can markets for health insurance work?
DESCRIPTION
Can Markets for Health Insurance Work?. Jonathan Levin Gaston Eyskens Lectures November 2013. Roadmap. Lectures Technology and Asymmetric Information High Risk Consumer Credit Markets Measuring Inefficiencies from Adverse Selection Can Markets for Health Insurance Work?. Introduction. - PowerPoint PPT PresentationTRANSCRIPT
Can Markets for Health Insurance Work?
Jonathan Levin
Gaston Eyskens LecturesNovember 2013
Roadmap
Lectures
1. Technology and Asymmetric Information2. High Risk Consumer Credit Markets3. Measuring Inefficiencies from Adverse Selection4. Can Markets for Health Insurance Work?
Introduction Arrow (1960): healthcare as an industry where traditional
notions of competition and efficiency are problematic.
Why? Uncertainty and asymmetric information. – Needs are concentrated => must have insurance– Potential for adverse selection in insurance– Moral hazard for insured individuals– Physicians asymmetrically informed about care
How well can markets for health insurance work? Should there be a major role for private market competition in a national healthcare system?
Private Market Solutions Of developed countries, US has perhaps the most “privatized”
healthcare / health insurance system.
US OECD$0
$1,000$2,000$3,000$4,000$5,000$6,000$7,000$8,000$9,000
Health Spending per capita, US$ PPP
US OECD0%
10%
20%
30%
40%
50%
60%
70%
80%
Public Spending, % of Total Spending
It also has the most expensive system.
Source: OECD
What Role for Markets? Major debate in the United States, where healthcare is 18 percent of
GDP, and 21 percent of federal budget.
Medicare (over 65)– Private market insurance now offered as a substitute for government
insurance. Republicans propose to shift entirely to “premium support”.
Obama Health Insurance Reform (under 65)– Attempts to provide universal coverage by mandating the purchase of
(subsidized) insurance, and establishing state insurance exchanges.
Some parallel debates in European countries.
Outline of Lecture Describe research with Jay Bhattacharya, Vilsa Curto, Liran Einav on
Medicare’s experiment with managed competition.
Motivation: current structure of US Medicare program is opportunity to compare two public / private systems operating head-to-head, fixing the population and set of healthcare providers.
Outline of lecture– Medicare and managed competition– Risk selection– Imperfect competition– Market design
Caveat: work in progress….
Background on US Medicare Program
Medicare program is government provided health insurance for all those over 65, plus disabled, ESRD, etc.– Individuals pay modest (≈$100) monthly premium.– Insurance covers hospital and outpatient services.– Privatized drug coverage introduced in 2006.
Traditional “fee for service” insurance (FFS)– Beneficiaries can use any doctor or hospital that accepts
Medicare. Most do. Medicare reimburses for services.– Price for each service is set administratively, plus there are a
complex set of hospital / facility subsidies.
Medicare is Expensive
Concerns about Medicare
Widely viewed as very inefficient– Why? Low copayments + physician incentives to “do more” =>
overuse of services. No constraints on spending / utilization.– Many documented distortions due to administrative pricing, e.g.
bias towards procedures, specialists.
Striking geographic variation in costs– Medicare spends $13,824 to provide healthcare to someone in
McAllen, TX, versus $7,766 in Rochester.– Common interpretation: indicates wide scope to provide
different services, and lack of cost-efficient “standards” of care.
Medicare Advantage Program
US introduced private plans in 1985: beneficiaries could opt out of traditional insurance, and enroll in a private insurance plan.
Approved plans receive annual payment for each enrollee, initially 95% of local FFS costs, and updated periodically.
Hope was that private plans would allow consumer flexibility, save taxpayer dollars, and preserve quality.
But there were problems … Low take-up – less than 15% of beneficiaries as of 2003 Evidence of significant risk selection (cream-skimming) Projected cost savings were not realized
Managed Competition
In 2003, new legislation introduced competitive bidding.– Attempt to avoid setting administrative capitation rates.– Risk adjustment added to address selection.
Based on proposals that date back to the 1970s. Some similarities to proposed state exchanges.
Risk adjustment developed between 2003-2006.‒ Every Medicare beneficiary scored based on disease history.‒ Regression model used to map diseases into predicted cost.
Expansion of Private Medicare
19851986
19871988
19891990
19911992
19931994
19951996
19971998
19992000
20012002
20032004
20052006
20072008
20092010
20112012
20130%
5%
10%
15%
20%
25%
30%
0
200
400
600
800
1000
1200
1400
1600
Number of ContractsEnrollment
Managed Competition
Pre-Legislation
Questions
Selection – Do private plans manage to select relatively healthy
enrollees? Is risk-adjustment successful in addressing this?
