Summarizing the AHCA CBO score

The Congressional Budget Office (CBO) released their analysis of the American Health Care Act (AHCA).  The AHCA is the bill that passed the House that is a combination repeal of significant elements of the Affordable Care Act, partial replacement of some of the health care provisions, and a major tax cut.  The CBO scored a draft of the bill in March before it was pulled from the House floor.  They did not have the opportunity to score the amended bill that passed the House.

• Medicaid is still getting changed from an entitlement that is responsive to changing needs to a block grant
• 23 million people will lose coverage compared to current law projections
• The MacArthur/Upton waivers are expected to destroy the individual markets that cover 15% of the country
• Most of the premium decreases are due to older and sicker people being priced out of the market
• Pre-existing condition protection is effectively destroyed by splitting the risk pool.

The MacArthur/Upton waivers were the amendments that significantly changed the bill.  These waivers allow states to fully opt out of the insurance regulations of the ACA and allow for full underwriting of health premiums.  The CBO believes that these incentives will split the pool and make it virtually impossible for individuals with expensive illnesses to be able to afford the premium.

The AHCA is a policy choice that will, in some states, effectively restore the 2009 status quo in the individual insurance markets even if there is language that prohibits denial of offering a plan due to health status.  It will not contain any ability to make that offered plan affordable.  It is a de facto underfunded high cost risk pool instead of a de jure denial of coverage.

Network construction in an underwritten world

If you loved credit card companies and how they all were headquartered in South Dakota to take advantage of their very lender friendly laws before the passage of the CARD Act, you will love the health care vision of Rep. Jim Jordan.

Jim Jordan lays out his understanding: allowing people to buy across state lines means if one state guts pre-existing conditions…

— Matt Fuller (@MEPFuller) April 26, 2017

The vision is that the House Freedom Caucus/MacArthur Amendment is attached to the AHCA. The AHCA passes the House and Senate without major modification. At least one state opts for the waiver to allow for medical underwriting and gutting essential health benefits by replacing a federal standard with a state standard. A minimal essential health benefit package could still include prescription drugs but only require generic drugs and very common, low cost brand drugs. Specialty drugs including most chemotherapy agents, cystic fibrosis treatments, Hep-C cures and coagulation disorder treatments could be excluded from the minimally required list.

A second bill would also be passed that would allow for insurance to be sold across state lines. And that gets us to the individual health insurance market looking like the pre-CARD Act credit card market. Most states may still require guarantee issue and community rating but the states mandating these restrictions either will not have any in-state insurers offering products in the individual market or their locally regulated individual market effectively work as a high risk pool.

Matthew Fiedler at Brookings has a good analysis on the race to the bottom in an opt-out state:

In brief, healthy people would have a strong incentive to “opt out” of the community-rated pool and instead pay a premium based on health status. With healthy enrollees opting out of the community-rated pool, community-rated premiums would need to be extremely high, forcing sicker individuals—including those with continuous coverage—to choose between paying the extremely high community-rated premium or being underwritten themselves. Either way, people with serious health conditions would face prohibitively high premiums. As a result, community rating would be eviscerated—and with it any meaningful guarantee that seriously ill people can access coverage.

And if a state elects to operate a waiver in an environment where insurers can choose the state of regulation, that state will effectively gut community issue across the country.

But.. but… but… what about networks?

That is a common argument as to why selling across state lines would not be attractive. In the current world of  guarantee issue and community rating, this is a strong defense. Networks are tough to assemble and expensive to build. We know there is a chicken or an egg problem. Large membership is needed to get good provider rates.  Good provider payment levels are needed to offer attractive premiums that leads to large membership numbers.  A new insurer trying to move into a new state has to build a network. And it has to build a network by either going super skinny or by being willing to lose significant money for several years to buy membership.

But that is under guarantee issue/community rating rules. Networks are not a blocking force for cream skimming carriers.

If an insurer wants to expand out of its home region in Rep. Jordan’s vision, they can either build a network organically or they can rent a network. Rental networks are very common. They are how regional carriers offer national emergency room coverage. They are how smaller carriers offer very high end specialty care. The provider are paid at a very high level. Some rental network contracts are full usual and customary, others are full billed charges, some offer a discount on one of those two benchmarks and others are 500% of Medicare. These are expensive networks where regional carriers work very hard to minimize the number of claims paid to that network.

