Assessing the Exchanges

By  Yuval Levin

Thursday, October 17, 2013

Over the last few days, I have spoken in some detail about the state of the federal Obamacare exchanges with several officials of the Center for Medicare and Medicaid Services (the HHS agency that is running the exchanges), and with a number of reasonably well placed insurance company officials in Washington. The picture they paint of how the rollout of the exchanges has gone is similar in its broad strokes to what has emerged in other reports in recent days, so I don’t think I’ll be breaking much news here, though some of the details have (I think) not been reported. For what it’s worth, I offer below the basics of what they had to say and some reflections on its implications. This is a long post, with apologies, but I thought some of the particulars would be of interest.

First, a couple of words of caution: I do not present this as a broad sampling of people involved in the rollout or a comprehensive overview. The CMS people I spoke with are people I know (from having worked on health policy for some time in and out of government), and who in turn know me, which means they know that (unlike all of them) I am an opponent of Obamacare. This may have led them to tell me some particular things and not others, or it may not—I have no way of knowing. The assessment below summarizes conversations with five CMS officials and three insurance-industry insiders, all of which took place on the understanding that I would publish such an overview, without their names attached. (I approached several additional people at CMS who politely declined to discuss the exchanges on these terms.) The CMS officials are all career agency personnel, not political appointees; they are fairly senior people but not so senior as to be routinely privy to political discussions. All are involved in the exchange project in different ways. What they see is the nitty gritty operations of the program. But they are policy and management people, not information-technology experts. 

This latter point turns out to be quite important. The reaction of these individuals to what has happened in the last two weeks is the reaction of people who are coming to realize that their expectations and understanding of web development were mistaken. They believed (as I did too, I admit) that whatever technical problems the exchange sites encountered at first could be cleared up quickly and simply once things got going—that the contractors developing the websites could just respond to problems on the fly, as they became apparent. It is now increasingly obvious to them that this is simply not how things work, that building a website like this is a matter of exceedingly complex programming and not “design,” and that the problems that plague the federal exchanges (and some state exchanges) are much more severe and fundamental than anything they imagined possible. That doesn’t mean they can’t be fixed, of course, and perhaps even fixed relatively quickly, but it means that at the very least the opening weeks (and quite possibly months) of the Obamacare exchanges will be very different from what either the administration or its critics expected. 

Some blame the contractors involved for not being upfront about the potential for such fundamental difficulties, but some say the contractors did offer warnings, and especially that the contractors believed the time they were given for development was totally inadequate. It seems clear, though, that the administration was not warned to expect quite what has happened here, and was not prepared for it.

What has happened, at least so far, presents itself in several layers. One key problem, which to date has been the most prominent in public, has to do with a late-in-the-game decision to require users to go through a complex account-creation process before even reaching any coverage options. Administration officials apparently went back and forth several times on this question, and the ultimate decision required the creation of a series of patches over an already developed site in a very short time. Most of the problems people have faced so far are a function of that decision, and have had to do with creating user accounts and so getting through the very first steps involved in purchasing coverage. Some journalists and analysts have speculated that this decision was made in order to prevent people from seeing premium costs before they could also see any subsidies they might be eligible for, so that the shock of higher prices could be contained and so that simply curious observers and journalists couldn’t get a picture of premium costs in the various states. This explanation strikes me as plausible, and it struck several of the people I spoke with as plausible, but none of them could confirm it. It may be true, but it’s surely not the only possible explanation. Whatever the cause, that decision has created crippling problems that are still largely unresolved.

Many of these problems were reported in an October 12 New York Times story that detailed some serious dysfunction in the development process. The people I spoke with all confirmed that nearly all of the details in the story were correct, though several of them did strenuously deny one claim—that CGI Federal, the biggest contractor involved in building the site, was not provided with the information it needed to start writing code for the site until the spring of this year. This detail in the story aroused some shock and surprise among outside web developers, and these CMS officials say it’s just not true.

The people I spoke with did all confirm the importance of one other detail in the Times story: that CMS did not hire a general contractor to manage the exchange project but handled that overall technical management task itself. None of the people I spoke with wanted to get into how this decision was made or at what level, but all of them agreed that it was a very bad idea and was at the core of the disaster they have so far experienced.

