Skip to content

The Role of Behavioral Frictions in Health Insurance Marketplace Enrollment and Risk: Evidence from a Field Experiment

icpsr_125801

Richard Domurat, Isaac Menashe, Wesley Yin (2021). American Economic Review

DOI: 10.1257/aer.20190823

field value
venue American Economic Review
paper 10.1257/aer.20190823
replication data openICPSR 125801
data license CC BY 4.0 — “This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) License.”
regression datasets (in this corpus) 3
graduated / eligible / exported 3 / 9 / 52 (policy claims_manual_regression_cap)
routed / gap / residue (at source) 28 / 150 / 0

Claims & tasks

Every claim the paper makes in its main body about a parameter, joined to the benchmark task (dataset · coefficient) that captures it. The effect, s.e. and p are the captured regression's own reported values (the paper's Stata figures, carrying the paper's SE method); the row shows the paper's verbatim quote (page · exhibit), so you can confirm the regression matches what the paper says. The prominence column is the within-study inclusion signal (⭐ headline → primary → secondary). The sensitivity column is the privacy level of the most-sensitive variable in the regression (🔴 high / 🟠 medium / ⚪ low).

# prominence sensitivity claim effect s.e. p task instance (dataset · coef)
C1 ⭐ primary 🔴 high Among low-income consumers, letters making personalized subsidy and plan comparison information more salient raise enrollment slightly, indicating that some consumers are not fully aware of the magnitude of subsidy benefits.
p. 1552 · Table 4, col 1 (Open enrollment (2))
+0.049 0.019 0.010 Linear Regression reg_13 · arm345
n=32698 · d=78
C2 ⭐ primary 🔴 high “In the open enrollment sample, we observe a larger negative coefficient on the income × subsidy-reporting (Arm345) interaction, indicating that take-up falls more steeply when receiving subsidy information than the Basic Letter.” — p. 1564
“The difference in the coefficient estimates on Arm2 × FPL (× 100) and Arm345 × FPL (× 100) is only significant at the 0.169 level for the full sample, and 0.158 level for the open enrollment sample.” — p. 1564
Table 4, col 2 (OLS Open enrollment (2))
-0.004 0.010 0.686 Linear Regression reg_13 · arm2_subsidy_fplx100
n=32698 · d=78
C3 ⭐ primary 🔴 high “In the open enrollment sample, we observe a larger negative coefficient on the income × subsidy-reporting (Arm345) interaction, indicating that take-up falls more steeply when receiving subsidy information than the Basic Letter.” — p. 1564
“As subsidies fall with income, the relative benefit of providing subsidy information also falls, further indicating that consumers misperceive their subsidies.” — p. 1552
“The difference in the coefficient estimates on Arm2 × FPL (× 100) and Arm345 × FPL (× 100) is only significant at the 0.169 level for the full sample, and 0.158 level for the open enrollment sample.” — p. 1564
Table 4, col 2 (OLS Open enrollment (2))
-0.015 0.008 0.053 Linear Regression reg_13 · arm345_subsidy_fplx100
n=32698 · d=78
C4 primary 🔴 high Using either unadjusted or regression-adjusted models, we find no differences in overall take-up across the letter interventions, a result we revisit in subgroup analyses.
p. 1561 · Table 3, col 2 (All (2))
+0.012 0.003 3.9e-05 Linear Regression reg_5 · arm2
n=87394 · d=4
C5 primary 🔴 high Using either unadjusted or regression-adjusted models, we find no differences in overall take-up across the letter interventions, a result we revisit in subgroup analyses.
p. 1561 · Table 3, col 2 (All (2))
+0.015 0.003 1.2e-06 Linear Regression reg_5 · arm3
n=87394 · d=4
C6 primary 🔴 high Using either unadjusted or regression-adjusted models, we find no differences in overall take-up across the letter interventions, a result we revisit in subgroup analyses.
p. 1561 · Table 3, col 2 (All (2))
+0.010 0.003 5.7e-04 Linear Regression reg_5 · arm4
n=87394 · d=4
C7 primary 🔴 high Using either unadjusted or regression-adjusted models, we find no differences in overall take-up across the letter interventions, a result we revisit in subgroup analyses.
p. 1561 · Table 3, col 2 (All (2))
+0.013 0.003 2.3e-05 Linear Regression reg_5 · arm5
n=87394 · d=4
C8 primary 🔴 high While the differences across the Basic Letter and subsidy-reporting arms are not statistically different, the patterns are consistent with a hypothesis that providing more information beyond the uniform reminder, about personalized premium subsidies and plan options, further reduces information frictions for a population with low baseline awareness.
p. 1563 · Table 3, col 6 (County referral (6))
+0.004 0.003 0.191 Linear Regression reg_9 · arm2
n=43146 · d=4
C9 primary 🔴 high “For the County referral sample, we find that providing additional subsidy and plan information (Arms 3–5) beyond the Basic Letter (Arm 2) leads to higher take-up (column 6, and in column 8, where the percent impact of a given treatment arm on county referrals is obtained by adding the coefficients on its uninteracted and county referral-interacted terms).” — p. 1563
“While the differences across the Basic Letter and subsidy-reporting arms are not statistically different, the patterns are consistent with a hypothesis that providing more information beyond the uniform reminder, about personalized premium subsidies and plan options, further reduces information frictions for a population with low baseline awareness.” — p. 1563
Table 3, col 6 (County referral (6))
+0.007 0.003 0.014 Linear Regression reg_9 · arm3
n=43146 · d=4
C10 primary 🔴 high “For the County referral sample, we find that providing additional subsidy and plan information (Arms 3–5) beyond the Basic Letter (Arm 2) leads to higher take-up (column 6, and in column 8, where the percent impact of a given treatment arm on county referrals is obtained by adding the coefficients on its uninteracted and county referral-interacted terms).” — p. 1563
“While the differences across the Basic Letter and subsidy-reporting arms are not statistically different, the patterns are consistent with a hypothesis that providing more information beyond the uniform reminder, about personalized premium subsidies and plan options, further reduces information frictions for a population with low baseline awareness.” — p. 1563
Table 3, col 6 (County referral (6))
+0.006 0.003 0.032 Linear Regression reg_9 · arm4
n=43146 · d=4
C11 primary 🔴 high “For the County referral sample, we find that providing additional subsidy and plan information (Arms 3–5) beyond the Basic Letter (Arm 2) leads to higher take-up (column 6, and in column 8, where the percent impact of a given treatment arm on county referrals is obtained by adding the coefficients on its uninteracted and county referral-interacted terms).” — p. 1563
“While the differences across the Basic Letter and subsidy-reporting arms are not statistically different, the patterns are consistent with a hypothesis that providing more information beyond the uniform reminder, about personalized premium subsidies and plan options, further reduces information frictions for a population with low baseline awareness.” — p. 1563
Table 3, col 6 (County referral (6))
+0.007 0.003 0.026 Linear Regression reg_9 · arm5
n=43146 · d=4