If a minimally viable network can be rented even at an extremely high per unit rate, and the plan can medically underwrite to only offer coverage to people who will never use the skimpy, practically inadequate, high cost network, this works as a business model.

If an individual has a complex medical condition during the contract year, the insurer has significant claims expense but since the network is extremely unattractive to individuals with complex care needs, the one time catastrophic expense will leave the plan at the next open enrollment if they can afford to do so or if they are in a policy that does not have guarantee renewability.

This is effectively the Assurant business model from the pre-ACA status quo. They aggressively underwrote policies to only include healthy people, they offered very low rates and access to a very expensive to them network and seldom payed a claim as their covered population just did not use services.

Networks are not a barrier to entry for carriers that think they can aggressively underwrite.

Cost clustering and concentrating risks in state high cost risk pools

Let’s imagine a hypothetical high cost risk pool of the top 1,000 individual claim years in the country.  The average claim will be $5,000,000 for the year. Let’s simplify things and say 990 are randomly distributed by population and 10 are a non-random cluster that we can insert into any state at any time.  The first run through is with fifty one state (and DC) based high cost risk pools.  We’ll look at two states, California and Wyoming, for this run.

California has about 10% of the population.  California should expect to see 99 people in this hypothetical pool plus an expectation that one of the ten non-random people would be expected to be in California.  Their expected high cost risk pool budget is $500 million.  Now if all ten of the non-randomly clustered people are in California, they increase the expected pool costs by 9%.  California is big enough and rich enough that a surprise $45 million dollar medical expense does not destroy their budget.

Now Wyoming should expect to see between 1 and 2 people qualify for the high cost risk pool.  Let’s assume the Wyoming state government is very cautious and they allocate $10 million for the high cost risk pool.  That works great in a normal year.  But if the travelling roadshow of catastrophic medical expenses arrive in Cheyenne, the state is now on the hook for twelve qualified individuals.  They are 500% over budget now and the state budget is underwater.

This thought experiment is amazingly unrealistic.

Even if we are to assume that extreme medical cost cases are randomly distributed, we should expect several states to be surprised at the number and expense that they face as Pennylsvania could reasonably expect to see anywhere from 45 to 51 qualifying individuals from the scenario above in any given year just do to random chance.

More importantly, we know that diseases are not randomly distributed.  My ongoing freak-out about Zika is based on the fact that this is a concentration of very high need and high cost individuals on states with low Medicaid funding.  Genetic disorders are tightly clustered due to both the combination of most people live near their families rather than being randomly distributed and localized clusters of diseases have led to local medical-industrial clusters of medical knowledge and treatment.  For instance, maple syrup urine disease is a common genetic disorder among Amish families, so there is a good deal of knowledge on treating that disease clustered in Lancaster County, Pennyslvania and Holmes County, Ohio.  Sickle cell disorders are overwhelmingly a disease of African Americans, so it is more common in Mississippi than Montana.

From a financial perspective, there is a chance that there is enough sample size that although one state will have more of one genetic disorder it washes out as another disorder it is light in is dis-proportionally prevalent in another state so the cash flows balance out.  That is an empirical question that I don’t know enough to answer.  But even if genetic disorders balance out, localized outbreaks like Zika won’t balance out.

State based high cost risk pools would remove some of the falling knife incentives that I described in Iowa but they will be underfunded and overwhelmed at times of high need.  National level pooling is far more efficient and effective.

Rebranding reinsurance in the AHCA

The House Rules Committee introduced a new amendment to the AHCA this morning. What does it do and does it matter?

SEC. 2205. FEDERAL INVISIBLE RISK SHARING PROGRAM.
(a) IN GENERAL.—There is established within the Patient and State Stability Fund a Federal Invisible Risk Sharing Program (in this section referred to as the ‘Program’), to be administered by the Secretary of Health and Human Services, acting through the Administrator of the Centers for Medicare & Medicaid Services (in this section referred to as the ‘Administrator’), to provide payments to health insurance issuers with respect to claims for eligible individuals for the purpose of lowering premiums for health insurance coverage offered in the individual market.

(b) FUNDING.—
(1) APPROPRIATION.—For the purpose of providing funding for the Program there is appropriated, out of any money in the Treasury not other- wise appropriated $15,000,000,000 for the period beginning on January 1, 2018, and ending with December 31, 2026.