The problems people are now facing with the basic interface have taken up most of the time that CMS and its contractors have devoted to troubleshooting so far, and although things have improved a little on this front quite serious problems remain. But there are very serious problems beyond that, which are more like the sorts of problems people were predicting before the launch: database problems at the nexus of several federal and industry data sources. The federal data hub itself is so far doing reasonably well at its basic tasks, and that has come as a relief to CMS. But some of the site functions that rely on the hub, both in the federal exchanges and a number of the state exchanges, remain highly problematic. The calculation of subsidies continues to fail tests, and it’s pretty clear that some actual consumers have made actual purchases with bad information, which will become apparent to them when they get their first bills. If the interface problems are addressed and the volume of purchases increases, this calculation problem could become a huge concern.

Meanwhile, the back-end communication between the exchanges and the insurers has been terrible, as is increasingly being reported. The extent of these problems has also been a surprise to CMS, and here too an increase in volume if the user interface issues are solved could lead to huge problems that would be very difficult to correct. CMS officials and the large insurers thought at first that the garbled data being automatically sent to insurers must be a function of some very simple problems of format incompatibility between the government and insurer systems, but that now seems not to be the case, and the problem appears to be deeper and harder to resolve. It is a very high priority problem, because the system will not be able to function if the insurers cannot have some confidence about the data they receive. At this point, insurers are trying to work through the data manually, because the volume of enrollments is very, very low. But again, if that changes, this could quickly become impossible.

In a couple of ways, then, the severe user-interface problems at the front end of the federal exchange has actually had some advantages from CMS’s point of view, because by keeping enrollment volume low it has kept some other huge problems from becoming instantly uncontrollable.

But that low volume is mostly a very bad thing for Obamacare, of course, since the viability of the exchanges depends on a certain size and demographic mix which cannot be attained unless these problems are resolved very quickly. I couldn’t get enrollment numbers from any of the people I spoke with, but I was told that the uptake model that HHS built (using CBO projections) to predict how the exchanges would work made a low-end estimate that just under half a million people would enroll nationwide by October 31st, and that enrollment would then accelerate dramatically between November 15 and December 30th. The October 31 target, which was thought to be modest, now looks essentially impossible to reach, but their bigger worry is that period in November and December.

If the problems now plaguing the system are not resolved by mid-November and the flow of enrollments at that point looks like it does now, the prospects for the first year of the exchanges will be in very grave jeopardy. Some large advertising and outreach campaigns are also geared to that crucial six-week period around Thanksgiving and Christmas, so if the sites are not functional, all of that might not happen—or else might be wasted. If that’s what the late fall looks like, the administration might need to consider what one of the people I spoke with described as “unthinkable options” regarding the first year of the exchanges. 

All of the CMS people I spoke with thought the state-run exchanges are in far better shape than the federal system under their purview. But the insurers do not seem that much happier with many of those state exchanges. Back-end data issues seem to be a problem everywhere, and some of the early enrollment figures being released by the states are not matching up with insurance company data about enrollments in those states, which suggests a breakdown in communication that is only beginning to be understood. The insurers believe that only Nevada, Colorado, Washington state, and Kentucky have what could reasonably be described as working systems at this point. Still, there is no question that on the whole the states with state-run exchanges are in better shape than those with federal ones.

The tone of the CMS officials who spoke with me was a kind of restrained panic. Among the insurance company officials (who, I should stress again, work in the Washington offices of some large insurers, and so are basically policy people and lobbyists), there was much less restraint. The insurers are very, very worried about the viability of the exchange system—especially but not exclusively at the federal level.

One key worry is based on the fact that what they’re facing is not a situation where it is impossible to buy coverage but one where it is possible but very difficult to buy coverage. That’s much worse from their point of view, because it means that only highly motivated consumers are getting coverage. People who are highly motivated to get coverage in a community-rated insurance system are very likely to be in bad health. The healthy young man who sees an ad for his state exchange during a baseball game and loads up the site to get coverage—the dream consumer so essential to the design of the exchange system—will not keep trying 25 times over a week if the site is not working. The person with high health costs and no insurance will. The exchange system is designed to enable that sick person to get coverage, of course, but it can only do that if the healthy person does too. The insurers don’t yet have a clear overall sense of the risk profile of the people who are signing up, but the circumstantial evidence they have is very distressing to them. The danger of a rapid adverse selection spiral is much more serious than they believed possible this summer. They would love it if the administration could shut down the exchange system, at least the federal one, until the interface problems can be addressed. But they know this is impossible.