Regression datasets

The 3 regression dataset(s) graduated into the benchmark corpus from this study's replication package — each backs one or more claims above. Expand each for its features, response, (sound, data-independent) public bounds, and reproduction grade.

icpsr_125801_reg_13 — any_takeup ~ arm2 + arm345 + subsidy_fplx100 + arm2_subsidy_fplx100 + arm345_subsidy_fplx…

Sensitivity proposal: 🔴 high maximum across columns (advisory; reviewed manually). Reproduction grade:match (benchmark-eligible).

Features

name description type sensitivity bound lo bound hi estimate s.e.
_cons intercept continuous 1 1 +0.1776 0.0345
arm2 Basic Letter treatment arm indicator continuous ⚪ low 0.0 1.0 +0.02451 0.0235
arm345 subsidy-reporting arms (3-5) combined indicator continuous ⚪ low 0.0 1.0 +0.04947 0.0192
subsidy_fplx100 household income as percent of poverty line continuous 🔴 high -0.03622 0.00683
arm2_subsidy_fplx100 interaction of arm2 and income-to-FPL ratio continuous ⚪ low -0.003948 0.00978
arm345_subsidy_fplx100 interaction of arm345 and income-to-FPL ratio continuous ⚪ low -0.01539 0.00795
age_mean10 mean age scaled by dividing by 10 continuous 🟠 medium 0.0 12.0 +0.01512 0.00353
funnel_round enrollment funnel stage or round categorical ⚪ low
latino Latino/Hispanic ethnicity indicator continuous 🔴 high 0.0 1.0 -0.0291 0.0118
asian Asian race/ethnicity indicator continuous 🔴 high 0.0 1.0 -0.003175 0.0162
black Black race/ethnicity indicator continuous 🔴 high 0.0 1.0 -0.04988 0.0218
otherrace other/unclassified race indicator continuous 🔴 high 0.0 1.0 -0.03621 0.0131
subsidy_hh_size household size used for subsidy calculation continuous 🟠 medium -0.01848 0.00499
hh_members total number of household members continuous 🟠 medium +0.03714 0.00862
hh_kids number of children in household continuous 🟠 medium -0.005296 0.014
married_infer inferred marital status of household head continuous 🟠 medium 0.0 1.0 -0.01539 0.0135
hoh_email_ok indicator household head email deliverable continuous ⚪ low 0.0 1.0 +0.02734 0.00976
x7 unnamed covariate, meaning unclear continuous ⚪ low -0.09798 0.00922
region geographic region of household residence categorical 🟠 medium
arm2_age_mean10 interaction of arm2 and scaled age continuous ⚪ low -0.004254 0.00504
arm2_funnel_round1 interaction of arm2 and funnel round dummy continuous ⚪ low -0.01312 0.0287
arm2_latino interaction of arm2 and Latino dummy continuous ⚪ low +0.01632 0.0169
arm2_asian interaction of arm2 and Asian dummy continuous ⚪ low -0.01252 0.0231
arm2_black interaction of arm2 and Black dummy continuous ⚪ low +0.01175 0.0319
arm2_otherrace interaction of arm2 and other-race dummy continuous ⚪ low +0.04406 0.0193
arm2_subsidy_hh_size interaction of arm2 and subsidy household size continuous ⚪ low -0.0003609 0.00754
arm2_hh_members interaction of arm2 and household size continuous ⚪ low +0.0193 0.0126
arm2_hh_kids interaction of arm2 and count of kids continuous ⚪ low -0.0207 0.0196
arm2_married_infer interaction of arm2 and inferred marriage continuous ⚪ low -0.001573 0.0195
arm2_hoh_email_ok interaction of arm2 and email-valid flag continuous ⚪ low +0.02222 0.014
arm2_x7 interaction of arm2 and covariate x7 continuous ⚪ low +0.001242 0.0137
arm2_region1 interaction of arm2 and region 1 dummy continuous ⚪ low -0.01619 0.0402
arm2_region2 interaction of arm2 and region 2 dummy continuous ⚪ low +0.02035 0.0423
arm2_region3 interaction of arm2 and region 3 dummy continuous ⚪ low +0.08286 0.0355
arm2_region4 interaction of arm2 and region 4 dummy continuous ⚪ low +0.03644 0.0506
arm2_region5 interaction of arm2 and region 5 dummy continuous ⚪ low +0.03371 0.0469
arm2_region6 interaction of arm2 and region 6 dummy continuous ⚪ low -0.04446 0.0419