These two paragraphs are the meat of the four page amendment.

So what does it do. First it renames reinsurance to “Invisible Risk Sharing Program.” Secondly, it authorizes HHS/CMS to inject $15 billion dollars of non-premium dollars into the claims payment of the individual market over nine years.

Renaming reinsurance to Invisible Risk Sharing Program is a nothingburger. The key is the injection of money from outside of the premium pool into claims payment. Assuming that HHS elects to use this money as if it is reinsurance, that means HHS will use general revenue to pay some portion of very high cost claims. Since the money is coming from outside of the premium pool, it will lower the premiums paid as the premiums no longer to have to cover full claims expenses.

The question is by how much?

Not much is the answer. The appropriation is $1.67 billion dollars per year. In 2015, the ACA spent $7.8 billion dollar on reinsurance. This was approximately 10% of the total premiums collected so reinsurance under the ACA reduced premiums by 10% in 2015 compared to what they otherwise would have been. So a very rough guess is a $1.67 billion dollar externally funded reinsurance pool would lead to a 2% reduction in premiums.

So we get a rebranding of a basic insurance concept and a 2% reduction in premiums. Not much of any real substance actually is happening in this amendment.

Molina’s Death spiral model and the AHCA

Molina Insurance, a fairly successful Exchange carrier that has usually been profitable on and off Exchange, issued a statement yesterday afternoon concerning the House Repeal bill. The statement is ugly and it illustrates a very important point of the Republican bill. It will create an on-Exchange death spiral in many counties.

Just out: As a practical matter, this renders the #ACA replacement D.O.A.https://t.co/YdE0H1fiOt pic.twitter.com/jjIujan2Cy

— Andy Slavitt (@ASlavitt) March 7, 2017

So what is the model that leads Molina to believe the Exchanges would see mass disenrollment and high price hikes under the Republican plan? It is a combination of the flat age based subsidies and the late enrollment penalty being a proportion of premiums. Some markets and counties would not be affected. Portions of Texas and Pennsylvania for instance would see 21 year olds be able to buy Bronze plans at zero cost. Those counties should have healthy risk pools. But risk pools are not county specific. They are state specific single pools. High cost counties will have a very different dynamic.

High cost counties with a flat age based credit will see young people face significant out of pocket monthly premium expenses if they maintain continual enrollment. A 21 year old is highly unlikely to need much care over the course of a year so most healthy 21 year olds will leave the market. They pay no penalty at the time of their decision to leave the market. And they will make the same decision the following year assuming no new information about their health status arises. Now a 22 year old can jump into the market with a 30% penalty which they will only pay if they know that they are going to be expensive in the following year. The ACA Exchanges average about 34% enrollment under age 35. This system will dramatically reduce enrollment as a number and as a percentage of enrolled for Under 35. The composition of the young enrollment will also be significantly sicker in high premium counties.

On the other hand, moderately expensive but still profitable 60 year olds know that their uncertainty zone for future costs are higher. They could have a year that looks a lot like the current year or they could have a catastrophic year. The odds of a catastrophic year are much higher for a 60 year old than a 21 year old. Further more the penalty for being unenrolled and then enrolling in the future period is much larger. For somewhat expensive but profitable members, the 30% penalty is much more stringent than the current $650 or 2.5% of income penalty. At the same time, the really healthy members who have low utilization at age 60 will drop coverage as the relative costs of the 30% one time late enrollment penalty is less than the cost of premiums for the current year.

Very low cost enrollees will flee unless there is a minimal cost plan out there.  The risk pools will become very old and very sick very quickly.  That is how Molina is modeling their future.

AHCA impact in North Carolina

North Carolina is a high cost state for health insurance.  Under the Affordable Care Act, people who receive subsidies on Healthcare.gov are shielded from price increases because the subsidy is tied to the cost of the second least expensive Silver plan on Exchange and the individual pays a fixed amount dependent on their income.  The personal contribution amount is fixed based on the person’s income defined by the Federal Poverty Level.  That means a 21 year old who is subsidized will pay the same post-subsidy premium to the insurer as a 64 year old with the same income.

