Randomized Trial- RAND Health Insurance Experiment (HIE)

•summary from Angrist and Pischke (Chapter 1)

How Elastic is the Demand for Medical Care?

moral hazard- individuals use more medical care once they are insured because they know that their
insurer will pay part or all of their bill

What is the price elasticity of demand for medical care? (Recall that the elasticity of demand is expressed
as a positive number. It is understood that there is a negative relationship between price and quantity
demanded.)

Consider this model: 𝑄𝑢𝑎𝑛𝑡𝑖𝑡𝑦 ℎ𝑒𝑎𝑙𝑡ℎ 𝑐𝑎𝑟𝑒𝑖 = 𝛽0 + 𝛽1 × 𝐶𝑜𝑖𝑛𝑠𝑢𝑟𝑎𝑛𝑐𝑒 𝑟𝑎𝑡𝑒𝑖 + 𝑢𝑖

Quantity health care is the amount of health care the individual consumed over the course of the year.
Coinsurance rate is the coinsurance rate of the individual’s health insurance plan. A researcher who
̂1 < 0. Because of omitted variable bias, is it likely the case
collects data and runs a regression finds that 𝛽
̂ ̂
that 𝛽1 > β1 or 𝛽1 < β1?

Initial studies compared individuals in different types of health plans and found that individuals in plans
with higher coinsurance used less care. These studies, however, were likely biased. Individuals who
spend a lot on health care are likely to buy plans with low coinsurance, so the correlation between low
coinsurance and high use of medical care is not a causal effect of coinsurance on utilization. Instead, the
correlation reflects that high-utilization individuals have chosen low-coinsurance plans.

Does Health Insurance Improve Health?

Consider this model: 𝐻𝑒𝑎𝑙𝑡ℎ 𝑠𝑡𝑎𝑡𝑢𝑠𝑖 = 𝛽0 + 𝛽1 × 𝐼𝑛𝑠𝑢𝑟𝑒𝑑𝑖 + 𝑢𝑖

Health status is a measure of health status where a higher number means better health. Insured is a
dummy that is equal to 1 if the individual is insured and equal to 0 otherwise. A researcher who collects
̂1 > 0. Because of omitted variable bias, is it likely the case that 𝛽
data and runs a regression finds that 𝛽 ̂1
̂1 < β1?
> β1 or 𝛽
 2
Randomized Trials

Comparing Group Averages

Y = health and Yi is the health of individual i

person i has two potential outcomes, only one of which is observed

Y0i = health of person i if they take the road without health insurance
Y1i = health of person i if they take the road with health insurance

Y1i – Y0i = causal effect of insurance on health for individual i

Avgn[Y1i – Y0i] = average causal effect of insurance on health for group of n people

The problem is that only one of the two hypothetical scenarios is observed for each individual.

Let’s start by comparing the average health of groups of insured and uninsured people.

1 if 𝑖 is insured
𝐷𝑖 = { }
0 if 𝑖 is not insured

Difference in group means = Avgn[Yi | Di = 1] – Avgn[Yi | Di = 0]

Difference in group means = Avgn[Y1i | Di = 1] – Avgn[Y0i | Di = 0]

The average Yi for the insured is necessarily an average of outcome Y1i, but contains no information about
Y0i. Likewise, the average Yi among the uninsured is an average of outcome Y0i, but this average is devoid
of information about the corresponding Y1i. We’re after Avgn[Y1i – Y0i], an average causal effect involving
everyone’s Y1i and everyone’s Y0i, but we see average Y1i only for the insured and average Y0i only for the
uninsured.

Let’s imagine that health insurance makes everyone healthier by a constant amount, κ (“kappa”). The
constant-effects assumption allows us to write: Y1i = Y0i + κ. Note that κ is both the individual and
average causal effect of insurance on health.

Difference in group means = Avgn[Y1i | Di = 1] – Avgn[Y0i | Di = 0]

Difference in group means = {κ + Avgn[Y0i | Di = 1]} – Avgn[Y0i | Di = 0]

Difference in group means = κ + {Avgn[Y0i | Di = 1] – Avgn[Y0i | Di = 0]}

Difference in group means = Average causal effect + Selection bias

Selection bias is defined as the difference in average Y0i between the groups being compared. We expect
Avgn[Y0i | Di = 1] > Avgn[Y0i | Di = 0]. Table 1.1 documents important noninsurance differences between
the insured and the uninsured. The insured are healthier for all sorts of reasons, including, perhaps, the
causal effects of insurance. But the insured are also healthier because they are more educated, richer, and
so on.