arm2_region7 interaction of arm2 and region 7 dummy continuous ⚪ low +0.02395 0.0397
arm2_region8 interaction of arm2 and region 8 dummy continuous ⚪ low +0.08631 0.0521
arm2_region9 interaction of arm2 and region 9 dummy continuous ⚪ low -0.06338 0.0516
arm2_region10 interaction of arm2 and region 10 dummy continuous ⚪ low +0.005626 0.0349
arm2_region11 interaction of arm2 and region 11 dummy continuous ⚪ low -8.93e-05 0.0425
arm2_region12 interaction of arm2 and region 12 dummy continuous ⚪ low +0.009812 0.04
arm2_region13 interaction of arm2 and region 13 dummy continuous ⚪ low +0.01272 0.0623
arm2_region14 interaction of arm2 and region 14 dummy continuous ⚪ low +0.01379 0.0409
arm2_region15 interaction of arm2 and region 15 dummy continuous ⚪ low -0.02281 0.0307
arm2_region16 interaction of arm2 and region 16 dummy continuous ⚪ low +0.01417 0.0287
arm2_region17 interaction of arm2 and region 17 dummy continuous ⚪ low +0.00412 0.0285
arm2_region18 interaction of arm2 and region 18 dummy continuous ⚪ low +0.06222 0.0304
arm345_age_mean10 interaction of arm345 and scaled age continuous ⚪ low -0.004828 0.00411
arm345_funnel_round1 interaction of arm345 and funnel round dummy continuous ⚪ low +0.006033 0.0236
arm345_latino interaction of arm345 and Latino dummy continuous ⚪ low +0.01512 0.0136
arm345_asian interaction of arm345 and Asian dummy continuous ⚪ low -0.009458 0.0188
arm345_black interaction of arm345 and Black dummy continuous ⚪ low +0.0144 0.0256
arm345_otherrace interaction of arm345 and other-race dummy continuous ⚪ low +0.0196 0.0153
arm345_subsidy_hh_size interaction of arm345 and subsidy household size continuous ⚪ low +0.0001751 0.00582
arm345_hh_members interaction of arm345 and household size continuous ⚪ low +0.01928 0.0101
arm345_hh_kids interaction of arm345 and count of kids continuous ⚪ low -0.02565 0.0161
arm345_married_infer interaction of arm345 and inferred marriage continuous ⚪ low -0.008004 0.0156
arm345_hoh_email_ok interaction of arm345 and email-valid flag continuous ⚪ low +0.008148 0.0114
arm345_x7 interaction of arm345 and covariate x7 continuous ⚪ low -0.0009506 0.0108
arm345_region1 interaction of arm345 and region 1 dummy continuous ⚪ low +0.03477 0.0335
arm345_region2 interaction of arm345 and region 2 dummy continuous ⚪ low +0.05366 0.0337
arm345_region3 interaction of arm345 and region 3 dummy continuous ⚪ low +0.05729 0.0277
arm345_region4 interaction of arm345 and region 4 dummy continuous ⚪ low +0.02097 0.0404
arm345_region5 interaction of arm345 and region 5 dummy continuous ⚪ low +0.05891 0.0387
arm345_region6 interaction of arm345 and region 6 dummy continuous ⚪ low +0.01602 0.0345
arm345_region7 interaction of arm345 and region 7 dummy continuous ⚪ low +0.07378 0.0315
arm345_region8 interaction of arm345 and region 8 dummy continuous ⚪ low +0.0476 0.0405
arm345_region9 interaction of arm345 and region 9 dummy continuous ⚪ low -0.02604 0.0423
arm345_region10 interaction of arm345 and region 10 dummy continuous ⚪ low +0.01042 0.0278
arm345_region11 interaction of arm345 and region 11 dummy continuous ⚪ low -0.01166 0.0344
arm345_region12 interaction of arm345 and region 12 dummy continuous ⚪ low +0.03206 0.0323
arm345_region13 interaction of arm345 and region 13 dummy continuous ⚪ low +0.05625 0.0538
arm345_region14 interaction of arm345 and region 14 dummy continuous ⚪ low +0.06772 0.0332
arm345_region15 interaction of arm345 and region 15 dummy continuous ⚪ low +0.01612 0.025
arm345_region16 interaction of arm345 and region 16 dummy continuous ⚪ low +0.02537 0.0231
arm345_region17 interaction of arm345 and region 17 dummy continuous ⚪ low +0.02573 0.023
arm345_region18 interaction of arm345 and region 18 dummy continuous ⚪ low +0.04953 0.024