And they believe, as the CMS officials I spoke with do, that all of these problems will not be addressed immediately. No one wants to say how long it might take, and no one would share with me what estimates they might be getting from their contractors (whom they no longer trust anyway), but there has so far been relatively little progress and it seems like everyone involved is preparing for a process that will take months, not weeks. An extension of the enrollment period for coverage, now set to end on March 31, seems to be almost taken for granted. A delay of the individual mandate penalty—which effectively begins in the middle of February—is not thought to be a crazy idea (though the people I spoke with said they have not seen internal preparations for such a move at this point).

The nightmare scenarios, the “unthinkable options,” involve larger moves than that—like putting enrollment on hold or re-starting the exchange system from scratch at some point. No one seems to know how this could work or what it would mean, but everyone involved is contending with a far worse set of circumstances than they were prepared for. This is a major disaster from their point of view, not a set of glitches, and they simply do not know how long it will take to fix. They dearly want to see progress day by day, but they are generally not seeing it.

The fate of these sites is the fate of Obamacare, for reasons that may not be immediately obvious. Health insurance is highly sensitive to the integrity and robustness of the market in which it is sold: though we don’t often think of it this way, health insurance is a financial service, a protection against risk, so the nature and structure of a given insurance plan is highly responsive to the scope and the character of the demand for it at any given time. It is in this sense rather different from most consumer products. This means it is not possible to think of the exchange websites as just sites where products are sold, and to believe that the product is fine but the site has some glitches. If the site doesn’t work, the product cannot work, and the insurance market created by the law cannot be sustained. So a great deal is at stake here, and it now seems a great deal is at risk.

All of that said, I want to end with a caveat. The character of the conversations I had with these very knowledgeable individuals in the last few days reminded me of something: It reminded me of the daily intra-governmental video conferences and calls in the wake of hurricane Katrina in 2005. I was witness to many of those, as a White House staffer. What I saw in the first days of the disaster quickly fell into a pattern: local, state, and federal officials on the ground would report on what they knew directly—which was often grim—and then they would pass along information they’d heard but hadn’t gotten first hand, which was often much more grim but almost always ultimately turned out not to be true. Some of these stories went public (remember the shootings at the Superdome? They never happened). Some didn’t. They were often reported with a kind of detached authority that made them believable, and they were a function of living in panic amid an unbelievable situation over time.

Obviously what’s happening here is nowhere near that scale or significance, but for the people involved—for the officials in charge of running this system—this is a category 5 nightmare, and in a number of instances they traded in stories they’d heard from others which struck this outsider as basically impossible. They have been witness to problems in recent days that they would not have believed a month ago, and so they believe things about the extent and depth of their problems that may not be true.

The combination of these conversations over a week has therefore left me thinking that it may not be clear to anyone exactly how deep and lasting these problems will prove to be, which could mean they’re worse than they seem but could mean they’re not as bad as they seem. The technical architecture of the federal exchanges and to a lesser extent the state ones has been very badly screwed up. The problem may be so bad as to render Obamacare’s rollout impossible in practice at this point. But it may not be. And right now no one knows if it will or will not. My gut sense after listening to these insiders, for what little it’s worth, is that it’s not likely that the situation will prove to be much worse than it now seems, and it’s more likely that it will prove to be less bad than it now seems.

But I don’t know, and no one else does either. The administration believes it will be possible to roll through a difficult period, get as many people as they can into the system, and just hold out until things stabilize. The insurers are not sure this will be possible. Everyone involved is guessing.

For me, and for other critics of Obamacare, the problem with the law was never about these technical matters. I didn’t think the system wouldn’t work because the government couldn’t build a website, but because the basic health economics involved is deeply misguided and would take the (badly inadequate) American  health-financing system in the wrong direction. So these problems only seem like a prelude to other, larger problems. But Obamacare was also always going to be a test of the sheer capacity of the administrative state to actually do what it claims the authority and ability to do. At this point, it looks as though we may be witnessing a failure of the administrative state on a level unimagined even by its staunchest critics. We may be. But we’ll have to see.