eco question

Health Economics ECON 5860 PROF. KURT LAVETTI

PMC

Medical Care

DMC_No Ins.

SupplyMC

Health Insurance and Social Welfare

50%xP*

Market Price (P*)

• Suppose our goal is to reduce DWL from moral hazard • What do we need to know empirically about healthcare

markets in order to measure the effects of policies or insurance generosity on social welfare?

Challenge to Estimating Demand: Price vs Quality Differences • Quality may also vary across observations (different demand for

different observations) • Ideally, η is measured along a demand curve, not between

demand curves. Result: Estimate of η is biased upward (even positive)

Demand (High Quality)

Demand (Low Quality)

Price

Quantity

Perceived “Demand” curve (incorrect)

Challenge to Estimating Demand: Price Changes over Time

• Even within an observation, many other factors may also vary across time (different demand for different time periods)

• Result: Estimate of η can be biased in either direction

Demand (Time X)

Demand (Time Y)

Price

Quantity

Estimating Demand Curves from Insurance Variation

 Terminology: A “Coinsurance” rate is the % of medical spending paid for by the consumer

 I.e. if the insurer pays 80% of costs, then the coinsurance rate is the remaining 20%

 People choose insurance policies with different coinsurance rates

 Suppose one insurance plan has a 50% coinsurance rate and another has a 0% rate. We could test whether individuals in the 50% plan purchase less healthcare than those in the 0% plan

 But what if sicker people are more likely to buy the plan with 0% coinsurance?

 Endogeneity of plan choice is an important source of bias in estimating the elasticity of demand

 This was a common problem with many early studies

Negative Bias and Insurance Selection

Demand (Sick) η = -0.2 (B)

Demand (Healthy) η = -1.2 (A)

“Perceived” Demand (incorrect) η = -3.8 (A)

Price

Quantity

• Suppose sicker patients with greater (and more inelastic) demand for medical care choose more generous insurance plans

• Because of this selection into insurance plans based on unobserved health status, the estimated elasticity will be too negative

• Will cause us to overestimate the degree to which healthy consumers respond to prices.

A B

Estimating the Elasticity of Demand  Randomized experiments:

Definition: a study that assigns treatments randomly to different groups of study participants

Includes: A control group (no treatment) Placebo group

Helps generate experimental groups that are statistically similar to each other

Two Randomized Experiments

 RAND Health Insurance Experiment (HIE), 1974

 Oregon Medicaid Experiment, 2008

RAND HIE

 Randomly assigned 2,000 families from six US cities to different insurance coverage plans  Coinsurance groups:

Free, 25%, 50%, and 95%  Tracked utilization of health care (Q) in each

copayment plan (P)  Coinsurance acts as the marginal cost that

each family faces when buying care

Oregon Medicaid Experiment

 Compared two groups of low-income adults  Medicaid lottery winners vs. lottery losers

 Lottery winners got to apply for public health insurance through Medicaid  So they faced lower out-of-pocket prices

for care  Lottery losers could not get Medicaid (but

might have purchased outside insurance)

Results?

 Health care demand curves are downward sloping Price changes affect demand for

health care The law of demand works!

Different measures of care

 Outpatient Care  Definition: any medical care that does not involve an

overnight hospital stay  E.g. runny noses, twisted ankles, minor broken bones

 Inpatient Care  Def: medical care requiring overnight stays

 E.g. More serious surgeries or conditions that require overnight recovery or monitoring

 ER Care  Def: care involving the emergency room

 E.g. heart attacks, strokes

Outpatient care

 RAND HIE  As patient cost-sharing (P) increases, number of episodes (Q) of

outpatient care decreases

 Holds for both acute and chronic conditions

Source: Keeler et al. (1988)

Outpatient care

 Oregon Medicaid Study  Lottery winners have more outpatient visits than lottery losers

Both the RAND HIE and the Oregon Medicaid Study find downward- sloping demand for outpatient care.

Inpatient care

 RAND HIE  Oregon Medicaid Study

No significant difference in usage rates between lottery winners and lottery losers

Demand is still downward-sloping but less elastic than demand for outpatient care

(Source: Keeler, 1988)

Emergency Room Care  RAND HIE

 Oregon Medicaid Study

 Gaining Medicaid insurance increased probability of going to ER by 20% (7 pp increase relative to mean of 35%)

 (Note: textbook discussion is outdated)

Even for emergency room care – likely the most urgent kind – those on the highest copayment plan in the RAND HIE were less likely to buy care!