The American Health Care Act (AHCA) bill that was released on Monday night changes the subsidy formula.  Subsidies are no longer tied to the cost of insurance or the individual’s income.  Instead any qualified individual who makes less than $75,000 per year receives a fixed subsidy amount determined by age.  A 21 year old receives a $2,000 subsidy.  A 64 year old receives a $4,000 subsidy.  In the ACA premiums are allowed to be three times higher for a 64 year old compared to a 21 year old.  In the proposed AHCA, premiums are allowed to be five times as high for a 64 year old than for a 21 year old.

Since the subsidy grows far slower than the premium, this means the 64 year old, for a given deductible, will pay far more for their coverage.  I’ve used the 2017 Exchange data to see how much a 60 year old in each county in North Carolina would have to pay after their subsidy every month to buy the least expensive Bronze plan currently offered.  Bronze plans tend to have deductibles of at least $6,500 with out of pocket maximums of $7,150. The circles are larger for counties with more enrollment as of 1/31/17.

60 year old residents in Nash County are the best off.  They would only pay $439 per month after the newly revised subsidy is applied.  Currently, someone earning $20,000 a year in Nash County would pay nothing for the least expensive Bronze plan and only $31 a month for a low deductible Silver plan.

60 year old residents in Bladen, Cumberland, Harnett, Hoke, Richmond, Robeson and Scotland counties are the worst off by this change.  They would pay $754 for a Bronze plan under the AHCA fixed age based subsidies.  Under the ACA, someone earning $20,000 a year would pay the same as a Nash county resident.  Someone who is age 60 and earning $40,000 a year would pay $110 a month for a Bronze plan and $283 per month for a Silver plan under the current ACA income and plan cost based subsidies.

High cost states with a large number of older residents will be significantly worse off under fixed age based subsidies.

There is one caveat. I’ve intentionally underestimated the costs for all counties.  I used the current 3:1 age band for premiums instead of the 5:1 premiums as that change is unlikely to survive in the Senate under current legislative rules.  If 5:1 premiums are used, all figures should increase by approximately 20 to 30%

Data: Enrollment data from Charles Gaba  sourced by Kaiser Family Foundation

Pricing data from CMS

Subsidy data from the legislative text

Distributional consequences of widening allowed Actuarial Value bands

One of the major proposals in the draft Center for Medicare and Medicaid Services (CMS) rule that was released on the 15th is to increase the de minimas allowed actuarial value band.  Currently the regulation allows a plan to be included in a metal band if it is within two Actuarial Value (AV) points of the target band for normal bands, and within a point in either direction for the targeted Cost Sharing Reduction Silver plans.  So that means a standard Silver plan which should be a 70% AV could be anywhere from 68% AV to 72% AV.  The proposed modification would allow for a plan to qualify for a band if it was no more than four points below or two points above the target.  A Silver plan would be anywhere from 66% AV to 72% AV.

All else being equal, a lower AV means a slightly lower premium.  It also means higher out of pocket spending for patients. But all else is seldom equal so things can get messy.

This has significant distributional consequences.  And these consequences are not entirely straightforward as the individual market is a complex market.  Let’s start looking at the easiest scenarios and then build in complexity.  We will need to divide the analytical units into four groups.  The vertical split of a 2×2 grid are people who either have met their out of pocket limit or have not incurred sufficient claims to meet their out of pocket limit.  The horizontal split is between people who receive premium tax credits that are keyed to the price of the second Silver and people who are not receiving premium tax credits and thus pay the full premium out of pocket.

We will only look at individual beneficiary consequences.

The simplest scenario to analyze is a market that has converged with multiple carriers.  Indianopolis, Indiana is a good example.  There are two carriers that currently offer Silver plans with 68% AV with similarly narrow networks and very similar pricing.  The strategic logic of that situation will have both carriers offer Silver plans that would be near 66% AV as soon as they could.

That produces a nice simple outcome matrix:

Individuals who are receiving subsidies and have significant claims that matched their out of pocket maximum under a 68% AV Silver are indisputably worse off. They will face higher cost sharing. If all cost sharing is from deductibles (an oversimplification), they will go from having a $4,400 deductible to a $4,850 deductible. They do not benefit from lower premiums as the federal government is the risk bearer and reward recipient of lower premiums as the subsidy formula is based on the federal government filling in the gap between the calculated individual contribution as determined by income and the cost of the second least expensive Silver.

Individuals who have not met their out of pocket maximum and are subsidized for a 68% Silver will still not meet their out of pocket maximum and their post-subsidy premium will not change. They are indifferent.

