Session 30 PD, New research in Behavioral Economics in Health Care. Moderator/Presenter: Christopher James Coulter, FSA, MAAA, CERA

(1)

Session 30 PD, New research in Behavioral Economics in Health Care

Moderator/Presenter:

Christopher James Coulter, FSA, MAAA, CERA

Presenters:

Keith Ericson, Ph.D. Ben Handel, Ph.D.

(2)

Health Care

(3)

(4)

(5)

(6)

• Loewenstein et al. “Consumers’ misunderstanding of health insurance.”

Journal of Health Economics 32 (2013) 850– 862

• Sydnor, J., 2010. (Over)insuring modest risks. American Economic Journal:

Applied Economics 2, 177–199.

• Kate Ho and Ariel Pakes (2014) “Hospital Choices, Hospital Prices and

Financial Incentives to Physicians,” American Economic Review, 104(12):

(7)

Society of Actuaries Annual Meeting

Panel Discussion:

Behavioral Economics in Health Care

Ben Handel

Department of Economics

Berkeley

(8)

Behavioral Economics and Health Insurance

• Choosing Health insurance for most consumers is:

•

Complicated

•

Boring

•

Important

• What does this imply for choice quality?

•

Active decision-making errors

•

Default options, inertia, and passive decision-making errors

•

Preponderance of evidence that both kind of errors are widespread

• Why should we care?

•

Consumer welfare

(9)

Roadmap

• Active Decision Errors: Limited Information

• Passive Decision Errors: Inertia

(10)

Limited Information and Active Decisions

Handel & Kolstad (2015)

•

Study of consumer information in health insurance decisions for

employees of large firm (~50,000 domestic employees)

•

Choice between:

• Broad network PPO option with zero premium and zero cost-sharing

• HDHP option with same network, substantial subsidy into HSA

•

Study value of insurance choices with:

• 5 years of claims data for employees and dependents

• Ex ante medical metrics using diagnostic information

• Insurance choices made

• Demographics and employment data

(11)

Limited Information and Active Decisions

Handel & Kolstad (2015)

• Despite evident financial value in HDHP, most people (85%) choose PPO

• Desire for risk protection by fully-informed consumers unlikely explanation

(12)

Linked Survey Data

Handel & Kolstad (2015)

•

One method: ask consumers a series of simple questions near open

enrollment about health insurance options

•

Electronic survey randomly given to consumers after open enrollment in

November 2011. 25 multiple choice questions covering:

• Knowledge of plan financial characteristics

• Knowledge of own total health expenditures

• Knowledge of plan provider networks

• Knowledge of HSA advantages

• Knowledge of hassle costs associated with cost-sharing / HSA

•

Survey randomly assigned to 4500 employees, response rate of

approximately 40%, with limited selection on observables

(13)

Example: Provider Networks

Handel & Kolstad (2015)

•

Both plans offered differ on financial characteristics but provide

access to exact same provider network / covered services

•

Hypothesis

: Some consumers will associate better financial

protection with broader provider access

• Evidence

:

(i)

< 50% know networks same

(ii) many ‘not sure’

(iii) 20% in line with hypothesis

(iv) Correlation w/ choices

(14)

Information Types

Handel & Kolstad (2015)

•

Can classify consumers

with information on one

dimension, or aggregrate

information across

dimensions into index

•

Index includes:

• Deductible

• Coinsurance

• OOP Max

• Provider Network

• Own Recent Expenses

• HSA Benefits

(15)

Quantifying Impact of Limited Information

Handel & Kolstad (2015)

• Mathematical decision model quantifying:

•

Ex ante health risk

•

Risk aversion

•

Willingness-to-pay for PPO under limited information

•

Perceived HDHP time and hassle costs

•

Method to identify limited information impact on

willingness-to-pay for more insurance:

•

Identify ex ante health risk with statistical model using claims data

•

Identify risk aversion using informed consumers

•

Construct relative value for PPO and HDHP choices

•

Compare choices of informed consumers and consumers with

limited information, using survey responses

(16)

Impact of Limited Information

Handel & Kolstad (2015)

•

Provider Networks:

• $2,200 against HDHP for those who think it has fewer providers •

Hassle Costs:

• Those who strongly dislike give up $220 per

hour of perceived

additional HDHP time and hassle costs

•

Types:

• Direct ordering of information types and WTP for PPO

(17)

Policy: Risk Protection and Forced HDHP Switch

Handel & Kolstad (2015)

•

Firm switched all

employees to HDHP in

years following

empirical study

•

What is true value of

risk protection

consumers lost?