The subtlety in Table 1.1 arises because when observed differences proliferate, so should our suspicions
about unobserved differences. The fact that people with and without health insurance differ in many
visible ways suggests that even were we to hold observed characteristics fixed, the uninsured would likely
differ from the insured in ways we don’t see. In other words, even in a sample consisting of insured and
uninsured people with the same education, income, and employment status, the insured might have higher
values of Y0i.
 3
Table 1.1- Health and demographic characteristics of insured and uninsured couples in the NHIS

Husbands
Some HI (1) No HI (2) Difference (3)
A. Health
Health index 4.01 3.70 0.31***
[0.93] [1.01] (0.03)
B. Characteristics
Nonwhite 0.16 0.17 -0.01
(0.01)
Age 43.98 41.26 2.71***
(0.29)
Education 14.31 11.56 2.74***
(0.10)
Family size 3.50 3.98 -0.47***
(0.05)
Employed 0.92 0.85 0.07***
(0.01)
Family income 106,467 45,656 60,810***
(1,355)
Sample size 8,114 1,281
***statistically significant at 1%

Conducting a randomized trial

To study the effects of health insurance in a randomized trial, we’d start with a sample of people who are
currently uninsured. We’d then provide health insurance to a randomly chosen subset of this sample, and
let the rest go to the emergency department if the need arises. Later, the health of the insured and
uninsured groups can be compared. Random assignment makes this comparison ceteris paribus: groups
insured and uninsured by random assignment differ only in their insurance status and any consequences
that follow from it.

Conditional expectation

The conditional expectation of a variable Yi, given a dummy variable, Di = 1, is written E[Yi | Di = 1].
This is the average of Yi in the population that has Di equal to 1. Likewise, the conditional expectation of
a variable, Yi, given Di = 0, written, E[Yi | Di = 0], is the average of Yi in the population that has Di equal
to 0. If Yi and Di are variables generated by a random process, such as throwing a die under different
circumstances, E[Yi | Di = d] is the average of infinitely many repetitions of this process while holding the
circumstances indicated by Di fixed at d. If Yi and Di come from a sample survey, E[Yi | Di = d] is the
average computed when everyone in the population who has Di = d is sampled.

Random assignment eliminates selection bias

Because randomly assigned treatment and control groups come from the same underlying population,
they are the same in every way, including their expected Y0i. In other words, the conditional expectations,
E[Y0i | Di = 1] and E[Y0i | Di = 0], are the same. Hence, the difference in expectations by treatment status
captures the causal effect of treatment:

E[Yi | Di = 1] – E[Yi | Di = 0] = E[Y1i | Di = 1] – E[Y0i | Di = 0]

= E[Y0i + κ | Di = 1] – E[Y0i | Di = 0]
= κ + E[Y0i | Di = 1] – E[Y0i | Di = 0]
= κ
 4
RAND Health Insurance Experiment (HIE)

The RAND Health Insurance Experiment (HIE), which ran from 1974 to 1982, was one of the most
influential social experiments in research history. The HIE enrolled 3,958 people aged 14 to 61 from six
areas of the country. The HIE sample excluded Medicare participants and most Medicaid and military
health insurance subscribers. HIE participants were randomly assigned to one of 14 insurance plans.
Participants did not have to pay insurance premiums, but the plans had a variety of provisions related to
cost sharing, leading to large differences in the amount of insurance they offered.

The most generous HIE plan offered comprehensive care for free. At the other end of the insurance
spectrum, three “catastrophic coverage” plans required families to pay 95% of their health-care costs,
though these costs were capped as a proportion of income (or capped at $1,000 per family, if that was
lower)… A second insurance scheme (the “individual deductible” plan) also required families to pay 95%
of outpatient charges, but only up to $150 per person or $450 per family. A group of nine other plans had
a variety of coinsurance provisions, requiring participants to cover anywhere from 25% to 50% of
charges, but always capped at a proportion of income or $1,000, whichever was lower. Participating
families enrolled in the experimental plans for 3 or 5 years and agreed to give up any earlier insurance
coverage in return for a fixed monthly payment unrelated to their use of medical care.