Competition– Is bidding competitive? Do bids reflect underlying costs, or
does the program create substantial insurer rents?
Further questions about quality, innovation, cost growth, efficacy of consumer choices, etc..
Selection: enrollment by risk score
.290.460
.643.851
1.0521.249
1.4501.649
1.8512.383
4.1915.995
7.7979.588
11.377
13.1540
0.05
0.1
0.15
0.2
0.25
Risk Score
Med
icar
e Ad
vant
age
Enro
llmen
t
High risk beneficiaries are less likely to enroll in private plans.
Selection: conditional on risk score
.290.344
.452.541
.642.748
.850.949
1.0521.148
1.2491.350
1.4501.548
1.6481.749
1.8511.950
0
0.02
0.04
0.06
0.08
0.1
0.12
-100
-80
-60
-40
-20
0
20
40
60
80
100
Lagged Cost Differential (TM-MA)TM MortalityMA Mortality (New Enrollees)
Risk Score
Mor
talit
y Ra
te
Lagg
ed M
onth
ly C
osts
($)
Mortality notably lower among new MA enrollees.
Disease Classification Incentives
00 00 01 01 02 0200
00
01
01
02
02
Traditional Medicare
Medicare Advantage
Risk Score in Year t
Risk
Sco
re in
Yea
r t+1
Private plans are more aggressive in coding diseases.
Adjusting for Selection & Up-coding For each private plan enrollee, we observe plan’s bid and payment,
and can impute an FFS cost based on county and risk score. Up-coding and selection adjustments (still somewhat ad hoc):
– Disease classification: MA risk scores higher by 2-5% (?)– Selection: based on mortality + lagged MA costs => 1-4% (?)
Avg. across MA enrollees
As fraction of imputed FFS
Imputed FFS Cost $ 9,021 100
Private Plan “Adj.” Bids $ 8,336 92.4
Private Plan Payment $ 9,311 103.2
Adjusted FFS Costs - 91-97
Competition and Market Concentration
Medicare Advantage market is highly concentrated
Fraction of counties where market share is > 90% > 85% > 75%
Top two insurers 0.26 0.40 0.66
Top three insurers 0.59 0.75 0.94
We need to take imperfect competition seriously.
Medicare Advantage Rules Country divided into local markets (3,034 counties). Medicare sets benchmark rate B for each county. Plans bid to provide “standard” insurance for a “standard” individual. Individuals choose a plan in their market, or standard Medicare.
B b
CMS pays plan BEnrollee pays b-B
$
b
CMS pays plan bEnrollee pays 0
c
Plan Cost Benchmark
And CMS pays plan a “rebate” ¾*(B-b) that it passes to consumer in as higher benefits.
Consumer “price”:
Benchmarks and Bidding Incentives Plan demand depends on .
Plan profits
Benchmark plays the role of a subsidy.
Under “perfect” competition with , .
Cross-Market Bid/Benchmark Relationship
Empirical Model of Plan Bids
Empirical model of plan j bid in year t.
A single plan bid may apply to several counties => define
– Plan benchmark = average of county benchmarks in a plan’s service area (using plan enrollment to weight).
– Plan costs = average FFS costs for an r=1 individual in plan’s service area (again, plan enrollment to weight).
Plan Bid EstimatesDependent Variable: Plan-Level Bid (mean 82.83)
(1) (2) (3) (4) (5)
Plan benchmark 0.488 0.503 0.537 0.499 0.602(0.0206) (0.0207) (0.0221) (0.0173) (0.0649)
Plan FFS cost 0.0910 0.0318 0.0520 0.234 0.0790
(0.0123) (0.0154) (0.0156) (0.0144) (0.0318)
Year FEs Y Y Y Y
Insurer FEs Y Y Y
Contract FEs Y Y
Plan FEs Y
R-squared 0.284 0.308 0.578 0.750 0.906
Observations 10436 10436 10436 10436 10436
Implications of Bid Estimates
Bids are highly responsive to benchmarks, but not closely related to FFS costs (conditional on B).
Suggestive of market power…
Ideally, would like to estimate private plan margins .– Local TM costs don’t appear be a good proxy for . – Directly measure utilization? Might be possible … not today.– Estimate demand and infer from profit-maximization.