•

Value of risk protection

lost much smaller

when you take limited

information into

account (context

specific)

(18)

Roadmap

• Active Decision Errors: Limited Information

• Passive Decision Errors: Inertia

(19)

Inertia in Insurance Choice

• Strong evidence that consumer inertia in insurance plan

choice leads to “money left on the table” that is quite a bit

larger than when consumers make active choices

•

Default option is typically previously chosen plan

•

Markets / prices / benefits change over time

•

Factors contributing to inertia:

•

Provider lock-in / provider relationships

•

Inattention combined with search costs

•

Status-quo bias

• Examples:

•

Handel (2013)

(20)

Handel (2013): “When Nudging Hurts”

• Large self-insured employer changed entire menu of

insurance options, forced all 10,000 employees to make

active insurance choices from new set of 5 options

• Observe employees making active decisions in that year,

then observe subsequent choices

in presence of default

options

with

substantial pricing / benefit changes

• Plans in new menu are

financially differentiated

but the

same in terms of provider networks. We can thus identify:

•

Actual preferences for insurance in active choice year

•

Impact of inertia on consumer choice value, netting out persistent

preferences for providers / plans

(21)

Handel (2013): Descriptive Inertia Evidence

(22)

Handel (2013): Descriptive Inertia Evidence

• Large change in premiums for next year, when people default into previously chosen option

• 30% of families had plan become completely dominated • 89% of those families continue to choose that plan!

(23)

(24)

(25)

(26)

Econometric Model to Quantify Inertia Impact

• Develop econometric model of consumer decision-making

under uncertainty to quantify impact of:

•

Risk aversion

•

Ex ante health risk (individual and families)

•

Inertia

• Key idea: use preferences expressed in ‘active’ choices

to model what choices should be in passive choice years

• Inertia matters a lot:

•

Consumers leave an average of $2,000 on table due to inertia

•

Consumer who make other active choices (e.g. FSA) leave less

(27)

Policies to Reduce Inertia

• Natural next step is to reduce inertia with available policies

• Possible policies:

• Reminders

• Information Provision

• Targeted Information Provision

• Targeted Recommendations

• Active choice only

• Targeted Defaults / Smart Defaults

• Product Standardization

(28)

Policies to Reduce Inertia

• As policies to reduce inertia become more effective, consumers make better choices given market structure

(29)

Roadmap

• Active Decision Errors: Limited Information

• Passive Decision Errors: Inertia

(30)

Policies to Reduce Inertia

Endogenous Insurance Offerings

• Adverse Selection

: Insurer costs linked to who buys insurance.

• Sicker consumers = higher costs = higher premiums

• Higher premiums = less comprehensive insurance

• Widespread consumer choice issue has direct / strong impact

on the extent of adverse selection in insurance markets

• Consequently, policies to aid consumer decision-making have

(31)

Policies to Reduce Inertia: Handel (2013)

Endogenous Insurance Offerings

• Insurer pricing model: prices this year equal last year’s average costs • Reducing impact of inertia by 75% substantially increases adverse

selection, and reduces consumer welfare by 6% 6% of cons. Premiums • Context-specific, key implication is that choice policy matters for

(32)

Other Policies: Handel et al. (2015)

• Paper uses estimates of active choice mistakes from Handel and Kolstad (2015) to show impact of choice frictions for:

-- Price subsidies -- Choice policies -- Risk adjustment

-- Actuarial value regulation

• Assumed competitive insurance market, similar to ACA exchange

• Key insight: Effective risk-adjustment transfers

complementary to choice

policies. Matters more (potentially a lot more) when consumers

(33)

ACA and Decision

Interventions

• Key element of ACA exchanges (and many

insurance markets) is insurers competing to attract

consumers, who in turn “discipline” the market • Empirical evidence on

behavioral decision-making (as in “Choose to Lose?”) suggests inherent tension

between managed competition paradigm and interventions to fix consumer mistakes [see e.g. Einav and Levin (2015)]

• As more aggressive

paternalism is implemented, consumers aren’t the ones disciplining the market

Consumer Agency Choice Effectiveness in Market Healthcare.gov Decision Tools General Recommendations Targeted Recommendations Defaults – Non-Inertial Threshold-Based Targeted Defaults Aggressive “Smart” Defaults

(34)

(35)

Lessons

• Consumers make mistakes when it’s possible to do so in

insurance plan choice, and can lose a lot of money

• Difficulties with active decision-making

• Difficulties with defaults and inertia

• Consumer decision-making has important implications for the

extent of risk-based selection in the market

• Interventions to improve choices could make all consumers worse off

• Effective risk-adjustment transfers complement decision-making policies

• There is still a lot to learn about:

