Good Morning
It’s a beautiful morning in Philadelphia. Sunny, clear skies, and the birds are chirping. It’s going to be hot today but for now it’s idyllic. Why I spent such a beautiful morning writing a post that defends the healthcare industry is beyond me, but I did, and it’s going out into the ether now, and please don’t be mad at me for saying my opinion on the internet.
There’s an easy story to tell about insurance companies and vertical integration – that it’s bad bad bad and these evil organizations are rent seeking cannibals weaved into the fabric of society by shadowy backroom deals with the federal government. Tons of people hate their insurance, and targeting these orgs is generally an easy win for talking heads. It’s also not generally been my experience having worked on both “sides of the aisle,” so to speak, in healthcare. There genuinely are really fine people on both sides. I can totally see why people get sketched out about integration, and at the end of the day these are businesses with both top and bottom lines they need to care about, but people with smiles and hopes and aging parents and grandparents run those businesses and often center their daily activities around ensuring what’s best happens for those people. This is just my experience, as someone who is not a rent seeking cannibal trying to drive up the cost of hospital visits, but I think it’s worth centering around.
I usually try to keep my finger on the NBER pulse in IO, particularly discussions of healthcare markets, so this paper stood out to me – Kakani et al.’s NBER working paper on strategic transfer pricing in Medicare Part D. The people who wrote this are really smart and I would be surprised if the story were wrong on its face. But I’m really not convinced about the mechanism as described, and that’s what I want to work through here.
The Paper
The argument is this: when the ACA introduced minimum Medical Loss Ratio (MLR) requirements for Medicare Part D insurers in 2014 – requiring that at least 85% of premium revenue go toward health care claims – it accidentally created an arbitrage opportunity for vertically integrated firms. Insurer profits are capped, pharmacy profits are not, and payments from an insurer to its affiliated pharmacy count as claims costs toward the MLR numerator. So if you’re a conglomerate that owns both an insurer and a pharmacy, you can inflate the transfer price from the insurer to the pharmacy, book it as a claims cost, keep your MLR above 85%, and preserve total profits within the corporate family. The paper calls this “strategic tunneling.”
They test it with a difference-in-differences design using Medicare Part D claims from 2010-2016, comparing price changes at affiliated versus unaffiliated pharmacies before and after MLR implementation. The headline result: prices at affiliated pharmacies increased by 9.5% per claim relative to unaffiliated pharmacies after 2014. The dose-response is what really gets your attention – plans with MLRs below 90% in 2014 (closer to the binding 85% threshold) moved prices about six times more than plans comfortably above it. They estimate roughly $1.2 billion in additional Part D spending from 2014-2016, with about 25% of that borne by parties outside the insurer-pharmacy conglomerate – mostly the federal government through reinsurance payments.
It’s a clean paper and the result is robust across a bunch of specification checks. I believe the prices moved, just not sure I buy that it was a centrally-driven unilateral decision to move them.
The Mechanism Problem
The entire “strategic tunneling” mechanism implies someone made a strategic decision – that there’s a person, or at least a decision-making node, somewhere in the corporate structure who looked at the consolidated P&L and said “increase transfer prices to the pharmacy arm by X percent.” I genuinely don’t know if that person exists, and I’ve spent enough time adjacent to how these products actually get built to be pretty skeptical.
Think critically for a minute about what an insurance company is trying to do on its face: premiums come in, claims come in, payments go out. They want to maximize the ratio of premium to outgoing payments. Now add the wrinkle of federal programs like Medicare Stars, designed to ensure insurance companies’ incentives are aligned to member outcomes. Now add the wrinkle that your costs cannot fall below 85% of your operating revenue, and you’ve got finance’s seat at the table. There are hundreds of staff and external brokers trying to get members enrolled every year – so sales has a seat at the table, trying to ensure the products they’re selling are aligned with member demand. Clinical has a seat at the table, ensuring the offerings are aligned with outcomes. All of these different groups with occasionally competing, occasionally overlapping interests are in the room deciding how a product should be built. The final decisions around product design are the result of hours if not days of debate with supporting empirical evidence between these parties.
Insurers are never asking “what does the PBM want” – it’s “what does our org within the insurance company need for our plans to be successful?” I would be really surprised if once a year Jim from the PBM sat in a room with a bunch of Medicare heads who run the business day in and day out and don’t know him at all, just to lobby for their needs in product design. The same holds for the pharmacy side. So does it seem likely that the output of all of that is overwritten after the fact by some central org? Why pay for these groups and subsidize this time in the first place? I would think CVS would have been chargeing more than four dollars a pack for Pokemon cards in the 2026 Christmas season while Target was completely sold out if they were as on top of pricing as the story would suggest.
Half of the problem with the proposed mechanism is coordination amongst many entities and leaders with different and often competing priorities. The other half is accurate MLR forecasting. At minimum, a central agent executing this strategy would need to know the formulary tiers for each plan prior to their own submission to CMS – so if plan design conversations kick off in Q2, that information would need to flow somewhere before the PBM submits their own data (would imagine on similar timelines). They’d also need a really tight estimation of the subsequent year’s MLR, which is really the part where I have a hard time with this. MLR forecasting is not easy. The resultant projections might be shared with the corporate finance function, but it’s not especially likely that the PBM or pharmacy org retain a point of view on the insurance arm’s MLR, and even less likely that the timelines they’d have them on would be sufficient for simultaneous pricing submissions. And if the MLR forecast is wrong it can have an absolutely devastating effect on the stock even if you don’t juice it! Look at 2023, when older people started going back to the hospital more – MLRs skyrocketed and a bunch of plans freaked out, and the subsequent 2024 reporting caused a significant downswing in equity markets for the companies with high MLRs. You need to have a really good understanding of where you’re going to land in a year if you want to be sneaky – and also not tank your stock. There’s actually a disincentive to implement this kind of thing exactly because you could do both at once.
What I Think Is Actually Going On
A gentler reading might be more like: vertical integration creates pricing distortions that interact badly with MLR regulation, and we can’t really tell from claims data whether anyone planned it that way. It doesn’t jump off the page as much but it’s probably more honest about what a difference-in-differences on administrative claims can actually establish. The empirical strategy is great at telling you that prices moved and when – not so much on the who and why.
I wouldn’t consider myself a staunch defender of our healthcare system who thinks it needs no change – and most of the leaders in the industry feel the same. But there’s a distinction between “prices are high as a result of decades of misaligned incentives that we’re actively working on fixing with things like value-based contracts (which maybe don’t work as well as we’d hope) or active care management, but change is slow and policy often has unintended consequences” and “these healthcare conglomerates are running a shell game, juicing prices by manipulating MLR.” I could be totally naive here, but I’ve never met anyone in all of my time on either the insurance side or the provider side who says “what I really care about is the PBM’s bottom line.” And maybe that’s bad! Maybe conglomerate-owned insurers should be more cognizant of margin and cash flows at the corporate level – but the way legal and compliance is designed almost intentionally ensures single-lane focus by any of the involved organizations. I’d argue that most of the people who suspect the brand of malfeasance suggested in the paper would prefer the structure as it is today to the trad micro \(\pi\)-maxxers, but are likely too far removed from the industry to be cognizant of the gap between ‘is’ and ‘should be’.
I would reiterate that the people who wrote this are really smart – albeit not especially charitable, and potentially have their own incentives to overstate their findings to drive up eyeballs and cites – and I am pretty dumb and likely fairly naive. It’s a well-written and well-designed study, but as one of the bright eyed and bushy tailed in the industry who wants the best for my aging family, I have a hard time believing that the people I engage with every day are a buncha sneaks who are up to no good.