(“The RAND HIE involved making a $1,000 payment to all participants to ensure that no one was made
worse off. This payment did not cause any bias, however, because it was made equally to treatments and
controls. So any income effect from the money was the same across both groups, and they could still be
compared to assess the impact of coinsurance variation.”- Gruber)

Most analyses of the HIE data start by grouping subjects who were assigned to similar HIE plans together
(because some of the treatment groups are small). A natural grouping scheme combines plans by the
amount of cost sharing they require. In the three catastrophic coverage plans, subscribers shoulder almost
all of their medical expenses up to a fairly high cap. The individual deductible plan provided more
coverage but only by reducing the cap on total expenses that plan participants were required to shoulder.
The nine coinsurance plans provided more substantial coverage by splitting subscribers’ health-care costs
with the insurer, starting with the first dollar of costs incurred. Finally, the free plan constituted a radical
intervention that might be expected to generate the largest increase in health-care usage and, perhaps,
health.

This categorization leads us to four groups of plans: catastrophic, deductible, coinsurance, and free
(characterized by increasing levels of coverage).

●Notice that the RAND HIE did not study the effects of having insurance versus not having insurance
(everyone in the study was insured). It studied the effects of insurance generosity, conditional on being
insured, on outcomes of interest.

•if n large→ t ~ N(0,1)

Confidence Level (C) 90% 95% 99%

α 0.10 0.05 0.01
z* 1.645 1.960 2.576
 5
Table 1.4- Health expenditure and health outcomes in the RAND HIE

Means Differences between plan groups

Catastrophic plan (1) Free – catastrophic (4)
A. Health-care use
Face-to-face visits 2.78 1.66
[5.50] (0.25)
Outpatient expenses 248 169
[488] (20)
Hospital admissions 0.099 0.029
[0.379] (0.010)
Inpatient expenses 388 116
[2,308] (60)
Total expenses 636 285
[2,535] (72)
B. Health outcomes
General health index 68.5 -0.78
[15.9] (0.87)
Cholesterol (mg/dl) 203 -1.83
[42] (2.39)
Systolic blood pressure (mm Hg) 122 -0.52
[19] (0.93)
Mental health index 75.5 0.43
[14.8] (0.83)
Number enrolled 759 1,295

Let’s indicate in the table which results are statistically significant.

What are the important takeaways from Table 1.4?

 6
Gruber’s summary of the RAND HIE results (Chapter 15)

●The findings of the HIE were striking. First, medical care is price sensitive. The implied elasticity
across the entire study was 0.2, meaning that each 10% rise in the price of medical care to individuals led
them to use 2% less care.

●Second, those who used more health care due to the lower price did not, on average, see a significant
improvement in their health. This finding does not imply, however, that insurance isn’t valuable at all
because everyone in this experiment was insured; after a family’s health spending reached $1,000, it had
full insurance. The RAND results imply that after individuals are insured for large expenditures, varying
the coinsurance for small expenditures does not seem to affect their health on average.

●Third, for those who are chronically ill and don’t have sufficient income to easily cover copayments,
there was some deterioration in health. In particular, low-income individuals who were hypertensive (had
high blood pressure) saw dangerous increases in their blood pressure arising from lack of care. More
recent studies have confirmed this finding, concluding that those who have treatable buy chronic illnesses
may be made worse off by copayments. Indeed, some studies find that for the chronically ill, raising
copayments actually raises total medical costs because the reduced use of prescription drugs and
physician visits results in more expensive hospitalizations due to health deterioration.

Optimal Health Insurance

Summary view from Angrist and Pischke

“These HIE findings convinced many economists that generous health insurance can have unintended and
undesirable consequences, increasing health-care usage and costs, without generating a dividend in the
form of better health.”

Summary view from Gruber

●“Optimal health insurance policy is one in which individuals bear a large share of medical costs within
some affordable range and are only fully insured when costs become unaffordable.”

●“Results from the RAND HIE and subsequent studies, however, suggest and important caveat to that
general structure: there is a benefit to targeting consumer cost sharing to encourage appropriate care and
discourage inappropriate care. For example, reduced cost sharing can be cost effective for prescription
drugs for those with chronic illness. Indeed, it may even be optimal to subsidize preventive care for that
population in order to avoid more expensive care down the road.”