• Specific mechanisms underlying consumer choice of complex products

• Firm responses to consumer choice inadequacy

(36)

Consumer Decision Making on

Health Insurance Exchanges

Keith Marzilli Ericson

Society of Actuaries, June 15 2015

Boston University and NBER

(37)



ACA relies on health insurance exchanges (HIX)



≈ 20 million enrollees, $100 billion/year in subsidies



HIX differ from existing markets new questions



_{Insurers sell directly to consumers, public prices}



_{Pre ACA: individual market small (5%), serving as}

gap coverage with short duration, limited

generosity



Examine interaction of consumer choice, insurer

(38)

Mandates: Entering the Market

Price Sensitivity and Markups on Massachusetts HIX Standardization and Choice Architecture on HIX

(39)

Renewal Decisions and Defaults

(40)



_{Adverse selection into insurance was major concern if}

only sick enter, “death spirals”



ACA contains “individual mandate” to buy insurance



Raises price of being uninsured: pay max of 2.5% of

income or $695



Major political, supreme court fight over constitutionality



_{Other changes in relative prices less controversial: $695}

subsidy to buy insurance has same net effect



_{Mortgage subsidy not described as “mandate to buy a home”}



_{Moral suasion of the mandate? Could the same effect have}

(41)



1670 subjects recruited from online labor market



Experiment took place on four different dates



Policy is in flux, suppose gov’t decided:



Mandate: “to mandate everyone purchase

insurance, or else pay a fine of $700 each year”



Tax: “to recommend that everyone purchase health

insurance, and charge people without insurance an

uninsurance tax of $700 each year”



How likely are you to buy insurance @ $3000/year,

(42)

0 2 00 4 00 6 00 T ot al # o f A rt ic le s 2010 J anuar y 2010 M arch 2010 M ay 2010 J uly 2010 S epte mber 2010 No vem ber 2011 J anuar y 2011 M arch 2011 M ay 2011 J uly 2011 S epte mber 2011 No vem ber 2012 J anuar y 2012 M arch 2012 M ay 2012 J uly 2012 S epte mber 2012 No vem ber 2013 J anuar y Month

News Articles about the Mandate



Developed after

waves 1 & 2



End of March

2012, Supreme

Court oral

arguments



_{How did}

controversy

affect mandate?

(43)

-10 -5 0 5 10 15 20 Wave 1

(Dec 2011) (Mar 2012)Wave 2 (June 2012)Wave 3 (Nov 2012)Wave 4

Di ffe re nce in P er ce nt ag e Li ke lih ood of P ur ch as e at $3000/ ye ar

(44)

-10 -5 0 5 10 15 20 Wave 1

(Dec 2011) (Mar 2012)Wave 2 (June 2012)Wave 3 (Nov 2012)Wave 4

Di ffe re nce in P er ce nt ag e Li ke lih ood of P ur ch as e at $3000/ ye ar

(45)



Large effect of articulating the policy as a mandate

rather than an tax in the pre-controversy period



After the controversy surrounding the mandate

this effect is gone.



Disappears in all demographic groups



Not due to changes in information, etc



Moral suasion of mandate lost due to controversy



May have undermined enrollment in HIX

(46)

Price Sensitivity and Markups on Massachusetts HIX

Standardization and Choice Architecture on HIX Renewal Decisions and Defaults

Ericson and Starc. Forthcoming. Pricing Regulation and Imperfect Competition. Review of Economics and Statistics Ericson and Starc. 2012. Heuristics and Heterogeneity in Health Insurance Exchanges. American Economic Review

(47)



Established as a result

of state reform



Guaranteed issue



_{Modified community}

rating: price by age



2007-2013: no risk

adjustment, no

subsidies here

(48)



How do consumers substitute among plans when

insurers raise premiums?



Prices = cost + markup



_{Determines how insurers set price markups}



Model: Discrete choice



_{Plan j has utility: 𝑢𝑢}

_{𝑖𝑖𝑖𝑖}

_{= 𝛽𝛽𝑋𝑋}

_𝑖𝑖

_{− 𝛼𝛼}

_𝑖𝑖

_𝑝𝑝

_𝑖𝑖

_{+ 𝜀𝜀}

_{𝑖𝑖𝑖𝑖}



Logit market share: 𝑠𝑠

_𝑖𝑖

₌

exp(𝛽𝛽𝑋𝑋𝑗𝑗−𝛼𝛼𝑝𝑝𝑗𝑗)

∑_𝑗𝑗 exp(𝛽𝛽𝑋𝑋_𝑗𝑗−𝛼𝛼𝑝𝑝_𝑗𝑗)



