimates of nongroup health
insurance in the American Community Survey. (SEHSD Working Paper Number 2011-28).
Plotzke, M.R., Klerman, J. A., & Davern, M. (2010). How does Medicaid-managed care impact reporting of Medicaid
status? Health Services Research, 45(5p1), 1310-1323.

Funded by a grant from The Robert Wood Johnson Foundation
2014 Regents of the University of Minnesota. All rights reserved. The University of Minnesota is an Equal Opportunity Employer.
Bridging the gap between research and policy @ www.shadac.org
Sign up to receive
our newsletter
and updates at
www.shadac.org
05082014