Estimating Plan Demand
Assume i’s utility for plan j in county-year (market) m is
Imposing a logit error gives us the demand model
‒ where Xs are plan and market characteristics and we use in place of premium, with effect varying on
Demand Elasticity EstimatesDependent Variable: ln(Plan Risk Q) - ln(TM Risk Q)
(1) (2) (3) (4) (5)
Bid minus benchmark ($00s) -0.871 -1.03 -1.76 -1.88 -1.06
× I{Bid > benchmark} (0.046) (0.047) (0.044) (0.043) (0.0425)
Bid minus benchmark ($00s) -0.353 -0.479 -0.791 -0.806 -0.496
× I{Bid ≤ benchmark} (0.012) (0.012) (0.013) (0.014) (0.0161)
ln(Total market size) -0.352 -0.361 -0.343 -0.374 -0.294
(0.00309) (0.00301) (0.00291) (0.00291) (0.00262)
Year FEs Y Y Y Y
Insurer FEs Y Y Y
Contract FEs Y Y
Plan FEs Y
R-squared 0.131 0.171 0.32 0.369 0.603
Observations 224797 224797 224797 224797 224797
Includes characteristics: plan star rating, indicators for part D & supplementary benefits.
Plan Demand and Margins𝑏 𝑗
𝐵
ln (𝑄 𝑗)
𝐵−100
𝐵+100
𝐵−200
𝑏 𝑗=𝑐 𝑗+(− 𝑑𝑙𝑛𝑄 𝑗𝑚
𝑑𝑏 𝑗𝑚)− 1
≈𝑐 𝑗+$1 25
𝑏 𝑗=𝑐 𝑗+(− 𝑑𝑙𝑛𝑄 𝑗𝑚
𝑑𝑏 𝑗𝑚)− 1
≈𝑐 𝑗+$55
𝑏
𝑏
Inferring Margins Profit-maximization
Average bid is , so margins are
Cutting margins in half would save $10 bn a year.
Could these savings be realized? How?– Better rebate design (current DWL) – Lower benchmarks (competition with TM) – Better choice guidance for consumer (star ratings!)– Entry and greater competition
Market Design: Rebates Insurer that bids receives rebate of .
Typically used to provide more generous insurance.
Why not pass the rebate directly to enrollees in cash?‒ Logistically challenging under current rules.‒ Estimates say elasticity increases => margins of $71, not $125.
What about increasing/decreasing the rebate?‒ If rebate then margins = $62, but taxpayers always pay … this is the
premium support model Republicans want.
‒ By lowering rebate, e.g. to 50%, reduces taxpayer cost with fixed bids, but reduces incentives for insurers to lower their bids.
Market Design: Benchmarks County benchmarks are main design parameter
‒ If bids track benchmark, then lower B means lower margins.– However, entry is also responsive to market benchmarks.
Back of envelope calculation. A $100 decrease in benchmark:‒ directly decreases bids by $50‒ but maybe plans would exit?
Benchmarks and Plan Entry
Plan Entry and Bids
Choosing Benchmarks County benchmarks are main design parameter
‒ If bids track benchmark, then lower B means lower margins.– However, entry is also responsive to market benchmarks.
Back of envelope calculation. A $100 decrease in benchmark:‒ directly decreases bids by $50– induces exit of 3.5 plans, which increases bids, but by < $5
Based on our estimates (including margins of > $100 per month per enrollee) desirable to reduce benchmarks.
We are about to try this under Obamacare.
Prospects for Competition
Die or Exit Sample
SwitchMA/TM
Stay MA/TM
Stay with Insurer
MA Enrollees 4.48% 4.57% 90.95% 81.40%
TM Enrollees 5.72% 1.89% 92.39% -
Are private plans competing with each other, or the government?
MA/TM enrollment transitions, 2006-2010
In markets with low private plan penetration, traditional government insurance is the primary competition.
(Note: choices also are pretty persistent … switching costs?)
Public “Option” as Constraint
Option to use TM appears to be an important constraint on plan bids, especially in areas with low penetration / few plans.
Interestingly, the Obamacare reform does not include a public option for insurance, although some lobbied for it.
One hypothesis is that private plan competition may not work all that well in low entry (likely rural) markets.
Conclusions
US is engaging in a huge national experiment with managed competition provision of health insurance.
Ongoing Medicare private plan program has grown rapidly, but at what appears to be fairly significant cost to taxpayers.
There do appear to be potential cost savings; the question is whether it is possible to realize them.
We have started to look a few possible mechanisms, but project is still at a fairly early stage…
Thank You!