_{Firm’s first-order condition for price setting}

𝑝𝑝

𝑖𝑖

= 𝑐𝑐 −

_𝑠𝑠

𝑠𝑠

𝑖𝑖

𝑖𝑖′

= c +

1

(49)



First result: big gain to being the cheapest plan



Equivalent to $300-$550/year premium decrease



_{Consistent with heuristic “choose the cheapest”;}

cheapest plan is listed first



Competition at bottom of the market may be quite

different from top of the market



Consistent with heuristics or heterogeneity



_{Heuristic: “choose the cheapest”, is 1}

st

_{plan listed}



_{Heterogeneity: some people just want the cheapest}

(50)

20 0 25 0 30 0 35 0 40 0 Ave ra g e Pre mi u m D if fe re n ce , G o ld - Bro n ze 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 F ra ct io n C h o o si n g Bro n ze <30 30-34 35-39 40-44 45-49 50-54 >55 Age Category

Chose Bronze Premium Difference

Reduced form fact



_{Marginal cost of more}

generous plan rises

with age



_{Older individuals don’t}

substitute toward less

generous plans



_{Or less generous}

(51)



Prices jump at round #s



Age 49 v 50: Similar people,

different prices



45+ are half as price

sensitive as under 45

implies firms would charge

twice the markup over cost

200

300 400 500 600 700 M ont hl y P rem ium i n $ 25 30 35 40 45 50 55 60 65 Age

(52)

Compared to no price

regulation, age-bands:



raise consumer surplus by

about $600/enrollee-year



lower firms’ profits by

$300/enrollee-year

Implications for risk

adjustment as well

Age-bands in perfectly competitive market

• move price toward average cost

• transfers $ from the young to the old

Age-band regulation: Prices for oldest

can only be twice price for youngest

Imperfectly competitive market

• still transfers $ from the young to the old • firms price to the marginal consumer

• here: young, cheap, and price sensitive • lowers markups, increases efficiency • lowers firms’ profits

(53)

Price Sensitivity and Markups on Massachusetts HIX

Standardization and Choice Architecture on HIX

Renewal Decisions and Defaults

(54)



Competition relies on informed consumer choice



_{Confusing choices in many markets}



Health insurance: many complex attributes



Coinsurance, deductible, maximum OOP, doctor

visit copay, Rx drug tiers 1, 2, and 3, ER copay,



Abaluck and Gruber 2012, Kling et al 2012, Kolstad

and Handel 2013



_{Choice architecture: tiers facilitate comparison}

(55)

(56)

(57)

(58)



"Consumers don't have to worry that there's some

sort of 'gotcha' ... They are comparing equivalent

products and so make better informed decisions

based on premium and provider differences".



_{Nancy Turnbull (HIX Board)}



_{Intuition #1: Standardization makes consumers}

more price sensitive, increases price competition ∙



_{Intuition #2: Increase weight on dimensions now}

(59)



_{More generous plans}

chosen



_{Major shift in brand}

choices



Little change in price

(60)



Sources of the effect:



Availability: What products were available changed



Valuation: weights consumers place on attributes

changed (e.g. context-dependent preferences)



We decompose via a discrete choice model

σ brand 0.968 (0.155)

σ cost-sharing 4.43

(0.382)

σ premium 1.11 (0.097)

(61)

Renewal Decisions and Defaults

Ericson. Consumer Inertia and Firm Pricing in the Medicare Part D Prescription Drug Insurance Exchange. American Economic Journal: Economic Policy

Ericson. When Consumers Don’t Make an Active Decision: Equilibrium Effects of Dynamic Defaults.

(62)

Medicare Part D



_{Initial assignment}

default:



_{Low Income Subsidy (LIS)}

recipients assigned to

plan below threshold



_{Matters in year 1 and}

beyond



Automatic switching

default



If firm raises price in year

2, LIS switched to cheaper

plan



High income enrollees

(63)

0 .02 .04 .06 .08 De ns it y 0 20 40 60 80 Monthly Premium 2008+ Cohorts

Medicare Part D Premiums in 2010



_{Insurers respond by using}

“invest-then-harvest”

pricing



_{Offer low prices in early}

years, capture enrollees



_{Costly to switch}



_{Raise prices in later years}



_{Unnecessary churn}

between plans, lower

investment in enrollee

health



_{Early years: good deal}



_{Reenrollment default v.}

active decision

0 .02 .04 .06 .08 De ns it y 0 20 40 60 80 Monthly Premium 2006-7 Cohorts 2008+ Cohorts