(Source: Newhouse, 1993)

Pediatric care

 Pediatric care  Def: care for infants or children usually paid for by a parent or

guardian

 Data from RAND HIE:

Mental Health & Dental Care RAND HIE Results

Prescription drugs

 Data from RAND HIE

Quasi-Random Non-Experimental Evidence  U.S. Medicare

 Citizens are eligible for health insurance through Medicare when they turn 65 but not before

 If demand for health care is downward-sloping, we expect a jump in health care usage at age 65

 This is known as a discontinuity study  There is a discontinuity in health insurance at age 65

Card et al. (2009)

 Increase in Medicare insurance coverage on exactly the 65th birthday

Card et al. (2009)

 Increase in coverage leads people to use the emergency room more right after their 65th birthday

Card et al. (2009)

 Card et al. have two main findings:  Unplanned emergency department admissions follow a linear

trend around the age of 65  Other hospital admissions jump up at the age of 65

 There is a discontinuity in medical usage at the same point of discontinuity in Medicare coverage!

 This is further evidence that demand for health care is sensitive to price

 Statistics Lesson: the assumption behind this analysis is that the error term in the regression is on average the same when someone is 64.999 years as it is when they are 65.001 years

Comparing demand curves

 How can we use data on coinsurance rates and quantities purchased to calculate the elasticity of demand?

Source: Keeler et al. (1988)

Arc Elasticity  One issue is that some plans have coinsurance rate of 0%.

We can’t calculate the percent difference in price when comparing a 0% plan to a 25% plan

 Instead of the typical elasticity measure, it’s common in healthcare to use an Arc Elasticity:

Example:

 Let’s go back to the RAND results on outpatient care

 Using only the two data points in the red boxes, what is the arc elasticity of demand for outpatient care?

Health care has inelastic demand

Does price for care affect health?  Mortality rates

 RAND HIE: no difference between treatment groups

 Looking specifically at high-risk study participants (with chronic conditions), those in the free care plan were 10% less likely to die than those on cost-sharing plans!

 Oregon Medicaid: no significant difference between lottery winners and losers

 Caveats: Both studies only examine short-run effects (1-3 years after the experiment)

DEMAND FOR HEALTH: THE GROSSMAN MODEL

The 3 Roles of Health (H)

Health plays three roles in the Grossman model:

1. A consumption good 2. An input into production 3. A form of stock/capital (an

investment)

Health as a consumption good

Health as a direct input into utility

 Health as a consumption good enters directly into utility

 Single-period Utility at time t Ut= U(Ht, Zt)

 Ht = level of health

 Zt= “home good”

 Everything non-health that contributes to utility

 E.g. video games, time with friends, movie tickets  Note: health is not the same as health care

 Health care is not in the utility function

 i.e. Getting vaccines does not provide utility, but staying healthy does

Time constraints in the Grossman model

 In a single period, there are only 24 hours in a day to contribute to your utility:

Θ = 24 = TW + TZ + TH + TS

 Divide total time Θ between:  Working TW

 Leisure Time TZ

 Investing in health TH

 Lost due to sickness TS

Health as an input into production

Producing H and Z

Both Health and Home good Z must be produced with time and market inputs

Ht = H (Ht-1, TtH, Mt)

Zt = Z (TtZ, Jt)

 Mt= market inputs for health H  Ex: doctor visit, gym membership

 Jt= market inputs for home goods Z  Ex: video games, opera tickets

 Today’s health Ht also depends on yesterday’s health Ht-1  This is health’s third role as a stock which we discuss later

Production function of “Healthy Days”

TP (“healthy

days”)

H (Health Stock)

365

• The ultimate health outcome in the Grossman model is “healthy days”. It is reasonable that health stock contributes to healthy days in a decreasing fashion

• As with most production functions, the marginal contribution of inputs declines with each additional unit of input

TS

TW + TZ + TH

Production Possibilities Frontier

 Point A Hmin: no productive time for work, play, or improvement of health

 Point B  “free-lunch zone”  Small improvements in

health yield large increases in productive time; can increase Z without giving up H

A CORRECT PPF

Choosing optimal H* and Z*

 Someone who values both H and Z chooses a point between C and E in order to maximize their utility

 Chooses point F  U2 is unattainable given

PPF constraints  At U0, an individual can

attain more utility  At F: U1 and PPF are

tangent  H* and Z* are optimal

levels of health and home goods

Health as an investment

The three roles of health (H)

Health plays three roles in the Grossman

Model: 1. A consumption good 2. An input into production 3. A form of stock/capital (an investment)

Lifetime of utility

 On any day, an individual considers not only today’s utility U(H0,Z0) but all future utility as well!

