CNBC got a quote from Aetna on their Pennsylvania strategy:
“We don’t discuss our performance on a state level,” Aetna spokesman T.J. Crawford told CNBC after the insurer’s profitable Pennsylvania picture was first highlighted by the website Balloon Juice on Wednesday.
Crawford declined to say whether Aetna’s plans were profitable in Pennsylvania this year, or identify which, if any, other states where the insurer is exiting were profitable in 2015 and for the current year.
He did note, however, that “past results are not indicative of future performance. Crawford also pointed out that 55 percent of Aetna’s Obamacare customers in 2016 are new customers, and that as a group they tend to have higher medical costs.
That is a very odd statement. It says a lot without saying as much as it looks to be.
Past performance is definitely not indicative of current or future performance. However when I have participated in budgeting exercises for new products, new populations or new geographic areas, we’ve always used either relevant past experience or proxy experience to attempt to budget.
Proxy experience, expert consultants and goat liver reading is not perfect but it is how we budgeted for Exchange in 2014. It is not how my company budgeted for Exchange for 2017.
Instead, we attempted to use as much actual experience as possible to project future expenses. Members were split into different streams by age, gender, geography and past insurance history. People who had multiple years of coverage through Mayhew Exchange had different expense profiles than people who had a prolonged stretch of non-coverage before they came on exchange. The 2015 not previously covered population looked a bit like the cohort who first got coverage in 2016 who also looked reasonably closely like the people who first got coverage in March 2014.
Projections can blow up. But what blows up projections are either completely unknown shocks, assumptions that prove not to be true or variance that consistently falls in a non-random location. I understand that.
When Aetna was running their pricing models for the Pennsylvania rate filings for the 2017 policy year, they had decent data. They had 99% of the claims for their 2015 policy year, 90% or more of their claims for their January and February 2016 policy months, 70% of their claims for March, and 50% of their claims for April. If their newly covered populations were going to be unexpectedly high cost utilizers, it would have shown up in their claims data. It might not have completely been there as more claim months and more importantly, more run out for incurred but not reported (IBNR) to go through the system could show catastrophic claims but this is very odd if there was an unanticipated shock in policy year 2016 that only became apparent between June 1 and August 1 when the previous relevant experience period had data that allowed their model to spit out profitable pricing.
Is it plausible that new claims data dramatically changed their future profit projections?
Yes.
Is it likely?
No.