Individuals who are not subsidized and who meet their out of pocket maximum are almost always worse off. They have a small gain in lower premiums (2% drop in AV leads to a 2.4% premium drop on first estimate) but higher cost sharing. There is a small sliver of individuals whose costs above current cost sharing is less than the premium drop. But this is a sliver of people whose total costs are within $100 of the current out of pocket maximum.

The big winners of this change from a beneficiary point of view are individuals who are not subsidized and who are under the out of pocket maximum. They have no incremental cost sharing and they have lower premiums. If the non-subsidized market is extremely price sensitive this will bring in more healthy individuals as prices will fall slightly.

This is the simplest scenario. This intuition should serve people well, but things will get complicated.

More on Advance Care Planning in Medicare

This article starts with the Death Panel nonsense, reignited by a Republican Party official in Florida last week, but the second half of the piece is a fairly good discussion of the issues around Advance Care Planning that the Medicare program began paying for January 1, 2016. Some of our recent work shows that the comment period held during 2015 for this pending change was not particularly controversial.

Forget about the nonsense, and read the discussion of the policy reality at stake.

A snippet:

The CMS rule requires no specific diagnosis and sets no guidelines for the end-of-life discussions. Conversations center on medical directives and treatment preferences, including hospice enrollment and the desire for care if patients lose the ability to make their own decisions. The conversations may occur during annual wellness exams, in separate office visits or in hospitals. Nurse practitioners and physicians’ assistants may also seek payment for end-of-life talks.

End-of-life conversations have occurred in the past, but not as often as they should, said Paul Malley, president of Aging with Dignity, a Florida nonprofit. Many doctors aren’t trained to have such discussions and find them difficult to initiate.

“For a lot of health providers, we hear the concern that this is not why patients come to us,” he said. “They come to us looking to be cured, for hope. And it’s sensitive to talk about what happens if we can’t cure you.”

Further on the impact:

Proponents of advance care planning cheered evidence of the program’s early use as a sign of growing interest in late-stage life planning. Being able to bill makes a difference, Malley said.

The new reimbursement led Dr. Peter Sutherland, a family medicine physician in Morristown, Tenn., to schedule more end-of-life conversations with patients last year.

“They were very few and far between before,” he said. “They were usually hospice-specific.”

Now, he said, he has time to have thorough discussions with patients, including a 60-year-old woman whose recent complaints of back and shoulder pain turned out to be cancer that had metastasized to her lungs. In early January, he talked with an 84-year-old woman with Stage IV breast cancer.

“She didn’t understand what a living will was,” Sutherland said. “We went through all that. I had her daughter with her and we went through it all.”

Exchange Actuarial Value spreads and de minimis exploits

The ACA simplifies insurance by restricting the actuarial value of the plans that can be sold.  Bronze covers roughly 60% of expected pool costs, Silver covers roughly 70% of the expected pool costs, Gold covers roughly 80% of the expected pool costs and Platinum covers roughly 90% of the projected pool costs.  Cost Sharing Reduction (CSR) adds three more layers of coverage at roughly 73%, 87% and 94% of the projected pool costs to be covered by the insurer.  

However the law allows for the Secretary of Health and Human Services (HHS) through the Center for Medicare and Medicaid Services (CMS) to make rules that allow for de minimas variation from the targeted actuarial value.  CMS has had a consistent rule that a 2% variation in actuarial value is the maximum allowable wiggle room for a policy to be included in a band.  Therefore a Silver policy with no CSR could cover anywhere from 68% to 72% of the pool’s expected cost.  

This is an area of exploitation for premium tax credit benchmark strategic manipulation. The benchmark for all premium tax credits is the cost of the second least expensive Silver plan.  A large spread between the least expensive Silver plan and the benchmark Silver plan advantages subsidized buyers.  Most of my analysis of the Silver Gap strategy space has focused on network manipulation as an obvious source of creating a usable gap between the first and second least expensive Silver plans in a region.  My background is in network and provider configuration so this was my prism.  

However, the de minimis actuarial value variation in a band is another opportunity for aggressive gap creation.  In this scenario, a carrier with a low cost network and product type can use two different cost sharing structures to create two plans.  The first plan would have a minimal actuarial value of 68% while the second, benchmark plan would have an actuarial value of 72%.  