 Health is a stock; some of it carries over each new period  Home good Z is a flow (it lasts for only 1 period)

 δ = individual’s discount rate  A person values utility now more than in the future

 Ω = individual’s lifespan (total number of periods)

Health depreciates over time

Some of yesterday’s health lasts to today but not all of it

Ht = H ( (1- γ)Ht-1, TtH, Mt )  γ = rate of depreciation  Recall:

 Ht = health at time period t

 Ht-1 = health from previous period

 TtH = time spent on health in period t

 Mt = market inputs for health (like checkups and prescription pills)

MEC curve and investments in health

 Marginal Efficiency of Capital (MEC) curve:

 indicates how efficient each unit of health capital is in increasing lifetime utility

 When level of H is low, small investments have high returns to productive time

The MEC Curve is the Marginal Benefit of Health Capital

H HA HB

• The marginal benefit of additional health capital is high when health capital is low (HA).

• Marginal benefit is low when health capital is high (HB). • Graph 2 traces out the shape of the marginal benefit curve, which

Grossman calls the “Marginal Efficiency of Capital (MEC)”

Marginal Benefit

HA HB

TP

TW + TZ + TH

365 TS

Marginal Cost of Health Capital • Marginal cost has two components:

• Opportunity cost (r)—rather than investing in health, you could save the money instead, and earn a rate of interest

• The foregone interest on savings is an opportunity cost that equals the market interest rate

• Depreciation (γ)—health capital always will depreciate in each period

• If you invest in health capital today you will lose some of that investment by tomorrow due to depreciation

• This is a cost of investing in health capital • Rate of depreciation could depend on things like age

• Both γ and r are exogenous assumptions in the Grossman model (they are not affected by the individual’s decisions).

• If γ and r are constant percentages, what will the marginal cost function look like?

Marginal Cost of Health Capital

Marginal Cost

HA HB

γ +r Marginal Cost

If γ and r are constant the marginal cost of health capital does not depend on H!

Optimal Health Capital

Combining these graphs, optimal health capital (H*) occurs where marginal benefit (MEC) = marginal cost.

Rate of Return

HA HB

δ+r

H*

MEC

Marginal Cost

MB>MC

MB=MC

MB<MC

H

Predictions of the Grossman model

The Grossman model helps explain why we observe:

1. Better health among the educated 2. Declining health among the aging

Health and education

 Suppose well-educated individuals are more efficient producers of health What hypothesis will this

model generate about the relationship between health status (H) and education?

MEC and efficiency of health investment

 Better educated are more efficient at each level of health investment

 MECCollege > MECHS  H*College is higher than H*HS

 MECC = college graduate  MECH = high school

dropout

Predictions of the Grossman model

The Grossman model helps explain why we observe:

1. Better health among the educated 2. Declining health among the aging

Depreciation of health  Recall:

Ht = H ( (1- γ)Ht-1, TtH, Mt )  Depreciation γ is not

constant  γ increases with age  As γ increases, costs

(r + γ) increase and it takes more resources to maintain same level of health

As a result of increasing depreciation γ over time, optimal health H* also declines over time!

Optimal death in the Grossman model

 Because of rising depreciation, there are better investments in the market than the individual’s health

 H* eventually reaches Hmin

 Why would anyone choose Hmin?  How is Hmin utility-

maximizing?

Question:

 Does this mean the level of health spending should decrease as people get older?

Age and the derived demand for medical care (or health inputs)

H

Health Inputs (Medical Care)

HYoung HOld

• Not necessarily: suppose the efficiency of investment also falls as people age

• Then the elderly may have a lower stock of H, but a higher demand for medical care • Model can be

easily adapted to fit reality

Health PFOld

Health PFYoung

Conclusion

 Is health something that happens to us or is chosen?  Grossman model says it is chosen

 In fact, we even choose when we die  While that may seem far-fetched, Grossman model

a useful tool for understanding the roles and tradeoffs of health