Response

name description type sensitivity bound lo bound hi estimate s.e.
any_takeup indicator household enrolled in health insurance continuous 🔴 high 0.0 1.0

Public bounds (data-independent) sourced from: age_mean/10, age range 0-120; binary indicator; binary indicator; health-insurance enrollment outcome; binary indicator; marital status; binary indicator; race/ethnicity.

n = 32,698 samples.

icpsr_125801_reg_5 — any_takeup ~ arm2 + arm3 + arm4 + arm5

Sensitivity proposal: 🔴 high maximum across columns (advisory; reviewed manually). Reproduction grade:match (benchmark-eligible).

Features

name description type sensitivity bound lo bound hi estimate s.e.
_cons intercept continuous 1 1 +0.08102 0.00207
arm2 Basic Letter treatment arm indicator continuous ⚪ low 0.0 1.0 +0.01243 0.00302
arm3 subsidy/plan-information letter arm indicator continuous ⚪ low 0.0 1.0 +0.01475 0.00304
arm4 subsidy/plan-information letter arm indicator continuous ⚪ low 0.0 1.0 +0.01036 0.003
arm5 subsidy/plan-information letter arm indicator continuous ⚪ low 0.0 1.0 +0.0128 0.00302

Response

name description type sensitivity bound lo bound hi estimate s.e.
any_takeup indicator household enrolled in health insurance continuous 🔴 high 0.0 1.0

Public bounds (data-independent) sourced from: binary indicator; binary indicator; health-insurance enrollment outcome.

n = 87,394 samples.

icpsr_125801_reg_9 — any_takeup ~ arm2 + arm3 + arm4 + arm5

Sensitivity proposal: 🔴 high maximum across columns (advisory; reviewed manually). Reproduction grade:match (benchmark-eligible).

Features

name description type sensitivity bound lo bound hi estimate s.e.
_cons intercept continuous 1 1 +0.03575 0.00199
arm2 Basic Letter treatment arm indicator continuous ⚪ low 0.0 1.0 +0.003782 0.00289
arm3 subsidy/plan-information letter arm indicator continuous ⚪ low 0.0 1.0 +0.007246 0.00295
arm4 subsidy/plan-information letter arm indicator continuous ⚪ low 0.0 1.0 +0.006296 0.00294
arm5 subsidy/plan-information letter arm indicator continuous ⚪ low 0.0 1.0 +0.006569 0.00294

Response

name description type sensitivity bound lo bound hi estimate s.e.
any_takeup indicator household enrolled in health insurance continuous 🔴 high 0.0 1.0

Public bounds (data-independent) sourced from: binary indicator; binary indicator; health-insurance enrollment outcome.

n = 43,146 samples.