UPMC Health Plan in Pittsburgh is an example of the foregone opportunity.  Plan 16322PA0050104 is the least expensive Silver in zip code 15219.  It has an actuarial value 71.53%.  Plan 16322PA0050103 Is the benchmark Silver.  It has an actuarial value of 71.5%.   These values are from the 2017 PUF URRT Worksheet 2.   Both plans are at the high end of the allowed actuarial value range.  They are both using the same network and the same EPO plan type.  The extreme similarity means the premiums minimally differ.  Healthy, low income individuals will not be seeing a significantly better deal on the least expensive Silver after the subsidy than the same value proposition for the least expensive Silver.  

If UPMC Health Plan elected to offer a their lowest priced offering a 68% Silver** plan with a comparatively higher cost sharing structure and concurrently lower actuarial value  and then offered 16322PA0050103 as the benchmark Silver, the almost four percent actuarial value spread would, all else being equal, lead to 4% to 5% reduction in premium for the least expensive Silver plan compared to the current case.  The lower premium will mildly advantage most non-subsidized buyers as their option space will have expanded.  It will significantly advantage low income, subsidized buyers who previously were marginally deciding to not buy at the current price points.  These individuals are highly likely to be comparatively healthy and profitable for a carrier and their decision to opt out of the market and pay either the individual mandate tax or claim an exemption leads to a less healthy risk pool.  

Carriers who have a dominant position at the lowest price Silver Plans without proximate competition due to either their sole carrier status or the lack of low priced, narrow network competition should seek to offer a significant actuarial value spread as allowed by the de minimis variation in order to improve the risk pool by including more comparatively healthy and low cost subsidized buyers.  

** Plan ID 76179IN0110008 (68.1% AV) and 54192IN0040001 (68.8%)   are good examples of 68% AV Silver plans.

 

Paying college players-cost of attendance

The Raleigh, N.C. News and Observer has a front pager today on the changes that are coming to college sports regarding paying players. There are so many issues, and so many questions, but a key one is understanding a key University concept, “the cost of attendance (COA).” Duke University’s COA for 2014 is shown below:

Historically, the NCAA has prevented University’s from covering the full COA via an athletic scholarship, but the ruling in the Ed O’Bannon case stated that this could not continue. An athletic scholarship covers Tuition and mandatory fees, as well as room and board, but not other expenses. However, personal expenses such as books are not covered by an athletic scholarship (under Duke’s need based financial aid system, it is possible an athlete could get aid for these expenses, depending upon family income).

The News and Observer article terms the covering of what above is called “books and personal expenses” as a stipend. Note the differences in the Triangle of the costs of “books and personal expenses.”

• At Duke, $3,466
• At N.C. State, $3,828 for both in and out of state students
• At UNC, $4,382 for in state, and $6,118 for out of state

These differences reflect policy choices by the schools, not just for their athletic programs. Merit based scholarship programs such as the B.N Duke and A.B. Duke don’t cover the full cost of attendance, but Duke’s need based financial aid program does (potentially, depending upon income) cover the full cost of attendance. At Duke there is also sensitivity about how large the “full cost of attendance” figure is, so there may be extra incentive to keep the “books and expenses” figure as low as possible.*

Calculating the full cost of attendance figure at a University is a highly idiosyncratic process, and there are many competing incentives. A quick look at the spread in the “books and expenses” aspect of Duke, UNC and NC State’s demonstrates. At UNC and NC State, around 8 in 10 undergrads have to be from the State of North Carolina. UNC shows a different figure for out of state as compared to in state, while NC State does not. Duke has about as many undergrads in a given class from California as there are from North Carolina, so it is hard to imagine how this amount in the expenses component at Duke could be lower than the in state figure at UNC.

I am not saying there is anything nefarious going on, but it is also hard to see how these figures at these three schools located within 30 miles of one another represent an externally validated cost to students. The Ed O’Bannon ruling is bringing attention to the cost of attendance figure for athletes, but this figure is important for all students at a University’s as a whole, and is set amidst a sea of many competing influences. The process of setting these figures seems murky, especially to families trying to compare across University’s. Because of increased attention due to athletics, the bright lights are coming.

*Duke has robust need based financial aid; the diddy I have learned is that a year at Duke costs $90k, we charge $63k and collect $30k. Around half of the undergrads pay full freight, and the other half is sliding based on income.