The Affordable Care Act relies on risk adjustment to make insurers agnostic about the type of people that they insure. Risk adjustment moves money from insurers that cover a population that codes as having lower than average predictable medical costs and sends that money to insurers that cover a population that codes as having a higher than average predictable medical cost. A perfect risk adjustment system should make insurers absolutely indifferent if the next person who signs up is a 23 year old Iron Man triathlete or a 63 year old with congestive heart failure, diabetes and metastatic cancer. We don’t have perfect risk adjustment but we hope for good enough risk adjustment where the residuals are both tightly clustered around the average transfer amount and hard to a priori predict.
The Centers for Medicare and Medicaid Services makes a key assumption in how the ACA risk adjustment system works. They fundamentally assume that all claims are paid for by premiums. Their recent technical paper lays this out nicely on page 9.
The denominators of the risk and rating terms in the transfer equation express statewide average required revenue and allowable premium, respectively. The statewide average required revenue and allowable premium include the same component variables from the numerator multiplied by each plan’s share of statewide enrollment…
This is a reasonable assumption for most insurers most of the time.
Most insurers most of the time set their premiums to cover their costs on something like the following concept:
Premiums = Claims + Overhead + Profit/Surplus Accumulation/Required Reserves
Sometimes there is a miss. There was an overwhelming systemic miss in 2014-2016 on premiums being set too low as the insurers systemically thought the claims levels were going to be lower and they were responding to strong winners curse incentives of price linked subsidies. However, insurers figured out how to price the market and policy has stabilized so most of the time, most insurers are pricing premiums to fully fund claims.
If all insurers were doing this, then insurers that have a lot of people who code as sick and expensive will get risk adjustment transfers based on state wide premiums that are sufficient to cover state wide claims. There may be local discrepancies because some insurers have really expensive contracts with given hospitals OR an insurer gets horrendous bad luck and is covering someone who risk adjusts for half a million dollars but has a ten million dollar claim year but structurally things are mostly balanced.
HOWEVER there are a class of insurers that don’t care about profitability. Insurers that were initially backed by tech venture capital and some are now funded by an initial public offering (IPO) or other investment vehicles are currently profit indifferent. Instead they are attempting to get to scale. OSCAR, BRIGHT and FRIDAY are all trying to get big enough that there is a chance in hell that there are sufficient economies of scale and significant claims flow to run good analytics to actually become profitable in several years. Part of this strategy is to buy membership by pricing low. Their pricing strategy thus looks like the following:
Premiums + Investor Money = Claims +Overhead + Required reserves
Their premiums don’t cover expenses.
Their premiums are low relative to what the actually provide.
This is great for the federal government as it lowers federal subsidies.
This is pretty good for non-subsidized buyers as it lowers the price level.
This is not a bad deal for subsidized buyers who are buying only on price.
This is pretty good for other insurers whose covered populations code as healthier than state wide average. The lower state wide average premium versus the fairly priced counterfactual likely means that the net payables that they owe are lower.
HOWEVER, this is may be really bad for insurers with populations that code as substantially more costly and ill than state wide average. Risk adjustment is supposed to transfer sufficient funds to pay for claims based on state wide premiums that are supposed to be sufficient to pay for state wide claims. In some states, this is what happens. In other states where the buy scale/membership insurers are operating and lighting investor money on fire to lower premiums, state wide average premiums are artificially low and claims are being partially funded by external money.
If this is happening, then we should expect insurers that are covering a disproportionate share of the risk and are not getting fully compensated for that risk to change strategies. Instead of being mostly risk agnostic, they become risk sensitive which means they have a strong motivation to compete on their ability to avoid certain sub-classes of risk where the risk adjustment transfers in the current reality make some people with significant health insurance conditions unprofitable to cover who otherwise would have been break even or profitable to cover. If these cohorts are reasonably predictable, insurers should respond by decreasing network breadth and depth as well as increasing prior-authorizations and prescription tiering as well as designing benefits that put more cost sharing on the highest utilizers.
I’m not sure what the fix is, or if there is a need for a fix as I don’t know if this is a Top 5 problem or a Top 50 problem or a Top 500 problem. But I think this is a problem.
Victor Matheson
Tech bros losing a ton of money and therefore subsidizing health care for the rest of the country because they think they are smarter than they are is not a Top 5, Top 50, or Top 500 problem for me. In fact, that might even be a feature not a bug.
Keaton Miller
My guess is this is second order relative to other gaming of the risk adjustment system, but would be interesting to investigate.
I don’t know what data you have access to, but a first-pass test: treat local entry of these investor-backed insurers as exogenous (for the time being – there are a couple of instruments that could work) and see how other insurers in the local markets change their products relative to changes in their products in other places.
David Anderson
@Keaton Miller: Let’s chat off line.
David Anderson
@Victor Matheson: In the context of Medicare Advantage where there are no first order effects of artificially low premiums on selection/screening, I agree with you.
In the context of zero sum insurer premium funded risk adjustment in the ACA, there are negative effects off-loaded to both other insurers and sick enrollees